{"id": "2606.14510", "title": "PepALD: Macrocyclic Peptide Generation via Autoregressive Latent Diffusion", "abstract": "Macrocyclic peptides are promising therapeutic candidates for intracellular targets, but their design requires simultaneous control over non-natural monomer chemistry, ring topology, membrane permeability, and target binding. Existing SMILES- or HELM-string generative models either operate in long atom-level sequence spaces or treat monomers as symbolic tokens with limited chemical grounding. We introduce PepALD, an Autoregressive Latent Diffusion (ALD) foundation model for \\textit{de novo} macrocyclic peptide generation. The model represents HELM monomers with structured chemical embeddings, generates each residue through context-conditioned diffusion in chemically informed latent space, predicts R-group-aware ring closures during autoregressive generation, and aligns the denoiser to affinity rewards using winner-protected diffusion-adapted preference optimization. In silico experiments demonstrate PepALD's generation quality and reward-optimization performance against representative peptide generation baselines.", "date": "2026-06-12", "categories": []}
{"id": "2606.19377", "title": "Emyx: Fast and efficient all-atom protein generation", "abstract": "Computational enzyme design requires generating proteins that scaffold catalytic residues and ligands, a task that demands both geometric accuracy and structural diversity from the underlying generative model. Current all-atom generators inherit expensive architectures from structure prediction, leading to high training costs and limited sample diversity. We argue that much of this complexity is unnecessary for generators, which condition on sparse geometric constraints rather than rich co-evolutionary signals. Emyx is a 140M-parameter conditional flow matching model that concentrates capacity within standard transformer blocks, replacing heavy embedding stacks with lightweight conditional representations and sparse connectivity. We additionally derive an exact reparametrisation of the flow matching interpolant into the EDM noise-level framework, bridging flow matching training efficiency with state-of-the-art sampling methods designed for diffusion models without retraining. Despite being the smallest model, Emyx outperforms both Prote\\'ina-Complexa and RFdiffusion3 against the AME enzyme design benchmark across success rate under strict evaluation requiring both global fold recovery and catalytic geometry accuracy, structural novelty, scaffold diversity, and geometric validity, while training in just $682$ GPU-hours, roughly $4\\times$ less than RFdiffusion3.", "date": "2026-06-12", "categories": []}
{"id": "2606.20174", "title": "Computational Methods and Challenges in Cell-Free DNA Analysis for Multi-Cancer Early Detection", "abstract": "Cell-free DNA (cfDNA) is a promising avenue for non-invasive multicancer early detection (MCED), in that, it can enable multiple cancer detection simultaneously from a single blood draw, with particular sensitivity to cancers that currently lack established screening programs. Here we review the computational methods developed between 2022 and 2025 for cfDNA-based MCED. We focus on how fragmentomics and epigenetic features are extracted and analyzed to detect cancer at early stages. We first briefly outline the biological basis of cfDNA signals, then review classical statistical and machine learning approaches alongside deep learning frameworks including autoencoder-based models. For each method we discuss biological interpretability, validation strategy, and readiness for clinical integration. Furthermore, we categorize the current challenges into technical, computational, and methodological while outlining open problems in the field. This review shows that multimodal ensemble approaches have the strongest promise for clinical integration and the highest readiness. However, for better assessment of future work and side-by-side comparison, standardization of evaluation protocols and reporting results will be crucial.", "date": "2026-06-18", "categories": []}
{"id": "2606.20329", "title": "Constrained hybrid modelling to predict microbial dynamics and organic matter turnover in soil systems", "abstract": "Soil microorganisms control organic matter cycling and largely determine how soil systems can cope with and mitigate climate change and environmental threats. Representing microbial dynamics in process-based soil models is therefore critical to predict carbon cycling in soils, albeit highly challenging to inform from data. One promising approach to improve their parametrisation is the integration of genomic data, yet modelling the complex and unknown relationship between genomes and the processes the microbes are driving is an unsolved problem. In this work, we present the first hybrid modeling framework for deriving biokinetic parameter values of a process-based soil organic matter turnover model from metagenome-inferred functional traits based on DNA sequencing data. Our model predicts biokinetic parameters of the process-based model from genomic trait data with a neural network and integrates constraints from ecological theory and literature to ensure realistic behavior, even of non-observed state variables. We evaluate our method on synthetic genomic trait datasets of varying complexity and on real data, showing that our approach improves performance over multiple baselines and learns the dynamics of unmeasurable components of the process-based model effectively, even for small training datasets.", "date": "2026-06-18", "categories": []}
{"id": "2506.13196", "title": "KEPLA: A Knowledge-Enhanced Deep Learning Framework for Accurate Protein-Ligand Binding Affinity Prediction", "abstract": "Accurate prediction of protein-ligand binding affinity is critical for drug discovery. While recent deep learning approaches have demonstrated promising results, they often rely solely on structural features of proteins and ligands, overlooking their valuable biochemical knowledge associated with binding affinity. To address this limitation, we propose KEPLA, a novel deep learning framework that explicitly integrates prior knowledge from Gene Ontology and ligand properties to enhance prediction performance. KEPLA takes protein sequences and ligand molecular graphs as input and optimizes two complementary objectives: (1) aligning global representations with knowledge graph relations to capture domain-specific biochemical insights, and (2) leveraging cross attention between local representations to construct fine-grained joint embeddings for prediction. Experiments on two benchmark datasets across both in-domain and cross-domain scenarios demonstrate that KEPLA consistently outperforms state-of-the-art baselines. Furthermore, interpretability analyses based on knowledge graph relations and cross attention maps provide valuable insights into the underlying predictive mechanisms.", "date": "2025-06-16", "categories": []}
{"id": "2601.12805", "title": "SciHorizon-GENE: Benchmarking LLM for Life Sciences Inference from Gene Knowledge to Functional Understanding", "abstract": "Large language models (LLMs) have shown growing promise in biomedical research, particularly for knowledge-driven interpretation tasks. However, their ability to reliably reason from gene-level knowledge to functional understanding, a core requirement for knowledge-enhanced cell atlas interpretation, remains largely underexplored. To address this gap, we introduce SciHorizon-GENE, a large-scale gene-centric benchmark constructed from authoritative biological databases. The benchmark integrates curated knowledge for over 190K human genes and comprises more than 540K questions covering diverse gene-to-function reasoning scenarios relevant to cell type annotation, functional interpretation, and mechanism-oriented analysis. Motivated by behavioral patterns observed in preliminary examinations, SciHorizon-GENE evaluates LLMs along four biologically critical perspectives: research attention sensitivity, hallucination tendency, answer completeness, and literature influence, explicitly targeting failure modes that limit the safe adoption of LLMs in biological interpretation pipelines. We systematically evaluate a wide range of state-of-the-art general-purpose and biomedical LLMs, revealing substantial heterogeneity in gene-level reasoning capabilities and persistent challenges in generating faithful, complete, and literature-grounded functional interpretations. Our benchmark establishes a systematic foundation for analyzing LLM behavior at the gene scale and offers insights for model selection and development, with direct relevance to knowledge-enhanced biological interpretation.", "date": "2026-01-19", "categories": []}
{"id": "2604.14906", "title": "Unraveling the Mechanism of Drug Binding to SARS-CoV-2 RNA Pseudoknot with Thermodynamics-Driven Machine Learning", "abstract": "The pseudoknot secondary structure in SARS-CoV-2 RNA is essential for regulating protein synthesis through $-$1 programmed ribosomal frameshifting ($-1$ PRF), a mechanism that allows the virus to generate both structural and non-structural proteins from overlapping reading frames. This pseudoknot exhibits both threaded and unthreaded long-lived topologies. The influence of ligand binding on its folding is a process critical for the development of $-$1 PRF small-molecule inhibitors. Understanding this process through unbiased molecular dynamics (MD) simulations can be facilitated by introducing collective variables (CVs) that capture the corresponding slowest dynamical modes. Here, we use spectral map (SM), a thermodynamics-driven machine learning technique, to learn such CVs directly from all-atom MD trajectories of the SARS-CoV-2 RNA pseudoknot in complex with the $-$1 PRF inhibitor merafloxacin and its two structural analogs in neutral and ionized forms. Free-energy landscapes (FELs) derived from the learned CVs indicate that ligand-induced destabilization is topology-selective. In the threaded pseudoknot, the inhibitors destabilize the S2 stem, while in the unthreaded pseudoknot, destabilization occurs in the S1 and S3 stems. Furthermore, the extent to which each ligand reshapes the FEL matches experimentally reported antiviral potency, whereas the protonation state qualitatively alters dynamics within the same RNA topology. Overall, our results show how pseudoknot topology, ligand type, and protonation state collectively influence the slow conformational dynamics of viral RNA and establish physiological protonation as a critical factor for modeling RNA-targeted drug action.", "date": "2026-04-16", "categories": []}
{"id": "2606.18302", "title": "Protein-Based Fish Species Identification: Dataset, Models, and Insights from Native Bangladeshi Fish", "abstract": "Correct identification of fish species is highly significant for food security, economic development, and climate resilience in Bangladesh. Protein sequences directly reflect functional and evolutionary constraints which are important for species authentication and biodiversity monitoring. Yet there exists no benchmark for native Bangladeshi fish species identification from protein sequence. In this study, we addressed this gap by introducing the first curated dataset for nine native Bangladeshi fish species of 2845 high quality protein sequences. We also established the first protein sequence classification baseline for this domain through a systematic benchmarking of seven architectural paradigms. Moreover, we propose a realistic deployable novel hybrid architecture of MotifCNN and Transformer with Terminal-Aware Positional-Encoding (MotifCNN-Transformer+TA-PE). Our novel architecture achieves 79.80% accuracy with macro-F1 of 0.80. The highest 83.04% accuracy is achieved by finetuned protein language model ProtBERT that has 420M parameters and requires dual 16GB GPUs for inference. According to McNemar's test, ProtBERT's 3.24% accuracy gain over our MotifCNN-Transformer+TA-PE is statistically insignificant (p = 0.1120). Our novel architecture beats it among six of the nine classes in per class identification. Also our MotifCNN-Transformer+TA-PE is approximately 5x faster, 42x smaller, and supports 16x larger batch size than ProtBERT and has GPU free inference, making it more practical for deployment in resources constrained areas such as rural Bangladesh. Beyond this, our foundational work shows effects of phylogenetic relationships on sequence similarity and establishes pathways for fisheries management, food authentication and biodiversity conservation in South Asia's protein dependent economy.", "date": "2026-06-16", "categories": []}
{"id": "2606.18672", "title": "scGTN: Deep Siamese Graph Transformer Network for Single-cell RNA Sequencing Clustering", "abstract": "Single-cell RNA sequencing (scRNA-seq) serves a pivotal role in characterizing gene expression at the cellular level, enabling the identification of cell types and advancing the understanding of cellular heterogeneity. Despite the significant progress in scRNA-seq data clustering, we argue that current methods always ignore the sparsity and noise, as well as the complex intercellular structural information inherent in scRNA-seq data. Toward this end, in this paper, we propose a novel single-cell RNA-seq clustering framework via deep Siamese Graph Transformer Network (termed scGTN), which explicitly integrates gene expression profile and intercellular structural dependencies for cell clustering. In particular, we formulate scRNA-seq data as a graph and construct two augmented graph views that serve as dual views to capture complementary intercellular information. Then, a Siamese graph transformer network is employed to explicitly incorporate shortest-path information and node-wise distances for capturing richer structural relationships between cells. Finally, we employ an optimal transport strategy to guide the cell clustering in a self-supervised manner. Extensive experiments on multiple benchmark scRNA-seq datasets demonstrate that our scGTN consistently outperforms existing methods. Our code is available at https://github.com/W-RMSL/scGTN.", "date": "2026-06-17", "categories": []}
{"id": "2507.04704", "title": "SPATIA: Multimodal Generation and Prediction of Spatial Cell Phenotypes", "abstract": "Understanding how cellular morphology, gene expression, and spatial context jointly shape tissue function is a central challenge in biology. Image-based spatial transcriptomics technologies now provide high-resolution measurements of cell images and gene expression profiles, but existing methods typically analyze these modalities in isolation or at limited resolution. We address the problem by introducing SPATIA, a multi-level generative and predictive model that learns unified, spatially aware representations by fusing morphology, gene expression, and spatial context from the cell to the tissue level. SPATIA also incorporates a spatially conditioned generative framework with confidence-aware OT reweighting and morphology-profile alignment for modeling target-state morphology distributions. Specifically, we propose a confidence-aware flow matching objective that reweights weak optimal-transport pairs based on uncertainty. We further apply morphology-profile alignment to encourage biologically meaningful image generation, enabling the modeling of microenvironment-dependent phenotypic transitions. We assembled a multi-scale dataset consisting of 25.9 million cell-gene pairs across 17 tissues. We benchmark SPATIA against 18 models across 12 tasks, spanning categories such as phenotype generation, annotation, clustering, gene imputation, and cross-modal prediction. SPATIA achieves improved performance over state-of-the-art models, improving generative fidelity by 8% and predictive accuracy by up to 3%.", "date": "2025-07-07", "categories": []}
{"id": "2602.04901", "title": "Beyond Independent Genes: Learning Module-Inductive Representations for Single-Cell Gene Perturbation Prediction", "abstract": "Predicting transcriptional responses to genetic perturbations is a central problem in functional genomics. In practice, perturbation responses are rarely gene-independent but instead manifest as coordinated, program-level transcriptional changes among functionally related genes. However, most existing methods do not explicitly model such coordination, due to gene-wise modeling paradigms and reliance on static biological priors that cannot capture dynamic program reorganization. To address these limitations, we propose scBIG, a module-inductive perturbation prediction framework that explicitly models coordinated gene programs. scBIG induces coherent gene programs from data via Gene-Relation Clustering, captures inter-program interactions through a Gene-Cluster-Aware Encoder, and preserves modular coordination using structure-aware alignment objectives. These structured representations are then modeled using conditional flow matching to enable flexible and generalizable perturbation prediction. Extensive experiments on multiple single-cell perturbation benchmarks show that scBIG consistently outperforms state-of-the-art methods, particularly on unseen and combinatorial perturbation settings, achieving an average improvement of 6.7% over the strongest baselines. The code is available at https://github.com/ttruan2426-dot/scBIG.", "date": "2026-02-03", "categories": []}
{"id": "2606.17127", "title": "Agentic Discovery of Non-Canonical Antimicrobial Peptides with AMPGAN v3", "abstract": "Antimicrobial resistance causes to over a million deaths annually. Antimicrobial peptides (AMPs) are a promising solution, but generative AMP models are not yet ready to design peptides with non-natural amino acids and/or chemical modifications, which are essential for real-world peptide drugs. We present AMPGAN v3, a multi-objective conditional GAN that expands the generative vocabulary to D-amino acids and N/C-terminus modifications such as amidation. By separating adversarial and activity-aware supervision across two specialized discriminators, AMPGAN v3 substantially improves training stability and outperforms prior generative AMP models on external classifiers. We validated five candidates spanning three structural classes in vitro; two showed activity against Gram-positive strains, with the best candidate reaching MIC 8 {\\mu}g/mL against B. subtilis. To support downstream curation, we further present PepCraft, a multi-agent framework for end-to-end AMP discovery in which a Planning Agent orchestrates specialized executors for generation, filtering, and verification. Its prioritization recommendations align with our in vitro outcomes. Together, these contributions let us examine, on a small but real scale, how generative and agentic AI compose in therapeutic peptide discovery. Code: https://github.com/marszzibros/AMPGANv3", "date": "2026-06-15", "categories": []}
{"id": "2407.07357", "title": "A polarity-aware multi-relational model for the signed interaction prediction in biological networks", "abstract": "Predicting signed interactions in biological networks is crucial for understanding drug mechanisms and facilitating drug repurposing. While deep graph models have demonstrated success in modeling complex biological systems, existing approaches often fail to distinguish between positive and negative interactions, limiting their utility for precise pharmacological predictions. In this study, we propose a novel deep graph model, PAMR (polarity-aware multi-relational model), designed to predict both polar (e.g., activation, inhibition) and non-polar (e.g., binding, affect) chemical-gene interactions. Our model integrates graph convolutional networks with tensor decomposition to enhance feature representation and incorporates a conflict-aware sampling strategy to resolve polarity ambiguities. We introduce new evaluation metrics, polarity discrimination score (PDS) and CP@100, to assess the model's ability to differentiate interaction types. Experimental results demonstrate that PAMR outperforms baseline models, achieving superior classification accuracy and improved discrimination of polar edges. Specifically, PAMR-CL attains a Macro AUROC of 0.9072 and CP@100 of 0.974, surpassing RGCN, GraphSAGE, TransE, and BioNet baselines. A case study on nicotine further identifies two novel chemical-gene suppression links, S100A6 and SPP1, that are corroborated by independent experimental literature. Furthermore, we analyze the impact of subgraph components on predictive performance, revealing that additional network structures do not always enhance accuracy. These findings highlight the importance of polarity-aware modeling in drug discovery and network pharmacology, providing a scalable computational framework for polarity-aware chemical-gene interaction prediction and network pharmacology analysis.", "date": "2024-07-10", "categories": []}
{"id": "2510.24987", "title": "scMRDR: A scalable and flexible framework for unpaired single-cell multi-omics data integration", "abstract": "Advances in single-cell sequencing have enabled high-resolution profiling of diverse molecular modalities, while integrating unpaired multi-omics single-cell data remains challenging. Existing approaches either rely on pair information or prior correspondences, or require computing a global pairwise coupling matrix, limiting their scalability and flexibility. In this paper, we introduce a scalable and flexible generative framework called single-cell Multi-omics Regularized Disentangled Representations (scMRDR) for unpaired multi-omics integration. Specifically, we disentangle each cell's latent representations into modality-shared and modality-specific components using a well-designed $\\beta$-VAE architecture, which are augmented with isometric regularization to preserve intra-omics biological heterogeneity, adversarial objective to encourage cross-modal alignment, and masked reconstruction loss strategy to address the issue of missing features across modalities. Our method achieves excellent performance on benchmark datasets in terms of batch correction, modality alignment, and biological signal preservation. Crucially, it scales effectively to large-scale datasets and supports integration of more than two omics, offering a powerful and flexible solution for large-scale multi-omics data integration and downstream biological discovery.", "date": "2025-10-28", "categories": []}
{"id": "2603.15952", "title": "Protein Design with Agent Rosetta: A Case Study for Specialized Scientific Agents", "abstract": "Large language models (LLMs) are capable of emulating reasoning and using tools, creating opportunities for autonomous agents that execute complex scientific tasks. Protein design provides a natural testbed: although machine learning (ML) methods achieve strong results, these are largely restricted to canonical amino acids and narrow objectives, leaving unfilled need for a generalist tool for broad design pipelines. We introduce Agent Rosetta, an LLM agent paired with a structured environment for operating Rosetta, the leading physics-based heteropolymer design software, capable of modeling non-canonical building blocks and geometries. Agent Rosetta iteratively refines designs to achieve user-defined objectives, combining LLM reasoning with Rosetta's generality. We evaluate Agent Rosetta on design with canonical amino acids, matching specialized models and expert baselines, and with non-canonical residues -- where ML approaches fail -- achieving comparable performance. Critically, prompt engineering alone often fails to generate Rosetta actions, demonstrating that environment design is essential for integrating LLM agents with specialized software. Our results show that properly designed environments enable LLM agents to make scientific software accessible while matching specialized tools and human experts.", "date": "2026-03-16", "categories": []}
{"id": "2604.25371", "title": "PhyloSDF: Phylogenetically-Conditioned Neural Generation of 3D Skull Morphology via Residual Flow Matching", "abstract": "Generating novel, biologically plausible three-dimensional morphological structures is a fundamental challenge in computational evolutionary biology, hampered by extreme data scarcity and the requirement that generated shapes respect phylogenetic relationships among species. In this work, we present PhyloSDF, a phylogenetically-conditioned neural generative model for 3D biological morphology that integrates two innovations: (1) a DeepSDF auto-decoder regularized by a novel Phylogenetic Consistency Loss that structures the latent space to correlate with evolutionary distances (Pearson r=0.993); (2) a Residual Conditional Flow Matching (Residual CFM) architecture that factorizes generation into analytic species-centroid lookup and learned residual prediction, enabling generation from as few as ~4 specimens per species. We evaluate PhyloSDF on 100 micro-CT-scanned skulls of Darwin's Finches and their relatives across 24 species. The model generates novel meshes achieving 88-129% of real intra-species variation at the code level, with all 180 generated meshes verified as non-memorized. Residual CFM surpasses denoising diffusion (which fails entirely at this scale), standard flow matching (which mode-collapses to 3-6% variation), and a Gaussian mixture baseline in both fidelity (Chamfer Distance 0.00181 vs. 0.00190) and morphometric Fr\\'{e}chet distance (10,641 vs. 13,322). Leave-one-species-out experiments across 18 species demonstrate phylogenetic extrapolation capability, and smooth latent interpolations produce biologically plausible ancestral skull reconstructions.", "date": "2026-04-28", "categories": []}
{"id": "2606.06834", "title": "The Dark Regulome: Disentangling Predictability from Regulation in Genomic Foundation Models", "abstract": "High-grade gliomas integrate into neural circuits through functional synapses with neurons, raising the question of which noncoding elements shape synaptogenic gene expression in tumor cells. The regulatory program written across the dark genome, what we call the $\\textit{dark regulome}$, is the natural substrate to probe, and sequence foundation models offer a zero-shot route through in-silico mutagenesis (ISM); yet likelihood-based scoring is tautologically coupled to local sequence predictability, leaving the regulatory interpretation underdetermined. Across three architecturally distinct foundation models (Caduceus-Ph, HyenaDNA, Enformer) and 30,448 dark genome elements at 92 glioma-relevant loci, we introduce a residualization-and-permutation diagnostic that separates predictability-driven from regulation-driven RIS variance. A sharp 10kb proximal-regulatory horizon survives every control we apply, but the LM-derived element-class hierarchy does not: a six-feature linear baseline matches Caduceus top-decile membership at AUC $= 0.985$. Cross-architecture decomposition cleanly separates a sequence-predictability layer (the two language models co-rank long well-predicted transposable elements) from a regulatory-output layer (Enformer alone retains residual cCRE-discriminative signal), with literally zero overlap between the two top-100 lists. Conservation, brain cis-eQTL, and STRING-PPI cross-checks then anchor what biology survives: top-100 elements across all three models are $3.3\\times$ enriched per model for matching brain eQTLs ($p_\\mathrm{emp} < 5\\times 10^{-3}$), while a tempting transposable-element regulatory layer and a striking NRXN1+NLGN1 protein-pair convergence both fail proper permutation tests once those tests are constructed. We deliver the diagnostic as a general methodological tool for any ISM-based regulatory study.", "date": "2026-06-05", "categories": []}
{"id": "2606.14734", "title": "BRIDGE: Biological Evidence Refinement and Heterogeneous Dynamic Gating for Gene Regulatory Networks", "abstract": "Motivation: Gene regulatory network inference from single-cell RNA sequencing (scRNA-seq) data is important for uncovering cell-state-specific transcriptional programs. However, scRNA-seq measurements are sparse and noisy, and experimentally validated TF-target interactions remain limited, making reliable inference challenging. Although graph neural networks have advanced GRN prediction, existing methods often rely on biologically unconstrained graph augmentation, such as random edge perturbation, and insufficiently control information transfer between genes and cells. These limitations may distort regulatory structures and weaken robustness under noisy and weakly supervised settings. Results: To address these issues, we propose an innovative framework named Biological Evidence Refinement and Heterogeneous Dynamic Gating for Gene Regulatory Networks (BRIDGE). BRIDGE extracts gene and cell representations from the expression matrix and its matrix dual, and performs contrastive learning in the gene space and cell space between self and neighbors across the co-expression-refined regulatory view and the original graph. It then applies heterogeneous gated encoding to adaptively regulate information transfer between genes and cells, enabling robust transcription factor-to-target gene prediction. Experiments on benchmark datasets spanning three network types and seven cell types show that BRIDGE achieves state-of-the-art AUROC and AUPRC in most settings. In particular, on Specific networks, BRIDGE improves average AUPRC by 5% over the second-best baseline, GCLink. In cross-cell-type few-shot transfer, BRIDGE consistently outperforms GCLink and GENELink across all six target cell types. A case study on hESC further supports the biological relevance of the predictions, with 9 of the top 10 and 46 of the top 100 novel TF-target interactions validated by ChIPBase.", "date": "2026-06-02", "categories": []}
{"id": "2606.14737", "title": "Learning Topological Representations for Molecular Dynamics", "abstract": "Molecular dynamics (MD) simulations generate trajectories in a high-dimensional configuration space whose analysis critically depends on molecular descriptors, typically handcrafted observables or learned kinetic embeddings. Designing descriptors that are both expressive and broadly applicable, however, remains challenging. We study persistent homology (PH) as a general-purpose representation for MD and introduce the masked Flood complex, a protein-tailored modification of a recently introduced simplicial complex construction that emphasizes inter-residue structure at low computational cost. Vectorized persistence diagrams then provide information-rich, geometry-aware summaries of protein conformations, which we evaluate on protein class prediction, frame-level observable regression, and Markov state model (MSM) estimation from learned low-dimensional coordinates in a single shared representation space. Results on the mdCATH dataset show that PH-based descriptors are competitive across tasks, with masked Flood PH yielding the most consistent overall performance. Further, when using topologically-informed MSMs as a drop-in replacement within the recent MarS-FM framework for generative modeling of protein conformations, we obtain consistently better ensemble statistics than MSMs based on physical observables. Finally, we explore the transferability of the generative model to qualitatively different, fast folding, proteins.", "date": "2026-06-02", "categories": []}
{"id": "2606.15422", "title": "Pepti-Agent: An AI Agent for Peptide Design and Optimization", "abstract": "Therapeutic peptides occupy a valuable design space between small molecules and biologics, but their development requires satisfying several competing constraints at once: solubility, hemolytic activity, and nonspecific surface fouling are governed by overlapping sequence features, so improving one property often degrades another. Computational design addresses this by pairing generative models with sequence-based property predictors, iteratively proposing and refining candidates. However, these components are typically wired together as monolithic scripts that are difficult to inspect, extend, or reuse, and they often refine sequences by natural-language reasoning rather than by tracking the evolving multi-property state of each candidate. We present Pepti-Agent, a closed-loop, peptide-specific framework that exposes generation, property prediction, and single-residue mutation as independently inspectable Model Context Protocol (MCP) tools. A large language model controller invokes these tools and consults live predictor output between calls, so refinement is guided by each sequence's current property profile rather than by language reasoning alone. Task-specific PeptideGPT models generate candidates, ProtBERT-based classifiers score solubility, hemolysis, and non-fouling, and two interchangeable mutation operators propose sequence edits. By recording a per-step trace of controller decisions, predictor outputs, and accepted mutations, Pepti-Agent offers a reproducible substrate for benchmarking multi-objective design strategies and for prioritizing candidates for experimental validation.", "date": "2026-06-13", "categories": []}
{"id": "2606.16044", "title": "Circuit Tracing in Autoregressive Protein Language Models", "abstract": "Protein language models (pLMs) can generate novel protein sequences with properties beyond those observed in nature, yet the mechanisms underlying protein generation remain poorly understood. Existing mechanistic interpretability methods based on sparse autoencoders and transcoders primarily focus on protein representation learning models and do not capture the computation required for autoregressive generation. Here, we introduce ProGenMech, a mechanistic interpretability framework for generative protein language models that extends cross-layer transcoders (CLTs) to ProGen3, a sparse Mixture-of-Experts model trained for both causal generation and span infilling. Unlike per-layer approaches, CLTs reconstruct each layer using sparse latent variables from all preceding layers, enabling faithful recovery of inter-layer generative computation. We further develop a zero-shot circuit discovery framework to identify sparse latent circuits responsible for protein generation and fitness prediction. In causal generation and zero-shot fitness estimation tasks, ProGenMech outperforms local transcoder baselines in recovering ProGen3's probability distribution and functional scoring behavior, while matching the original model's generative distribution in span infilling tasks. Moreover, the recovered circuits reveal biologically meaningful motifs and functional regions associated with conserved sequence patterns and protein fitness landscapes, establishing a foundation for interpretable and steerable protein generation.", "date": "2026-06-14", "categories": []}
{"id": "2606.16477", "title": "AURA: Active-Response Attribution under Treatment Ambiguity in Bacterial Cytological Profiling", "abstract": "When a bacterial sample is exposed to several antibiotics, not every applied drug necessarily acts: if the organism is resistant to one of them, that drug leaves no morphological trace. The clinically meaningful quantity is therefore not which antibiotics were applied, but which ones were active. We show that these two are sharply decoupled in real E. coli microscopy - naively assuming the applied combination equals the active one is correct only about 37% of the time - yet existing computational tools are ill-suited to recovering the active set. Forward perturbation models such as scGen, CPA, and IMPA are designed to predict appearance from treatment, not the reverse, and inverting them degrades sharply; discriminative image classifiers tend to memorise strain- and batch-specific texture and fail to transfer across experimental replicates. We introduce AURA, which reframes the task as constrained, energy-based inverse attribution. Its central inductive bias is that the active set must be a subset of the applied set; this collapses the candidate space and lets AURA infer the active subset of applied antibiotics by decomposing residual morphology into antibiotic response atoms and selecting the subset with the lowest reconstruction energy, using no strain label at test time. AURA-E adds evidence-aware abstention, withholding a prediction when candidate explanations remain near-equally plausible. On cross-replicate transfer in an E. coli cytological profiling dataset, AURA recovers the active antibiotic combination with 95.47% exact-match accuracy.", "date": "2026-06-15", "categories": []}
{"id": "2606.16540", "title": "MultiMolecule: a modular ecosystem for biomolecular sequence-model workflows", "abstract": "Biomolecular sequence models are increasingly reused outside the studies in which they were introduced, but public checkpoints rarely preserve the execution context needed to inspect source-defined behavior, adapt models to new assays, compare models under shared task definitions or deploy biological predictions. MultiMolecule is an open-source Python ecosystem that turns heterogeneous RNA, DNA and protein sequence-model releases into complete, source-checked model-family implementations with shared loading, workflow and prediction interfaces. The Resource state reported here includes 53 complete model-family implementations with 112 standardized model checkpoints, together with 16 curated dataset resources released through 39 public dataset repositories and 10 user-facing prediction pipelines. Standardized components are linked to source provenance, conversion or preparation code, source-reference checks, Extended Data summaries and public documentation, allowing users to inspect what was standardized, what behavior was checked and how each component enters training, evaluation, inference or deployment. By shifting reuse from repository-specific checkpoints to executable implementations connected to standardized checkpoints, curated datasets, Runner workflows and biological prediction pipelines, MultiMolecule provides common infrastructure for preserving source-defined model behavior, adapting models to new assays, enabling controlled evaluation and deploying biomolecular predictions.", "date": "2026-06-15", "categories": []}
{"id": "2605.29228", "title": "Traditional machine learning vs. deep learning from dynamic graph representations of proteins' 3D folds in the task of protein structure classification", "abstract": "Protein structure classification (PSC) uses supervised learning to predict a protein's CATH/SCOP(e) class from the protein's sequence or 3D structural feature(s). We already modeled 3D structures as (static) protein structure networks (PSNs), demonstrating the competitiveness of PSN-based features to sequence or direct (i.e. non-network) 3D structural features in the PSC task. More recently, we demonstrated the power of features extracted from dynamic PSNs over features extracted from static PSNs (and thus by transitivity over sequence and direct 3D structural features) in the same task. That dynamic PSN approach used traditional machine learning (ML), combining manual (pre-engineered) features with an off-the-shelf classifier. Here, we evaluate whether automatic deep learning (DL) from the dynamic PSNs yields improvements. Our evaluation on 72 datasets spanning ~44,000 CATH- or SCOPe-labeled dynamic PSNs reveals that in terms of PSC accuracy, traditional ML and DL are (close to) tied for a large majority of the datasets, while DL is on average 10+ times slower. We are the first to evaluate traditional ML vs. DL in the dynamic PSN-based PSC task.", "date": "2026-05-28", "categories": []}
{"id": "2606.13713", "title": "CisTransCell: Single-Cell Perturbation Prediction via Gene Function, Regulatory Control, and Cellular Context", "abstract": "Predicting cellular transcriptional responses to genetic perturbations is a central problem in single-cell biology, especially in the zero-shot setting where the perturbed gene or gene combination is unseen during training. A major difficulty is that perturbation effects are not determined by expression state alone: they depend on how the perturbed gene product influences other genes and proteins, how those downstream factors act on cis-regulatory elements, and which regulatory programs are active in the current cell state. To better capture this biological complexity, we propose CisTransCell, a cell-conditioned multi-modal framework for single-cell perturbation prediction that augments each gene with two complementary priors: a regulatory-sequence prior that captures how the gene is controlled, and a coding-sequence prior that captures what the gene product does. By integrating these priors with cellular expression state, CisTransCell models perturbation response as a cascade from gene function to regulatory control to downstream transcriptional change. Experiments on benchmark single-cell perturbation datasets show that CisTransCell achieves strong performance in zero-shot perturbation prediction.", "date": "2026-06-10", "categories": []}
{"id": "2606.14159", "title": "Curvature-Guided Geometric Representation for Protein-Ligand Binding Affinity Prediction", "abstract": "Protein-ligand binding affinity (PLA) prediction is critical in drug discovery. Despite the notable advancements in machine learning-based approaches, existing methods struggle to jointly characterize local geometric organization and globally coordinated cross-molecular interactions, limiting their ability to model complex binding mechanisms. Here, we propose RicciBind, a geometric representation framework that integrates curvature-guided hierarchical structure learning with optimal transport (OT)-based cross-domain alignment to model molecular interactions. Specifically, RicciBind leverages Ricci curvature to capture local interaction tightness within molecular structures, enhancing structural awareness and organizing atomic interactions into curvature-aware hierarchical representations. An OT-based cluster matching mechanism then aligns protein and ligand clusters across heterogeneous domains under geometric constraints, enabling globally consistent correspondences and revealing higher-order interaction patterns beyond local neighborhoods. By coupling curvature-guided structure encoding with OT-driven cross-domain alignment, RicciBind effectively models complex interaction semantics and substantially improves both the accuracy and interpretability of binding affinity prediction. Extensive experiments demonstrate that RicciBind achieved superior predictive performance and generalization across PLA benchmarks and virtual screening tasks. Ablation studies further confirmed the essential role of Ricci curvature in enhancing molecular interaction representations.", "date": "2026-06-12", "categories": []}
{"id": "2606.14217", "title": "Curvature-Informed Potential Energy Surface for Protein-Ligand Binding Affinity Prediction", "abstract": "Accurate prediction of protein-ligand binding affinity is essential for structure-based drug discovery. Recent geometric deep learning methods have achieved promising performance by representing protein-ligand complexes as three-dimensional graphs. However, most existing approaches mainly rely on static interaction geometry from a single bound conformation, while neglecting molecular flexibility and binding-induced conformational changes. To address this limitation, we propose a curvature-informed potential energy surface (CPES) graph neural network for protein-ligand binding affinity prediction, which incorporates physics-informed curvature representations to model conformational flexibility. CPES first derives curvature spectral descriptors from the Hessian of the potential energy surface evaluated at equilibrium configurations, whose eigenvalues define the local principal curvatures of the potential energy surface. It then uses spectral cross-attention to compare the unbound ligand and protein with the bound complex, thereby capturing binding-induced changes in conformational dynamics. In parallel, hierarchical protein-ligand interaction representations are learned from static structural features through geometry-aware message passing, soft clustering, and bidirectional cross-attention. Finally, CPES fuses the curvature-informed dynamic representations with static interaction representations for affinity regression. Extensive evaluations on multiple benchmark datasets demonstrate that CPES achieves improved predictive performance and offers physical interpretability.", "date": "2026-06-12", "categories": []}
{"id": "2505.23823", "title": "RAGPPI: RAG Benchmark for Protein-Protein Interactions in Drug Discovery", "abstract": "Retrieving the biological impacts of protein-protein interactions (PPIs) is essential for target identification (Target ID) in drug development. Given the vast number of proteins involved, this process remains time-consuming and challenging. Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) frameworks have supported Target ID; however, no benchmark currently exists for identifying the biological impacts of PPIs. To bridge this gap, we introduce the RAG Benchmark for PPIs (RAGPPI), a factual question-answer benchmark of 4,420 question-answer pairs that focus on the potential biological impacts of PPIs. Through interviews with experts, we identified criteria for a benchmark dataset, such as a type of QA and source. We built a gold-standard dataset (500 QA pairs) through expert-driven data annotation. We developed an ensemble auto-evaluation LLM that incorporates expert labeling characteristics, average fact-abstract similarity (F1), and low-similarity fact counts (F2), enabling the construction of a silver-standard dataset (3,720 QA pairs). We are committed to maintaining RAGPPI as a resource to support the research community in advancing RAG systems for drug discovery QA solutions.", "date": "2025-05-28", "categories": []}
{"id": "2512.15133", "title": "HD-Prot: A Protein Language Model for Joint Sequence-Structure Modeling with Continuous Structure Tokens", "abstract": "Proteins inherently possess a consistent sequence-structure duality. The abundance of protein sequence data, which can be readily represented as discrete tokens, has driven fruitful developments in protein language models (pLMs). A key remaining challenge, however, is how to effectively integrate continuous structural knowledge into pLMs. Current methods often discretize protein structures to accommodate the language modeling framework, which inevitably results in the loss of fine-grained information and limits the performance potential of multimodal pLMs. In this paper, we argue that such concerns can be circumvented: a sequence-based pLM can be extended to incorporate the structure modality through continuous tokens, i.e., high-fidelity protein structure latents that avoid vector quantization. Specifically, we propose a hybrid diffusion protein language model, HD-Prot, which embeds a continuous-valued diffusion head atop a discrete pLM, enabling seamless operation with both discrete and continuous tokens for joint sequence-structure modeling. It captures inter-token dependencies across modalities through a unified absorbing diffusion process, and estimates per-token distributions via categorical prediction for sequences and continuous diffusion for structures. Extensive results demonstrate that HD-Prot achieves competitive performance in unconditional sequence-structure co-generation, motif-scaffolding, protein structure prediction, and inverse folding tasks. Furthermore, our method can perform on par with state-of-the-art multimodal pLMs, despite being developed under limited computational resources (i.e., less than one-tenth the budget for modality extension fine-tuning). It highlights the viability of simultaneously estimating categorical and continuous distributions within a unified language model architecture, offering a promising alternative direction for multimodal pLMs.", "date": "2025-12-17", "categories": []}
{"id": "2603.02274", "title": "Contextual Invertible World Models: A Neuro-Symbolic Agentic Framework for Colorectal Cancer Drug Response", "abstract": "Precision oncology is currently limited by the small-N, large-P paradox, where high-dimensional genomic data is abundant but pharmacological response samples are sparse. While deep learning achieves predictive accuracy, it frequently fails to provide the mechanistic clarity required for clinical adoption. We present the Contextual Invertible World Model (CIWM), a Neuro-Symbolic Agentic Framework that bridges this gap by integrating a quantitative machine learning emulator with a Large Language Model reasoning layer. Utilising a stringently curated, high-fidelity data engineering pipeline on the Sanger GDSC dataset (\\( N=83 \\)), we isolate true biological signals from in vitro artifacts to establish a rigorous baseline predictive correlation for complex transcriptomics (\\( r=0.268 \\)). Through Inverse Reasoning, we perform in silico CRISPR perturbations across the colorectal landscape. The framework autonomously overturns classical mechanistic assumptions, identifying a hierarchical dominance of mutant KRAS over the APC/Wnt-axis in driving 5-fluorouracil resistance (\\( \\Delta=-0.0469 \\)) via a \"KRAS Shield\" mapped to MAPK/PI3K networks. Furthermore, the agentic layer identified a \"PIK3CA Paradox\", revealing that repairing PIK3CA inadvertently increases chemoresistance (\\( \\Delta=+0.0085 \\)) by triggering a compensatory feedback loop that hyperactivates the dominant MAPK survival pathway.", "date": "2026-03-01", "categories": []}
{"id": "2603.26705", "title": "PI-Mamba: Linear-Time Protein Backbone Generation via Spectrally Initialized Flow Matching", "abstract": "Motivation: Generative models for protein backbone design have to simultaneously ensure geometric validity, sampling efficiency, and scalability to long sequences. However, most existing approaches rely on iterative refinement, quadratic attention mechanisms, or post-hoc geometry correction, leading to a persistent trade-off between computational efficiency and structural fidelity.\n  Results: We present Physics-Informed Mamba (PI-Mamba), a generative model that enforces exact local covalent geometry by construction while enabling linear-time inference. PI-Mamba integrates a differentiable constraint-enforcement operator into a flow-matching framework and couples it with a Mamba-based state-space architecture. To improve optimisation stability and backbone realism, we introduce a spectral initialization derived from the Rouse polymer model and an auxiliary cis-proline awareness head. Across benchmark tasks, PI-Mamba achieves 0.0\\% local geometry violations and high designability (scTM = $0.91\\pm 0.03$, n = 100), while scaling to proteins exceeding 2,000 residues on a single A5000 GPU (24 GB).", "date": "2026-03-17", "categories": []}
{"id": "2606.04525", "title": "GENEB: Why Genomic Models Are Hard to Compare", "abstract": "Progress in genomic foundation models is difficult to assess due to fragmented benchmarks, incompatible evaluation protocols, and task-specific reporting. As a result, claims of superiority or generality across models are often not directly comparable. We introduce GENEB, a large-scale diagnostic benchmark that evaluates frozen representations from 40 genomic foundation models across 100 tasks spanning 13 functional categories under a unified probing-based protocol, including few-shot regimes. GENEB enables controlled comparison across model scale, architecture, tokenization, and pretraining data while explicitly exposing task-level trade-offs. Our analysis shows that aggregate leaderboards are unstable: model rankings vary sharply across task categories, scale provides only modest and inconsistent gains, and architectural and pretraining alignment frequently outweigh parameter count. These results highlight limitations of current evaluation practices and position GENEB as a reference framework for principled comparison and category-aware model selection in genomic machine learning.", "date": "2026-06-03", "categories": []}
{"id": "2606.12639", "title": "The Metric Picks the Winner: Evaluation Choice Flips Model Rankings for Drug-Response Prediction in Unseen Chemistry", "abstract": "Predicting how a cell's transcriptome responds to a drug it has never seen is a core, hard problem in computational cell biology: recent benchmarks show complex models often fail to beat trivial baselines once test compounds are held out by chemistry. We study one cell line and assay, THP-1 cells profiled by DRUG-seq, scored by the active-compound weighted MSE(wMSE) of the VCPI prediction contest. We propose a staged approach: dumb baselines (untreated control and mean training-compound response) that the field keeps failing to beat; non-parametric retrieval (a Tanimoto-weighted average of a held-out compound's nearest training compounds); and a fusion stage combining a frozen chemistry embedding with retrieval-support features to predict the residual over the mean, with an uncertainty head and gene programs. On the released VCPI THP-1 drug-seq data (14,026 training compounds), under a Bemis-Murcko scaffold split, the model ranking inverts depending on the metric. Under an inverse-variance per-gene proxy, a regularized linear regression on Morgan fingerprints appears to win over the deep models, retrieval, and ChemBERTa -- the textbook \"simple baselines win\" result. But under the contest's true active-set metric (per-(gene, compound) Mejia weights, validated against the official scorer; mean baseline 0.535 vs the organizers' 0.507 reference), that reverses: the deep models win, our fusion decoder significantly beats the linear fingerprint baseline (-0.012 wMSE, paired bootstrap p < 10^-4), and the proxy's winner becomes the worst chemistry-aware predictor. Picking the metric picks the winner -- to our knowledge the first demonstration on real held-out drug chemistry of the metric-calibration effect established largely on genetic perturbation. We release a reproducible pipeline\n  wired to the official scorer that emits a valid submission over the real 1064 x 12,995 grid.", "date": "2026-06-10", "categories": []}
{"id": "2606.12838", "title": "OCOO-T : A Simple and Scalable Virtual Cell Model for Transcriptional Perturbation Response Prediction", "abstract": "Predicting single-cell transcriptional responses to genetic, chemical and cytokine perturbations is a fundamental challenge in computational biology and AI Virtual Cell (AIVC) modeling, with direct implications for drug discovery and the elucidation of gene regulatory networks. Existing approaches often rely on auxiliary cell-state encoders, hierarchical variational autoencoders, dedicated Transformer encoder-decoder modules, or gene-interaction priors to compress high-dimensional expression profiles into latent representations. While effective, these designs increase architectural complexity and may limit scalability and generalizability. This paper introduces OCOO-T, a minimalist flow-matching-based AIVC model for transcriptional perturbation response prediction. OCOO-T utilizes a vanilla Transformer stack that operates directly on continuous gene expression profiles and formulates perturbation response prediction as a continuous-time denoising process. Perturbation embeddings, dosage information, and cell-line/cell-type specificity are integrated through adaptive layer normalization and in-context tokens. Comprehensive evaluations on Tahoe100M, Replogle, and PBMC benchmarks demonstrate that OCOO-T achieves state-of-the-art performance across diverse perturbations and cell types while effectively scaling to long transcriptional profiles through patching and depatching of cellular contexts. By leveraging the simplicity of Transformer-based denoising for single-cell omics, OCOO-T provides an effective and scalable framework for in-silico cellular simulation.", "date": "2026-06-11", "categories": []}
{"id": "2606.12991", "title": "APCyc: Property-Informed Design of Cyclic Peptides via Automated Cyclization", "abstract": "Cyclic peptides represent a promising class of therapeutic compounds in modern drug discovery, often offering improved stability and binding affinity. However, the de novo design of cyclic peptides remains challenging because methods must identify pocket-adaptive cyclization patterns and linkage sites while simultaneously controlling drug-relevant properties. This challenge is particularly pronounced for recent generative models trained predominantly on linear peptide data, which may fail to capture cyclization-specific constraints. To address the limitation, we introduce APCyc, a target-aware de novo cyclic peptide generation framework that explicitly models cyclization and jointly optimizes multiple essential physicochemical properties. By using an expanded residue vocabulary and explicitly encoding cyclization-site and linkage-type information, APCyc learns cyclization-aware representations and leverages Bayesian posterior guidance to steer sampling toward cyclic peptides satisfying multiple property objectives. Experimental results demonstrate that our model learns target-dependent cyclization preferences, and enables effective and controllable multi-property optimization for cyclic peptide design. The source code of this paper is available at https://github.com/HKUSTGZ-ML4Health-Lab/APCyc.", "date": "2026-06-11", "categories": []}
{"id": "2606.13007", "title": "scLLM-DSC: LLM-Knowledge Enhanced Cross-Modal Deep Structural Clustering for Single-Cell RNA Sequencing", "abstract": "Clustering is fundamental to scRNA-seq analysis, serving as a cornerstone for identifying cell populations and resolving tissue heterogeneity. However, existing methods focus on mining numerical statistical patterns, suffering from semantic agnosticism by neglecting the intrinsic biological functions encoded by genes. While Large Language Models (LLMs) offer promising semantic capabilities, their direct adaptation to cell clustering is hindered by the structural mismatch between generative pre-training objectives and discriminative downstream tasks. To bridge this gap, we propose scLLM-DSC, a novel LLM-Knowledge Enhanced Cross-Modal Deep Structural Clustering framework. Diverging from data-driven paradigms, scLLM-DSC establishes a semantically-grounded representation by synergizing two views: a Knowledge-Driven Semantic View derived from NCBI gene priors and contextualized Cell2Sentence embeddings, and a Structure-Aware Topological View extracted via a graph-guided encoder. Crucially, we introduce a cross-modal contrastive alignment mechanism to enforce consistency between biological semantics and transcriptomic features within a unified latent space. Extensive benchmarks demonstrate that scLLM-DSC significantly outperforms eleven state-of-the-art baselines in clustering accuracy.", "date": "2026-06-11", "categories": []}
{"id": "2606.08493", "title": "Querying Counterfactuals on Tissue Graphs with Supervised Disentanglement", "abstract": "Tissue graph counterfactuals ask how a cell's expression would change under altered spatial neighbor contexts. Such queries are central to predicting cell behavior in tissues, but lack a unified definition, with existing methods targeting specific intervention types or treating cells as i.i.d. In this work, we first formalize tissue graph counterfactuals as a class of spatial interventions that either rewire connections between cells (edge perturbation) or modify the expression of their neighbors (node perturbation). We then introduce Cellina (https://cellina.readthedocs.io) - a framework that uses supervised disentanglement to decompose a cell's intrinsic state from its spatial context, using the latter as a conditioning input for counterfactual predictions. Across benchmarks spanning over 2.5 million spatially-resolved cells in colorectal cancer and mouse brain, Cellina outperforms spatially-informed and non-spatial competitors in in-silico graph perturbations, disentanglement, and scalability. Additionally, we show that Cellina reveals biologically distinct cancer subdomains in an unsupervised manner and enables targeted neighbor perturbation simulations.", "date": "2026-06-07", "categories": []}
{"id": "2606.11243", "title": "ProHiFlo: Hierarchical Flow Matching with Functional Guidance for De Novo Protein Generation", "abstract": "De novo protein generation has transformative potential in therapeutic design, enzyme engineering, and synthetic biology. While diffusion-based and flow matching approaches have achieved progress, they typically operate at single resolution and lack mechanisms for incorporating functional constraints. We introduce ProHiFlo, a hierarchical flow matching framework with three innovations: (1) coarse-to-fine generation that models backbone geometry before refining to all-atom coordinates, reducing computational cost while maintaining accuracy; (2) functional guidance leveraging pretrained predictors to steer generation toward desired properties without retraining; (3) adaptive SE(3)-equivariant architecture for efficient multi-scale processing. Experiments on unconditional generation, motif scaffolding, and functional design demonstrate state-ofthe-art performance while requiring 4 fewer sampling steps. On enzyme active site scaffolding, ProHiFlo achieves 58.9% success rate compared to 41.2% for RFDiffusion.", "date": "2026-06-03", "categories": []}
{"id": "2606.11286", "title": "FreeBridge: Variational Schr\\\"odinger Bridges for Cellular Transition Dynamics", "abstract": "High-content imaging assays quantify cellular responses to chemical and genetic perturbations, yet continuous trajectories of individual cells are unobservable because cells are chemically fixed at acquisition. Perturbation modeling therefore reduces to inferring stochastic transport between control and treated populations observed only as separate marginals. While recent generative models achieve strong end-point alignment, boundary consistency does not determine intermediate evolution: multiple stochastic processes may connect identical marginals while traversing regions unsupported by observed single-cell morphologies. We introduce \\textbf{FreeBridge}, a Schr\\\"odinger Bridge formulation for single-cell transition modeling under endpoint-only supervision. FreeBridge defines atomic states as instance-segmented single-cell representations, establishing a fixed cellular manifold, and learns stochastic transport constrained within this geometry via empirical latent support regularization. Across BBBC021, RxRx1, and JUMP, FreeBridge maintains competitive or improved endpoint fidelity and mechanism-of-action retention under a unified evaluation protocol; on BBBC021, it further reduces intermediate support violations. These findings highlight the importance of geometric grounding for biologically interpretable perturbation dynamics. Project page: https://y-research-sbu.github.io/FreeBridge/.", "date": "2026-06-09", "categories": []}
{"id": "2606.11651", "title": "DeepRHP: A Hybrid Variational Autoencoder for Designing Random Heteropolymers as Protein Mimics", "abstract": "Synthetic random heteropolymers (RHPs), consisting of a predefined set of monomers, offer an approach toward the design of protein-like materials. These RHPs, if designed appropriately, can mimic protein behavior and function. As such, there is a need for computational tools to efficiently guide RHP design. We bridge this gap by developing DeepRHP, a modified variational autoencoder (VAE) model under a semi-supervised framework. By equipping a classical VAE with an additional feature-based VAE, DeepRHP forces the latent space to capture structures of critical chemical features as well as individual RHP sequence patterns. In this sense, our method is versatile by allowing any relevant features to be incorporated in a hybrid manner. We demonstrate the effectiveness of DeepRHP by suggesting potential monomer compositions that stabilize membrane proteins (e.g. Aquaporin Z) in non-native environments and cross-validating our prediction with published results. The concordance between our model and true RHP function suggests strong potential in utilizing hybrid autoencoder architectures to guide RHP design for proteins and other biological compounds.", "date": "2026-06-10", "categories": []}
{"id": "2606.11868", "title": "MemNovo: Look Back at the Spectrum for Balanced De Novo Peptide Sequencing from Mass Spectrometry", "abstract": "De novo peptide sequencing from tandem mass spectrometry is pivotal in proteomics, enabling identification of novel peptides without reference databases. While recent Transformer-based encoder-decoder models have achieved remarkable performance, we uncover a critical pathology in their inference dynamics. Through comprehensive feature scaling experiments, we demonstrate that existing auto-regressive peptide decoders tend to over-rely on generated-sequence priors while progressively under-utilizing fine-grained physical evidence from the input mass spectrum. This phenomenon leads to suboptimal results, where generated peptide sequences are biologically plausible yet not faithful to the input spectrum. To rectify this, we propose MemNovo, a training-free and plug-and-play mechanism that re-balances peptide and spectral contributions at inference time. MemNovo alleviates the information bottleneck by establishing a persistent spectral memory bank and injecting retrieved features directly into the final decoding stage via an ultra-conservative residual connection. Theoretical analysis confirms that this mechanism restores the mutual information between the decoder state and the raw spectrum. Extensive experiments on the Nine Species benchmark with two representative baselines, Casanovo and InstaNovo, demonstrate that MemNovo consistently improves both amino acid precision and peptide precision, achieving up to 39.1% relative improvement in peptide precision for Casanovo and up to 3.9% for InstaNovo, with negligible computational overhead.", "date": "2026-06-10", "categories": []}
{"id": "2604.04287", "title": "Entropy, Disagreement, and the Limits of Foundation Models in Genomics", "abstract": "Foundation models in genomics have shown mixed success compared to their counterparts in natural language processing. Yet, the reasons for their limited effectiveness remain poorly understood. In this work, we investigate the role of entropy as a fundamental factor limiting the capacities of such models to learn from their training data and develop foundational capabilities. We train ensembles of models on text and DNA sequences and analyze their predictions, static embeddings, and empirical Fisher information flow. We show that the high entropy of genomic sequences -- from the point of view of unseen token prediction -- leads to near-uniform output distributions, disagreement across models, and unstable static embeddings, even for models that are matched in architecture, training and data. We then demonstrate that models trained on DNA concentrate Fisher information in embedding layers, seemingly failing to exploit inter-token relationships. Our results suggest that self-supervised training from sequences alone may not be applicable to genomic data, calling into question the assumptions underlying current methodologies for training genomic foundation models.", "date": "2026-04-05", "categories": []}
{"id": "2606.02386", "title": "AgentPLM: Agentic Protein Language Models with Reasoning-Augmented Decoding for Protein Sequence Design", "abstract": "Protein language models (PLMs) are passive oracles: they generate sequences in a single forward pass with no mechanism to consult external biophysical feedback or redirect generation when a candidate violates thermodynamic or structural constraints. We introduce AgentPLM, which addresses this by equipping a pre-trained PLM with i) Reasoning-Augmented Decoding (RAD), which interleaves autoregressive generation with tool calls (ESMFold, FoldX, AutoDock Vina), and ii) Contrastive Agent Policy Optimisation (CAPO), a trajectory-level extension of direct preference optimisation that trains the policy end-to-end to learn when oracle feedback is informative rather than merely imitating high-fitness sequences. We evaluate AgentPLM on benchmark tasks spanning de novo enzyme design, antibody optimisation, thermostability, PPI interface design, and zero-shot fitness prediction with standardised oracle APIs and controlled sequence-identity splits. AgentPLM achieves state-of-the-art results with a gain in antibody top-10% hit rate over the strongest passive baseline, providing mechanistic evidence of online error correction without explicit backtracking.", "date": "2026-06-01", "categories": []}
{"id": "2606.09898", "title": "TRAPS: Therapeutic Response Analysis via Pathway-informed Stratification", "abstract": "Cancer treatment planning requires decisions across multiple clinical dimensions at once. Clinicians must determine whether a patient should receive targeted molecular therapy, radiation therapy, and whether they are likely to survive beyond six months. Existing pathway-informed deep learning models have been developed and tested in isolation, making fair comparison across architectures impossible. We present the first unified benchmark for pathway-guided therapy response modeling, evaluating three biologically informed architectures, BINN, GraphPath, and PATH, across five cancer cohorts drawn from The Cancer Genome Atlas, representing 2,622 patients encoded using Reactome pathway activity scores. Each model is trained jointly on all three clinical outcomes under identical data and evaluation conditions, the first study to treat pathway-structured deep learning as a combined therapy and survival prediction problem. Our results show that no single architecture wins across all tasks: PATH performs best for targeted molecular therapy prediction overall, BINN is most reliable for survival prediction, and no model produces useful predictions for radiation therapy, as the key drivers of that decision are clinical variables not captured in gene expression data. Most strikingly, GraphPath achieves an AUROC of 0.92 on prostate targeted molecular therapy prediction, the highest score in the entire benchmark, demonstrating that lateral co-regulation structure produces exceptional discriminative power when matched to a cohort with a narrow targetable driver programme, even under conditions of extreme class imbalance at only 11\\% positive prevalence.", "date": "2026-06-05", "categories": []}
{"id": "2606.10255", "title": "POPSICLE: Benchmark Datasets for Segmentation and Localization in CryoET", "abstract": "Cryo-electron tomography (cryoET) has emerged as a powerful tool in structural and cellular biology by enabling direct visualization of macromolecular structures within intact cells, thereby linking molecular architecture to cellular organization in a native context. Realizing the full potential of cryoET, however, increasingly depends on advances in computational analysis, particularly machine learning (ML), to interpret its complex and information-rich data. Despite rapid progress, ML development for cryoET remains bottlenecked by the lack of standardized, well-annotated benchmarks. Existing evaluations are typically small, task-specific, and are assembled in isolation, limiting robust comparisons across methods. Here, we present POPSICLE, a benchmark suite for cryoET segmentation and macromolecular localization built from the CryoET Data Portal - an open, ML-ready repository of tomographic data, metadata, and annotations. POPSICLE spans eukaryotic and prokaryotic systems, both purified and fully in situ samples, and dense voxel-wise segmentation as well as sparse localization tasks. Built on a living data resource, it can expand as new datasets and annotations become available. Baseline experiments reveal substantial variation in model rankings across tasks, underscoring the need for benchmarks tailored to the unique characteristics of cryoET rather than evaluation practices adapted from adjacent biomedical imaging domains. POPSICLE thus provides an open and extensible foundation for reproducible ML evaluation in cryoET.", "date": "2026-06-08", "categories": []}
{"id": "2606.10543", "title": "Flexible Flows for Biological Sequence Design", "abstract": "Designing functional biological sequences requires navigating vast discrete spaces under strict evolutionary and biophysical constraints. Discrete Flow Matching (DFM) offers a generative framework over such spaces, but existing approaches rely on biologically uninformative couplings and offer limited flexibility for variable-length sequence generation and fine-grained control. We propose a structured coupling that encodes domain-specific preferences among sequence elements, biasing the source distribution toward plausible regions without modifying the flow objective or training procedure. Building on this, we introduce a latent edit-based rate parameterization that models variable-length generation via edit operations conditioned on a shared global latent, akin to a latent variable model, while remaining tractable. We further introduce a latent classifier-free guidance mechanism that steers generation coherently in continuous latent space, along with Dirichlet-prior temperature scaling for test-time control over edit operations. Our method achieves state-of-the-art performance across diverse biological sequence tasks, including density estimation, unconditional and conditional DNA sequence generation, and peptide sequence generation.", "date": "2026-06-09", "categories": []}
{"id": "2507.08920", "title": "AMix-1: A Pathway to Test-Time Scalable Protein Foundation Model", "abstract": "We introduce AMix-1, a powerful protein foundation model built on Bayesian Flow Networks and empowered by a systematic training methodology, encompassing pretraining scaling laws, emergent capability analysis, in-context learning mechanism, and test-time scaling algorithm. To guarantee robust scalability, we establish a predictive scaling law and reveal the progressive emergence of structural understanding via loss perspective, culminating in a strong 1.7-billion model. Building on this foundation, we devise a multiple sequence alignment (MSA)-based in-context learning strategy to unify protein design into a general framework, where AMix-1 recognizes deep evolutionary signals among MSAs and consistently generates structurally and functionally coherent proteins. This framework enables the successful design of a dramatically improved AmeR variant with an up to $50\\times$ activity increase over its wild type. Pushing the boundaries of protein engineering, we further empower AMix-1 with an evolutionary test-time scaling algorithm for in silico directed evolution that delivers substantial, scalable performance gains as verification budgets are intensified, laying the groundwork for next-generation lab-in-the-loop protein design.", "date": "2025-07-11", "categories": []}
{"id": "2602.15253", "title": "Scaling Laws for Masked-Reconstruction Transformers on Single-Cell Transcriptomics", "abstract": "Neural scaling laws -- power-law relationships between loss, model size, and data -- have been extensively documented for language and vision transformers, yet their existence in single-cell genomics remains largely unexplored. We present the first systematic study of scaling behaviour for masked-reconstruction transformers trained on single-cell RNA sequencing (scRNA-seq) data. Using expression profiles from the CELLxGENE Census, we construct two experimental regimes: a data-rich regime (512 highly variable genes, 200,000 cells) and a data-limited regime (1,024 genes, 10,000 cells). Across seven model sizes spanning three orders of magnitude in parameter count (533 to 3.4 x 10^8 parameters), we fit the parametric scaling law to validation mean squared error (MSE). The data-rich regime exhibits clear power-law scaling with an irreducible loss floor of c ~ 1.44, while the data-limited regime shows negligible scaling, indicating that model capacity is not the binding constraint when data are scarce. These results establish that scaling laws analogous to those observed in natural language processing do emerge in single-cell transcriptomics when sufficient data are available, and they identify the data-to-parameter ratio as a critical determinant of scaling behaviour. A preliminary conversion of the data-rich asymptotic floor to information-theoretic units yields an estimate of approximately 2.30 bits of entropy per masked gene position. We discuss implications for the design of single-cell foundation models and outline the additional measurements needed to refine this entropy estimate.", "date": "2026-02-16", "categories": []}
{"id": "2603.13431", "title": "CHIMERA-Bench: A Benchmark Dataset for Epitope-Specific Antibody Design", "abstract": "Computational antibody design has seen rapid methodological progress, with dozens of deep generative methods proposed in the past three years, yet the field lacks a standardized benchmark for fair comparison and model development. These methods are evaluated on different SAbDab snapshots, non-overlapping test sets, and incompatible metrics, and the literature fragments the design problem into numerous sub-tasks with no common definition. We introduce CHIMERA-Bench: (CDR Modeling with Epitope-guided Redesign), a unified benchmark built around a single canonical task: epitope-conditioned CDR sequence-structure co-design. CHIMERA-Bench provides three components. The first is a curated, deduplicated dataset of 2,922 antibody-antigen complexes with epitope and paratope annotations. The second is a set of three biologically motivated splits that test generalization to unseen epitopes, unseen antigen folds, and prospective temporal targets. The third is a comprehensive evaluation protocol with five metric groups, including novel epitope-specificity measures. We benchmark eleven methods spanning six generative paradigms and report results across all splits. CHIMERA-Bench is the largest dataset of its kind for the antibody design problem, allowing the community to develop and test novel methods and evaluate their generalizability.", "date": "2026-03-13", "categories": []}
{"id": "2604.26498", "title": "Do Larger Models Really Win in Drug Discovery? A Benchmark Assessment of Model Scaling in AI-Driven Molecular Property and Activity Prediction", "abstract": "The rapid growth of molecular foundation models and large language models (LLMs) has encouraged a scale centred view of AI in drug discovery, in which larger pretrained models are expected to supersede compact cheminformatics models. We test this assumption across 26 ADME, toxicity and bioactivity endpoints, covering 165,541 endpoint level compound label records. The benchmark contains 78 endpoint and split entries evaluated under random, Murcko scaffold and structure separated 5-fold cross validation protocols, representing increasing chemical generalization difficulty. Across 156 task and metric comparisons, classical machine learning (ML) provides the largest share of best performing entries (47.4%), followed by pretrained molecular sequence models (28.8%), graph neural networks (21.8%) and LLM based SAR baselines (1.9%). Classical ML dominates random split interpolation and remains the largest winner family overall. GNN and sequence models are competitive in selected harder splits, but their strict winner shares decrease under a fixed final-window readout, indicating sensitivity to training settings and model selection. Paired bootstrap analyses show that small numerical differences between individual models should not be read as decisive victories. SAR knowledge from training folds improves GPT5.5-SAR and Opus4.7-SAR metrics but does not make rule based reasoning a universal substitute for supervised predictors. Compact specialized models remain highly effective, and predictive performance depends on the fit among model, task and validation scenario, not on scale alone.", "date": "2026-04-29", "categories": []}
{"id": "2605.31014", "title": "SDM-Q: Cost-Aware Staged Decision-Making for Multi-Omics Classification with Deep Q-Learning", "abstract": "Multi-omics data provide complementary molecular characterizations of disease phenotypes and play an important role in disease diagnosis and subtype classification in precision medicine. However, acquiring complete multi-omics profiles is expensive and time-consuming, while most existing deep learning methods assume full modality availability during inference, resulting in substantial redundancy and limited practicality in clinical settings. To address this issue, we propose SDM-Q, a reinforcement learning framework for adaptive and cost-aware multi-omics classification. Specifically, multi-omics diagnosis is reformulated as a finite-horizon sequential decision problem, where the currently acquired omics modalities define the diagnostic state at each stage. An action--value function determines whether to acquire an additional modality or terminate the decision process and output the final prediction. To balance diagnostic utility and acquisition cost, the reward is defined only at the terminal stage and jointly determined by classification correctness and cumulative modality acquisition cost. A backward stage-wise optimization strategy is introduced to improve policy consistency and training stability. Experiments on four public multi-omics datasets, including ROSMAP, LGG, BRCA, and KIPAN, demonstrate that SDM-Q effectively reduces redundant modality acquisition while maintaining competitive classification performance compared with methods using complete multi-omics inputs. In the BRCA and KIPAN datasets, more than 99\\% and 95\\% of subjects, respectively, achieve accurate classification using only a single omics modality, while the average number of acquired modalities remains below two for ROSMAP and LGG. These results suggest that cost-aware sequential decision-making provides an effective paradigm for improving the efficiency of precision medicine workflows.", "date": "2026-05-29", "categories": []}
{"id": "2606.00568", "title": "On the Recoverability of Causal Relations from Bulk Gene Expression Data", "abstract": "Bulk gene expression profiling, which aggregates pooled RNA across cells within a biological sample, remains important in the single-cell era because it is typically less noisy, more sensitive, and more cost-effective than single-cell assays. Accordingly, a growing body of computational methods seeks to recover causal relations among genes from bulk expression data. However, aggregation is a lossy, non-invertible coarsening of the underlying cellular system, and it remains unclear whether and under what conditions causal relations are recoverable from aggregated bulk gene expression data. To answer this, we formalize recoverability under aggregation through two notions of consistency: functional-form consistency and conditional-independence consistency. We then derive necessary and sufficient conditions for recoverability, showing that these properties are preserved only under linear aggregations (e.g., sum/mean) coupled with affine structural equations. To assess the practical plausibility of these conditions, analyses of four bulk and four single-cell gene expression datasets further reveal that the estimated pairwise regulatory functions among genes deviate from linearity in both data types, providing limited empirical support for the linearity assumptions required for recoverability. Together, these results caution against recovering causal relations from aggregated bulk expression data without strong additional assumptions.", "date": "2026-05-30", "categories": []}
{"id": "2606.07562", "title": "The Montparnasse Algorithm for RNA Design", "abstract": "RNA design consists of discovering a nucleotide sequence that optimizes predefined criteria, such as secondary structure. It is useful for synthetic biology, medicine, and nanotechnology. We propose Montparnasse, a Monte Carlo search framework based on Generalized Nested Rollout Policy Adaptation, augmented with a problem-specific prior, slow and long adaptation at level 1, and a lexicographic multicriteria evaluation. Montparnasse solves all 100 puzzles of the Eterna100 V1 benchmark consistently faster than DesiRNA, the previous state of the art, across all time limits, reaching full coverage more than three times faster overall. On messenger RNA secondary structure optimization for hemoglobin alpha, it identifies sequences with more paired bases than the MFE-optimal solution of LinearDesign.", "date": "2026-05-25", "categories": []}
{"id": "2606.07567", "title": "SurfDesign: Effective Protein Design on Molecular Surfaces", "abstract": "Protein function is largely determined by molecular surface geometry and physicochemical complementarity, yet most protein design methods condition only on backbone structure. We introduce SurfDesign, a surface-conditioned protein design framework that models molecular surfaces as continuous geometric manifolds and integrates them with pretrained protein language models. SurfDesign employs surface-based equivariant message passing to capture surface normals, curvature, and directional geometry, together with a parameter-efficient fine-tuning strategy. Focusing on functional protein design, we show that SurfDesign consistently outperforms prior surface-conditioned and backbone-only methods on de novo binder and enzyme design benchmarks. We also report strong performance on inverse-folding benchmarks as a diagnostic of structural compatibility. Our results highlight manifold-aware surface representations as a principled foundation for functional protein and enzyme design. Code is available at https://github.com/smiles724/SurfDesign.", "date": "2026-05-25", "categories": []}
{"id": "2606.07686", "title": "Knowledge-Inclusive Adaptive Physics-Informed Neural Network for Microbial Interaction Modelling", "abstract": "Physics-Informed Neural Network (PINN) is a way of including knowledge in the form of equations in Machine Learning methods. Beyond equations, knowledge exists in other forms, such as text and network structure. While existing PINN-based approaches discover equation parameters from data, they rely solely on experimental measurements. We propose a new PINN framework that enriches parameter discovery by incorporating auxiliary knowledge sources. We instantiate our framework for microbiology, where generalised Lotka-Volterra (gLV) serves as a biological foundation for modelling microbial communities. We demonstrate that incorporating knowledge improves microbial community modelling. Our framework enriches the gLV parameters using peer-reviewed metagenomics literature, as text provides biological context on external influences that gLV alone cannot capture. We combine this knowledge with experimental measurements of microbial abundance using a data-driven integration approach. We integrate network-based structural knowledge by explicitly modelling microbial interactions. Our knowledge-inclusive framework infers microbial networks, revealing ecological insights. We validate these findings against ecological roles documented in the literature. We evaluate on real and simulated datasets spanning human- and plant-associated microbial communities. Our framework improves over the state-of-the-art by up to 53%, even without knowledge. Knowledge addition yields gains of up to 23% in Bray-Curtis Dissimilarity-based accuracy and 47% in $\\mathrm{R}^2$.", "date": "2026-06-05", "categories": []}
{"id": "2606.07700", "title": "EssentialGIN: a new approach for gene essentiality prediction based on graph isomorphism neural networks", "abstract": "Background: Prediction of essential genes (proteins), is a basic and challenging problem but at the same time very costly and time-consuming in wet-lab experiments. Predicting essential genes, only based on computational methods (to introduce wet-lab candidates) using centrality measures are not accurate and result in large number of false positives; therefore, more complex models such as deep learning and also integration of biological information are used in recent research to identify essential genes.\n  Methods: In this work we focus on graph isomorphism networks, in order to embed proteins as a node in PPI network to conserve topological features of PPI network, and also integrate biological data such as gene expression data, gene orthology information and gene subcellular localization information, and introduced a deep architecture for predicting essential genes. Graph isomorphism network architecture is modified in this work for embedding node information.\n  Results: Our experiments proved that the proposed method outperforms baseline centrality-based methods and also machine learning based methods such as Node2Vec, MLP, and also graph attention networks (GAT).\n  Conclusion: In this paper we observed that using graph isomorphism networks that integrate biological data (as node attributes) and preserve network topology can significantly improve the essential gene prediction accuracy. In simpler organisms such as E. coli and D. melanogaster, methods such as multi-layer perceptron using Node2Vec embedding also performs very good, but in H. sapiens the introduced architecture significantly outperforms deep learning and other graph neural network solutions.\n  Keywords: Essential gene prediction, graph neural network, graph isomorphism network, PPI network, node embedding", "date": "2026-06-05", "categories": []}
{"id": "2606.07760", "title": "scCBGM: Interpretable Single-Cell Counterfactual Editing", "abstract": "Understanding cellular phenotypes and how they respond to perturbations is critical for disease biology and therapeutic design. Single-cell RNA sequencing enables characterization at cellular resolution, yet the combinatorial space of conditions makes exhaustive experimental mapping infeasible. We introduce single-cell Concept Bottleneck Generative Models (scCBGM), a framework for interpretable and precise counterfactual editing of individual cells. scCBGM adapts concept bottleneck architectures for single-cell data through decoder skip connections and a cross-covariance penalty that promotes disentanglement without dimensional constraints. We extend the framework to flow matching models, enabling concept-guided editing in both encoding-decoding and generation regimes. To enable rigorous evaluation, we develop a synthetic benchmark with ground-truth counterfactuals. Across multiple real datasets, scCBGM demonstrates superior performance in combinatorial generalization and counterfactual prediction, supported by cell-level validation on synthetic data and population-level benchmarks on real datasets.", "date": "2026-06-05", "categories": []}
{"id": "2606.08147", "title": "Biological Reasoning-Informed Regression for Interpretable Regulatory DNA Activity Prediction", "abstract": "DNA cis-regulatory elements (CREs) such as enhancers control gene expression levels. Accurately predicting regulatory activity from DNA sequences is valuable but challenging, as it requires understanding complex biological regulatory processes. Existing methods typically regress activity scores from sequences in a black-box manner, limiting both interpretability and regression performance. Meanwhile, large language models (LLMs) benefit from explicit reasoning processes, yet directly applying LLMs to raw DNA sequences performs poorly. In this paper, we bridge this gap by introducing R3LM, a framework that teaches LLMs reasoning-informed regression on regulatory DNA through structured biological knowledge. Specifically, we design a biologically grounded data format that structures DNA's regulatory information for improved LLM understanding, and construct CRE-ReasonBench, the first dataset that associates DNA sequences and activity scores with mechanistic reasoning traces. Through two-stage training that first teaches LLMs reasoning over structured biological information then performs regression, R3LM achieves state-of-the-art performance on enhancer prediction across three cell types, outperforming both LLMs with raw sequence input and specialized DNA models while providing interpretable mechanistic explanations. We expect R3LM as an interpretable reward model that can effectively assist biologists in CRE design. Code is available at https://github.com/DuanYi516/R3LM.", "date": "2026-06-06", "categories": []}
{"id": "2606.08375", "title": "Few-step Cofolding with All-Atom Flow Maps", "abstract": "All-atom generative modeling of 3D biomolecular complexes has emerged as the dominant paradigm for predicting the structure of proteins and protein-ligand systems. Generating structures at the atomic level of fidelity, however, typically requires expensive iterative diffusion rollouts, making both conventional deployment and inference-time search techniques computationally costly. In this paper, we introduce the Denoiser Cofolding All-Atom Flowmap (DeCAF) framework for distilling state-of-the-art all-atom cofolding models into all-atom flow maps that produce high-quality samples in only a few inference steps. We build DeCAF on a denoiser-based formulation of flow maps with endpoint losses that naturally support SE(3) rigid alignment, which we show is critical for training accurate models. We further derive a simple change of variables that lets DeCAF operate in the {\\sigma}-space noise schedule of EDM-style architectures, enabling direct distillation from pretrained cofolding diffusion models. Equipped with DeCAF's flowmap lookahead, we introduce a purpose-built inference-time framework that improves sampling through reward-guided search. Empirically, DeCAF-Boltz statistically improves over Boltz-1x in both accuracy (RMSD) and physical validity scores of protein-ligand poses at strict NFE budgets on the challenging Runs N' Poses, while also showing a more optimal Pareto frontier across all inference compute budgets on PoseBusters. Distilling the state-of-the-art Pearl cofolding model, DeCAF-Pearl outperforms diffusion-based cofolding models and matches its teacher on success rate while using 5x fewer NFEs. We release our code at https://github.com/genesistherapeutics/decaf.", "date": "2026-06-07", "categories": []}
{"id": "2606.08816", "title": "Knowledge Graphs and Reasoning LLMs for Finding Simple Yet Effective Transcriptomic Perturbation Predictors", "abstract": "Predicting the effect of an unseen gene knockout perturbation on transcriptomic gene expression remains a highly challenging problem for virtual cell models. Recent progress has been made by leveraging biological knowledge graphs to provide a notion of similar perturbation, allowing for improved extrapolation beyond the set of training perturbations. In this work, we demonstrate that the simplest model to leverage these assumptions - a K-nearest neighbour from the knowledge graph - achieves highly competitive performance on this task, and that this can be improved further using LLMs optimised via reinforcement learning (RL) for predictive performance. Specifically, we find that the K-nearest neighbour approach beats almost all methods on out-of-distribution perturbation prediction, and when a reasoning LLM is trained via RL to make changes to the neighbourhood, it obtains equivalent performance to current state of the art methods on the cell lines from Replogle et al. (2022). We also demonstrate that the RL training improves the LLM's performance on the downstream task of differential expression prediction, despite not being trained on this directly. Overall, these findings demonstrate the efficacy of knowledge graphs as model priors, and show early signs that RL can refine LLMs into generalizable tools for predicting complex biological responses.", "date": "2026-06-07", "categories": []}
{"id": "2606.09558", "title": "Integrating gene regulatory priors into Transformer attention with scTransformer for interpretable scRNA-seq analysis", "abstract": "Motivation: Transformer-based models are increasingly applied to large-scale single-cell transcriptomics, showing strong performance through self-supervised learning on millions of cells. However, most existing approaches treat genes as independent features, and largely ignore prior biological knowledge, which limits interpretability and robustness. In this paper, we explore whether explicitly incorporating gene regulatory information can improve both model performance and biological insight. Results: We present scTransformer, the first Transformer-based approach that builds a priori knowledge of biological mechanisms into the model's attention patterns. By constraining information flow according to known regulatory structures, the model learns representations that are more biologically meaningful. We evaluate scTransformer on a disease-relevant single-nucleus RNA-seq dataset using supervised cell-type classification. Compared to standard Transformers, our approach improves classification accuracy, enhances separation of cell types in embedding space, and produces attention patterns consistent with known regulatory programs. Overall, our results demonstrate that embedding biological structure into Transformer models can enhance interpretability without sacrificing performance, offering a principled step toward biologically grounded foundation models for single-cell omics.", "date": "2026-06-08", "categories": []}
{"id": "2602.02014", "title": "Rethinking Genomic Modeling Through Optical Character Recognition", "abstract": "Recent genomic foundation models largely adopt large language model architectures that treat DNA as a one-dimensional token sequence. However, exhaustive sequential reading is structurally misaligned with sparse and discontinuous genomic semantics, leading to wasted computation on low-information background and preventing understanding-driven compression for long contexts. Here, we present OpticalDNA, a vision-based framework that reframes genomic modeling as Optical Character Recognition (OCR)-style document understanding. OpticalDNA renders DNA into structured visual layouts and trains an OCR-capable vision--language model with a visual DNA encoder and a document decoder, where the encoder produces compact, reconstructible visual tokens for high-fidelity compression. Building on this representation, OpticalDNA defines prompt-conditioned objectives over core genomic primitives-reading, region grounding, subsequence retrieval, and masked span completion-thereby learning layout-aware DNA representations that retain fine-grained genomic information under a reduced effective token budget. Across diverse genomic benchmarks, OpticalDNA consistently outperforms recent baselines; on sequences up to 450k bases, it achieves the best overall performance with nearly 20$\\times$ fewer effective tokens, and surpasses models with up to 985$\\times$ more activated parameters while tuning only 256k trainable parameters.", "date": "2026-02-02", "categories": []}
{"id": "2606.06717", "title": "ShallowBench: Benchmarking Generative Drug Design Models on Shallow-Pocket Targets", "abstract": "While generative AI models have demonstrated remarkable success in structure-based drug design, they predominantly rely on deep binding pockets and struggle to sample effective ligands for challenging low-pocketability targets, such as the historically \"undruggable\" oncology targets KRAS and MYC. To address this gap, we introduce ShallowBench, a strictly curated benchmark of 5,780 shallow-pocket targets extracted from CrossDocked2020. By computing the difference between an Alpha Shape \"lid\" volume and the underlying protein atom voxel volume, we successfully isolated targets with low concavity while ensuring sufficient surface area for binding. Evaluating various state-of-the-art generative models reveals weaker predicted binding affinity on these low-concavity interfaces. ShallowBench therefore provides a rigorous benchmark for generative biology models and highlights the necessity of new architectural innovations or loss functions capable of navigating these challenging targets.", "date": "2026-06-04", "categories": []}
{"id": "2506.11152", "title": "HEIST: A Graph Foundation Model for Spatial Transcriptomics and Proteomics Data", "abstract": "Single-cell transcriptomics and proteomics have become a great source for data-driven insights into biology, enabling the use of advanced deep learning methods to understand cellular heterogeneity and gene expression at the single-cell level. With the advent of spatial-omics data, we have the promise of characterizing cells within their tissue context as it provides both spatial coordinates and intra-cellular transcriptional or protein counts. Proteomics offers a complementary view by directly measuring proteins, which are the primary effectors of cellular function and key therapeutic targets. However, existing models either ignore the spatial information or the complex genetic and proteomic programs within cells. Thus they cannot infer how cell internal regulation adapts to microenvironmental cues. Furthermore, these models often utilize fixed gene vocabularies, hindering their generalizability unseen genes. In this paper, we introduce HEIST, a hierarchical graph transformer foundation model for spatial transcriptomics and proteomics. HEIST models tissues as hierarchical graphs. The higher level graph is a spatial cell graph, and each cell in turn, is represented by its lower level gene co-expression network graph. HEIST achieves this by performing both intra-level and cross-level message passing to utilize the hierarchy in its embeddings and can thus generalize to novel datatypes including spatial proteomics without retraining. HEIST is pretrained on 22.3M cells from 124 tissues across 15 organs using spatially-aware contrastive and masked autoencoding objectives. Unsupervised analysis of HEIST embeddings reveals spatially informed subpopulations missed by prior models. Downstream evaluations demonstrate generalizability to proteomics data and state-of-the-art performance in clinical outcome prediction, cell type annotation, and gene imputation across multiple technologies.", "date": "2025-06-11", "categories": []}
{"id": "2606.05198", "title": "An accurate nucleic acid-small molecule docking framework via geometric deep learning with large-scale pretraining", "abstract": "Nucleic acids are increasingly recognized as therapeutic targets beyond conventional protein-centered drug discovery, yet accurate and efficient docking of small molecules to nucleic acid structures remains challenging. Physics-based docking methods often show limited accuracy and efficiency, whereas deep learning approaches are constrained by the scarcity of experimentally resolved nucleic acid-ligand complexes. Here, we present NucleoDock, a deep learning framework for nucleic acid-small molecule docking. To address data scarcity, NucleoDock combines physics-guided large-scale pretraining on millions of docking-generated synthetic complexes with fine-tuning on curated experimental co-crystal structures. It further integrates sequence- and structure-informed nucleotide representations with atomistic three-dimensional features to capture both biological context and binding-site geometry. A mixture density network-based geometric scoring head is used to model conditional interaction-distance distributions for pose ranking. On an external benchmark of 125 nucleic acid-ligand complexes, NucleoDock achieved a top-1 success rate of 56 percent at an RMSD cutoff of 2.0 Angstrom, outperforming rDock with 29 percent, while generating 100 poses in approximately 5 seconds per complex. Retrospective virtual screening on the ROBIN benchmark further showed improved early enrichment. NucleoDock represents a step toward bridging the methodological gap between protein- and nucleic acid-directed computational drug discovery.", "date": "2026-05-22", "categories": []}
{"id": "2606.05474", "title": "AlloGen: Conformation-Selective Binder Generation with Differential State Scoring", "abstract": "Protein binder design has largely optimized for affinity alone, leaving conformational selectivity unaddressed: for allosteric targets such as kinases, nuclear receptors, and GPCRs, a binder that engages both active and inactive states provides no functional specificity regardless of how tightly it binds. We introduce AlloGen, a modular framework that decouples backbone generation from a learned state-selectivity scorer $Q_\\theta$, an SE(3)-invariant interface graph transformer trained via a two-phase curriculum that first learns interface geometry before imposing conformational discrimination. Because $Q_\\theta$ is fully differentiable and generator-agnostic, it integrates with any backbone generator as a passive reranker or an active gradient-based guide without retraining. Across a diverse benchmark of proteins spanning multiple families and conformational mechanisms, AlloGen consistently identifies binders that preferentially recognize desired structural states while rejecting alternative conformations. Experimental validation on calmodulin further demonstrates that these computational selectivity signals translate to physical molecules, yielding de novo peptides that bind the desired holo conformation while exhibiting no detectable binding to the apo state. Together, these results establish conformational selectivity as a learnable property and provide a general framework for state-selective protein binder design.", "date": "2026-06-03", "categories": []}
{"id": "2606.06117", "title": "$p$-adic Bi-Filtrations for Topological Machine Learning on Genomic Sequences", "abstract": "We introduce pVR, a topological machine learning framework for alignment-free genomic sequence classification that combines $p$-adic numbers with topological data analysis. Each DNA sequence is encoded along two complementary axes: a $p$-adic distance on $k$-mer prefixes, which captures hierarchical positional structure, and a compositional $L_1$ distance on $k$-mer frequencies, which captures local sequence content. The two distances jointly parameterise a bi-filtered Vietoris--Rips complex, and per-sequence topological summaries from this bi-filtration serve as features for standard machine learning classifiers. We establish theoretical guarantees for the construction: stability under metric perturbations and invariance to the choice of prime, alongside a result that explains why a single $p$-adic axis is topologically uninformative and why the bi-filtration recovers nontrivial homology. On twelve genomic benchmarks ($28$ to $500$ sequences, $3$ to $7$ classes), pVR outperforms four established alignment-free baselines on three of six low-sample datasets, with gains of up to $21$ percentage points; it underperforms only on a SARS-CoV-2 variant benchmark whose point-mutation divergence violates the hierarchical assumption, and all methods saturate in the large-sample regime. pVR also outperforms zero-shot frozen embeddings from the 500M-parameter Nucleotide Transformer v2 by $6.7$ to $11.4$ percentage points on three low-sample benchmarks. The pVR codebase is publicly available at https://github.com/MAHI-Group/pVR.", "date": "2026-06-04", "categories": []}
{"id": "2509.03351", "title": "epiGPTope: A machine learning-based epitope generator and classifier", "abstract": "Epitopes are short antigenic peptide sequences which are recognized by antibodies or immune cell receptors. These are central to the development of immunotherapies, vaccines, and diagnostics. However, the rational design of synthetic epitope libraries is challenging due to the large combinatorial sequence space, $20^n$ combinations for linear epitopes of n amino acids, making screening and testing unfeasible, even with high throughput experimental techniques. In this study, we present a large language model, epiGPTope, pre-trained on protein data and specifically fine-tuned on linear epitopes, which for the first time can directly generate novel epitope-like sequences, which are found to possess statistical properties analogous to the ones of known epitopes. This generative approach can be used to prepare libraries of epitope candidate sequences. We further train statistical classifiers to predict whether an epitope sequence is of bacterial or viral origin, thus narrowing the candidate library and increasing the likelihood of identifying specific epitopes. We propose that such combination of generative and predictive models can be of assistance in epitope discovery. The approach uses only primary amino acid sequences of linear epitopes, bypassing the need for a geometric framework or hand-crafted features of the sequences. By developing a method to create biologically feasible sequences, we anticipate faster and more cost-effective generation and screening of synthetic epitopes, with relevant applications in the development of new biotechnologies.", "date": "2025-09-03", "categories": []}
{"id": "2510.17064", "title": "BRAINCELL-AID: An Agentic AI Created Brain Cell Type Resource for Community Annotation", "abstract": "Single-cell RNA sequencing has transformed our ability to identify diverse cell types and their transcriptomic signatures. However, annotating these signatures-especially those involving poorly characterized genes-remains a major challenge. Traditional methods, such as Gene Set Enrichment Analysis (GSEA), depend on well-curated annotations and often perform poorly in these contexts. Large Language Models (LLMs) offer a promising alternative but struggle to represent complex biological knowledge within structured ontologies. To address this, we present BRAINCELL-AID (BRAINCELL-AID: https://biodataai.uth.edu/BRAINCELL-AID), a novel multi-agent AI system that integrates free-text descriptions with ontology labels to enable more accurate and robust gene set annotation. By incorporating retrieval-augmented generation (RAG), we developed a robust agentic workflow that refines predictions using relevant PubMed literature, reducing hallucinations and enhancing interpretability. Using this workflow, we achieved correct annotations for 77% of mouse gene sets among their top predictions. Applying this approach, we annotated 5,322 brain cell clusters from the comprehensive mouse brain cell atlas generated by the BRAIN Initiative Cell Census Network, enabling novel insights into brain cell function by identifying region-specific gene co-expression patterns and inferring functional roles of gene ensembles. BRAINCELL-AID also identifies Basal Ganglia-related cell types with neurologically meaningful descriptions. Hence, we create a valuable resource to support community-driven cell type annotation.", "date": "2025-10-20", "categories": []}
{"id": "2512.17678", "title": "You Only Train Once: Differentiable Subset Selection for Omics Data", "abstract": "Selecting compact and informative gene subsets from single-cell transcriptomic data is essential for biomarker discovery, improving interpretability, and cost-effective profiling. However, most existing feature selection approaches either operate as multi-stage pipelines or rely on post hoc feature attribution, making selection and prediction weakly coupled. In this work, we present YOTO (you only train once), an end-to-end framework that jointly identifies discrete gene subsets and performs prediction within a single differentiable architecture. In our model, the prediction task directly guides which genes are selected, while the learned subsets, in turn, shape the predictive representation. This closed feedback loop enables the model to iteratively refine both what it selects and how it predicts during training. Unlike existing approaches, YOTO enforces sparsity so that only the selected genes contribute to inference, eliminating the need to train additional downstream classifiers. Through a multi-task learning design, the model learns shared representations across related objectives, allowing partially labeled datasets to inform one another, and discovering gene subsets that generalize across tasks without additional training steps. We evaluate YOTO on two representative single-cell RNA-seq datasets, showing that it consistently outperforms state-of-the-art baselines. These results demonstrate that sparse, end-to-end, multi-task gene subset selection improves predictive performance and yields compact and meaningful gene subsets, advancing biomarker discovery and single-cell analysis.", "date": "2025-12-19", "categories": []}
{"id": "2602.11189", "title": "MuCO: Generative Peptide Cyclization Empowered by Multi-stage Conformation Optimization", "abstract": "Modeling peptide cyclization is critical for the virtual screening of candidate peptides with desirable physical and pharmaceutical properties. This task is challenging because a cyclic peptide often exhibits diverse, ring-shaped conformations, which cannot be well captured by deterministic prediction models derived from linear peptide folding. In this study, we propose MuCO (Multi-stage Conformation Optimization), a generative peptide cyclization method that models the distribution of cyclic peptide conformations conditioned on the corresponding linear peptide. In principle, MuCO decouples the peptide cyclization task into three stages: topology-aware backbone design, generative side-chain packing, and physics-aware all-atom optimization, thereby generating and optimizing conformations of cyclic peptides in a coarse-to-fine manner. This multi-stage framework enables an efficient parallel sampling strategy for conformation generation and allows for rapid exploration of diverse, low-energy conformations. Experiments on the large-scale CPSea dataset demonstrate that MuCO significantly and consistently outperforms state-of-the-art methods in physical stability, structural diversity, secondary structure recovery, and computational efficiency, making it a promising computational tool for exploring and designing cyclic peptides. The demo of the proposed method can be found at https://github.com/mianqiu00/MuCO.", "date": "2026-01-30", "categories": []}
{"id": "2605.00182", "title": "Towards A Generative Protein Evolution Machine with DPLM-Evo", "abstract": "Proteins are shaped by gradual evolution under biophysical and functional constraints. Protein language models learn rich evolutionary constraints from large-scale sequences, and discrete diffusion-based protein language models~(\\eg, DPLMs) are promising for both understanding and generation. However, existing DPLMs typically rely on masked diffusion that contradicts a simple biological intuition: proteins evolve through accumulated edits, not by emerging from masks. Consequently, these frameworks lack explicit pretraining objectives for substitution and insertion/deletion (indel) operations, limiting both optimization-style post-editing and flexible guided generation. To address these limitations, we present DPLM-Evo, an evolutionary discrete diffusion framework that explicitly predicts substitution, insertion, and deletion operations during denoising. DPLM-Evo decouples an upsampled-length latent alignment space from the variable-length observed sequence space, which makes indel-aware generation tractable. To better align substitutions with real evolution, we further introduce a contextualized evolutionary noising kernel that produces biologically informed, context-dependent mutation patterns. Across tasks, DPLM-Evo improves sequence understanding and achieves state-of-the-art mutation effect prediction performance on ProteinGym in the single-sequence setting. It also enables variable-length simulated evolution, and post-editing/optimization of existing proteins via explicit edit trajectories.", "date": "2026-04-30", "categories": []}
{"id": "2605.16331", "title": "Retrieval and competition: how a protein foundation model starts a protein", "abstract": "Protein language models are increasingly used to guide experimental and clinical decisions, yet it is often unclear whether a confident prediction reflects recognition of biological evidence or retrieval of a statistical default. We examine this distinction for a near-universal biological rule, that proteins begin with methionine, by tracing the computational pathway through which ESM2-8M produces this prediction. The model does not detect methionine at the masked position. Instead, it retrieves a methionine-favouring signal from a reference representation at the beginning-of-sequence token via a position-specific query assembled across layers, with the final output emerging through competition with context-dependent circuits. To understand how positional information reaches the readout, we introduce a norm-direction decomposition of attention scores within rotary frequency bands. Positional encoding operates through coupled changes in query norm and angular alignment distributed across these bands. On sequences whose true N-terminus is not methionine, where the biological question matters, the model predicts methionine anyway. This is not a correct prediction produced by an unexpected mechanism, but the output of a positional-prior retrieval circuit that matches the statistical average and fails where biology diverges from it. Distinguishing the two requires resolution at the level of individual circuits, frequency bands, and query composition, suggesting that mechanistic verification will be necessary, and challenging, for predictions where the biological stakes are higher. Even for the simplest biological rule, the model's prediction is mediated by a distributed computational circuit rather than direct recognition, suggesting that increasing task complexity will further obscure the relationship between model confidence and underlying biological evidence.", "date": "2026-05-05", "categories": []}
{"id": "2606.04021", "title": "Structure-Aware Prediction of PROTAC-Mediated Protein Degradability via Graph Neural Networks", "abstract": "Proteolysis-targeting chimeras (PROTACs) can selectively degrade disease-causing proteins, yet predicting which targets are amenable to degradation remains a critical bottleneck: existing computational methods require the complete PROTAC molecular structure, information unavailable before synthesis. We present DegradoMap, a graph neural network that predicts PROTAC-mediated degradability from protein structure and E3 ligase identity alone -- the minimal information available at the target selection stage. The model encodes biophysical priors through lysine-weighted graph pooling with per-protein normalization, models protein-E3 compatibility via cross-attention, and integrates cellular context from the Cancer Dependency Map. On the PROTAC-8K benchmark (3,101 samples, 155 targets, 10 E3 ligases), DegradoMap achieves 0.646+-0.124 AUROC on target-unseen evaluation (best seed: 0.7449) and 0.811 AUROC on CRBN->VHL E3-unseen transfer, outperforming GNN and machine learning baselines. The model additionally recommends optimal E3 ligases with 74% Hit@3 accuracy. Two findings carry broader implications: E(3)-equivariant architectures underperform the simpler invariant design for this scalar prediction task, and ESM-2 embeddings improve peak performance only with careful regularization -- naive integration fails. DegradoMap provides pre-synthesis computational guidance for degradability assessment; its well-calibrated confidence scores (ECE = 0.029, target-unseen) enable practitioners to prioritize high-confidence predictions for experimental follow-up. However, the high seed variance (std = 0.124) and limited E3 coverage require ensembling for reliable deployment.", "date": "2026-06-01", "categories": []}
{"id": "2606.04154", "title": "EpiFormer: Learning Antigen-Antibody Interactions for Epitope Prediction via Geometric Deep Learning", "abstract": "Antibodies neutralize foreign antigens by binding to specific surface regions called epitopes. Computational epitope prediction is critical for understanding immune recognition and guiding antibody engineering. However, existing methods face three fundamental challenges: antibody-aware models encode each chain independently and combine them only at a late stage, failing to capture co-dependent structural features that define binding interfaces, whereas severe class imbalance and scarcity of known antibody-antigen complexes render standard training objectives ineffective. We propose EpiFormer, a general encoder-decoder framework that addresses these challenges jointly. Our key design principle is interleaved cross-attention within GNN encoding layers, enabling bidirectional antigen-antibody information flow throughout representation learning rather than only at the output. This early-fusion principle is backbone-agnostic, providing consistent gains across GNN architectures from simple GCNs to equivariant models. We further show that sparsity-aware objectives are effective when paired with early-fusion architectures for the epitope prediction task. EpiFormer improves over the previous best method by over 40% in F1 score on standard benchmarks, demonstrating generalizability and cross-dataset transferability. Notably, EpiFormer discovers known biological principles as emergent behaviors of end-to-end training, where the learned cross-attention gates favor antigen-to-antibody information flow, consistent with the asymmetric roles of the two chains at the binding interface, and the model's preference for geometric over evolutionary features aligns with the established finding that epitope residues are not evolutionarily conserved. The source code is available at: https://github.com/mansoor181/epiformer.git", "date": "2026-06-02", "categories": []}
{"id": "2606.04994", "title": "New Benchmarking Shows Limited Generalization Power of TCR Antigenic Epitope Prediction Models", "abstract": "Accurate computational prediction of T cell receptor (TCR) antigen specificity would transform the study of T cell biology and enable scalable immune engineering, yet existing models lack sufficient sensitivity and specificity for broad applications. A major limitation is the absence of rigorously defined, unseen benchmark datasets that allow unbiased evaluation of model performance and generalizability. Here, we describe two complementary classes of datasets that meet this criterion and argue that they provide both a robust framework for model assessment and a foundation for next-generation TCR-antigen prediction algorithm development.", "date": "2026-06-03", "categories": []}
{"id": "2606.05139", "title": "BBOmix: A Tabular Benchmark for Hyperparameter Optimization of Unsupervised Biological Representation Learning", "abstract": "The rapid advancement of high-throughput sequencing has led to large, high-dimensional omics datasets. Deep unsupervised learning architectures, particularly Autoencoders (AEs), are increasingly used for dimensionality reduction and representation learning in this domain. However, AEs are highly sensitive to architectural choices and hyperparameters, and unsupervised optimization typically relies on reconstruction loss, which may be a poor proxy for downstream utility. Exhaustive hyperparameter optimization (HPO) is computationally expensive, leading researchers to frequently rely on suboptimal default configurations. To democratize access to large-scale unsupervised HPO research, we introduce $\\textbf{BBOmix}$, the first open-source tabular benchmark for unsupervised representation learning on real-world biological data. Our benchmark includes 105,000 evaluations across four AE architectures and seven multi-omics modalities from the TCGA and SCHC datasets. We quantify the correlation between reconstruction loss and downstream task performance and provide an extensive evaluation of state-of-the-art single-fidelity, multi-fidelity, and transfer learning HPO methods, establishing a rigorous baseline for future research in unsupervised biological representation learning.", "date": "2026-06-03", "categories": []}
{"id": "2509.08707", "title": "Tokenizing Loops of Antibodies", "abstract": "The complementarity-determining regions of antibodies are loop structures that are key to their interactions with antigens, and of high importance to the design of novel biologics. Since the 1980s, categorizing the diversity of CDR structures into canonical clusters has enabled the identification of key structural motifs of antibodies. However, existing approaches have limited coverage and cannot be readily incorporated into protein foundation models. Here we introduce ImmunoGlobulin LOOp Tokenizer, Igloo, a multimodal antibody loop tokenizer that encodes backbone dihedral angles and sequence. Igloo is trained using a contrastive learning objective to map loops with similar backbone dihedral angles closer together in latent space. Igloo can efficiently retrieve the closest matching loop structures from a structural antibody database, outperforming existing methods on identifying similar H3 loops by 5.9\\%. Igloo assigns tokens to all loops, addressing the limited coverage issue of canonical clusters, while retaining the ability to recover canonical loop conformations. To demonstrate the versatility of Igloo tokens, we show that they can be incorporated into protein language models with IglooLM and IglooALM. On predicting binding affinity of heavy chain variants, IglooLM outperforms the base protein language model on 8 out of 10 antibody-antigen targets. Additionally, it is on par with existing state-of-the-art sequence-based and multimodal protein language models, performing comparably to models with $7\\times$ more parameters. IglooALM samples antibody loops which are diverse in sequence and more consistent in structure than state-of-the-art antibody inverse folding models. Igloo demonstrates the benefit of introducing multimodal tokens for antibody loops for encoding the diverse landscape of antibody loops, improving protein foundation models, and for antibody CDR design.", "date": "2025-09-10", "categories": []}
{"id": "2511.02986", "title": "Scalable Single-Cell Gene Expression Generation with Latent Diffusion Models", "abstract": "Computational modeling of single-cell gene expression is crucial for understanding cellular processes, but generating realistic expression profiles remains a major challenge. This difficulty arises from the count nature of gene expression data and complex latent dependencies among genes. Existing generative models often impose artificial gene orderings or rely on shallow neural network architectures. We introduce a scalable latent diffusion model for single-cell gene expression data, which we refer to as scLDM, that respects the fundamental exchangeability property of the data. Our VAE uses fixed-size latent variables leveraging a unified Multi-head Cross-Attention Block (MCAB) architecture, which serves dual roles: permutation-invariant pooling in the encoder and permutation-equivariant unpooling in the decoder. We enhance this framework by replacing the Gaussian prior with a latent diffusion model using Diffusion Transformers and linear interpolants, enabling high-quality generation with multi-conditional classifier-free guidance. We show its superior performance in a variety of experiments for both observational and perturbational single-cell data, as well as downstream tasks like cell-level classification.", "date": "2025-11-04", "categories": []}
{"id": "2602.00423", "title": "scBatchProx: Federated-Inspired Refinement for Stable Cell-Type Discriminability under Heterogeneous Batch Compositions", "abstract": "Single-cell integration workflows often construct low-dimensional cell embeddings and then refine them with post-hoc methods to reduce batch effects. This refinement process can become unstable when cell-type compositions vary across batches, with some populations underrepresented or absent in particular batches. The problem becomes more consequential in dynamic single-cell data systems, where newly acquired batches can change both technical conditions and cell-type composition. Such instability can reduce downstream cell-type classification performance and weaken stability under imbalance perturbations. We introduce scBatchProx, a lightweight post-hoc refinement method for stabilizing single-cell latent embeddings in these heterogeneous and evolving settings. scBatchProx operates directly on precomputed embeddings and treats each batch or study as a client in a federated-inspired optimization procedure. A batch-conditioned FiLM adapter learns local latent updates, while proximal and identity-preserving regularization keep these updates conservative. Experiments on multi-batch and cross-study single-cell datasets show that scBatchProx improves downstream cell-type classification across different upstream embeddings. In controlled imbalance perturbations, scBatchProx maintains more stable affected-cell-type F1 when selected populations are downsampled or ablated from one batch. In cumulative retraining and continual integration settings, scBatchProx remains effective as new datasets arrive over time. Together, these results suggest that conservative, federated-inspired refinement can help maintain stable single-cell embeddings as batch compositions change across datasets and over time.", "date": "2026-01-31", "categories": []}
{"id": "2606.00555", "title": "Probe Before You Edit: Probing-Guided Molecular Optimization for LLM Agents in Structure-Based Drug Design", "abstract": "Structure-based drug design increasingly employs LLM agents to iteratively refine ligands against a target pocket, yet a viable ligand must satisfy two often-conflicting objectives -- binding affinity and druggability -- which single optimization steps rarely improve together. To quantify this difficulty, we introduce two diagnostic metrics: the first measures how often a single edit improves both objectives, and the second measures how often a gain on one objective comes with a loss on the other. Applying these diagnostics to current LLM-agent pipelines exposes a consistent failure mode: the agent performs molecular editing without knowing how the pocket-ligand complex responds to local modifications, thus rarely achieving joint improvement. Inspired by medicinal chemists, who probe the pocket-ligand complex with controlled analog edits before choosing an optimization direction, we propose \\textbf{PROBE}, an optimization framework built around edit-response probing. PROBE first decomposes the ligand into editable sites and builds a pocket-specific \\textbf{site map} that flags where joint gains are plausible, where the two objectives are likely in tension, and where liability substructures should be changed; it then performs controlled probe edits whose responses are distilled into an \\textbf{EditManual}. Guided by the site map and EditManual, PROBE runs an iterative multi-agent loop in which an affinity agent, a druggability agent, and a co-optimization agent jointly produce edits. On the CrossDocked2020 benchmark, PROBE achieves state-of-the-art performance and substantially mitigates the failure modes exposed by our diagnostics metrics.", "date": "2026-05-30", "categories": []}
{"id": "2606.02624", "title": "TadA-Bench: A Million-Variant Benchmark for Future-Round Discovery Toward Agentic Protein Engineering", "abstract": "AI for scientific discovery is entering an agentic era, where protein-engineering systems are expected to prioritize future wet-lab experiments rather than merely fit static measurements. We introduce TadA-Bench, a million-variant wet-lab replay benchmark from 31 TadA directed-evolution rounds for future-round discovery toward agentic protein engineering. TadA-Bench preserves the campaign chronology and defines a fixed-data replay task: given earlier experimental rounds, models rank variants that appear only in later rounds. It provides aligned DNA, RNA, and protein views, and uses Seq2Graph, a graph-based label-unification pipeline, to reconcile noisy enrichment measurements into consistent cross-round activity labels. Random-split controls show strong interpolation, but future-round ranking and finite-budget candidate selection are much weaker. Controlled analyses suggest that evolutionary coverage is more informative than local data density, positioning TadA-Bench as a reproducible wet-lab replay substrate for future-round discovery toward agentic protein engineering; the data and code are released on Hugging Face and GitHub.", "date": "2026-05-29", "categories": []}
{"id": "2606.02629", "title": "Enhancing Protein-Protein Interaction Prediction with Hierarchical Motif-based Multimodal Protein Embedding", "abstract": "Protein-protein interactions (PPIs) are essential for many biological processes. However, existing PPI prediction approaches suffer from two major limitations: they overlook the hierarchical organization of proteins, particularly meso-scale motifs that critically regulate PPIs, and fail to effectively integrate sequence, structure, and function modalities. To address these limitations, we propose MMM-PPI, a Hierarchical Motif-based Multi-Modal protein Encoder for PPI Prediction that constructs PPI embeddings in a bottom-up multi-modal manner across three scales. At the micro-scale, we encode three modal residue features; at the meso-scale, a novel multimodal motif encoder aggregates residues into spatially-informed motif embeddings; at the macro-scale, a multimodal protein encoder integrates motifs into protein embeddings by jointly modeling motif importance and inter-modal correlations. The pre-trained encoder can be used off-the-shelf for large-scale PPI prediction. Extensive experiments on multiple PPI datasets show that MMM-PPI outperforms state-of-the-art multi-label PPI prediction models, particularly under challenging data partitions and limited data scenarios. Codes are in https://github.com/yzf-code/MMM-PPI.", "date": "2026-05-30", "categories": []}
{"id": "2606.03435", "title": "CP-Agent: Context-Aware Multimodal Reasoning for Cellular Morphological Profiling under Chemical Perturbations", "abstract": "Cell Painting combines multiplexed fluorescent staining, high-content imaging, and quantitative analysis to generate high-dimensional phenotypic readouts to support diverse downstream tasks such as mechanism-of-action (MoA) inference, toxicity prediction, and construction of drug-disease atlases. However, existing workflows are slow, costly and difficult to interpret. Approaches for drug screening modeling predominantly focus on molecular representation learning, while neglecting actual experimental context (e.g., cell line, dosing schedule, etc.), limiting generalization and MoA resolution. We introduce CP-Agent, an agentic multimodal large language model (MLLM) capable of generating mechanism-relevant, human-interpretable rationales for cell morphological changes under drug perturbations. At its core, CP-Agent leverages a context-aware alignment module, CP-CLIP, that jointly embeds high-content images and experimental metadata to enable robust treatment and MoA discrimination (achieving a maximum F1-score of 0.896). By integrating CP-CLIP outputs with agentic tool usage and reasoning, CP-Agent compiles rationales into a structured report to guide experimental design and hypothesis refinement. These capabilities highlight CP-Agent's potential to accelerate drug discovery by enabling more interpretable, scalable, and context-aware phenotypic screening -- streamlining iterative cycles of hypothesis generation in drug discovery.", "date": "2026-06-02", "categories": []}
{"id": "2501.02409", "title": "Interpretable Neural ODEs for Gene Regulatory Network Discovery under Perturbations", "abstract": "Modern high-throughput biological datasets containing thousands of perturbations enable large-scale discovery of causal graphs that represent regulatory interactions between genes. Differentiable causal graphical models and regression-based methods have been developed to infer gene regulatory networks (GRNs) from interventional datasets. However, existing approaches fail to capture the non-linear dynamics of biological processes such as cellular differentiation. To address this limitation, we propose PerturbODE, a novel framework that employs interpretable neural ordinary differential equations (neural ODEs) to model cell state trajectories under perturbations and derive the underlying causal GRN from the neural ODE parameters, enabling downstream simulation of unseen genetic interventions. The GRN is encoded via a single-hidden-layer feedforward network, implicitly grouping genes into interpretable co-regulated modules. We demonstrate PerturbODE's efficacy in GRN inference and extension to perturbation response prediction across both simulated and real overexpression datasets.", "date": "2025-01-05", "categories": []}
{"id": "2503.22939", "title": "Interpretable Graph Kolmogorov-Arnold Networks for Multi-Cancer Classification and Biomarker Identification using Multi-Omics Data", "abstract": "The integration of heterogeneous multi-omics datasets at a systems level remains a central challenge for developing analytical and computational models in precision cancer diagnostics. This paper introduces Multi-Omics Graph Kolmogorov-Arnold Network (MOGKAN), a deep learning framework that utilizes messenger-RNA, micro-RNA sequences, and DNA methylation samples together with Protein-Protein Interaction (PPI) networks for cancer classification across 31 different cancer types. The proposed approach combines differential gene expression with DESeq2, Linear Models for Microarray (LIMMA), and Least Absolute Shrinkage and Selection Operator (LASSO) regression to reduce multi-omics data dimensionality while preserving relevant biological features. The model architecture is based on the Kolmogorov-Arnold theorem principle and uses trainable univariate functions to enhance interpretability and feature analysis. MOGKAN achieves classification accuracy of 96.28 percent and exhibits low experimental variability in comparison to related deep learning-based models. The biomarkers identified by MOGKAN were validated as cancer-related markers through Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. By integrating multi-omics data with graph-based deep learning, our proposed approach demonstrates robust predictive performance and interpretability with potential to enhance the translation of complex multi-omics data into clinically actionable cancer diagnostics.", "date": "2025-03-29", "categories": []}
{"id": "2512.02328", "title": "Molecular Embedding-Based Algorithm Selection in Protein-Ligand Docking", "abstract": "Selecting an effective docking algorithm is highly context-dependent, and no single method performs reliably across structural, chemical, and protocol regimes. MolAS is a lightweight algorithm-selection model that predicts per-algorithm performance from pretrained protein and ligand embeddings using attentional pooling and a shallow residual decoder. With hundreds to a few thousand labelled complexes, MolAS achieves up to a 15 percentage-point absolute improvement over the single-best solver (SBS) and closes 17--66\\% of the Virtual Best Solver (VBS)--SBS gap across five docking benchmarks. Analyses of selection frequencies, margin-conditioned reliability, and benchmark-level oracle structure indicate that MolAS is most effective when the workflow-defined oracle landscape has low winner entropy and a reasonably separable top-solver region, but degrades under protocol mismatch that shifts solver rankings and changes the induced labels. These results suggest that, in the evaluated regime, robustness is limited less by representational capacity than by workflow- and protocol-induced instability in solver hierarchies, positioning MolAS as an in-domain selector for fixed pipelines and as a diagnostic tool for assessing when docking algorithm selection is well-posed.", "date": "2025-12-02", "categories": []}
{"id": "2606.00008", "title": "Agents on a Tree: Pathwise Coordination for Multi-Objective Molecular Optimization", "abstract": "Multi-objective molecular optimization requires searching vast chemical spaces under conflicting objectives, where early design decisions strongly constrain downstream outcomes. Existing methods typically rely on a single policy or fixed scalarization, which limits their ability to represent diverse trade-offs and to explore multiple promising design trajectories. We propose ATOM, a multi-agent framework that formulates molecular optimization as a tree-structured search. Each node corresponds to an atomic operation and hosts an agent specialized for a particular objective or decision context. Agents coordinate along different paths of the tree rather than enforcing a global consensus, enabling the method to maintain and compare alternative molecular evolution trajectories. A global memory of past optimization behaviors further supports balanced exploration and exploitation across objectives. This tree-structured interaction enables reasoning over long-horizon dependencies inherent in molecular design. Experiments on challenging multi-objective benchmarks involving activity, synthesizability, and ADMET-related properties show that ATOM consistently achieves improved Pareto coverage and hypervolume over strong baselines. These results demonstrate the effectiveness of pathwise multi-agent coordination for molecular optimization. Code is available at https://anonymous.4open.science/r/ATOM-41CE.", "date": "2026-03-27", "categories": []}
{"id": "2606.00483", "title": "Annotation-Informed Block-Sparse Bayesian Modeling for cis-Expression Prediction", "abstract": "Genotype-based cis-expression prediction depends on accurately modeling local regulatory architecture. We present block-sparse Bayesian sparse linear mixed model (bsBSLMM), an extension of Bayesian sparse linear mixed model (BSLMM) that incorporates linkage disequilibrium (LD)-block spike-and-slab sparsity and a transcription start site (TSS)-informed SNP inclusion prior. Across 23,098 genes from GEUVADIS European-ancestry lymphoblastoid cell lines, bsBSLMM retained more predictable genes than BSLMM, LASSO, BLUP, TIGAR elastic net, and TIGAR Dirichlet-process regression under matched evaluation criteria. Compared with BSLMM, bsBSLMM improved held-out prediction performance for most shared genes, with gains driven primarily by LD-block sparsity and further enhanced by the TSS-informed prior. Variants selected by bsBSLMM showed stronger enrichment in GM12878 DNase and H3K27ac regulatory regions than variants selected by BSLMM. In transcriptome-wide association study (TWAS) analysis, bsBSLMM recovered established inflammatory bowel disease signals, including IL23R, and identified additional genome-wide significant genes not detected by BSLMM. Independent validation in the Louisiana Osteoporosis Study reproduced the increased prediction yield across ancestries and recovered biologically relevant bone mineral density pathways in downstream TWAS and gene set enrichment analyses. These results demonstrate that incorporating LD-block structure and biologically informed SNP priors improves cis-expression prediction and enhances downstream TWAS discovery.", "date": "2026-05-30", "categories": []}
{"id": "2606.00685", "title": "Prior-Guided Multi-Omic Transformers for Single-Cell Gene Regulatory Network Inference", "abstract": "Gene regulatory networks (GRNs) capture transcription factor-target interactions and are central to understanding cell-state regulation and disease. Reconstructing GRNs from paired single-cell transcriptomic and chromatin accessibility data is promising but challenging: scATAC is extremely sparse, and most methods rely on fixed peak-to-gene links and weak supervision. We present EpiAwareNet, a prior-guided multi-omic Transformer framework that reconstructs GRNs from paired single-cell data using only lightweight biological priors. In Stage 1, EpiAwareNet learns joint gene-peak representations with a gene-peak cross-attention module, enabling data-driven, gene-specific aggregation of accessibility signals rather than hard-coded peak-to-gene assignments. In Stage 2, EpiAwareNet incorporates a bulk-derived GRN prior as noisy positive edges to provide weak supervision under label scarcity, refining regulatory scores while remaining robust to prior noise. In our experiments, EpiAwareNet improves GRN reconstruction over representative single- and multi-omic baselines and yields GRNs with greater biological plausibility, such as improved recovery of known regulatory interactions, suggesting that lightweight biological priors from bulk data can effectively guide single-cell GRN inference when combined with adaptive cross-modal representation learning. Code and data will be available at https://github.com/tianyang-x/EpiAwareNet_pub.", "date": "2026-05-30", "categories": []}
{"id": "2606.00955", "title": "CryoProt: A Protein Pretraining Framework with Cross-Box Interactions on Cryo-EM Density Maps", "abstract": "Despite the growing availability of cryo-electron microscopy (cryo-EM) density maps, effectively leveraging them for protein representation remains challenging. First, current methods lack a general-purpose protein pretraining framework tailored for cryo-EM density maps, designed for protein-related property prediction. Second, existing approaches typically partition density maps into local box regions and model them independently, overlooking interactions across boxes which are essential for capturing global structural context in cryo-EM density map. To address these challenges, we propose CryoProt, a protein pretraining framework designed for cryo-EM density maps. CryoProt introduces a Map Encoder based on multi-head latent attention (MLA), where box-level representations interact through a shared latent space, enabling explicit modeling of cross-box dependencies within the density map. Furthermore, we adopt a multi-task pretraining strategy to learn generalizable representations that can be effectively transferred to diverse downstream tasks, such as protein flexibility prediction, where cryo-EM density maps are not required and can be inferred implicitly by the pretrained model. Experimental results demonstrate that CryoProt consistently outperforms existing state-of-the-art methods across multiple benchmarks, achieving up to 12% improvement over the best-performing baselines, highlighting the importance of modeling cross-box interactions in cryo-EM data. The source code is publicly available at https://anonymous.4open.science/r/CryoProt.", "date": "2026-05-31", "categories": []}
{"id": "2606.01781", "title": "Structure-Guided Adaptive Propagation for Protein-Protein Interaction Site Prediction", "abstract": "Accurate prediction of protein-protein interaction sites (PPIS) is essential for understanding cellular processes, disease mechanisms, and therapeutic target discovery. Graph-based deep learning has advanced PPIS prediction by incorporating residue-level structural context. However, most graph-based models still rely on fixed propagation schemes that treat all residues similarly, despite the structural and functional heterogeneity of protein interfaces. Such propagation may limit the ability to adapt information diffusion to local geometric environments, making it difficult to distinguish true interaction sites from structurally similar non-interacting neighbors. We present SGAP-PPIS, a structure-guided adaptive propagation model for PPIS prediction. Rather than using a fixed propagation mechanism, SGAP-PPIS leverages multi-scale geometric states from an equivariant graph neural network to generate residue-wise propagation coefficients. This design allows each residue to adaptively balance local feature preservation and neighborhood diffusion according to its geometric microenvironment. Experimental results show that SGAP-PPIS achieves competitive performance among the state-of-the-art methods on Test\\_60. Ablation studies show that geometry-conditioned adaptive propagation, scale-aligned geometric guidance, and multi-step propagation-state representation jointly drive these improvements.", "date": "2026-06-01", "categories": []}
{"id": "2606.01816", "title": "Site4Drug: Predicting Drug-Binding Target Sites with an AI Agent", "abstract": "Selecting where to intervene on a protein (i.e., choosing a targetable site) is often a more ambiguous and failure-prone bottleneck than selecting what binds, especially for membrane proteins where accessibility, topology, and post-translational modifications (PTMs) constrain actionable regions. We present Site4Drug, a modality-aware site-finding agent that outputs a ranked list of targetable regions with explicit constraints, evidence summaries, risk flags, and a traceable decision log.\n  Rather than requiring users to specify the drug modality upfront, Site4Drug can recommend a binding modality (e.g., antibody/peptide-like vs small-molecule) from the same evidence used for site discovery, including topology, hydropathy, PTM propensity, disulfides, domain context, and sequence. Importantly, this evidence is applied consistently across modalities, including small-molecule pocket discovery, to avoid selecting chemically plausible but biologically occluded sites.", "date": "2026-06-01", "categories": []}
{"id": "2606.01846", "title": "Mos-Gen: A Generative Molecular Framework for Mosquito Insecticide Design", "abstract": "Mosquito-borne infectious diseases cause more than 700000 deaths worldwide each year. The long-term use of conventional chemical insecticides has induced serious resistance problems, creating an urgent need to develop novel, highly effective, and ecologically sustainable alternatives. While existing artificial intelligence approaches in this domain have focused primarily on activity prediction and classification, they leave a critical gap in the de~novo generation of novel molecular scaffolds. In this study, we propose Mos-Gen, a motif-aware generative collaborative framework that couples the pretrained molecular representation model Uni-Mol with a variational autoencoder (VAE), specifically tailored for the design of disulfide-containing allicin derivatives as mosquito insecticides. Among the generated candidates, fourteen compounds -- comprising nine predicted positives and five predicted negatives -- were selected for chemical synthesis and experimental validation. The hit rate among the predicted positives reached 78%, whereas none of the predicted negatives exhibited mosquitocidal activity. These experimental results fully validated the high-precision screening capability of the Mos-Gen framework.", "date": "2026-06-01", "categories": []}
{"id": "2606.02424", "title": "GC-MoE: Genomics-Guided Cell-Type-Specific Mixture of Experts for Histology-Based Single-Cell Spatial Transcriptomics", "abstract": "Histology-based single-cell spatial transcriptomics (ST) estimation aims to predict gene expression for individual cells from histopathological images and cell locations, reducing the need for costly single-cell ST measurements. Unlike existing histology-to-ST methods that mainly predict spot-level profiles for local regions containing multiple cells, this task requires modeling cell-to-cell expression variability, which is strongly structured by cell type. We propose Genomics-Guided Cell-Type-Specific Mixture-of-Experts (GC-MoE), which estimates cell-type probabilities with a routing network and softly combines cell-type-specific experts for gene expression prediction. To further encode cell-type-dependent gene programs, we introduce the Cell-Type-Specific Co-Expression-Aware Predictor (CAP), together with a lightweight Cell-to-Cell Interaction Attention (C2CA) module for neighboring-cell context. Experiments and ablations on public single-cell ST datasets show consistent improvements over existing single-cell and adapted spot-level baselines.", "date": "2026-06-01", "categories": []}
{"id": "2510.02578", "title": "FLOWR.root: A flow matching based foundation model for joint multi-purpose structure-aware 3D ligand generation and affinity prediction", "abstract": "We present FLOWR.root, an SE(3)-equivariant flow-matching model for pocket-aware 3D ligand generation with joint potency and binding affinity prediction and confidence estimation. The model supports de novo generation, interaction- and pharmacophore-conditional sampling, fragment elaboration and replacement, and multi-endpoint affinity prediction (pIC50, pKi, pKd, pEC50). Training combines large-scale ligand libraries with mixed-fidelity protein-ligand complexes, refined on curated co-crystal datasets and adapted to project-specific data through parameter-efficient finetuning. The base FLOWR.root model achieves state-of-the-art performance in unconditional 3D molecule and pocket-conditional ligand generation. On HiQBind, the pre-trained and finetuned model demonstrates highly accurate affinity predictions, and outperforms recent state-of-the-art methods such as Boltz-2 on the FEP+/OpenFE benchmark with substantial speed advantages. However, we show that addressing unseen structure-activity landscapes requires domain adaptation; parameter-efficient LoRA finetuning yields marked improvements on diverse proprietary datasets and PDE10A. Joint generation and affinity prediction enable inference-time scaling through importance sampling, steering design toward higher-affinity compounds. Case studies validate this: selective CK2$\\alpha$ ligand generation against CLK3 shows significant correlation between predicted and quantum-mechanical binding energies. Scaffold elaboration on ER$\\alpha$, TYK2, and BACE1 demonstrates strong agreement between predicted affinities and QM calculations while confirming geometric fidelity. By integrating structure-aware generation, affinity estimation, property-guided sampling, and efficient domain adaptation, FLOWR.root provides a comprehensive foundation for structure-based drug design from hit identification through lead optimization.", "date": "2025-10-02", "categories": []}
{"id": "2602.20176", "title": "Cross-Chirality Generalization by Axial Vectors for Hetero-Chiral Protein-Peptide Interaction Design", "abstract": "D-peptide binders targeting L-proteins have promising therapeutic potential. Despite rapid advances in machine learning-based target-conditioned peptide design, generating D-peptide binders remains largely unexplored. In this work, we show that by injecting axial features to $E(3)$-equivariant (polar) vector features, it is feasible to achieve cross-chirality generalization from homo-chiral (L--L) training data to hetero-chiral (D--L) design tasks. By implementing this method within a latent diffusion model, we achieved D-peptide binder design that not only outperforms existing tools in \\textit{in silico} benchmarks, but also demonstrates efficacy in wet-lab validation. To our knowledge, our approach represents the first wet-lab validated generative AI for the \\textit{de novo} design of D-peptide binders, offering new perspectives on handling chirality in protein design. Codes are available at https://github.com/YZY010418/PepMirror", "date": "2026-02-13", "categories": []}
{"id": "2605.30635", "title": "CellBRIDGE: Learning Cellular Trajectories via Interaction-Aware Alignment", "abstract": "Inferring dynamics from population snapshots is a fundamental challenge in machine learning and biology. In scRNA-sequencing (scRNA-seq), destructive measurements preclude direct tracking of individual cells across time, making trajectory inference underdetermined. Optimal Transport (OT) provides a principled framework for snapshot alignment, but a long-standing modeling question is which cost functions yield biologically meaningful couplings. Standard OT approaches rely on gene-expression distances, implicitly treating cells as independent points and neglecting structured cell-cell communication mediated by ligand-receptor signaling. We introduce CellBRIDGE (Cell-Based Regularized Interaction-Driven Gene Expression), which augments feature-based OT with a directed, typed interaction cost derived from ligand-receptor activity. By explicitly modeling cell-cell communication, CellBRIDGE improves cross-snapshot couplings and downstream trajectory estimates across synthetic and real scRNA-seq datasets relative to feature-only baselines. Notably, CellBRIDGE enables mechanistically interpretable in silico perturbations: on lung cancer data, silencing specific ligand-receptor pairs induces trajectory shifts that recapitulate expected effects of targeted pathway inhibition.", "date": "2026-05-28", "categories": []}
{"id": "2605.31296", "title": "mRNAutilus: Multi-Objective-Guided Discrete Generation of mRNA with Optimized Therapeutic Properties", "abstract": "Therapeutic mRNA design requires coordinating multiple interacting sequence features across the full transcript, where codon usage, untranslated regions (UTRs), and their coupling jointly determine stability, translation efficiency, and protein expression. Here, we present mRNA generation via unrolled trajectories and informed latent updates (mRNAutilus), a framework for simultaneous codon optimization and de novo UTR design directly from sequence. mRNAutilus combines a masked discrete diffusion model trained on millions of full-length mRNAs with Monte Carlo Tree Guidance to generate Pareto-efficient sequences under multiple functional objectives, using lightweight regressors over model embeddings to predict half-life, translation efficiency, and protein abundance. Unlike recent methods that design coding sequences and UTRs separately or rely on post hoc assembly and screening, mRNAutilus generates complete transcripts in a single process optimized across properties. Across diverse targets, zero-shot mRNAs encoding P. pyralis luciferase achieve over 400-fold higher expression than wild-type and outperform commercial and machine learning-designed baselines, including zero-shot generative approaches. Zero-shot SARS-CoV-2 Spike mRNAs exceed clinically used and commercial constructs and match or surpass lab-optimized designs with improved durability. We further demonstrate generality in therapeutic settings, including prime editing (PEMax) and programmable proteome modulation, where mRNAutilus-designed constructs enhance expression of peptide-guided E3 ligases (uAbs) for beta-catenin degradation. These results establish a sequence-based, multi-objective framework for generating functional mRNAs tailored to diverse biological applications.", "date": "2026-05-29", "categories": []}
{"id": "2507.03318", "title": "Structure-Aware Compound-Protein Affinity Prediction via Graph Neural Network with Group Lasso Regularization", "abstract": "Explainable artificial intelligence (XAI) approaches have been increasingly applied in drug discovery to learn molecular representations and identify substructures driving property predictions. However, building end-to-end explainable models for structure-activity relationship (SAR) modeling for compound property prediction faces many challenges, such as the limited number of compound-protein interaction activity data for specific protein targets, and plenty of subtle changes in molecular configuration sites significantly affecting molecular properties. We exploit pairs of molecules with activity cliffs that share scaffolds but differ at substituent sites, characterized by large potency differences for specific protein targets. We propose a framework by implementing graph neural networks (GNNs) to leverage property and structure information from activity cliff pairs to predict compound-protein affinity (i.e., half maximal inhibitory concentration, IC50). To enhance model performance and explainability, we train GNNs with structure-aware loss functions using group lasso and sparse group lasso regularizations, which prune and highlight molecular subgraphs relevant to activity differences. We applied this framework to activity cliff data of molecules targeting three proto-oncogene tyrosine-protein kinase Src proteins (PDB IDs: 1O42, 2H8H, 4MXO). Our approach improved property prediction by integrating common and uncommon node information with sparse group lasso, as reflected in reduced root mean squared error (RMSE) and improved Pearson's correlation coefficient (PCC). Applying regularizations also enhances feature attribution for GNN by boosting graph-level global direction scores and improving atom-level coloring accuracy. These advances strengthen model interpretability in drug discovery pipelines, particularly for identifying critical molecular substructures in lead optimization.", "date": "2025-07-04", "categories": []}
{"id": "2509.24895", "title": "Towards Understanding the Shape of Representations in Protein Language Models", "abstract": "While protein language models (PLMs) are one of the most promising avenues of research for future de novo protein design, the way in which they transform sequences to hidden representations, as well as the information encoded in such representations is yet to be fully understood. Several works have attempted to propose interpretability tools for PLMs, but they have focused on understanding how individual sequences are transformed by such models. Therefore, the way in which PLMs transform the whole space of sequences along with their relations is still unknown. In this work we attempt to understand this transformed space of sequences by identifying protein structure and representation with square-root velocity (SRV) representations and graph filtrations. Both approaches naturally lead to a metric space in which pairs of proteins or protein representations can be compared with each other.\n  We analyze different types of proteins from the SCOP dataset and show that the Karcher mean and effective dimension of the SRV shape space follow a non-linear pattern as a function of the layers in ESM2 models of different sizes. Furthermore, we use graph filtrations as a tool to study the context lengths at which models encode the structural features of proteins. We find that PLMs preferentially encode immediate as well as local relations between residues, but start to degrade for larger context lengths. The most structurally faithful encoding tends to occur close to, but before the last layer of the models, indicating that training a folding model ontop of these layers might lead to improved folding performance.", "date": "2025-09-29", "categories": []}
{"id": "2512.00283", "title": "BioArc: Discovering Optimal Neural Architectures for Biological Foundation Models", "abstract": "Foundation models have revolutionized various fields such as natural language processing (NLP) and computer vision (CV). While efforts have been made to transfer the success of the foundation models in general AI domains to biology, existing works focus on directly adopting the existing foundation model architectures from general machine learning domains without a systematic design considering the unique physicochemical and structural properties of each biological data modality. This leads to suboptimal performance, as these repurposed architectures struggle to capture the long-range dependencies, sparse information, and complex underlying ``grammars'' inherent to biological data. To address this gap, we introduce BioArc, a novel framework designed to move beyond intuition-driven architecture design towards principled, automated architecture discovery for biological foundation models. Leveraging Neural Architecture Search (NAS), BioArc systematically explores a vast architecture design space, evaluating architectures across multiple biological modalities while rigorously analyzing the interplay between architecture, tokenization, and training strategies. This large-scale analysis identifies novel, high-performance architectures, allowing us to distill a set of empirical design principles to guide future model development. Furthermore, to make the best of this set of discovered principled architectures, we propose and compare several architecture prediction methods that effectively and efficiently predict optimal architectures for new biological tasks. Overall, our work provides a foundational resource and a principled methodology to guide the creation of the next generation of task-specific and foundation models for biology.", "date": "2025-11-29", "categories": []}
{"id": "2605.18587", "title": "PACE: Geometry-Aware Bridge Transport for Single-Cell Trajectory Inference", "abstract": "Single-cell trajectory inference from destructive time-course snapshots is fundamentally ill-posed: neither cross-time cell correspondences nor continuous trajectories are observed, so the snapshot distributions alone do not uniquely determine the underlying dynamics. Existing optimal transport and flow-based methods typically couple cells by Euclidean proximity at observed clock times, which can misalign trajectories when development is asynchronous and cells sampled at the same experimental time occupy different latent pseudotime stages. We propose PACE, a trajectory inference framework that recovers geometry-consistent continuous transport dynamics from destructive time-course snapshots through three coupled components. First, PACE constructs a state- and time-dependent anisotropic Riemannian metric that assigns low transport cost along locally supported tangent directions while penalizing normal velocity components. Second, it alternates between refining cross-time couplings under the induced path-action cost and fitting endpoint-preserving neural bridges between adjacent snapshots. Third, it distills the learned bridge dynamics into a global continuous-time velocity field over cellular states. Across seven controlled and biological datasets covering nine held-out reconstruction experiments, PACE achieves the strongest overall reconstruction performance, reducing MMD, Wasserstein-1 distance, and Wasserstein-2 distance by 23.7% on average relative to the strongest competing baseline. PACE also improves RNA-velocity alignment by 15.4% on an embryoid body differentiation benchmark, without requiring explicit cell pairing, lineage tracing, or RNA-velocity supervision during training. Code is available at https://github.com/AI4Science-WestlakeU/PACE.", "date": "2026-05-18", "categories": []}
{"id": "2605.29158", "title": "PROTOCOL: Late Interaction Retrieval for Protein Homolog Search", "abstract": "Protein homology search underlies function annotation, structure prediction, and evolutionary analysis, but remains challenging in the \"twilight zone,\" where global sequence similarity is weak and classical alignment methods lose sensitivity. Protein language models provide context-aware representations that could improve alignment sensitivity in this regime. However, prior protein embedding-based retrieval pipelines often pool these representations into a single vector, potentially obscuring local motifs, domains, or conserved residues that reveal remote homology. We introduce ProtoCol, a model which represents proteins as sets of residue embeddings and uses ColBERT-style late interaction to test whether residue-level comparison improves homolog retrieval. ProtoCol encodes proteins independently, keeps candidate representations pre-computable, and scores candidates with MaxSim over residue embeddings. On SCOPe superfamily and Pfam clan benchmarks, ProtoCol outperforms sequence-composition, alignment-based, pooled PLM, and trained single-vector baselines, supporting late interaction as an effective retrieval layer for remote homology search.", "date": "2026-05-27", "categories": []}
{"id": "2507.09466", "title": "La-Proteina: Atomistic Protein Generation via Partially Latent Flow Matching", "abstract": "Recently, many generative models for de novo protein structure design have emerged. Yet, only few tackle the difficult task of directly generating fully atomistic structures jointly with the underlying amino acid sequence. This is challenging, for instance, because the model must reason over side chains that change in length during generation. We introduce La-Proteina for atomistic protein design based on a novel partially latent protein representation: coarse backbone structure is modeled explicitly, while sequence and atomistic details are captured via per-residue latent variables of fixed dimensionality, thereby effectively side-stepping challenges of explicit side-chain representations. Flow matching in this partially latent space then models the joint distribution over sequences and full-atom structures. La-Proteina achieves state-of-the-art performance on multiple generation benchmarks, including all-atom co-designability, diversity, and structural validity, as confirmed through detailed structural analyses and evaluations. Notably, La-Proteina also surpasses previous models in atomistic motif scaffolding performance, unlocking critical atomistic structure-conditioned protein design tasks. Moreover, La-Proteina is able to generate co-designable proteins of up to 800 residues, a regime where most baselines collapse and fail to produce valid samples, demonstrating La-Proteina's scalability and robustness.", "date": "2025-07-13", "categories": []}
{"id": "2602.18982", "title": "Conditionally Site-Independent Neural Evolution of Antibody Sequences", "abstract": "Common deep learning approaches for antibody engineering focus on modeling the marginal distribution of sequences. By treating sequences as independent samples, however, these methods overlook affinity maturation as a rich and largely untapped source of information about the evolutionary process by which antibodies explore the underlying fitness landscape. In contrast, classical phylogenetic models explicitly represent evolutionary dynamics but lack the expressivity to capture complex epistatic interactions. We bridge this gap with CoSiNE, a continuous-time Markov chain parameterized by a deep neural network. Mathematically, we prove that CoSiNE provides a first-order approximation to the intractable sequential point mutation process, capturing epistatic effects with an error bound that is quadratic in branch length. Empirically, CoSiNE outperforms state-of-the-art language models in zero-shot variant effect prediction by explicitly disentangling selection from context-dependent somatic hypermutation. Finally, we introduce Guided Gillespie, a classifier-guided sampling scheme that steers CoSiNE at inference time, enabling efficient optimization of antibody binding affinity toward specific antigens.", "date": "2026-02-21", "categories": []}
{"id": "2605.27413", "title": "Ligand-Conditioned Discrete Diffusion for Protein Sequence-Structure Co-Design", "abstract": "Proteins perform their biological functions through three-dimensional structures encoded by amino acid sequences, and ligand-binding protein co-design requires models that generate sequence-structure compatible proteins under explicit ligand constraints. Although continuous diffusion and flow-based models support ligand-aware design in coordinate or latent spaces, existing discrete diffusion protein language models mainly operate over sequence or structure tokens without direct small-molecule conditioning. We introduce \\textbf{ProtLiD$^2$}, a \\textbf{Prot}ein \\textbf{L}igand-conditioned \\textbf{D}iscrete \\textbf{D}iffusion model for protein sequence-structure co-design. ProtLiD$^2$ jointly generates amino-acid sequence and discrete structure tokens while incorporating ligand chemical and geometric information through geometry-aware cross-attention. Trained on over one million ligand-protein complexes, ProtLiD$^2$ extends masked discrete diffusion to ligand-aware functional protein design. We further propose maximum confidence-margin guided ReMask decoding, an inference-time self-correction strategy that retains confident predictions and remasks uncertain tokens. ProtLiD$^2$ improves global fold confidence over Complexa in whole-protein design, increasing TM-score from 0.672 to 0.802 and pLDDT from 64.55 to 73.00. In pocket co-design, ProtLiD$^2$ reduces active-site BB-RMSD from 3.46/3.40{\\AA} for FAIR/PocketGen to 1.97{\\AA}, and improves ligand-aware pass rates over PocketGen from 14.86% to 59.73% and from 6.08% to 23.49% under stricter docking thresholds. These results support ligand-conditioned discrete diffusion as an effective token-space framework for functional protein co-design. Code will be available at https://github.com/auroua/ProtLiD.", "date": "2026-05-15", "categories": []}
{"id": "2605.27986", "title": "An Evolutionary Approach for Designing Stable and Highly Expressible Low-Immunogenicity Therapeutic mRNA Sequences", "abstract": "Messenger RNA (mRNA) sequences as therapeutics require optimized design to ensure efficient translation, structural stability, and minimal immunogenicity. This study presents a two-stage in-silico framework that integrates deep learning and evolutionary computation for rational mRNA optimization instead of existing state-of-the-art models. In the first stage, a pretrained CodonTransformer (BERT-like Large Language Model) generates biologically coherent mRNA sequences encoding the target antigen. In the second stage, a genetic algorithm (GA) evolves these candidate sequences through codon-aware crossover and synonymous mutation guided by human codon usage preferences. Fitness functions for evaluation combined translation-related metrics (CAI, tAI, codon-pair bias), mRNA structural stability (local and global MFE via RNAfold, GC content), and reduced immunogenicity (CpG/UpA motif frequency). Over successive generations (38th, 40th, and 42nd), the GA improved (achieved CAI values of 0.73 to 0.74 and tAI values of 0.63 to 0.64) CAI and tAI by over 6% and codon-pair bias is high and consistent (0.97 ) and improved ribosomal accessibility at the 5' end, with an unpaired_30 fraction reaching 0.87; Global Minimum Free Energy (MFE) converged to a balanced range of -346 to -356 kcal/mol, achieving approximately 84% base-paired structural stability, and reduced immune-stimulatory motifs - lowering the average immune penalty to 27.3 in the final generation. Linear Design produces hyper-stable transcripts (MFE < - 2000 kcal/mol) that risk translation inefficiency due to extreme rigidity, and BiLSTM-CRF focuses solely on high CAI (0.96 to 0.98) without structural constraints, our framework achieves an optimal translation-stability equilibrium, highlighting the proposed BERT-GA framework as an effective, data-driven approach for the design and optimization of in-silico mRNA sequences.", "date": "2026-05-27", "categories": []}
{"id": "2605.28111", "title": "Chreode: A Cell World Model for One-Step Temporal Dynamics and Perturbation Prediction", "abstract": "Predicting how a cell will change its transcriptional state under a developmental signal or a genetic perturbation is the computational core of in-silico biology and the AI Virtual Cell program. Existing approaches either fit static control-to-treated maps that discard time, or solve multi-step ODE / Schr\\\"odinger-bridge problems on each dataset independently. We introduce Chreode, a one-step cell world model that predicts action-conditioned cell-state transitions through a structured residual transition operator. It shifts distributional evolution from inference time to training time, enabling single-pass generation while preserving a Waddington-inspired decomposition into downhill landscape flow, rotational in-tangent dynamics, and stochastic spread. The model is pretrained with a shared scVI encoder and a DiT-based dynamics backbone on a 2.4M-cell mouse embryonic atlas spanning 7 datasets. As a fine-tuning initialization, Chreode improves per-target Sinkhorn distance on Weinreb hematopoiesis and Veres islet differentiation over matched scratch models, PI-SDE, and PRESCIENT. As a transferable gene-state embedding for GEARS, the pretrained dynamics representation reduces shared-vocabulary DE20 mean squared error on Norman Perturb-seq from 0.2121 to 0.1858, a 12.4% relative improvement, without changing the GEARS training procedure. We interpret this transfer to perturbation prediction as evidence that pretrained developmental-trajectory dynamics encode differentiation primitives transferable to CRISPR-induced state shifts, since both involve cell-state transitions in a shared latent geometry. The pretrained backbone additionally produces zero-shot clonal fate scores on Weinreb that are competitive with strong dynamic-OT baselines.", "date": "2026-05-27", "categories": []}
{"id": "2512.05794", "title": "Mechanistic Interpretability of Antibody Language Models Using SAEs", "abstract": "Sparse autoencoders (SAEs) are a mechanistic interpretability technique that have been used to provide insight into learned concepts within large protein language models. Here, we employ TopK and Ordered SAEs to investigate autoregressive antibody language models, and steer their generation. We show that TopK SAEs can reveal biologically meaningful latent features, but high feature-concept correlation does not guarantee causal control over generation. In contrast, Ordered SAEs impose a hierarchical structure that reliably identifies steerable features, but at the expense of more complex and less interpretable activation patterns. These findings advance the mechanistic interpretability of domain-specific protein language models and suggest that, while TopK SAEs suffice for mapping latent features to concepts, Ordered SAEs are preferable when precise generative steering is required.", "date": "2025-12-05", "categories": []}
{"id": "2605.21617", "title": "$\\textit{BlockFormer}$ : Transformer-based inference from interaction maps", "abstract": "Inference from interaction maps, such as centromere identification from genome-wide chromosome conformation capture techniques -- notably Hi-C -- can be formulated as a generic inverse problem: infer a set of parameters given a map summarizing pairwise interactions between entities through blocks of variable numbers and sizes. In this work, we introduce a data-driven approach that leverages shared structure between these maps, such as global alignment between localized patterns, while handling the variability in number and size of entities arising in real-world data. Our approach relies on a transformer architecture capable of handling such variability and a custom simulator to generate abundant, yet computationally cheap synthetic data for training. Applied to the problem of centromere localization, the method accurately recovers their genomic positions across a wide range of species of various genome sizes.", "date": "2026-05-20", "categories": []}
{"id": "2605.22133", "title": "Atom-level Protein Representation Learning Improves Protein Structure Prediction", "abstract": "Recent advances in generative modeling show that pretrained representations can improve generation as conditioning features or alignment targets. Motivated by this, we study protein representations for predicting structures beyond conventional function annotation. We propose TriProRep, a structure-aware pretraining method that jointly models three aligned residue-level views: amino-acid identity, backbone geometry, and local full-atom geometry, discretely encoded via VQ-VAE tokenizers. By pretraining to recover original tokens from generator-corrupted views, TriProRep learns to distinguish plausible but incorrect cross-view augmentations from the original protein. We further introduce RepSP, a benchmark for evaluating protein representations in structure-predictive settings. RepSP tests three uses of representations: homodimer co-folding from apo-chain representations, residue-level prediction of homodimer-derived interaction properties, and representation-aligned monomer structure prediction. Across these tasks, TriProRep improves over sequence-only and prior structure-aware representation models, while maintaining competitive performance on conventional benchmarks.", "date": "2026-05-21", "categories": []}
{"id": "2410.10652", "title": "Querying structural and functional niches on spatial transcriptomics data", "abstract": "Cells in multicellular organisms coordinate to form structural and functional niches. With spatial transcriptomics (ST) enabling gene expression profiling in spatial contexts, it has been revealed that spatial niches serve as cohesive and recurrent units in physiological and pathological processes. These observations suggest universal tissue organization principles encoded by conserved niche patterns, and call for a query-based niche analytical paradigm beyond current computational tools. In this work, we defined the niche-query task, which is to identify similar niches across ST samples given a niche of interest (NOI). We further developed QueST, a specialized method for solving this task. QueST models each niche as a subgraph, uses contrastive learning to learn discriminative niche embeddings, and incorporates adversarial training to mitigate batch effects. In simulations and benchmark datasets, QueST outperformed existing methods repurposed for niche querying, accurately capturing niche structures in heterogeneous environments and demonstrating strong generalizability across diverse sequencing platforms. Applied to tertiary lymphoid structures in renal and lung cancers, QueST revealed functionally distinct niches associated with patient prognosis and uncovered conserved and divergent spatial architectures across cancer types. Applied to a combinatorial spatial perturbation dataset, QueST demonstrated a complete de novo discovery-oriented workflow, characterizing previously unresolved tumor nodules through querying. These results demonstrate that QueST enables systematic, quantitative profiling of spatial niches across samples, providing a powerful tool to dissect spatial tissue architecture in health and disease.", "date": "2024-10-14", "categories": []}
{"id": "2501.18278", "title": "ReactEmbed: A Plug-and-Play Module for Unifying Protein-Molecule Representations Guided by Biochemical Reaction Networks", "abstract": "State-of-the-art models represent proteins and molecules in separate embedding manifolds, limiting the modeling of systemic biological processes. We introduce ReactEmbed, a lightweight, plug-and-play module that bridges this gap. ReactEmbed leverages biochemical reaction networks as a source of functional context, based on the principle that co-participation in reactions defines a shared functional scope. The module aligns frozen embeddings from models like ESM-3 and MolFormer into a unified space using a weighted reaction graph and a specialized sampling strategy. This process enriches unimodal embeddings and enables strong performance on cross-domain benchmarks. ReactEmbed offers a practical method to unify biological representations without costly retraining. The code and database are available for open use\\footnote{https://github.com/amitaysicherman/ReactEmbeded}.", "date": "2025-01-30", "categories": []}
{"id": "2512.08508", "title": "Multi-Alignment Contrastive Learning for Enzyme--Reaction Retrieval", "abstract": "Identifying enzymes that catalyze target biochemical reactions is a key step in computational enzyme discovery and biocatalyst design. Recent representation-learning methods formulate this problem as enzyme--reaction matching, where paired enzymes and reactions are embedded into a shared space. However, most existing approaches primarily rely on pairwise enzyme--reaction supervision and make limited use of the relationships within reaction sets or enzyme families. This work introduces a multi-alignment contrastive learning framework for biochemical retrieval. The framework jointly models cross-domain compatibility between enzymes and reactions and within-domain relationships induced by functional annotations. In addition, a Gromov--Wasserstein-inspired regularization objective encourages geometric consistency between the learned enzyme and reaction representation spaces. By combining pairwise catalytic supervision with higher-order relational alignment, the model captures both direct enzyme--reaction associations and broader functional organization. We evaluate the approach on enzyme virtual screening and bidirectional enzyme--reaction retrieval tasks. Experiments on EnzymeMap show improved early-recognition performance under BEDROC and enrichment-factor metrics compared with strong contrastive baselines. On ReactZyme, the method achieves consistent gains across time-based, enzyme-similarity, and reaction-similarity splits, demonstrating robustness to unseen enzymes and unseen reactions. Ablation studies further indicate that within-domain alignment, functional supervision, and the geometric regularization term each contribute to the observed improvements. These results suggest that modeling multiple forms of alignment can improve contrastive retrieval models for enzyme discovery, reaction annotation, and related computational biology applications.", "date": "2025-12-09", "categories": []}
{"id": "2601.22466", "title": "EvoEGF-Mol: Evolving Exponential Geodesic Flow for Structure-based Drug Design", "abstract": "Structure-Based Drug Design (SBDD) aims to discover bioactive ligands. Conventional approaches construct probability paths separately in Euclidean and probabilistic spaces for continuous atomic coordinates and discrete chemical categories, leading to a mismatch with the underlying statistical manifolds. We address this issue by representing molecules using composite exponential-family distributions, where coordinates and categories are represented within a unified natural parameter space to evolve synchronously along exponential geodesics under the Fisher-Rao metric. To avoid the instantaneous trajectory collapse induced by geodesics directly targeting Dirac distributions, we propose Evolving Exponential Geodesic Flow for SBDD (EvoEGF-Mol), which replaces static Dirac targets with dynamically concentrating distributions and is trained with a progressive-parameter-refinement architecture. Our model approaches a reference-level PoseBusters passing rate (93.4%) on CrossDock, demonstrating remarkable geometric precision and interaction fidelity, while achieving superior performance over baseline methods on real-world MolGenBench tasks for bioactive scaffold recovery. Code is available at https://github.com/BLEACH366/EvoEGF-Mol.", "date": "2026-01-30", "categories": []}
{"id": "2602.17162", "title": "JEPA-DNA: Grounding Genomic Foundation Models through Joint-Embedding Predictive Architectures", "abstract": "Genomic Foundation Models (GFMs) typically rely on Masked Language Modeling (MLM) or Next-Token Prediction (NTP) to learn the \"Laws of Nature\". While effective at capturing local syntax, these generative paradigms prioritize token-level reconstruction over high-level functional context. We introduce JEPA-DNA, a model-agnostic continual training framework that integrates a Joint-Embedding Predictive Architecture (JEPA) with traditional generative objectives. By supervising global sequence embeddings in a latent space, JEPA-DNA forces models to predict the functional representations of masked genomic segments, shifting the learning signal from token recovery to semantic alignment. We evaluate JEPA-DNA on 17 diverse genomic benchmark tasks, demonstrating consistent gains in linear probing and zero-shot performance regardless of the underlying GFM architecture or generative objective. Our framework establishes a new state-of-the-art for GFMs, surpassing the best existing models by bridging generative precision with latent semantic grounding. Through extensive ablation studies, we further characterize the synergistic interplay between generative and latent objectives. Our code is publicly available at https://github.com/NVIDIA-Digital-Bio/JEPA-DNA.", "date": "2026-02-19", "categories": []}
{"id": "2605.20747", "title": "Multi-Modal Machine Learning for Population- and Subject-Specific lncRNA-Type 2 Diabetes Association Analysis", "abstract": "Long non-coding RNAs (lncRNAs) are emerging regulatory molecules implicated in chronic disease pathogenesis, including Type 2 Diabetes Mellitus (T2D). We investigated ten literature reported lncRNAs associated with T2D: MALAT1, MEG3, MIAT, ANRIL, GAS5, KCNQ1OT1, H19, BCYRN1, XIST, and HOTAIR across two independent population-based RNA-seq cohorts. Single-omics approaches provide an incomplete view of disease biology, therefore, an integrative multi-feature framework was developed, extracting expression, secondary-structure, and sequence features for each lncRNA. Eight machine learning (ML) classifiers were evaluated under stratified k-fold, leave-one-out cross-validation (LOOCV), and repeated hold-out schemes to ensure robust performance estimation. SHAP analysis was applied for subject-level association interpretation. In one cohort, GAS5 and XIST expression features, along with GAS5, MEG3, and ANRIL sequence features, were found to be associated with T2D, while MALAT1 expression and KCNQ1OT1, ANRIL, and MEG3 sequence features were found to be associated in the second cohort. MEG3 was identified by SHAP as the dominant lncRNA in both cohorts. ML results were consistent with established statistical methods while additionally providing population- and subject-level disease association profiles linked to specific molecular feature types. The proposed framework advances mechanistic understanding of T2D and supports lncRNA-based precision medicine.", "date": "2026-05-20", "categories": []}
{"id": "2605.23960", "title": "Learning Protein Structure-Function Relationships through Knowledge-guided Representation Decomposition", "abstract": "Proteins encode diverse functions within complex three-dimensional structures, yet most deep learning representations remain highly entangled, obscuring the biophysical signals that underlie function. Here we introduce ProtDiS, a knowledge-guided framework that decomposes pretrained protein micro-environment embeddings into biologically grounded and task-relevant dimensions. Inspired by the information bottleneck principle, ProtDiS learns representations that balance informativeness and compression, yielding structural features that are more specific, independent, and information-efficient, and achieving consistent improvements across twelve downstream tasks, with the largest gains under structure-based splits. Protein- and residue-level analyses further show that ProtDiS differentiates proteins with similar folds but divergent functions and captures fine-grained biophysical signals critical. These findings suggest that knowledge-guided decomposition provides a general and interpretable approach for structuring latent spaces in protein structural modeling. The source code and implementation details are publicly available at https://github.com/AI-HPC-Research-Team/ProtDiS.", "date": "2026-05-12", "categories": []}
{"id": "2605.23961", "title": "Multimodal Alignment and Preference Optimization for Zero-Shot Conditional RNA Generation", "abstract": "The design of RNA molecules that interact with specific proteins is a critical challenge in experimental and computational biology. Despite recent progress in natural language modeling and deep learning-based protein design, there remains significant room to improve the frequency of successful interactions and the authenticity of generated sequences for functional applications. In this work, we frame conditional RNA sequence generation as a multi-stage alignment problem, introducing Moirain: a suite of models optimized via multimodal supervised fine-tuning (SFT) and Direct Preference Optimization (DPO). Our approach begins with large-scale pretraining on diverse RNA corpora to capture the fundamental grammars of sequence plausibility. To achieve target-specific generation, we employ a multimodal SFT architecture that conditions RNA synthesis on protein structural and sequential features. Finally, we leverage DPO to refine the model using synthetic interaction data: taking advantage of DPO's unique ability to navigate non-aligned preference spaces, we improve functional fitness without collapsing the learned natural distribution. Extensive evaluation of the Moirain series (Moirain-Base, -Multi, and -DPO) demonstrates that our framework consistently produces novel, diverse, and biologically plausible RNA sequences with superior binding affinities compared to existing baselines.", "date": "2026-05-12", "categories": []}
{"id": "2605.24034", "title": "WTKO-CNN: Deep Learning Reveals Sequence Motifs Distinguishing Wild-Type and Knockout ATAC-seq Peaks", "abstract": "Chromatin regulators can alter transcriptional programs by modifying the accessibility of regulatory DNA elements. Understanding how regulatory sequences differ between wild-type (WT) and knockout (KO) conditions is crucial for deciphering transcriptional control. Here, we applied a convolutional neural network, \\textbf{WTKO-CNN} with an attention mechanism to classify DNA sequences as WT or KO, achieving high predictive performance. To interpret the model, we generated saliency maps to identify nucleotide positions most influential for the classification decision. From these high-saliency regions, we extracted and clustered k-mers, enabling de novo motif discovery. Sequence logos and consensus motifs derived from the CNN filters revealed biologically meaningful patterns, which are further validated using MEME, TOMTOM, and HOMER against known transcription factor binding sites. Our analysis identified motifs associated with transcription factor families that discriminate WT from KO sequences, demonstrating that CNN-guided saliency mapping is a powerful approach for uncovering functional sequence features.", "date": "2026-05-21", "categories": []}
{"id": "2605.24045", "title": "A Large-Scale Dataset and Benchmark: Do Protein-Ligand Models Learn Binding Sites or Just Binding Likelihood?", "abstract": "Protein-ligand modeling underpins computational drug discovery and molecular design. Existing protein-ligand benchmarks typically evaluate whether a protein and ligand interact and how strongly they bind, through tasks such as binary binding prediction and affinity regression. However, these evaluations provide limited evidence of whether models can localize binding sites or identify the non-covalent interactions underlying molecular recognition. To address this gap, we introduce InteractBind, a large-scale protein-ligand dataset comprising approximately 100k protein-ligand pairs, together with a benchmark for fine-grained evaluation. The core fine-grained task is that of binding-site localization, which uses protein-residue and ligand-atom interaction maps spanning six major types of non-covalent interactions to assess whether model-derived interaction maps localize binding sites. InteractBind further includes binding affinity and protein similarity-controlled splits to support realistic generalization assessment. Using InteractBind, we evaluate eight existing sequence-based and interaction-aware models, assessing binary binding prediction and binding-site localization. Results reveal limited binding-site localization despite strong binary binding prediction, with marked variation across non-covalent interaction types. Overall, InteractBind establishes a benchmark paradigm that encourages the development of more interpretable and physically grounded protein-ligand models.", "date": "2026-05-21", "categories": []}
{"id": "2605.24489", "title": "TIGER: Text-Informed Generalized Enzyme-Reaction Retrieval", "abstract": "Enzyme-reaction retrieval is a fundamental problem in computational biology, underpinning enzyme characterization, reaction mechanism elucidation, and the rational design of metabolic pathways and biocatalysts. As a bidirectional task, it entails both enzyme-to-reaction and reaction-to-enzyme mapping. However, existing approaches suffer from poor generalization across tasks and distributions, with performance highly sensitive to dataset splits and substantial asymmetry between retrieval directions. To address these challenges, we present TIGER, a Text-Informed Generalized Enzyme-Reaction Retrieval framework that leverages protein-to-text generation models to distill textual semantic knowledge from enzyme sequences, providing a generalized representation that bridges enzymes and biochemical reactions. To ensure the quality and reliability of textual semantics, we design a Dynamic Gating Network that adaptively fuses text-derived knowledge with sequence features, enabling more consistent and informative enzyme representations, while a Structure-Shared Feature Projector aligns enzyme and reaction representations within a unified latent space. Extensive experiments demonstrate that, under bidirectional retrieval supervision, TIGER significantly outperforms state-of-the-art baselines across diverse distributions and exhibits strong robustness and transferability across tasks.", "date": "2026-05-23", "categories": []}
{"id": "2605.24520", "title": "AnnotateMissense: a genome-wide annotation and benchmarking framework for missense pathogenicity prediction", "abstract": "Missense variant interpretation remains challenging because pathogenicity depends on heterogeneous evidence from population frequency, evolutionary conservation, transcript context, amino acid substitution severity, prior pathogenicity predictors and protein-language-model-derived features. We present AnnotateMissense, a scalable annotation, benchmarking and genome-wide prediction framework for missense variant interpretation. AnnotateMissense integrates hg38 missense variants derived from dbNSFP v5.1 with ANNOVAR annotations, dbNSFP transcript/protein descriptors, AlphaMissense scores, ESM-derived features, conservation metrics, population-frequency variables, established pathogenicity predictors and engineered amino acid/codon-context features. Using 132,714 ClinVar-labelled missense variants, we benchmarked machine-learning and deep-learning models under controlled feature configurations. The full 303-feature benchmark set achieved the strongest performance with XGBoost, reaching mean MCC = 0.9411 and ROC-AUC = 0.9950 across stratified five-fold cross-validation. Restricted naive and location-oriented feature sets achieved lower best MCC values of 0.4989 and 0.5113, respectively. Circularity-controlled ablations showed that removing prior-predictor, population-frequency and clinically overlapping evidence reduced performance, whereas excluding AlphaMissense and ESM-derived features alone had minimal effect. Temporal ClinVar validation on newly observed pathogenic/benign variants achieved MCC = 0.7613, accuracy = 0.8798 and F1-score = 0.8750. The final model was applied to 90,643,830 hg38 missense variants to generate AnnotateMissense pathogenicity scores and binary prediction labels. Code and outputs are available at https://github.com/MuhammadMuneeb007/CAGI7_Annotate_All_Missense and https://doi.org/10.5281/zenodo.19981867.", "date": "2026-05-23", "categories": []}
{"id": "2605.25250", "title": "LipoAgent: Coordinating Fine-Tuned LLM Agents for Safer Lipid Design", "abstract": "Lipid nanoparticles (LNPs) are among the most clinically mature platforms for nucleic acid delivery, yet designing lipids that are both effective and biologically safe remains a major bottleneck. In practical screening, toxicity is a decision-level constraint: if a lipid is toxic, its efficiency prediction is clinically irrelevant. We propose LipoAgent, a safety-aware multi-agent LLM framework for lipid discovery. LipoAgent combines domain-specific finetuning with a conditional prediction objective that enforces toxicity as a prerequisite for efficiency prediction, and further improves reliability via multi-agent verification with lightweight human oversight when disagreement persists. Across multiple foundation models, LipoAgent achieves an average 32% relative improvement in mRNA transfection efficiency prediction compared with other reported models for lipid design. Wet-lab validation confirms that virtual screening rankings reliably translate to biological transfection outcomes. The code is publicly available at https://github.com/SAI-Lab-NYU/LipoAgent.git.", "date": "2026-05-24", "categories": []}
{"id": "2605.25388", "title": "ViroBench: Benchmarking Nucleotide Foundation Models on Viral Genomics Tasks", "abstract": "Nucleotide sequences constitute the fundamental genetic basis of biological systems, rendering viral genomic analysis critical for biomedical advancement. Despite progress in biological foundation models, specifically nucleotide foundation models (NFMs), the field lacks a unified standard for viral genomics to facilitate community development and enforce biosecurity constraints. To address this, we introduce ViroBench, the first comprehensive and large-scale benchmark specifically designed for NFMs in viral settings. ViroBench evaluates models across two critical dimensions: biological understanding and latent biosecurity risk, covering 18 diverse scenarios within 4 task types. Extensive evaluation of 66 NFMs across diverse architectures yields three critical conclusions. Firstly, NFMs exhibit a performance degradation in biological understanding under phylogenetic and temporal shifts, indicating weak extrapolation capabilities. Secondly, generation tasks reveal a decoupling between statistical likelihood and biological functional validity, posing latent biosecurity risks. Thirdly, controlled ablation studies reveal that taxonomic diversity in pretraining data outweighs parameter scale. Specifically, a lightweight baseline trained on diverse data achieves a 67.5% performance gain over its original model. Overall, ViroBench provides interpretable, diagnostic evaluations and a reproducible measurement framework for future research on viral nucleotide foundation models. The datasets and code are publicly available at https://github.com/QIANJINYDX/ViroBench.", "date": "2026-05-25", "categories": []}
{"id": "2605.04118", "title": "ProtDBench: A Unified Benchmark of Protein Binder Design and Evaluation", "abstract": "Recent advances in de novo protein binder design have enabled increasing experimental validation, yet reported in silico metrics remain difficult to interpret or compare across studies due to non-standardized evaluation protocols. We introduce ProtDBench, a standardized and throughput-aware evaluation framework for protein binder design. ProtDBench defines unified benchmark tasks, evaluation protocols, and success criteria, enabling systematic analysis of how evaluation design influences observed performance. Using a large wet-lab annotated dataset, we analyze commonly used structure prediction models as evaluation verifiers, revealing substantial verifier-dependent bias and limited agreement under identical filtering protocols. We then benchmark representative open-source generative binder design methods across ten diverse protein targets under a fixed evaluation protocol. Beyond per-sequence success rates, ProtDBench incorporates throughput-aware metrics based on a fixed 24-hour budget, as well as cluster-level success criteria to account for structural diversity. Together, these results expose systematic differences induced by filtering rules, success definitions, and throughput-aware evaluation between computational efficiency, success rate, and structural diversity. Overall, ProtDBench provides a fair and reproducible evaluation pipeline that supports systematic and controlled comparison of protein binder design methods under realistic evaluation settings.", "date": "2026-05-05", "categories": []}
{"id": "2605.23559", "title": "PathNavigate: A Training-Free Pathology Agent with Surprise-Guided Scan and Shared Slide Memory for Whole-Slide Image VQA", "abstract": "Whole-slide image visual question answering (WSI-VQA) frames pathology as an extreme-context search problem: to answer a free-form clinical query, a system must first navigate a gigapixel slide under a strict inspection budget to locate sparse, high-resolution evidence. Existing approaches largely fall into two paradigms: i) supervised pathology multimodal large language models (MLLMs) and agents can absorb localization and reasoning into learned modules, but they often couple navigation to task-specific supervision and retraining, limiting their practicality; ii) training-free pathology agents avoid this cost by keeping core models frozen, but often follow a question-first design, constructing the initial candidate set mainly from query-conditioned relevance. This can miss decisive morphology that is not named in the question, and force heavier inference-time scaffolding. To address this challenge, we introduce PathNavigate, a training-free pathology agent built around a scan-search-readout routine. Before question matching, PathNavigate scans the current slide at low magnification with a shared online memory module over frozen pathology features, producing a slide-specific surprise field that marks an abnormal-region pool. It then applies question-conditioned PLIP relevance only within this pool to select high-magnification search targets. Finally, it extracts local high-magnification evidence and answers with a frozen perceptor-adjudicator stack, using the same online memory as slide-level context. Experiments on WSI-VQA and SlideBench-BCNB show that the proposed scan-search-readout design improves answer accuracy and yields more interpretable evidence-selection trajectories with higher efficiency.The code is available online.", "date": "2026-05-22", "categories": []}
{"id": "2603.21743", "title": "CellFluxRL: Biologically-Constrained Virtual Cell Modeling via Reinforcement Learning", "abstract": "Building virtual cells with generative models to simulate cellular behavior in silico is emerging as a promising paradigm for accelerating drug discovery. However, prior image-based generative approaches can produce implausible cell images that violate basic physical and biological constraints. To address this, we propose to post-train virtual cell models with reinforcement learning (RL), leveraging biologically meaningful evaluators as reward functions. We design seven rewards spanning three categories-biological function, structural validity, and morphological correctness-and optimize the state-of-the-art CellFlux model to yield CellFluxRL. CellFluxRL consistently improves over CellFlux across all rewards, with further performance boosts from test-time scaling. Overall, our results present a virtual cell modeling framework that enforces physically-based constraints through RL, advancing beyond \"visually realistic\" generations towards \"biologically meaningful\" ones.", "date": "2026-03-23", "categories": []}
{"id": "2605.21522", "title": "Protein Thoughts: Interpretable Reasoning with Tree of Thoughts and Embedding-Space Flow Matching for Protein-Protein Interaction Discovery", "abstract": "Protein-protein interactions (PPIs) govern nearly all cellular processes, yet computational methods for identifying binding partners typically produce ranked predictions without mechanistic justification. This creates a fundamental barrier to adoption because biologists cannot assess whether predictions reflect genuine biochemical insight or spurious correlations. We present \\textbf{Protein Thoughts}, a framework that reformulates PPI discovery as an interpretable search problem with explicit reasoning. The system decomposes binding evidence into four biologically meaningful signals: sequence similarity reflecting evolutionary relationships, structural complementarity capturing geometric fit, interface balance, and chemical compatibility encoding residue-level interactions. Rather than collapsing these signals into an opaque score, we preserve their individual contributions through a transparent value function that enables both ranking and auditing. To navigate large candidate spaces efficiently, we introduce hypothesis-guided entropy-regularized Tree-of-Thoughts search. A fine-tuned language model generates search directives from embedding-derived features, classifying candidates as high-priority, exploratory, or skippable. These directives condition a Boltzmann policy that balances exploitation with entropy-driven exploration, while hypothesis-aware pruning prevents premature abandonment of promising candidates. For candidates exhibiting score disagreement, hypothesis-conditioned embedding-space flow matching transports protein embeddings toward the binder manifold. On the SHS148k benchmark, Protein Thoughts achieves mean best-binder rank of 11.2 versus 47.7 for an entropic tree search baseline, a 76% improvement, and for binding prediction the trained value function achieves $91.08 \\pm 0.19$ Micro-F1, outperforming existing PPI methods on the same dataset.", "date": "2026-05-19", "categories": []}
{"id": "2605.21600", "title": "ConTact: Contact-First Antibody CDR Design via Explicit Interface Reasoning", "abstract": "Computational antibody CDR design methods condition on antigen structure to generate binding loops, yet existing architectures conflate two fundamentally distinct sub-problems: identifying which CDR positions will contact the antigen, and selecting amino acids at those positions. This conflation forces models to learn contact reasoning implicitly through uniform message passing, diluting antigen signal across all positions equally. We introduce ConTact, a contact-then-act architecture that explicitly decomposes CDR design into three cascaded stages: learning surface complementarity fingerprints, predicting CDR-antigen contacts, and injecting contact-gated antigen features into the sequence head. A distance-biased cross-attention module encodes geometric priors favoring spatial neighbors, while a contact-weighted cross-entropy loss concentrates gradient signal on binding-critical positions. On CHIMERA-Bench dataset, ConTact achieves the best structural quality (7% RMSD improvement over the next-best baseline), best epitope awareness (10% F1 score over GNN baselines), and competitive sequence recovery (AAR 0.38) among several CDR-H3 design baselines.", "date": "2026-05-20", "categories": []}
{"id": "2605.21610", "title": "AgForce Enables Antigen-conditioned Generative Antibody Design", "abstract": "Antibody design methods condition on antigen structure to generate complementarity-determining regions (CDR), yet a systematic evaluation of baseline methods reveals that they largely ignore the antigen input. We identify three failure modes that explain this behavior. Antigen blindness arises because models derive predictions from antibody framework context rather than antigen information, producing nearly identical CDRs regardless of the target. Vocabulary collapse reduces predicted amino acids to three to five per position, far below the ground truth distribution in native sequences. Moreover, any model trained with standard per-position cross-entropy converges to the positional marginal distribution, making it provably unable to produce antigen-specific sequence predictions. We propose a novel encoder-decoder architecture called AgForce, that uses a graph neural network (GNN) as the encoder and specialized decoders for sequence-structure co-design. Specifically, we apply framework dropout, gated bottlenecks, and hyperbolic cross attention that prevent the antibody shortcut path. In the decoder, a Mixture Density Network (MDN) sequence head with Potts-like pairwise coupling and annealed Multiple Choice Learning (aMCL) replaces the cross-entropy objective with a multi-component distribution whose optimal solution differs from the positional marginal. An antigen cycle consistency head routes gradients through the sequence decoder, forcing predicted distributions to encode antigen identity. AgForce achieves the best binding quality and sequence recovery simultaneously on the CHIMERA-Bench dataset, improving amino acid recovery by 8% over the strongest sequence baseline while surpassing the baselines across all interface metrics, and nearly doubling the effective vocabulary of GNN methods. The source code is available at: https://github.com/mansoor181/ag-force.git", "date": "2026-05-20", "categories": []}
{"id": "2605.22340", "title": "From Snapshots to Trajectories: Learning Single-Cell Gene Expression Dynamics via Conditional Flow Matching", "abstract": "Single-cell RNA sequencing (scRNA-seq) provides high-dimensional profiles of cellular states, enabling data-driven modeling of cellular dynamics over time. In practice, time-resolved scRNA-seq is collected at only a few discrete time points as unpaired snapshot populations, leaving substantial temporal gaps. This motivates trajectory inference at unmeasured time points. Existing methods mainly follow two directions, optimal-transport (OT) alignment provides distribution-level matching between observed snapshots, while continuous-time generative models support forecasting via learned dynamics. However, two challenges remain: (i) unpaired snapshots render local transitions between adjacent time points ambiguous, leading to unstable supervision; and (ii) long-horizon prediction relies on repeated integration, where small modeling errors compound and cause distribution drift. To address these challenges, we propose single-cell Flow Matching (scFM), a latent generative framework based on coupling-conditioned flow matching. First, we compute entropically regularized OT couplings between adjacent snapshots and use them to construct soft, weighted flow-matching targets for learning time-dependent velocity fields. Second, we learn bidirectional velocity fields and leverage their consistency to refine couplings and improve temporal coherence under sparse supervision. Third, we introduce distribution-level alignment and latent dynamic regularization to anchor long rollouts and mitigate drift. Experiments on real-world time-series scRNA-seq datasets show that scFM consistently improves distributional prediction performance for both temporal interpolation and extrapolation. Moreover, scFM yields more accurate trajectory reconstruction and temporally coherent visualizations where intermediate time points are absent, indicating a more faithful recovery of underlying temporal gene expression dynamics.", "date": "2026-05-21", "categories": []}
{"id": "2601.03019", "title": "DNACHUNKER: Learnable Tokenization for DNA Language Models", "abstract": "DNA language models are increasingly used to represent genomic sequence, yet their effectiveness depends critically on how raw nucleotides are converted into model inputs. Unlike natural language, DNA offers no canonical boundaries, making fixed tokenizations a brittle design choice under shifts, indels, and local repeats. We introduce DNAChunker, a masked DNA language model that incorporates a learnable adaptive segmentation module to produce context-dependent, variable-length units. Building on a dynamic segmentation procedure, DNAChunker learns to allocate finer granularity to functionally enriched regions while compressing repetitive or redundant sequence. We pretrain DNAChunker on the human reference genome and evaluate it across five benchmarks, where it consistently improves over strong fixed-tokenization baselines. Further analyses and ablations indicate that unlike fixed tokenizations, segmentation is learned in a biologically-informed, mutation-resilient manner.", "date": "2026-01-06", "categories": []}
{"id": "2602.04916", "title": "AFD-INSTRUCTION: A Comprehensive Antibody Instruction Dataset with Functional Annotations for LLM-Based Understanding and Design", "abstract": "Large language models (LLMs) have significantly advanced protein representation learning. However, their capacity to interpret and design antibodies through natural language remains limited. To address this challenge, we present AFD-Instruction, the first large-scale instruction dataset with functional annotations tailored to antibodies. This dataset encompasses two key components: antibody understanding, which infers functional attributes directly from sequences, and antibody design, which enables de novo sequence generation under functional constraints. These components provide explicit sequence-function alignment and support antibody design guided by natural language instructions. Extensive instruction-tuning experiments on general-purpose LLMs demonstrate that AFD-Instruction consistently improves performance across diverse antibody-related tasks. By linking antibody sequences with textual descriptions of function, AFD-Instruction establishes a new foundation for advancing antibody modeling and accelerating therapeutic discovery.", "date": "2026-02-04", "categories": []}
{"id": "2603.20420", "title": "CRANE: Correcting Errors in Raw Nanopore Signals Using Hidden Markov Models", "abstract": "Nanopore sequencing can read substantially longer sequences of nucleic acid molecules, called reads, than other sequencing methods, which has led to advances in genomic analysis such as the gapless human genome assembly. By analyzing the raw electrical signal reads that nanopore sequencing generates from molecules, existing works can map these reads without translating them into DNA characters (i.e., basecalling), allowing for quick and efficient analysis of sequencing data. However, raw signals often contain errors due to noise and processing errors, which limits the overall accuracy of raw signal analysis. Our goal in this work is to detect and correct errors in raw signals to improve the accuracy of raw signal analyses. To this end, we propose CRANE, a mechanism that trains and utilizes a Hidden Markov Model (HMM) to accurately correct signal errors. Our extensive evaluation on various datasets shows that CRANE 1) consistently improves the overall accuracy of the underlying raw signal analysis tools, 2) minimizes the burden of optimizing analysis pipelines for newer nanopore technologies, and 3) does not introduce substantial computational overhead. We conclude that CRANE provides an effective mechanism to systematically identify and correct the errors in raw nanopore signals before further analysis, which can enable the development of a new class of error correction mechanisms purely designed for raw nanopore signals. Source Code: CRANE is available at https://github.com/STORMgroup/CRANE. We also provide the scripts to fully reproduce our results on our GitHub page", "date": "2026-03-20", "categories": []}
{"id": "2604.11661", "title": "Towards Autonomous Mechanistic Reasoning in Virtual Cells", "abstract": "Large language models (LLMs) have recently gained significant attention as a promising approach to accelerate scientific discovery. However, their application in open-ended scientific domains such as biology remains limited, primarily due to the lack of factually grounded and actionable explanations. To address this, we introduce a structured explanation formalism for virtual cells that represents biological reasoning as mechanistic action graphs, enabling systematic verification and falsification. Building upon this, we propose VCR-Agent, a multi-agent framework that integrates biologically grounded knowledge retrieval with a verifier-based filtering approach to generate and validate mechanistic reasoning autonomously. Using this framework, we release VC-TRACES dataset, which consists of verified mechanistic explanations derived from the Tahoe-100M atlas. Empirically, we demonstrate that training with these explanations improves factual precision and provides a more effective supervision signal for downstream gene expression prediction. These results underscore the importance of reliable mechanistic reasoning for virtual cells, achieved through the synergy of multi-agent and rigorous verification.", "date": "2026-04-13", "categories": []}
{"id": "2605.20885", "title": "Training distribution determines the ceiling of drug-blind cancer sensitivity prediction", "abstract": "Precision oncology requires predicting which drugs will suppress a specific tumor from its molecular profile, but drug-blind sensitivity prediction has plateaued despite increasingly complex drug representations. Here we show that this stagnation reflects a metric artifact rather than a representational bottleneck. The standard benchmark, global Pearson r, is dominated by between-drug potency differences that a trivial drug-mean predictor captures without any cell-specific learning. Per-drug Pearson r, which isolates within-drug cell ranking, reveals that no drug encoding improves over cell-only features across four independent datasets. A controlled experiment channeling mechanism-of-action identity as either a drug feature or a training-distribution constraint identifies the cause. Supplying MoA as a feature yields negligible benefit, whereas using it to stratify training raises per-drug r substantially for targeted kinase inhibitors, because pan-cancer co-training suppresses pathway-specific sensitivity signals. Mechanism-stratified training and response matching from pilot observations provide two deployable strategies that together recover the principal sources of predictive gain in drug-blind sensitivity prediction.", "date": "2026-05-20", "categories": []}
{"id": "2605.21454", "title": "ProtoPathway: Biologically Structured Prototype-Pathway Fusion for Multimodal Cancer Survival Prediction", "abstract": "We introduce ProtoPathway, an interpretable-by-design multimodal framework for cancer survival prediction that unifies whole slide imaging and transcriptomics through encoders producing biologically grounded representations on both sides of the fusion. On the histopathology side, $K$ learnable morphological prototypes, trained end-to-end with the survival objective, serve as the slide representation itself: patches flow into prototype tokens via soft assignment, compressing variable-length patch sets into fixed task-adaptive tokens. On the genomic side, a bipartite graph neural network encodes gene expression within the Reactome pathway hierarchy, producing pathway embeddings that reflect both constituent genes and their broader biological context through bidirectional message passing over a shared gene--pathway graph. Cross-modal attention then operates over a compact prototype $\\times$ pathway matrix in which prototypes query pathways, modeling the biological direction in which molecular programs give rise to tissue morphology. Because both axes carry stable task-learned identity, the attention matrix is itself an interpretability output, yielding native inference-time attribution across the full biological hierarchy, from genes through pathways and prototypes to spatial tissue maps. We evaluate on five TCGA cancer cohorts, demonstrating competitive or superior survival prediction with substantially improved biological interpretability and reduced computational cost, with interpretability claims validated through fold-stratified rank-based population-level analysis. Our source code, model weights, and Reactome pathways, together with a unified codebase reimplementing all multimodal survival baselines under identical preprocessing and evaluation, are available at: https://github.com/AmayaGS/ProtoPathway.", "date": "2026-05-20", "categories": []}
{"id": "2605.21485", "title": "EvoStruct: Bridging Evolutionary and Structural Priors for Antibody CDR Design via Protein Language Model Adaptation", "abstract": "Equivariant graph neural network (GNN) methods for antibody complementarity-determining region (CDR) design achieve the highest sequence recovery but suffer from severe vocabulary collapse. The current best GNN methods over-predict very few amino acids, such as tyrosine and glycine, while ignoring functionally important residues. We trace this failure to GNN encoders learning amino acid distributions de novo from limited structural data, discarding substitution patterns encoded in evolutionary databases. To resolve this, we propose EvoStruct, which bridges a frozen protein language model (PLM) with 3D structural context from an E(3)-equivariant GNN via a cross-attention adapter. Unlike prior PLM-structure adapters for general protein design, EvoStruct targets the vocabulary collapse problem specific to CDR design through progressive PLM unfreezing and R-Drop consistency regularization. On the CHIMERA-Bench dataset, EvoStruct achieves the highest amino acid recovery and lowest perplexity among several antibody design methods, improving sequence recovery by 16% and reducing perplexity by 43% relative to the best GNN baselines, while recovering 2.3x greater amino acid diversity and the highest binding-pair correlation with ground truth.", "date": "2026-05-20", "categories": []}
{"id": "2510.11344", "title": "MMAP: A Multi-Magnification and Prototype-Aware Architecture for Predicting Spatial Gene Expression", "abstract": "Spatial Transcriptomics (ST) enables the measurement of gene expression while preserving spatial information, offering critical insights into tissue architecture and disease pathology. Recent developments have explored the use of hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) to predict transcriptome-wide gene expression profiles through deep neural networks. This task is commonly framed as a regression problem, where each input corresponds to a localized image patch extracted from the WSI. However, predicting spatial gene expression from histological images remains a challenging problem due to the significant modality gap between visual features and molecular signals. Recent studies have attempted to incorporate both local and global information into predictive models. Nevertheless, existing methods still suffer from two key limitations: (1) insufficient granularity in local feature extraction, and (2) inadequate coverage of global spatial context. In this work, we propose a novel framework, MMAP (Multi-MAgnification and Prototype-enhanced architecture), that addresses both challenges simultaneously. To enhance local feature granularity, MMAP leverages multi-magnification patch representations that capture fine-grained histological details. To improve global contextual understanding, it learns a set of latent prototype embeddings that serve as compact representations of slide-level information. Extensive experimental results demonstrate that MMAP consistently outperforms all existing state-of-the-art methods across multiple evaluation metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), and Pearson Correlation Coefficient (PCC).", "date": "2025-10-13", "categories": []}
{"id": "2602.04883", "title": "Protein Autoregressive Modeling via Multiscale Structure Generation", "abstract": "We present protein autoregressive modeling (PAR), the first multi-scale autoregressive framework for protein backbone generation via coarse-to-fine next-scale prediction. Using the hierarchical nature of proteins, PAR generates structures that mimic sculpting a statue, forming a coarse topology and refining structural details over scales. To achieve this, PAR consists of three key components: (i) multi-scale downsampling operations that represent protein structures across multiple scales during training; (ii) an autoregressive transformer that encodes multi-scale information and produces conditional embeddings to guide structure generation; (iii) a flow-based backbone decoder that generates backbone atoms conditioned on these embeddings. Moreover, autoregressive models suffer from exposure bias, caused by the training and the generation procedure mismatch, and substantially degrades structure generation quality. We effectively alleviate this issue by adopting noisy context learning and scheduled sampling, enabling robust backbone generation. Notably, PAR exhibits strong zero-shot generalization, supporting flexible human-prompted conditional generation and motif scaffolding without requiring fine-tuning. On the unconditional generation benchmark, PAR effectively learns protein distributions and produces backbones of high design quality, and exhibits favorable scaling behavior. Together, these properties establish PAR as a promising framework for protein structure generation.", "date": "2026-02-04", "categories": []}
{"id": "2603.06740", "title": "ViroGym: Realistic Large-Scale Benchmarks for Evaluating Viral Proteins", "abstract": "Protein language models (pLMs) have shown strong potential for zero-shot prediction of missense variant effects, yet systematic benchmarking on viral proteins remains limited, a critical gap given the need for proactive tools that can anticipate emerging mutations ahead of experimental validation. Here we introduce ViroGym, a comprehensive benchmark evaluating pLMs across three tasks: 79 deep mutational scanning (DMS) assays covering eukaryotic viruses with 552,065 mutated sequences across 7 phenotypic readouts, 21 influenza neutralisation tasks, and a real-world pandemic prediction task for SARS-CoV-2. We benchmark well-established pLMs on fitness landscapes, antigenic diversity, and pandemic forecasting, and find that the ProGen2 family consistently achieves the strongest performance across all three tasks. Crucially, DMS and neutralisation performance reliably identifies models that generalise to real-world emergence, even though the mutation sets they surface barely overlap, revealing that complementary in vitro benchmarks capture the evolutionary constraints needed for real-world mutation forecasting.", "date": "2026-03-06", "categories": []}
{"id": "2605.19902", "title": "Hierarchical Contrastive Learning for Multi-Domain Protein-Ligand Binding", "abstract": "Predicting protein-ligand binding affinity remains intractable for multi-domain proteins, where inter-domain dynamics govern molecular recognition. Existing geometric deep learning methods typically treat proteins as monolithic static graphs, suffering from rigid-body assumptions and aleatoric noise in flexible regions. To address this, we introduced HCLBind, a self-supervised framework that decouples geometric representation learning from affinity regression. HCLBind leverages a general-to-specific pre-training paradigm on the Q-BioLiP database to learn a robust physical grammar of binding. We propose a novel hierarchical decoy strategy: the model learns local physicochemical constraints through protein coordinate perturbation in single-domain proteins and global conformational geometry through inter-domain rotation in multi-domain complexes. Our hybrid architecture integrates a domain-gated graph attention network and cross-modal attention to explicitly prioritize domain interfaces. Furthermore, we employ LoRA on protein and ligand foundation models, ensuring efficient optimization while preserving evolutionary knowledge. Experiments on PDBBind demonstrate that HCLBind effectively learns discriminative interface features and provides robust uncertainty estimation, overcoming the limitations of standard supervised learning. The code is available at https://github.com/jiankliu/HCLBind.", "date": "2026-05-19", "categories": []}
{"id": "2506.23287", "title": "HDTree: Generative Modeling of Cellular Hierarchies for Robust Lineage Inference", "abstract": "In single-cell research, tracing and analyzing high-throughput single-cell differentiation trajectories is crucial for understanding biological processes. Key to this is the robust modeling of hierarchical structures that govern cellular development. Traditional methods face limitations in computational cost, performance, and stability. VAE-based approaches have made strides but still require branch-specific network modules, limiting their scalability and stability, while often suffering from posterior collapse. To overcome these challenges, we introduce HDTree, a generative modeling framework designed for robust lineage inference. HDTree captures tree relationships within a hierarchical latent space using a unified hierarchical codebook and employs a quantized diffusion process to model continuous cell state transitions. By aligning the generative process with the Waddington landscape, this method not only improves stability and scalability but also enhances the biological plausibility of inferred lineages. HDTree's effectiveness is demonstrated through comparisons on both general-purpose and single-cell datasets, where it outperforms existing methods in lineage inference accuracy, reconstruction quality, and hierarchical consistency. These contributions enable accurate and efficient modeling of cellular differentiation paths, offering reliable insights for biological discovery.\\footnote{Code is available at https://github.com/zangzelin/code\\_HDTree\\_icml.", "date": "2025-06-29", "categories": []}
{"id": "2507.21035", "title": "GenoMAS: A Multi-Agent Framework for Scientific Discovery via Code-Driven Gene Expression Analysis", "abstract": "Gene expression analysis holds the key to many biomedical discoveries, yet extracting insights from raw transcriptomic data remains formidable due to the complexity of multiple large, semi-structured files and the need for extensive domain expertise. Current automation approaches are often limited by either inflexible workflows that break down in edge cases or by fully autonomous agents that lack the necessary precision for rigorous scientific inquiry. GenoMAS charts a different course by presenting a team of LLM-based scientists that integrates the reliability of structured workflows with the adaptability of autonomous agents. GenoMAS orchestrates six specialized LLM agents through typed message-passing protocols, each contributing complementary strengths to a shared analytic canvas. At the heart of GenoMAS lies a guided-planning framework: programming agents unfold high-level task guidelines into Action Units and, at each juncture, elect to advance, revise, bypass, or backtrack, thereby maintaining logical coherence while bending gracefully to the idiosyncrasies of genomic data.\n  On the GenoTEX benchmark, GenoMAS reaches a Composite Similarity Correlation of 89.13% for data preprocessing and an F$_1$ of 60.48% for gene identification, surpassing the best prior art by 10.61% and 16.85% respectively. Beyond metrics, GenoMAS surfaces biologically plausible gene-phenotype associations corroborated by the literature, all while adjusting for latent confounders. Code is available at https://github.com/Liu-Hy/GenoMAS.", "date": "2025-07-28", "categories": []}
{"id": "2604.02511", "title": "Re-analysis of the Human Transcription Factor Atlas Recovers TF-Specific Signatures from Pooled Single-Cell Screens with Missing Controls", "abstract": "Public pooled single-cell perturbation atlases are valuable resources for studying transcription factor (TF) function, but downstream re-analysis can be limited by incomplete deposited metadata and missing internal controls. Here we re-analyze the human TF Atlas dataset (GSE216481), a MORF-based pooled overexpression screen spanning 3,550 TF open reading frames and 254,519 cells, with a reproducible pipeline for quality control, MORF barcode demultiplexing, per-TF differential expression, and functional enrichment. From 77,018 cells in the pooled screen, we assign 60,997 (79.2\\%) to 87 TF identities. Because the deposited barcode mapping lacks the GFP and mCherry negative controls present in the original library, we use embryoid body (EB) cells as an external baseline and remove shared batch/transduction artifacts by background subtraction. This strategy recovers TF-specific signatures for 59 of 61 testable TFs, compared with 27 detected by one-vs-rest alone, showing that robust TF-level signal can be rescued despite missing intra-pool controls. HOPX, MAZ, PAX6, FOS, and FEZF2 emerge as the strongest transcriptional remodelers, while per-TF enrichment links FEZF2 to regulation of differentiation, EGR1 to Hippo and cardiac programs, FOS to focal adhesion, and NFIC to collagen biosynthesis. Condition-level analyses reveal convergent Wnt, neurogenic, EMT, and Hippo signatures, and Harmony indicates minimal confounding batch effects across pooled replicates. Our per-TF effect sizes significantly agree with Joung et al.'s published rankings (Spearman $\\rho = -0.316$, $p = 0.013$; negative because lower rank indicates stronger effect). Together, these results show that the deposited TF Atlas data can support validated TF-specific transcriptional and pathway analyses when paired with principled external controls, artifact removal, and reproducible computation.", "date": "2026-04-02", "categories": []}
{"id": "2604.11852", "title": "Limitations of Sequence-Based Protein Representations for Parkinson's Disease Classification: A Leakage-Free Benchmark", "abstract": "The identification of reliable molecular biomarkers for Parkinson's disease remains challenging due to its multifactorial nature. Although protein sequences constitute a fundamental and widely available source of biological information, their standalone discriminative capacity for complex disease classification remains unclear. In this work, we present a controlled and leakage-free evaluation of multiple representations derived exclusively from protein primary sequences, including amino acid composition, k-mers, physicochemical descriptors, hybrid representations, and embeddings from protein language models, all assessed under a nested stratified cross-validation framework to ensure unbiased performance estimation. The best-performing configuration (ProtBERT + MLP) achieves an F1-score of 0.704 +/- 0.028 and ROC-AUC of 0.748 +/- 0.047, indicating only moderate discriminative performance. Classical representations such as k-mers reach comparable F1 values (up to approximately 0.667), but exhibit highly imbalanced behavior, with recall close to 0.98 and precision around 0.50, reflecting a strong bias toward positive predictions. Across representations, performance differences remain within a narrow range (F1 between 0.60 and 0.70), while unsupervised analyses reveal no intrinsic structure aligned with class labels, and statistical testing (Friedman test, p = 0.1749) does not indicate significant differences across models. These results demonstrate substantial overlap between classes and indicate that primary sequence information alone provides limited discriminative power for Parkinson's disease classification. This work establishes a reproducible baseline and provides empirical evidence that more informative biological features, such as structural, functional, or interaction-based descriptors, are required for robust disease modeling.", "date": "2026-04-13", "categories": []}
{"id": "2605.16529", "title": "Multiscale Supervised Unbalanced Optimal Transport Flow Matching", "abstract": "Unbalanced optimal transport (UOT) provides a principled framework for modeling single-cell transitions and birth-death dynamics, but its high computational cost limits scalability to large-scale datasets. Although single-cell data often contain hierarchical annotations and known transition priors, existing UOT approximations rarely exploit this multiscale structure or prior knowledge. We introduce Multiscale Supervised Unbalanced Optimal Transport Flow Matching (MUST-FM), a simulation-free framework that scales UOT by leveraging hierarchical data structure. MUST-FM further supports an optional supervised formulation that incorporates transition priors, such as cell lineages, to guide the learning of displacement fields and mass variations. Experiments show that MUST-FM reduces computational overhead while achieving robust and biologically meaningful trajectory inference, enabling dynamic modeling of atlas-scale single-cell datasets.", "date": "2026-05-15", "categories": []}
{"id": "2605.17693", "title": "Fine-tuning Pocket-Aware Diffusion Models via Denoising Policy Optimization", "abstract": "Structure-based drug design has been accelerated by pocket-aware 3D generative models, yet most methods primarily fit the training distribution and may fall short of satisfying multiple properties required in real-world therapeutic drug discovery. Recently, increasing attention has focused on structure-based molecule optimization (SBMO), which targets fine-grained control over multiple specified molecular properties. In this paper, we present DEPPA, a novel SBMO approach building upon Denoising Diffusion Policy Optimization for fine-tuning a pre-trained pocket-aware diffusion model via reinforcement learning. DEPPA enables optimization over multiple properties, including binding affinity, drug-likeness, synthesizability and diversity. We formulate the reverse denoising process of the pretrained pocket-aware diffusion model as a multi-step Markov Decision Process, where the desired properties that serve as reward signals are evaluated on the final generated ligand molecules. DEPPA incorporates a coarse denoising scheduler during the RL fine-tuning to achieve efficient and effective molecule optimization. Experimental results on the CrossDocked2020 benchmark demonstrate that DEPPA outperforms baselines in binding affinity (Vina Score -8.5 kcal/mol), drug-likeness and diversity while exhibiting competitive performance in synthesizability. The source code is available at https://github.com/xy9485/DePPA .", "date": "2026-05-17", "categories": []}
{"id": "2605.17899", "title": "DCFold: Efficient Protein Structure Generation with Single Forward Pass", "abstract": "AlphaFold3 introduces a diffusion-based architecture that elevates protein structure prediction to all-atom resolution with improved accuracy. This state-of-the-art performance has established AlphaFold3 as a foundation model for diverse generation and design tasks. However, its iterative design substantially increases inference time, limiting practical deployment in downstream settings such as virtual screening and protein design. We propose DCFold, a single-step generative model that attains AlphaFold3-level accuracy. Our Dual Consistency training framework, which incorporates a novel Temporal Geodesic Matching (TGM) scheduler, enables DCFold to achieve a 15x acceleration in inference while maintaining predictive fidelity. We validate its effectiveness across both structure prediction and binder design benchmarks.", "date": "2026-05-18", "categories": []}
{"id": "2605.18055", "title": "FLAG: Foundation model representation with Latent diffusion Alignment via Graph for spatial gene expression prediction", "abstract": "Predicting spatial gene expression from routine H\\&E enables large-scale molecular profiling, yet current models treat this as isolated pointwise tasks, thereby overlooking essential biological structures like gene coordination and spatial distribution. To preserve these relationships, we introduce \\textbf{FLAG}, a diffusion-based framework that redefines this task as structured distribution modeling. At the same time, we identify the critical \\textbf{Gene Dimension Curse}, where joint modeling gene expression and their spatial interactions fail in high-dimensional spaces, and FLAG solves this challenge by integrating a spatial graph encoder for topological consistency and utilizing Gene Foundation Model (GFM) alignment for gene-gene fidelity in the generation process. To rigorously assess model performance, we propose a set of novel structural evaluation metrics, including Gene Structural Correlation (\\textbf{GSC}) and Spatial Structural Correlation (\\textbf{SSC}). Our experiments demonstrate that FLAG is highly competitive in traditional accuracy (PCC/MSE) while achieving significantly enhanced structural fidelity in capturing both gene-gene and gene-spatial relationships. The code is available at https://github.com/darkflash03/FLAG.", "date": "2026-05-18", "categories": []}
{"id": "2605.18552", "title": "Protein Fold Classification at Scale: Benchmarking and Pretraining", "abstract": "Classifying protein topology is essential for deciphering biological function, but progress is held back by the lack of large-scale benchmarks that avoid duplicates and by models that do not scale well. We introduce TEDBench, a large-scale, non-redundant benchmark for protein fold classification constructed from the Encyclopedia of Domains (TED) and Foldseek-clustered AlphaFold structures. We show that on TEDBench, current protein representation learning methods either require very large models or fail to deliver strong performance. To address this challenge, we propose Masked Invariant Autoencoders (MiAE), a self-supervised framework for protein structure representation learning. MiAE uses an extremely high masking ratio of up to 90% with an $\\mathrm{SE(3)}$-invariant encoder and a lightweight decoder that reconstructs backbone coordinates from the latent representation and mask tokens. MiAE scales well and outperforms supervised counterparts and state-of-the-art baselines on TEDBench, establishing a strong recipe for protein fold classification. To test transfer beyond AlphaFold structures, we further benchmark on a curated dataset from experimental structures of CATH v4.4. TEDBench is available at https://github.com/BorgwardtLab/TEDBench.", "date": "2026-05-18", "categories": []}
{"id": "2605.18576", "title": "scHelix: Asymmetric Dual-Stream Integration via Explicit Gene-Level Disentanglement", "abstract": "A critical challenge in single-cell RNA sequencing (scRNA-seq) integration is resolving the tension between eliminating batch effects and maintaining biological fidelity. While recent evidence indicates that batch effects manifest heterogeneously across genes, most existing methods process the transcriptome uniformly, frequently resulting in over-correction and loss of subtle biological signals. To address this, we present scHelix, a dataset-adaptive framework that fundamentally changes how features are processed by explicitly partitioning genes into domain-invariant Anchors and domain-sensitive Variants at the input level. scHelix utilizes a dual-stream sparse diffusion encoder equipped with stop-gradient graph caching to efficiently learn multi-scale structural representations. The core of our approach is a novel asymmetric Align-Refine-Fuse protocol: the unstable Variant stream is first aligned to the robust topology of the Anchor stream, followed by a conservative refinement phase where the Anchor stream absorbs denoised details via bounded residual gating. This divide-and-conquer architecture prevents shortcut learning and ensures robust batch removal without compromising the integrity of biological clusters. Extensive benchmarking demonstrates that scHelix outperforms state-of-the-art methods.", "date": "2026-05-18", "categories": []}
{"id": "2504.09544", "title": "Integrating chemical structures as treatments improves representations of microscopy images for morphological profiling", "abstract": "Recent advances in self-supervised deep learning have improved our ability to quantify cellular morphological changes in high-throughput microscopy screens, a process known as morphological profiling. However, most current methods only learn from images, despite many screens being inherently multimodal, as they involve both a chemical or genetic perturbation as well as an image-based readout. We hypothesized that incorporating chemical compound structures during self-supervised pre-training could improve learned representations of images from high-throughput microscopy screens. We introduce a representation learning framework, MICON (Molecular-Image Contrastive Learning), that models chemical compounds as treatments that induce transformations of cell phenotypes. MICON significantly outperforms classical hand-crafted features such as CellProfiler and existing deep-learning-based representation learning methods in challenging evaluation settings where models must identify reproducible effects of drugs across independent replicates and data-generating centers. We demonstrate that incorporating chemical compound information into the learning process provides small, but consistent improvements in performance and that modeling compounds specifically as treatments outperforms approaches that directly align images and compounds in a single representation space. Our findings point to a new direction for representation learning in morphological profiling, suggesting that methods should explicitly account for the multimodal nature of microscopy screening data.", "date": "2025-04-13", "categories": []}
{"id": "2504.18522", "title": "Extrapolation Guarantees for Perturbation Modeling Under the Additive Latent Shift Assumption", "abstract": "We consider the problem of modeling the effects of perturbations like gene knockouts on measurements such as single-cell RNA counts. Given data for some perturbations, we aim to predict the distribution of measurements for new combinations of perturbations. To address this challenging extrapolation task, we posit that perturbations act additively in a suitable, unknown embedding space. We formulate the data-generating process as a latent variable model, in which perturbations amount to mean shifts in latent space and can be combined additively. We then prove that, given sufficiently diverse training perturbations, the representation and perturbation effects are identifiable up to orthogonal transformation and use this to derive extrapolation guarantees for unseen perturbations that can be expressed as linear combinations of seen ones. To estimate the model from data, we propose the perturbation distribution autoencoder (PDAE), which is trained by maximizing the distributional similarity between true and simulated perturbation distributions. The trained model can then be used to predict previously unseen perturbation distributions. In support of our theoretical results, we demonstrate through simulations that PDAE can accurately predict the effects of unseen but identifiable perturbations, and showcase the method on combinatorial gene perturbation data.", "date": "2025-04-25", "categories": []}
{"id": "2509.12266", "title": "Genome-Factory: A Library for Tuning, Deploying, and Interpreting Genomic Foundation Models", "abstract": "We introduce Genome-Factory, the first integrated Python library for tuning, deploying, and interpreting genomic foundation models. Our core contribution is to simplify and unify the workflow for genomic model development: data collection, model tuning, inference, benchmarking, and interpretability. For data collection, Genome-Factory offers an automated pipeline to download genomic sequences and preprocess them. For model tuning, Genome-Factory supports both full and parameter-efficient fine-tuning across diverse genomic models. For inference, Genome-Factory enables both embedding extraction and DNA sequence generation. For benchmarking, we include two existing benchmarks and provide a flexible interface to incorporate additional benchmarks. For interpretability, Genome-Factory introduces an open-source biological interpreter based on a sparse auto-encoder. We validate the utility of Genome-Factory across three dimensions: (i) Compatibility with diverse models and fine-tuning methods; (ii) Benchmarking downstream performance using two open-source benchmarks; (iii) Biological interpretation of learned representations with DNABERT-2. These results highlight its practical value for real-world genomic analysis. GitHub: https://github.com/WeiminWu2000/Genome_Factory.", "date": "2025-09-13", "categories": []}
{"id": "2510.01632", "title": "BioBlobs: Unsupervised Discovery of Functional Substructures for Protein Function Prediction", "abstract": "Protein function is driven by cohesive substructures, such as catalytic triads, binding pockets, and structural motifs, that occupy only a small fraction of a protein's residues. Yet existing pipelines built on protein encoders do not model proteins at the substructure level, leaving the central biological question unanswered: which substructure of a protein is responsible for its function? We introduce BioBlobs, an encoder-agnostic, end-to-end differentiable framework that compresses a protein into a small set of cohesive substructures (blobs) and predicts function from these blobs alone, so that each blob corresponds to a candidate functional region. Across diverse protein function prediction tasks and multiple sequence- and structure-based encoders, BioBlobs matches or exceeds strong baselines while operating on only a small fraction of residues. The discovered blobs adapt their spatial scale to the task, ranging from local catalytic sites to entire structural domains. Trained only on protein-level labels, BioBlobs recovers experimentally annotated catalytic sites in the M-CSA database, demonstrating unsupervised functional substructure discovery and opening a path to large-scale functional site discovery across the unannotated proteome.", "date": "2025-10-02", "categories": []}
{"id": "2510.02734", "title": "SAE-RNA: A Sparse Autoencoder Model for Interpreting RNA Language Model Representations", "abstract": "Deep learning, particularly with the advancement of Large Language Models, has transformed biomolecular modeling, with protein language models such as ESM inspiring emerging RNA language models such as RiNALMo. Recent work has begun applying sparse autoencoders (SAEs) to protein language model representations, exploring representation-level interpretability in biomolecular models. Here, we explore whether SAEs can provide interpretable feature decompositions of RNA language model representations, while also examining their limitations in this setting. We present SAE-RNA, interpretability model that analyzes RiNALMo representations and maps them to known human-level biological features. Rather than claiming definitive biological concept discovery, our study frames SAE-based analysis as a representation-level probe for characterizing how RNA language models organize biological information internally. More broadly, SAE-RNA provides a feature-level framework for comparing RNA groups and identifying sparse representation components associated with RNA family identity or structural context.", "date": "2025-10-03", "categories": []}
{"id": "2605.15383", "title": "MorphoHELM: A Comprehensive Benchmark for Evaluating Representations for Microscopy-Based Morphology Assays", "abstract": "Microscopy images contain rich information about how cells respond to perturbations, making them essential to applications like drug screening. To quantify images, researchers often use representation extraction methods, and recent years have seen a proliferation of deep learning methods. While measuring the quality of these representations is essential, evaluation remains fragmented, with each proposed model evaluated on different tasks and datasets, using custom pipelines and metrics, making it difficult to fairly compare models. Here, we introduce MorphoHELM, a comprehensive open benchmark for evaluating feature extraction methods for Cell Painting, the most widely-used morphological profiling assay. MorphoHELM consolidates evaluation standards in the field, extends and corrects them to be more robust, and evaluates on the widest range of methods to date. A defining feature of the benchmark is that each task is evaluated at different degrees of batch effects (or technical noise), directly quantifying how the ability of methods to detect biological signal degrades as noise increases. Together, these properties enable MorphoHELM to detect trade-offs between methods, and we demonstrate that models that excel at certain kinds of biological signal are weaker at others. We show that no existing model outperforms classic computer vision analytic strategies across all settings, which remain the strongest general use-case representations. All datasets, code, and evaluation tools are publicly available at https://github.com/microsoft/MorphoHELM.", "date": "2026-05-14", "categories": []}
{"id": "2605.15564", "title": "CrystalBoltz: End-to-End Protein Structure Determination via Experiment-Guided Diffusion for X-Ray Crystallography", "abstract": "Generative models trained on public databases of protein structures, most of which have been determined by X-ray crystallography, now provide powerful priors for structure prediction. However, they are not readily conditioned on the measurements from a new crystallographic experiment, limiting their use for X-ray structure determination. In crystallography, the measured structure-factor amplitudes do not by themselves determine an electron density map or atomic structure because the associated phases are unobserved and must be inferred. Structure determination therefore remains an inverse problem in which candidate models must be both structurally plausible and consistent with measured diffraction data, often requiring substantial manual refinement by human experts. Emerging methods aim to incorporate experimental information more directly into predictive and refinement workflows. We present CrystalBoltz, a generative framework that casts crystallographic refinement as Bayesian inference over atomic structures and operates directly on structure-factor amplitudes. CrystalBoltz moves from unguided generation with a pre-trained prior over protein structures to experiment-guided posterior sampling, followed by atomic coordinate and B-factor refinement. Across multiple protein crystallography datasets, CrystalBoltz attains lower coordinate RMSD and lower R-factors than the strongest baselines considered, while reducing runtime by a factor of 33 relative to existing experimentally guided refinement.", "date": "2026-05-15", "categories": []}
{"id": "2602.05285", "title": "Robust Inference-Time Steering of Protein Diffusion Models via Embedding Optimization", "abstract": "A core challenge in structural biophysics is generating biomolecular conformations that are both physically plausible and consistent with experimental measurements. While sequence-to-structure diffusion models provide powerful priors, posterior sampling methods steer generation by perturbing atomic coordinates with gradients from experimental likelihoods. However, when the target lies in a low-density region of the prior, these methods require aggressive upweighting of the likelihood that can destabilize sampling and be sensitive to hyperparameters. We propose EmbedOpt, an inference-time steering framework that introduces an orthogonal optimization axis: rather than performing posterior sampling under a fixed prior, EmbedOpt directly optimizes the prior by updating the model's conditional embedding. This embedding space encodes rich coevolutionary signals, so optimizing it shifts the structural prior to align with experimental constraints. Empirically, EmbedOpt matches coordinate-based posterior sampling baselines on sparse distance constraints and outperforms them on cryo-electron microscopy map fitting, including real, noisy experimental ones. Furthermore, EmbedOpt's smooth optimization behavior yields robustness to hyperparameters spanning two orders of magnitude and enables comparable performance with fewer diffusion steps. Code is available at https://github.com/rs-station/embedopt.", "date": "2026-02-05", "categories": []}
{"id": "2605.12784", "title": "ToolMol: Evolutionary Agentic Framework for Multi-objective Drug Discovery", "abstract": "Advances in large language models (LLMs) have recently opened new and promising avenues for small-molecule drug discovery. Yet existing LLM-based approaches for molecular generation often suffer from high rates of invalid and low-quality ligand candidates, a result of the syntactic limitations of current models with regard to molecular strings. In this paper, we introduce $\\texttt{ToolMol}$, an evolutionary agentic framework for de novo drug design. $\\texttt{ToolMol}$ combines a multi-objective genetic algorithm with an agentic LLM operator that iteratively updates the ligand population. We build a comprehensive toolbox of RDKit-backed functions that allows our agentic operator to consisently make precise ligand modifications. $\\texttt{ToolMol}$ achieves state-of-the-art performance on multi-objective property optimization tasks, discovering drug-like and synthesizable ligands that have $>10\\%$ stronger predicted binding affinity compared to existing methods, evaluated on three protein targets. $\\texttt{ToolMol}$ ligands additionally achieve state-of-the-art results in gold-standard Absolute Binding Free Energy scores, gaining over existing methods by over $35\\%$. By studying chain-of-thought reasoning traces, we observe that tool-calling enables the model to more faithfully execute its planned modifications, efficiently exploiting the strong chemical prior knowledge in LLMs.", "date": "2026-05-12", "categories": []}
{"id": "2605.13789", "title": "ENSEMBITS: an alphabet of protein conformational ensembles", "abstract": "Protein structure tokenizers (PSTs) are workhorses in protein language modeling, function prediction, and evolutionary analysis. However, existing PSTs only capture local geometry of static structures, and miss the correlated motions and alternative conformational states revealed by protein ensembles. Here we introduce Ensembits, the first tokenizer of protein conformational ensembles. Ensembits address challenges inherent to tokenizing dynamics: deriving informative geometric descriptors across conformations, permutation-invariance encoding of variable-size ensembles, and conquering sparsity in dynamics data. Trained with a Residual VQ-VAE using a frame distillation objective on a large molecular dynamics corpus, Ensembits outperforms all related methods on RMSF prediction, and is the strongest standalone structural tokenizer on an token-conditioned ANOVA test on per-residue motion amplitude. Ensembits further matches or exceeds static tokenizers on EC, GO, binding site/affinity prediction, and zero-shot mutation-effect prediction despite using far less pretraining data. Notably, the distillation objective enables Ensembits to predict dynamics token from one single predicted structure, which alleviates dynamics data sparsity. As the field moves from static structure prediction toward ensemble generation, Ensembits offer the discrete vocabulary needed to bring dynamics into protein language modeling and design.", "date": "2026-05-13", "categories": []}
{"id": "2605.14073", "title": "AttnGen: Attention-Guided Saliency Learning for Interpretable Genomic Sequence Classification", "abstract": "Deep neural networks have achieved strong performance in genomic sequence classification; however, relating their predictions to biologically meaningful sequence patterns remains challenging. In this work, we present AttnGen, an attention-guided training framework that embeds interpretability directly into the optimization process. AttnGen computes nucleotide-level importance scores using an attention mechanism and progressively suppresses low-contribution positions during training. This encourages the model to focus its predictions on a compact set of informative regions while reducing reliance on noisy sequence elements. We evaluate AttnGen on the standardized demo_human_or_worm benchmark, a binary classification task over 200-nucleotide sequences. With moderate masking, AttnGen achieves a validation accuracy of 96.73%, outperforming a conventional CNN baseline with 95.83% accuracy, while also exhibiting faster convergence and improved training stability. To assess whether the learned importance scores reflect functionally relevant signal, we conduct perturbation-based analysis by removing high-saliency nucleotides. This causes accuracy to drop from 96.9% to near chance level on a 3,000-sequence evaluation set, indicating that the model relies on a relatively small subset of informative positions. Our analysis shows that masking 10--20% of positions provides the most favorable trade-off between predictive performance and interpretability. These results suggest that attention-guided masking not only improves classification performance but also reshapes how models distribute importance across sequence positions. Although this study focuses on short genomic sequences, the proposed approach may extend to more complex interpretable sequence modeling settings.", "date": "2026-05-13", "categories": []}
{"id": "2605.14104", "title": "DUET: Dual-Paradigm Adaptive Expert Triage with Single-cell Inductive Prior for Spatial Transcriptomics Prediction", "abstract": "Inferring spatially resolved gene expression from histology images offers a cost-effective complement to spatial transcriptomics (ST). However, existing methods reduce this task to a simple morphology-to-expression mapping, where visual similarity does not guarantee molecular consistency. Meanwhile, single-cell data has amassed rich resources far surpassing the scale of ST data, yet it remains underexplored in vision-omics modeling. Furthermore, current approaches commit to a monolithic paradigm with bottlenecks, unable to balance expressive flexibility with biological fidelity. To bridge these gaps, we propose DUET, a novel dual-paradigm framework that synergizes parametric prediction and memory-based retrieval under cellular inductive priors. DUET implements a parallel regression-retrieval paradigm, adaptively reconciling the outputs of its complementary pathways. To mitigate aleatoric vision ambiguity, we incorporate large-scale single-cell references to impose molecular states as biological constraints for faithful learning. Building upon structural refinement, we further design a lightweight adapter to dynamically assign branch preference across spatial contexts to achieve optimal performance. Extensive experiments on three public datasets across varied gene scales demonstrate that DUET achieves SOTA performance, with consistent gains contributed by each proposed component. Code is available at https://github.com/Junchao-Zhu/DUET", "date": "2026-05-13", "categories": []}
{"id": "2605.14215", "title": "GenCircuit-RL: Reinforcement Learning from Hierarchical Verification for Genetic Circuit Design", "abstract": "Genetic circuit design remains a laborious, expert-driven process despite decades of progress in synthetic biology. We study this problem through code generation: models produce Python code in pysbol3 to construct genetic circuits in the Synthetic Biology Open Language (SBOL), a formal representation that supports automated verification. We introduce GenCircuit-RL, a reinforcement learning framework built around hierarchical verification rewards that decompose correctness into five levels, from code execution to task-specific topological checks, and a four-stage curriculum that shifts optimization pressure from code generation to functional reasoning. We also introduce SynBio-Reason, a benchmark of 4,753 circuits spanning six canonical circuit types and nine tasks from code repair to de novo design, with held-out biological parts for out-of-distribution evaluation. Hierarchical verification improves task success on functional reasoning tasks by 14 to 16 percentage points over binary rewards, and curriculum learning is required for strong design performance. The resulting models generate topologically correct circuits, generalize to novel biological parts, and rediscover canonical designs from the synthetic biology literature.", "date": "2026-05-14", "categories": []}
{"id": "2605.14327", "title": "AIM-DDI: A Model-Agnostic Multimodal Integration Module for Drug-Drug Interaction Prediction", "abstract": "Drug-drug interaction (DDI) prediction is a critical task in computational biomedicine, as adverse interactions between co-administered drugs can cause severe side effects and clinical risks. A key challenge is unseen-drug generalization, where interactions must be predicted for drugs not observed during training. Although multimodal DDI models exploit diverse drug-related information, their fusion mechanisms are often tied to specific prediction architectures, limiting their reuse across models. To address this, we propose AIM-DDI, an architecture-independent multimodal integration module that represents heterogeneous modality information as tokens in a shared latent space. By modeling dependencies across modality tokens through a unified fusion module, AIM-DDI enables model-agnostic integration of structural, chemical, and semantic drug signals across different DDI prediction architectures. Extensive evaluations across diverse DDI models and DrugBank-based settings show that AIM-DDI consistently improves prediction performance, with the strongest gains under the most challenging both-unseen setting where neither drug in a test pair is observed during training. These results suggest that treating multimodal integration as a reusable module, rather than a model-specific fusion component, is an effective strategy for robust unseen-drug DDI prediction.", "date": "2026-05-14", "categories": []}
{"id": "2602.12026", "title": "Protein Circuit Tracing via Cross-layer Transcoders", "abstract": "Protein language models (pLMs) have emerged as powerful predictors of protein structure and function. However, the computational circuits underlying their predictions remain poorly understood. Recent mechanistic interpretability methods decompose pLM representations into interpretable features, but they treat each layer independently and thus fail to capture cross-layer computation, limiting their ability to approximate the full model. We introduce ProtoMech, a framework for discovering computational circuits in pLMs using cross-layer transcoders that learn sparse latent representations jointly across layers to capture the model's full computational circuitry. Applied to the pLM ESM2, ProtoMech recovers 82-89% of the original performance on protein family classification and function prediction tasks. ProtoMech then identifies compressed circuits that use <1% of the latent space while retaining up to 79% of model accuracy, revealing correspondence with structural and functional motifs, including binding, signaling, and stability. Steering along these circuits enables high-fitness protein design, surpassing baseline methods in more than 70% of cases. These results establish ProtoMech as a principled framework for protein circuit tracing.", "date": "2026-02-12", "categories": []}
{"id": "2604.13986", "title": "PRiMeFlow: Capturing Complex Expression Heterogeneity in Perturbation Response Modelling", "abstract": "Predicting the effects of perturbations in-silico on cell state can identify drivers of cell behavior at scale and accelerate drug discovery. However, modeling challenges remain due to the inherent heterogeneity of single cell gene expression and the complex, latent gene dependencies. Here, we present PRiMeFlow, an end-to-end flow matching based approach to directly model the effects of genetic and small molecule perturbations in the gene expression space. The distribution-fitting approach taken by PRiMeFlow enables it to accurately approximate the empirical distribution of single-cell gene expression, which we demonstrate through extensive benchmarking inside PerturBench. Through ablation studies, we also validate important model design choices such as operating in gene expression space and parameterizing the velocity field with a U-Net architecture. Finally, by scaling PRiMeFlow to a broad perturbation data atlas spanning multiple datasets and employing a carefully designed pretraining-finetuning strategy, we demonstrate its outstanding performance on the H1 human embryonic stem cells from the ARC Virtual Cell Challenge benchmark.", "date": "2026-04-15", "categories": []}
{"id": "2605.07433", "title": "Inference of Qualitative Models from Steady-State Data via Weighted MaxSMT", "abstract": "Qualitative models provide crucial instruments for modelling complex biological systems. While advances in automated reasoning and symbolic encodings have enabled rigorous inference of these models from data, the process remains highly fragile. First, biological measurement errors inevitably propagate into formal model specifications. Second, when a specification becomes unsatisfiable, distinguishing between fundamental design flaws and minor technical errors is notoriously difficult. This uncertainty often leads to under-specification, as it is unclear which observations are still ``safe'' to incorporate. To overcome these challenges, we introduce a robust inference method based on weighted MaxSMT. By encoding uncertain biological observations as weighted soft constraints, our approach enables the solver to identify a model best reflecting the observations, even with some conflicting constraints. Our method allows for Boolean and multi-valued variable domains, alongside observations derived from discretisation (level constraints) and differential expression (ordering constraints). We show our approach can be used to successfully infer neural cell differentiation models from prior-knowledge networks with 200--1300 genes using ordering constraints on all included genes.", "date": "2026-05-08", "categories": []}
{"id": "2605.12662", "title": "scShapeBench: Discovering geometry from high dimensional scRNAseq data", "abstract": "High-dimensional point cloud data arise across many scientific domains, especially single-cell biology. The shapes or topologies of these datasets determine the types of information that can be extracted. For example, clustered data supports cell-type identification, trajectory structures support transition analysis, and archetypal structures capture continua of cellular behaviors. Existing analysis pipelines often assume a specific shape. The standard Seurat pipeline combines UMAP visualization with Louvain clustering and therefore assumes clustered data, while tools such as Monocle and SPADE assume tree-like structures, and flow-based models such as MIOFlow and Conditional Flow Matching target trajectories. Choosing which pipeline to apply is therefore often left to bioinformaticians who visually inspect datasets before selecting an analysis strategy. With the rise of agentic AI scientists, automating shape detection is increasingly important for selecting downstream analysis pipelines. To address this problem, we introduce scShapeBench, a benchmark dataset for shape detection containing both synthetic and expert-annotated single-cell datasets. Synthetic datasets are sampled from ground-truth skeleton graphs with controlled variance. Real single-cell datasets are curated from diverse sources and annotated by experts into four categories: clusters, single trajectory, multi-branching, and archetypal. We additionally introduce scReebTower, a baseline method that uses diffusion geometry to extract Reeb graphs and connect visualization with pipeline selection. We provide topology-aware evaluation metrics and compare scReebTower against PAGA and Mapper on synthetic and real data. Our results indicate that scReebTower outperforms existing baselines. Overall, our contributions span benchmarks, evaluation metrics, and a baseline for automated shape detection in single-cell data.", "date": "2026-05-12", "categories": []}
{"id": "2605.12823", "title": "Hessian Matching for Machine-Learned Coarse-Grained Molecular Dynamics", "abstract": "Coarse-grained (CG) molecular dynamics enables simulations of atomic systems such as biomolecules at timescales inaccessible to all-atom (AA) methods, but existing CG neural potentials trained via force matching capture only the gradient of the free-energy surface, leaving its curvature unconstrained. We introduce a framework that augments force matching with stochastic Hessian-vector product (HVP) matching, instilling second-order curvature information into CG potentials without constructing the full Hessian. We derive a decomposition of the target CG Hessian into a model-independent projected AA Hessian, precomputed once before training, and a model-dependent covariance correction computed online at negligible cost. We construct an unbiased stochastic estimator of the Hessian-matching objective by using random probe vectors. We evaluate our method by comparing against force matching on a benchmark of nine fast-folding proteins unseen during training. HVP matching outperforms plain force matching on 8 of 9 proteins on slow-mode metrics, with reductions of up to 85% in the Kullback--Leibler divergence between the CG and reference distributions along the slowest collective mode of the largest protein. Our results demonstrate that higher-order physical supervision is a practical path to more accurate and transferable CG potentials for biomolecular simulation.", "date": "2026-05-12", "categories": []}
{"id": "2602.02282", "title": "MoLF: Mixture-of-Latent-Flow for Pan-Cancer Spatial Gene Expression Prediction from Histology", "abstract": "Inferring spatial transcriptomics (ST) from histology enables scalable histogenomic profiling, yet current methods are largely restricted to single-tissue models. This fragmentation fails to leverage biological principles shared across cancer types and hinders application to data-scarce scenarios. While pan-cancer training offers a solution, the resulting heterogeneity challenges monolithic architectures. To bridge this gap, we introduce MoLF (Mixture-of-Latent-Flow), a generative model for pan-cancer histogenomic prediction. MoLF leverages a conditional Flow Matching objective to map noise to the gene latent manifold, parameterized by a Mixture-of-Experts (MoE) velocity field. By dynamically routing inputs to specialized sub-networks, this architecture effectively decouples the optimization of diverse tissue patterns. Our experiments demonstrate that MoLF establishes a new state-of-the-art, consistently outperforming both specialized and foundation model baselines on pan-cancer benchmarks. Furthermore, MoLF exhibits zero-shot generalization to cross-species data, suggesting it captures fundamental, conserved histo-molecular mechanisms.", "date": "2026-02-02", "categories": []}
{"id": "2602.17739", "title": "GeneZip: Region-Aware Compression for Long Context DNA Modeling", "abstract": "Long-context DNA models are limited by token-mixing cost and by how compression allocates representational budget across the genome. Existing approaches operate close to base-pair resolution, apply fixed downsampling, or learn content-dependent chunks without an explicit genomic budget, making long-context pretraining expensive and difficult to control. We introduce GeneZip, a region-aware DNA compression framework that combines H-Net-style dynamic routing with a Region-Aware Ratio (RAR) objective and bounded routing. GeneZip uses static gene-structure annotations during compression training to specify region-wise base-pairs-per-token (BPT) targets; at inference time, it compresses raw unseen DNA without annotations. GeneZip provides three main benefits. First, it is effective: GeneZip variants achieve the best validation PPL among encoder-based compressors, with GeneZip-70M operating at 137.6 BPT, and across four reproducible DNALongBench tasks--contact map prediction, eQTL prediction, enhancer-target gene prediction, and transcription-initiation signal prediction--GeneZip obtains the best average rank among compared sequence models. Second, it is redundancy-aware: a post-hoc RepeatMasker/TRF analysis shows that, without repeat supervision, GeneZip assigns higher local BPT to TE-derived interspersed repeats and tandem repeats, two major classes of repetitive DNA sequence redundancy. Third, it is efficient: by reducing the effective token-mixing length, GeneZip enables longer-context and larger-capacity pretraining, including 128K-context and 636M-parameter variants on a single A100 80GB GPU, and fine-tunes the eQTL task 50.4x faster than JanusDNA (50 vs. 2520 minutes). These results establish GeneZip as an effective, redundancy-aware, and efficient compression interface for long-context DNA modeling.", "date": "2026-02-19", "categories": []}
{"id": "2605.01625", "title": "PRIME: Protein Representation via Physics-Informed Multiscale Equivariant Hierarchies", "abstract": "Proteins are inherently multiscale physical systems whose functional properties emerge from coordinated structural organization across multiple spatial resolutions, ranging from atomic interactions to global fold topology. However, existing protein representation learning methods typically operate at a single structural level or treat different sources of structural information as parallel modalities, without explicitly modeling their hierarchical relationships. We introduce PRIME (Protein Representation via Physics-Informed Multiscale Equivariant Hierarchies), a unified framework that models proteins as a nested family of five physically grounded structural graphs spanning surface, atomic, residue, secondary-structure, and protein levels. Adjacent levels are connected through deterministic, physics-informed assignment operators, enabling bidirectional information exchange via bottom-up aggregation and top-down contextual refinement. Experiments on standard protein representation learning benchmarks demonstrate strong and competitive performance across diverse tasks, with particularly notable gains on the Fold Classification benchmark, where PRIME outperforms the strongest geometric GNN baseline by margins of 13.80 and 18.30 points on the harder Superfamily and Fold splits, and achieves a state-of-the-art accuracy of 84.10\\% on Reaction Class prediction, surpassing all baseline methods, including ESM. Ablation studies confirm that each structural level contributes complementary and non-redundant information, and adaptive cross-attention analysis reveals that PRIME autonomously identifies the most task-relevant structural resolutions at prediction time. Our source code is publicly available at https://github.com/HySonLab/PRIME", "date": "2026-05-02", "categories": []}
{"id": "2605.09964", "title": "Learning the Interaction Prior for Protein-Protein Interaction Prediction: A Model-Agnostic Approach", "abstract": "Protein-protein interactions (PPIs) are fundamental to cellular function and disease mechanisms. Current learning-based PPI predictors focus on learning powerful protein representations but neglect designing specialized classification heads. They mainly rely on generic aggregating methods like concatenation or dot products, which lack biological insight. Motivated by the biological \"L3 rule\", where multiple length-3 paths between a pair of proteins indicate their interaction likelihood, our study addresses this gap by designing a biologically informed PPI classifier. In this paper, we provide empirical evidence that popular PPI datasets strongly support the L3 rule. We propose an L3-path-regularized graph prompt learning method called L3-PPI, which can generate a prompt graph with virtual L3 paths based on protein representations and controls the number of paths. L3-PPI reformulates the classification of protein embedding pairs into a graph-level classification task over the generated prompt graph. This lightweight module seamlessly integrates with PPI predictors as a plug-and-play component, injecting the interaction prior of complementarity to enhance performance. Extensive experiments show that L3-PPI achieves superior performance enhancements over advanced competitors.", "date": "2026-05-11", "categories": []}
{"id": "2605.10985", "title": "Structural Interpretations of Protein Language Model Representations via Differentiable Graph Partitioning", "abstract": "Protein language models such as ESM-2 learn rich residue representations that achieve strong performance on protein function prediction, but their features remain difficult to interpret as structural $\\&$ evolutionary signals are encoded in dense latent spaces. We propose a plug-$\\&$-play framework that projects ESM-2 representations onto protein contact graphs $\\&$ applies $\\textbf{SoftBlobGIN}$, a lightweight Graph Isomorphism Network with differentiable Gumbel-softmax substructure pooling, to perform structure-aware message passing $\\&$ learn coarse functional substructures for downstream prediction tasks. Across enzyme classification, SoftBlobGIN achieves 92.8\\% accuracy $\\&$ 0.898 macro-F1. Unlike post hoc analysis of protein language models alone, our method produces directly auditable structural explanations: GNNExplainer recovers biologically meaningful active-site residues, spatially localized functional clusters, $\\&$ catalytic contact patterns. On binding-site detection, SoftBlobGIN improves residue AUROC from $0.885$ using an ESM-2 linear probe to $0.983$, indicating that these structural explanations are not recoverable from language-model features alone. Learned blob partitions provide an additional layer of interpretability by automatically grouping residues into functional substructures, with blobs containing annotated active-site residues showing $1.85\\times$ higher importance than other blobs ($\\rho{=}0.339$, $p{=}0.009$), without any active-site supervision. Our framework requires no retraining of the language model, adds only $\\sim$1.1M parameters, $\\&$ generalises across ProteinShake tasks, achieving $F_{\\max}$ of $0.733$ on Gene Ontology prediction $\\&$ AUROC of $0.969$ on binding-site detection. We position this as an interpretable structural companion to protein language models that makes their predictions more transparent $\\&$ auditable.", "date": "2026-05-09", "categories": []}
{"id": "2605.11022", "title": "SCOPE: Siamese Contrastive Operon Pair Embeddings for Functional Sequence Representation and Classification", "abstract": "Identifying operons is a fundamental step in understanding prokaryotic gene regulation, as classifying genes into operons supports the reconstruction of regulatory networks, functional annotation of unannotated genes, and drug candidate development. Experimental approaches such as RT-PCR and RNA-seq provide precise evidence of operon structure, but are laborious and largely limited to well-studied model organisms, making scalable computational methods essential for genome-wide operon identification. Prior computational approaches have employed traditional classifiers such as logistic regression and decision trees, motivating our use of these as physicochemical baselines. The DGEB benchmark evaluates operonic pair classification by embedding each sequence independently with a pre-trained protein language model and computing pairwise cosine similarity. In contrast, our Siamese MLP learns a classifier over the fused embedding space, which is theoretically better motivated for binary classification, as cosine similarity can yield meaningless scores depending on the regularization of the embedding model. While protein language model embeddings substantially outperform physicochemical features in ROC-AUC, a learned Siamese MLP head does not significantly improve over unsupervised cosine similarity in Average Precision, suggesting that the geometry of the embedding space already captures the functional relationships needed for this task. Nonetheless, our Siamese MLP achieves a ROC-AUC of 0.71, competitive with state-of-the-art models on the DGEB leaderboard. These findings indicate that protein language model embeddings are a viable, scalable foundation for operonic pair classification across diverse microbial genomes, with implications for automated genome annotation, regulatory network reconstruction, and characterization of organisms lacking experimental operon annotations.", "date": "2026-05-10", "categories": []}
{"id": "2605.11368", "title": "LPDP: Inference-Time Reward Control for Variable-Length DNA Generation with Edit Flows", "abstract": "We study the application of recent Edit Flows for inference-time reward control for DNA sequence generation. Unlike most reward-guided DNA generation frameworks, which operate on fixed-length sequence spaces, Edit Flows have a potential to generate variable-length DNA through biologically plausible insertion, deletion, and substitution operations. In particular, we propose Local Perturbation Discrete Programming (LPDP), a training-free, intermediate-state and action-aware local re-solving operator for variable-length DNA edit-action generators at inference time. More specifically, at each guided rollout step, LPDP scores one-step root edits, retains a near-best root band, and re-ranks each retained root by solving a bounded local discrete program around its child sequence. This local program uses the typed geometry of edit actions to focus on coherent substitution, insertion, or deletion subgraphs, and aggregates local continuations with either a hard Max backup or a soft log-sum-exponential (LSE) backup. We instantiate LPDP in two regimes: front-loaded reward tilting for enhancer optimization, where early edits are critical for establishing global regulatory sequence structure, and back-loaded reward tilting for exon-intron-exon inpainting, where late edits fine-tune splice-boundary contexts.", "date": "2026-05-12", "categories": []}
{"id": "2605.12286", "title": "Set-Aggregated Genome Embeddings for Microbiome Abundance Prediction", "abstract": "Microbiome functions are encoded within the genes of the community-wide metagenome. A natural question is whether properties of a microbial community can be predicted just from knowing the raw DNA sequences of its members. In this work, we employ set-aggregated genome embeddings (SAGE) to predict community-level abundance profiles, exploiting the few-shot learning capabilities of genomic language models (GLMs). We benchmark this approach to show improved generalization on novel genomes compared to classical bioinformatics approaches. Model ablation shows that community-level latent representations directly result in improved performance. Lastly, we demonstrate the benefits of intermediate transformations between latent representations and demonstrate the differences between GLM embedding choices.", "date": "2026-05-12", "categories": []}
{"id": "2406.12910", "title": "Human-level molecular optimization driven by mol-gene evolution", "abstract": "De novo molecule generation allows the search for more drug-like hits across a vast chemical space. However, lead optimization is still required, and the process of optimizing molecular structures faces the challenge of balancing structural novelty with pharmacological properties. This study introduces the Deep Genetic Molecular Modification Algorithm (DGMM), which brings structure modification to the level of medicinal chemists. A discrete variational autoencoder (D-VAE) is used in DGMM to encode molecules as quantization code, mol-gene, which incorporates deep learning into genetic algorithms for flexible structural optimization. The mol-gene allows for the discovery of pharmacologically similar but structurally distinct compounds, and reveals the trade-offs of structural optimization in drug discovery. We demonstrate the effectiveness of the DGMM in several applications.", "date": "2024-06-13", "categories": []}
{"id": "2410.19471", "title": "Improving Inverse Folding for Peptide Design with Diversity-regularized Direct Preference Optimization", "abstract": "Inverse folding models play an important role in structure-based design by predicting amino acid sequences that fold into desired reference structures. Models like ProteinMPNN, a message-passing encoder-decoder model, are trained to reliably produce new sequences from a reference structure. However, when applied to peptides, these models are prone to generating repetitive sequences that do not fold into the reference structure. To address this, we fine-tune ProteinMPNN to produce diverse and structurally consistent peptide sequences via Direct Preference Optimization (DPO). We derive two enhancements to DPO: online diversity regularization and domain-specific priors. Additionally, we develop a new understanding on improving diversity in decoder models. When conditioned on OpenFold generated structures, our fine-tuned models achieve state-of-the-art structural similarity scores, improving base ProteinMPNN by at least 8%. Compared to standard DPO, our regularized method achieves up to 20% higher sequence diversity with no loss in structural similarity score.", "date": "2024-10-25", "categories": []}
{"id": "2602.22551", "title": "Identifying Multi-Hit Cancer Drivers Without Massive Parallelization: A CP, MIP, and Column Generation Framework", "abstract": "Cancer is often driven by specific combinations of an estimated two to nine gene mutations, known as multi-hit combinations. Identifying these multi-hit combinations of gene mutations that drive cancer is critical for understanding carcinogenesis and designing targeted therapies. We formalize this challenge as the Multi-Hit Cancer Driver Set Cover Problem (MHCDSCP), optimizing the selection of gene combinations to maximize tumor coverage while strictly minimizing normal sample misclassification. While existing approaches rely on exhaustive enumeration and massive parallelization, we introduce fast heuristics based on constraint programming and mixed integer programming formulations. Evaluated on real-world cancer genomics data, our framework matches state-of-the-art supercomputing methods using a single commodity CPU in under a minute. We also propose a price-and-branch heuristic which, by solving the root node to optimality, provides the first provably optimal solutions for over half of the benchmark instances, thereby verifying the near-optimality of our fast heuristics. These findings demonstrate that on real-world problem instances, the MHCDSCP is far less computationally demanding than previously believed, providing an accessible baseline that enables the exploration of previously intractable multi-hit modeling assumptions.", "date": "2026-02-26", "categories": []}
{"id": "2603.25561", "title": "An Integrative Genome-Scale Metabolic Modeling and Machine Learning Framework for Predicting and Optimizing Single-Cell Protein Production in Saccharomyces cerevisiae", "abstract": "Saccharomyces cerevisiae is increasingly recognised as a key source for single-cell protein (SCP) production, a rising solution to global protein-supply challenges. This study presents a computational framework combining the Yeast9 genome-scale metabolic model (GEM) with machine learning and optimisation to predict and enhance biomass flux for SCP yield. The Yeast9 GEM, comprising 4,131 reactions, 2,806 metabolites, and 1,161 genes, was simulated using flux balance analysis (FBA) across 2,000 Latin Hypercube-sampled flux profiles. Random Forest and XGBoost regressors achieved R2 values of 0.9999760 and 0.9997702, respectively. A variational autoencoder (VAE) identified four metabolic clusters with mean biomass fluxes of 0.472, 0.493, 0.527, and 0.505 gDW/hr. SHAP-based feature attribution identified twenty key reactions in glycolysis, the TCA cycle, and amino-acid biosynthesis; 18/20 (90%) were confirmed essential by in silico knockout. Bayesian optimisation produced a 12.13-fold improvement in biomass flux (0.0858 to 1.041 gDW/hr) at glucose = -20.0, oxygen = -20.0, and ammonium = -8.9 mmol/gDW/hr. A generative adversarial network (GAN) generated novel flux configurations (variance = 0.124); stoichiometric feasibility verification returned 0/100 feasible profiles due to incomplete generator convergence, reported as a limitation. Pareto front analysis identified an optimal SCP operating point at 0.0858 gDW/hr biomass flux with amino-acid biosynthesis score of 1000.029 mmol/gDW/hr.", "date": "2026-03-26", "categories": []}
{"id": "2605.06644", "title": "Edge-specific signal propagation on mature chromophore-region 3D mechanism graphs for fluorescent protein quantum-yield prediction", "abstract": "Fluorescent protein quantum yield (QY) is governed by the mature chromophore and its three-dimensional microenvironment rather than sequence identity alone. Protein language models and emission-band averages capture global trends, but do not model how local physical signals act on specific chromophore regions.\n  We present a chromophore-centred mechanism graph algorithm for QY prediction. Each PDB structure is converted into a typed 3D residue graph, registered to a mature-CRO state, partitioned into phenolate, bridge and imidazolinone regions, and transformed by channel-signal-region propagation. The representation contains 121 enrichment features; after removing identity shortcuts, 52 non-identity features are used for band-specific ExtraTrees regression. Because each feature encodes a contact channel, seed signal and target CRO region, interpretation is intrinsic rather than post hoc. On a 531-protein benchmark, the method achieved the best random-CV performance among model-based baselines (R = 0.772 +/- 0.008, MAE = 0.131 +/- 0.002), exceeding Band mean (R = 0.632), ESM-C (R = 0.734) and SaProt (R = 0.731), and ranked first in bright screening (Bright P@5 = 0.704). Under homology control, the advantage was clearest in the remote bucket (<50% similarity; R = 0.697 versus 0.633, 0.575 and 0.408), with the strongest overall bright/dark Top-K screening. Stable selected features recovered band-specific mechanisms: aromatic packing and clamp asymmetry in GFP-like proteins, charge/clamp balance in Red proteins, and flexibility-risk/bulky-contact features in Far-red proteins.\n  Source code, feature tables and evaluation scripts are available from the first author upon request. Contact: yuchenak05@gmail.com", "date": "2026-05-07", "categories": []}
{"id": "2605.08128", "title": "Towards Universal Gene Regulatory Network Inference: Unlocking Generalizable Regulatory Knowledge in Single-cell Foundation Models", "abstract": "Gene Regulatory Network (GRN) inference is essential for understanding complex cellular mechanisms, rendered tractable through single-cell transcriptomic data. With the emergence of single-cell Foundation Models (scFMs), enhanced transcriptomic encoding is widely expected to revolutionize GRN inference. However, we observe that their performance remains far from satisfactory. The primary reason is that the standard reconstruction-based pre-training objectives often fail to explicitly capture latent regulatory signals. To bridge this gap, we first introduce a GRN generalization benchmark designed to evaluate regulatory predictions on unseen genes and datasets, which relies on the zero-shot capabilities of scFMs and is inherently challenging for traditional methods. Furthermore, to unlock the regulatory knowledge within the foundation models, we propose two novel methods, Virtual Value Perturbation and Gradient Trajectory, to distill implicit regulatory information from scFMs into highly generalizable inter-gene features. Extensive experiments demonstrate that our approach significantly outperforms existing methods, establishing a new paradigm for leveraging the potential of scFMs in universal GRN inference.", "date": "2026-05-01", "categories": []}
{"id": "2605.08815", "title": "MicroFuse: Protein-to-Genome Expert Fusion for Microbial Operon Reasoning", "abstract": "Predicting microbial operon co-membership requires integrating two complementary biological signals: protein-scale molecular identity and genome-context organization. While recent biological foundation models provide powerful representations of each view independently, naive concatenation of these modalities ignores a key biological property -- protein identity and genomic context may agree when adjacent genes form a coherent functional module, or conflict when sequence similarity is misleading but genomic layout indicates independent regulation. We present MicroFuse, a protein-to-genome expert fusion framework that integrates structure-aware protein representations from ProstT5 with genome-context representations from Bacformer through a four-expert Mixture-of-Experts module (protein, genome-context, agreement, and conflict experts) with a learned soft router. Training combines binary cross-entropy with symmetric cross-modal InfoNCE alignment and disagreement-weighted supervised contrastive shaping. We further construct OG-Operon100K, a 100,000-pair scaffold-level benchmark from the OMG metagenomic corpus with biologically grounded positive and negative criteria. On OG-Operon100K, MicroFuse achieves the strongest AUROC, AUPRC, mAP, and mAR among ProstT5-only, Bacformer-only, and Concat MLP baselines. Ablations identify cross-modal contrastive alignment as the dominant component, and a hard sequence-conflict subset reveals MicroFuse's largest gains precisely in biologically ambiguous cases where protein identity alone is misleading.", "date": "2026-05-09", "categories": []}
{"id": "2605.08954", "title": "MolWorld: Molecule World Models for Actionable Molecular Optimization", "abstract": "Molecular optimization in drug discovery aims to discover molecules with improved target properties, but practical lead optimization often requires more than high predicted scores. A useful candidate should also be actionable: it should be reachable from known molecules through valid local structural transformations, so that it can be interpreted as a plausible revision within an evolving chemical series. Existing de novo and single-molecule optimization methods do not explicitly model such reachability, especially when both the target molecules and the intermediate molecules connecting them to known compounds are unknown. In this work, we formulate actionable molecular optimization as sequential expansion of a molecule-transfer graph, where nodes are molecules and edges encode valid local transformations. We propose MolWorld, a molecule world model-guided framework that treats the current molecule-transfer graph as an evolving search state. At each iteration, MolWorld selects local anchor contexts, generates candidate molecules conditioned on these contexts, evaluates their properties, and uses a learned world model to update the evolving molecule world by retaining admissible candidates and inserting them into the molecule-transfer graph. The expanded molecule world then guides subsequent optimization. Experiments on property optimization and docking-based tasks show that MolWorld discovers high-property molecules while maintaining substantially stronger structural connectivity, supporting actionable and sequential molecular design.", "date": "2026-05-09", "categories": []}
{"id": "2605.09832", "title": "Modeling Atomic Conformational Ensembles of Proteins via Test-Time Supervision of Boltz-2 on Cryo-EM Density Maps", "abstract": "Knowledge of a protein's atomic conformational ensemble is critical to determining its function, yet state-of-the-art ensemble prediction models are limited by lack of high-quality conformational data from simulation or experiment. Recent advances in heterogeneous reconstruction for cryo-electron microscopy (cryo-EM) have enabled scientists to visualize ensembles of density maps for larger proteins and complexes not typically accessible through simulation, but building atomic models into these maps remains a challenge. Traditionally, ensemble prediction models are trained via a two-stage process: experimental density maps are converted into atomic structural ensembles through model building, after which these structures are used to train sequence-to-atomic ensemble predictors. In this work, we propose a new principle for fine-tuning pre-trained static structure prediction models such as Boltz-2 directly on raw cryo-EM maps, bypassing the two-stage process. We apply this technique to the problem of atomic model building by fine-tuning Boltz-2 to generate atomic conformations from an input ensemble of cryo-EM maps, achieving superior model building accuracy compared to prior work. Beyond overfitting to individual map ensembles, our method, CryoSampler, also shows preliminary evidence of in-domain generalization after fine-tuning, sampling diverse atomic conformations for an unseen sequences within the same protein family without requiring cryo-EM data. These capabilities indicate that CryoSampler holds the potential to train next-generation atomic ensemble prediction models directly on raw cryo-EM measurements.", "date": "2026-05-11", "categories": []}
{"id": "2605.09981", "title": "Yeti: A compact protein structure tokenizer for reconstruction and multi-modal generation", "abstract": "Multimodal models that jointly reason over protein sequences, structures, and function annotations within a unified representation hold immense potential for integrating multimodal data and generating new proteins with designed functional properties. To utilize transformer architectures, such models require a tokenizer that converts protein structure from continuous atomic coordinates into discrete representations suitable for scalable multimodal training. The quality of such models are fundamentally upper bounded by the fidelity and expressiveness of the underlying tokenized structure. However, existing tokenizers prioritize reconstruction over generative abilities. To address these gaps, we introduce Yeti, a simple and compact protein structure tokenizer based on lookup free quantization and trained end to end with a flow matching objective for multimodal learning. Compared to existing models, Yeti generally achieves the best codebook utilization and token diversity, and second best reconstruction accuracy (with 10x fewer parameters than ESM3) on diverse datasets. To validate Yeti's generative capability, we trained a compact multimodal model jointly over its structure tokens and amino acid sequence entirely from scratch, with no pretrained initialization. The resulting multimodal model generates plausible structures under unconditional cogeneration of protein sequence and structures, achieving comparable results to 10x larger models. Together, these results demonstrate that Yeti is a compact and expressive protein structure tokenizer suitable for training multimodal models that cogenerates highly plausible sequences and structures.", "date": "2026-05-11", "categories": []}
{"id": "2605.10035", "title": "From Single-Step Edit Response to Multi-Step Molecular Optimization", "abstract": "Conditional molecular optimization aims to edit a molecule to realize a specified property shift. In practice, structurally similar molecule data is scarce, while decisions are inherently action-level: at each step, the system must select one local structural edit from a candidate set that is strictly filtered by chemical feasibility rules. This level mismatch between supervision and decision makes oracle-in-the-loop search unstable in molecular optimization. Regressing on property differences between molecule pairs improves data efficiency but relies on oracle-in-the-loop search, entangling transformation effects with global context and providing limited guidance for selecting the next feasible edit, often resorting to oracle-in-the-loop search. For this reason, we propose a response-oriented discrete edit optimization approach comprising two tightly coupled components: a single-step molecular edit response predictor (SMER) and a multi-step planner that composes local predictions into optimization trajectories via guided tree search (SMER-Opt). The approach learns a directional evaluation model over edit actions to support constraint-aware planning. It mines weakly related molecule pairs and decomposes their structural differences into minimal edit units, turning endpoint property annotations into process-level supervision and yielding reusable, transferable action primitives. A directional edit evaluator then scores feasible candidate edits by their likelihood of moving the molecule toward the desired property change, substantially reducing dependence on external evaluator queries at decision time. Code is available at https://anonymous.4open.science/r/SMER.", "date": "2026-05-11", "categories": []}
{"id": "2605.10541", "title": "Bridging Sequence and Graph Structure for Epigenetic Age Prediction", "abstract": "Epigenetic clocks based on DNA methylation have emerged as powerful tools for estimating biological age, with broad applications in aging research, age-related disease studies, and longevity science. Despite advances across machine learning approaches to epigenetic age prediction, spanning penalised linear regression, deep feedforward networks, residual architectures, and graph neural networks, no existing method jointly models co-methylation graph structure and site-specific DNA sequence context within a unified framework. We propose a unified sequence--graph integration framework for epigenetic age prediction that addresses this gap, integrating eight-dimensional DNA sequence statistical features through a lightweight gated modulation mechanism that adaptively scales each site's methylation signal according to its sequence-determined biological relevance prior to graph convolution. Evaluated on 3,707 blood methylation samples against a comprehensive set of baselines, our method achieves a test MAE of 3.149 years, a 12.8\\% improvement over the strongest graph-based baseline. Biologically informed statistical features outperform CNN-based sequence encoding, demonstrating that handcrafted sequence features are more effective than end-to-end learned representations in this data regime. Post-hoc interpretability analysis identifies CpG density and local adenine frequency as features with age-dependent importance shifts, consistent with known mechanisms of age-related hypermethylation at CpG-dense promoter regions. Our code is at https://github.com/yaoli2022/graphage-seq.", "date": "2026-05-11", "categories": []}
{"id": "2605.10876", "title": "AssayBench: An Assay-Level Virtual Cell Benchmark for LLMs and Agents", "abstract": "Recent advances in machine learning and large-scale biological data collections have revived the prospect of building a virtual cell, a computational model of cellular behavior that could accelerate biological discovery. One of the most compelling promises of this vision is the ability to perform in silico phenotypic screens, in which a model predicts the effects of cellular perturbations in unseen biological contexts. This task combines heterogeneous textual inputs with diverse phenotypic outputs, making it particularly well-suited to LLMs and agentic systems. Yet, no standard benchmark currently exists for this task, as existing efforts focus on narrower molecular readouts that are only indirectly aligned with the phenotypic endpoints driving many real-world drug discovery workflows. In this work, we present AssayBench, a benchmark for phenotypic screen prediction, built from 1,920 publicly available CRISPR screens spanning five broad classes of cellular phenotypes. We formulate the screen prediction task as a gene rank prediction for each screen and introduce the adjusted nDCG, a continuous metric for comparing performance across heterogeneous assays. Our extensive evaluation shows that existing methods remain far from empirically estimated performance ceilings and zero-shot generalist LLMs outperform biology-specific LLMs and trainable baselines. Optimization techniques such as fine-tuning, ensembling, and prompt optimization can further improve LLM performance on this task. Overall, AssayBench offers a practical testbed for measuring progress toward in silico phenotypic screening and, more broadly, virtual cell models.", "date": "2026-05-11", "categories": []}
{"id": "2504.16559", "title": "Synergistic Benefits of Joint Molecule Generation and Property Prediction", "abstract": "Modeling the joint distribution of data samples and their properties allows to construct a single model for both data generation and property prediction, with synergistic benefits reaching beyond purely generative or predictive models. However, training joint models presents daunting architectural and optimization challenges. Here, we propose Hyformer, a transformer-based joint model that successfully blends the generative and predictive functionalities, using an alternating attention mechanism and a joint pre-training scheme. We show that Hyformer is simultaneously optimized for molecule generation and property prediction, while exhibiting synergistic benefits in conditional sampling, out-of-distribution property prediction and representation learning. Finally, we demonstrate the benefits of joint learning in a drug design use case of discovering novel antimicrobial~peptides.", "date": "2025-04-23", "categories": []}
{"id": "2510.24736", "title": "RNAGenScape: Property-Guided, Optimized Generation of mRNA Sequences with Manifold Langevin Dynamics", "abstract": "Generating property-optimized mRNA sequences is central to applications such as vaccine design and protein replacement therapy, but remains challenging due to limited data, complex sequence-function relationships, and the narrow space of biologically viable sequences. Generative methods that drift away from the data manifold can yield sequences that fail to fold, translate poorly, or are otherwise nonfunctional. We present RNAGenScape, a property-guided manifold Langevin dynamics framework for mRNA sequence generation that operates directly on a learned manifold of real data. By performing iterative local optimization constrained to this manifold, RNAGenScape preserves biological viability, accesses reliable guidance, and avoids excursions into nonfunctional regions of the ambient sequence space. The framework integrates three components: (1) an autoencoder jointly trained with a property predictor to learn a property-organized latent manifold, (2) a denoising autoencoder that projects updates back onto the manifold, and (3) a property-guided Langevin dynamics procedure that performs optimization along the manifold. Across three real-world mRNA datasets spanning two orders of magnitude in size, RNAGenScape increases median property gain by up to 148% and success rate by up to 30% while ensuring biological viability of generated sequences, and achieves competitive inference efficiency relative to existing generative approaches.", "date": "2025-10-14", "categories": []}
{"id": "2602.16548", "title": "RIDER: 3D RNA Inverse Design with Reinforcement Learning-Guided Diffusion", "abstract": "The inverse design of RNA three-dimensional (3D) structures is crucial for engineering functional RNAs in synthetic biology and therapeutics. While recent deep learning approaches have advanced this field, they are typically optimized and evaluated using native sequence recovery, which is a limited surrogate for structural fidelity, since different sequences can fold into similar 3D structures and high recovery does not necessarily indicate correct folding. To address this limitation, we propose RIDER, an RNA Inverse DEsign framework with Reinforcement learning that directly optimizes for 3D structural similarity. First, we develop and pre-train a GNN-based generative diffusion model conditioned on the target 3D structure, achieving a 9% improvement in native sequence recovery over state-of-the-art methods. Then, we fine-tune the model with an improved policy gradient algorithm using four task-specific reward functions based on 3D self-consistency metrics. Experimental results show that RIDER improves structural similarity by over 100% across all metrics and discovers designs that are distinct from native sequences.", "date": "2026-02-18", "categories": []}
{"id": "2605.06720", "title": "Conditional generation of antibody sequences with classifier-guided germline-absorbing discrete diffusion", "abstract": "Antibody therapeutics are among the most successful modern medicines, yet computationally designing antibodies with desirable binding and developability properties remains challenging. While protein language models (pLMs) have emerged as powerful tools for antibody sequence design, existing approaches largely suffer from two key limitations: they predominantly memorize germline sequences rather than modeling biologically meaningful somatic variation, and they offer limited support for flexible classifier-guided conditional generation. We address these challenges through two primary contributions. First, we demonstrate that discrete diffusion fine-tuning achieves strong language modeling performance on antibody sequences while allowing for generation conditioned on any off-the-shelf classifier. Second, we introduce germline absorbing diffusion, a novel modification of the discrete diffusion noise process in which the germline sequence - rather than a masked sequence - serves as the absorbing state. This biologically motivated inductive bias restricts the model to learning the trajectory from germline to observed sequence, effectively excluding genetic variation and V(D)J recombination statistics from the learned distribution and dramatically mitigating germline bias. We show that germline diffusion improves non-germline residue prediction accuracy from 26 percent to 46 percent, approaching the theoretical upper bound set by true biological variability. We then demonstrate the utility of our germline diffusion model on the conditional generation tasks of sampling antibodies with improved hydrophobicity and predicted binding affinity. On both tasks our model shows an improved tradeoff between class adherence and sample quality, significantly outperforming EvoProtGrad, a popular strategy to sample from pLMs with gradient-based discrete Markov Chain Monte Carlo.", "date": "2026-05-07", "categories": []}
{"id": "2605.06762", "title": "A Linear-Transformer Hybrid for SNP-Based Genotype-to-Phenotype Prediction in Grapevine", "abstract": "Robust genotype-to-phenotype (G2P) prediction is essential for accelerating breeding decisions and genetic gain. However, it remains challenging to measure complex traits under variable field conditions and across years. In this study, we propose a linear-Transformer approach, LiT-G2P (Linear-Transformer Genotype-to-Phenotype), an automated predictive framework that integrates additive genetic variance effects with Transformer-based nonlinear interactions using genome-wide single-nucleotide polymorphisms (SNPs) data. We evaluated LiT-G2P on a panel of diverse grape accessions, genotyped with SNP markers and measured for phenotypes across two consecutive years. Target phenotypic traits include leaf hair density and trichome density of grapevines. Across both single-year and cross-year testing scenarios, LiT-G2P consistently improves prediction performance compared with baseline models. For hair density, LiT-G2P achieves the lowest error in both single-year and cross-year evaluations, with RMSEs of 0.469 and 0.454, respectively, while maintaining strong tolerance accuracies of 79.2% and 74.6%, respectively. For trichome density, LiT-G2P also presents the best overall G2P performance. In addition, we extract model-prioritized SNPs from attention weights and apply genotype-stratified analysis to provide interpretable candidate marker for downstream validation. These results demonstrate that integrating stable additive effects with learned interaction patterns can enhance cross-year robustness and support practical SNP-based predictive modeling for genomic selection.", "date": "2026-05-07", "categories": []}
{"id": "2605.06879", "title": "Better Protein Function Prediction by Modeling Survivorship Bias", "abstract": "Protein sequence data from nature exhibits survivorship bias: we only observe data from those organisms that survive and reproduce, while non-functional protein mutations are eliminated by natural selection. Thus, predicting whether a protein sequence is functional often requires learning from positive examples alone. While positive-unlabeled (PU) learning frameworks offer a generic solution to this problem, existing PU methods ignore the evolutionary processes that shape sequence observability and cause survivorship bias. Consider a sequence that is one mutation away from a commonly-observed protein variant in a well-surveilled organism. If the sequence were functional, it would likely be observed. If it is not observed, this suggests non-functionality. In contrast, sequences that are unlikely to arise through mutation may be missing simply because they never arose. Thus, these two kinds of missing sequences should be treated differently when training models. In this work, we propose Evo-PU, a PU learning framework that uses a scientific understanding of nucleotide mutation to model survivorship bias for well-surveilled single-organism sequence data. On three prediction tasks using single-organism uniform-coverage surveillance data -- predicting results from held-out influenza and respiratory syncytial virus (RSV) mutagenesis studies, and predicting future SARS-CoV-2 variants -- Evo-PU outperforms standard PU learning, one-class classification (OCC), and protein language models (PLMs). On prediction tasks from multi-organism ProteinGym datasets with more heterogeneous surveillance coverage, we identify opportunities to generalize our approach.", "date": "2026-05-07", "categories": []}
{"id": "2605.07175", "title": "Learning Multi-Relational Graph Representations for DNA Methylation-Based Biological Age Estimation", "abstract": "Aging clocks aim to estimate biological age, a measure of physiological state distinct from chronological age, from observable biomarkers, and are widely used for health assessment and disease analysis. DNA methylation is a particularly informative biomarker due to its stability and strong association with aging, and recent learning-based approaches have improved predictive performance. However, most existing methods treat CpG sites as independent features, overlooking the complex and heterogeneous biological relationships among them. We propose RelAge-GNN, a multi-relational graph neural network framework for DNA methylation-based age prediction. Our method constructs three complementary graphs capturing co-methylation patterns, genomic co-localization, and gene-level associations among CpG sites. Each graph is modeled by an independent GNN branch, and a learnable gating mechanism adaptively fuses the resulting representations. Experiments on large-scale datasets show that RelAge-GNN achieves competitive accuracy and stronger correlation with chronological age compared to state-of-the-art methods. Moreover, the model exhibits improved sensitivity in detecting age acceleration across diverse disease cohorts, highlighting its potential utility for disease characterization. Finally, through post hoc interpretability analyses, we quantify the contributions of different relational structures and CpG sites, providing biologically meaningful insights and suggesting potential directions for aging-related research. Our code is available at: https://anonymous.4open.science/r/RelAge-GNN-F1E3/.", "date": "2026-05-08", "categories": []}
{"id": "2605.07335", "title": "CellScientist: Dual-Space Hierarchical Orchestration for Closed-Loop Refinement of Virtual Cell Models", "abstract": "Virtual Cell Modeling (VCM) requires models that not only predict perturbation responses, but also support targeted revision when predictions fail. Current LLM-assisted modeling workflows face a refinement-routing problem: prediction discrepancies are observed through executable implementations, but the relevant revision may involve the modeling assumption, representation design, implementation, or task constraint. Without structured feedback propagation across these levels, iterative refinement may repair code while failing to revise the assumption responsible for the discrepancy. We propose CellScientist, a dual-space hierarchical framework that couples a high-level hypothesis space with a low-level executable implementation space. CellScientist represents modeling decisions as structured states, realizes them as admissible programs under task and interface constraints, and routes execution discrepancies back to targeted hypothesis or implementation updates. This enables a closed Hypothesis -> Implementation -> Hypothesis loop where failures become structured signals for model refinement rather than debugging events. Across morphology and transcriptomic benchmarks, with additional single-cell perturbation evaluations, the final executable models selected by CellScientist improve over reference baselines under fixed split and evaluation protocols, while the workflow produces auditable refinement traces.", "date": "2026-05-08", "categories": []}
{"id": "2605.07554", "title": "ProteinJEPA: Latent prediction complements protein language models", "abstract": "Protein language models are trained primarily with masked language modeling (MLM), which predicts amino-acid identities at masked positions. We ask whether latent-space prediction can complement these token-level objectives under matched wall-clock budget. Across pretrained and random-init protein sequence encoders at 35--150M parameters, we find that the best protein-JEPA design is not all-position latent prediction but a variant: predicting latent targets only at masked positions, and retaining the MLM cross-entropy. We call this recipe masked-position MLM+JEPA. On a 16-task downstream suite (15 frozen linear probes plus SCOPe-40 zero-shot fold retrieval), under matched wall-clock budgets, this recipe wins more tasks than it loses against MLM-only continuation: 10 wins / 3 losses / 3 ties (hereafter W/L/T) on pretrained ESM2-35M, 11/2/3 on ESM2-150M while results in pretraining from scratch are mixed (6/8/2). Gains are seen for multiple models on 11 of 16 tasks, including stability, \\b{eta}\\beta \\b{eta}-lactamase fitness, variant effect, intrinsic disorder, remote homology, enzyme classification, and SCOPe-40 fold retrieval. Tasks with more losses than wins are Fluorescence (TAPE) and Peptide-HLA Binding. All-position MLM+JEPA matches MLM-only overall but does not reproduce the masked-position gains. JEPA-only (no MLM) collapses in nearly every experiment. We conclude that JEPA, when combined with MLM, is competitive and can outperform pure MLM in pretraining and continued training, even under matched wall-clock budgets.", "date": "2026-05-08", "categories": []}
{"id": "2605.07693", "title": "Toward Better Geometric Representations for Molecule Generative Models", "abstract": "Geometric representation-conditioned molecule generation provides an effective paradigm that decouples molecule representation modeling from structure generation. By decoupling molecule generation into two stages-first generating a meaningful molecule representation, and then generating a 3D molecule conditioned on this representation-the efficiency and quality of the generation process can be significantly enhanced. However, its effectiveness is fundamentally limited by the quality of the representation space: pretrained molecular encoders, such as UniMol, produce representations that are non-smooth and not fully exploited during the generative training process. In this work, we propose LENSEs, a framework that better exploits the potential of molecule representations in representation-conditioned generation methods. In particular, LENSEs introduces three complementary mechanisms: (1) a representation head, simultaneously trained during generative tasks, that extracts multi-level representations from the pretrained encoder; (2) a molecule perceptual loss that optimizes the generator in a semantic-informative representation space; and (3) a node-level representation alignment (REPA) loss that explicitly aligns the generator's hidden states with encoder representations, reducing the semantic gap between pretraining and generation. We demonstrate the effectiveness of these improvements through extensive molecule generation tasks. Specifically, on the challenging molecule generation dataset GEOM-DRUG, LENSEs achieves 97.28% validity and 98.51% molecule stability, surpassing existing advanced methods. Further analyses through Lipschitz constant reduction (4.6x) and QM9 probing tasks also demonstrate the smoother, more informative refined representations, establishing generative training with alignment objectives as a potential pretraining paradigm for molecular encoders.", "date": "2026-05-08", "categories": []}
{"id": "2605.07838", "title": "PPI-Net connects molecular protein interactions to functional processes in disease", "abstract": "Understanding how molecular alterations propagate across biological systems to drive disease remains a central challenge. Although high-throughput profiling enables comprehensive characterization of tumor states, most models neglect structured biological relationships or lack interpretability across scales. Here we present PPI-Net, a hierarchical graph neural network that integrates protein-protein interaction (PPI) networks with pathway-level representations to model disease from molecular interactions to functional processes. Patient-specific molecular profiles are embedded within a shared interaction network from STRING and propagated through a multi-layer Reactome hierarchy using graph attention, enabling aggregation of gene-level signals into higher-order biological programs. Across RNA-seq data from ten cancer types from The Cancer Genome Atlas, PPI-Net achieves robust predictive performance, with balanced accuracy exceeding 90% in multiple cohorts. Comparative analysis on RNA-Seq data from breast cancer demonstrated that PPI-Net's integration of the Reactome hierarchy improved balanced accuracy by 6.7% relative to a PPI-only model, while hierarchical multi-level supervision improved balanced accuracy by 12.3% relative to using only a single top-level prediction head. Applying a multi-omics approach using RNA-seq and methylation data improves model interpretation, recovering canonical oncogenic modules, including TP53-AKT signaling and stress response pathways, while revealing convergence onto coherent programs such as ion signaling and cellular responses to stimuli. These results demonstrate that integrating interaction networks with pathway hierarchies enables accurate prediction while providing mechanistic insight into cancer biology.", "date": "2026-05-08", "categories": []}
{"id": "2605.07938", "title": "Prototype Guided Post-pretraining for Single-Cell Representation Learning", "abstract": "Single-cell representation learning (SCRL) from gene expression data offers a way to uncover the complex regulatory logic underlying cellular function. Inspired by large language models in natural language modeling, several single-cell pretrained models have recently been proposed that treat genes as tokens and cells as sentences. However, these models are fundamentally limited by the long-tailed nature of cell-type distributions and struggle to generalize under covariate shifts in gene expression data. While fine-tuning is often used to mitigate these issues, we observe that performance remains bounded. To address this challenge, we introduce CellRefine, a post-pretraining method that operates between the pretraining and fine-tuning stages of a single-cell foundation model. CellRefine uses a multi-faceted objective that incorporates marker-gene sets as structural priors to guide post-pretraining and refine the latent embedding manifold of cells. Across multiple computational biology tasks, empirical results show that CellRefine consistently improves downstream performance, yielding gains up to 15%.", "date": "2026-05-08", "categories": []}
{"id": "2601.21092", "title": "MapPFN: Learning Causal Perturbation Maps in Context", "abstract": "Planning effective interventions in biological systems requires treatment-effect models that adapt to unseen biological contexts by identifying their specific underlying mechanisms. Yet single-cell perturbation datasets span only a handful of biological contexts, and existing methods cannot leverage new interventional evidence at inference time to adapt beyond their training data. To meta-learn a perturbation effect estimator, we present MapPFN, a prior-data fitted network (PFN) pre-trained on a synthetic biological prior with causal interventions, decoupling pre-training from limited wet-lab data. Unlike existing methods, MapPFN uses in-context learning to map a sequence of experiments to a post-perturbation distribution, enabling a single pre-trained model to adapt to new datasets and arbitrary gene sets at inference time. Zero-shot, MapPFN identifies differentially expressed genes on par with models trained on real single-cell data, and fine-tuning further improves predictions across biological contexts. Our code, model and data are available at https://marvinsxtr.github.io/MapPFN.", "date": "2026-01-28", "categories": []}
{"id": "2601.22382", "title": "Purely Agent-Driven Black-Box Optimization for Biological Design", "abstract": "Many key challenges in biological design -- such as small-molecule drug discovery, antimicrobial peptide development, and protein engineering -- can be framed as black-box optimization over vast, complex structured spaces. Existing methods rely mainly on raw structural data and struggle to exploit the rich scientific literature. While large language models (LLMs) have been added to these pipelines, they have been confined to narrow roles within structure-centered optimizers. We instead cast biological black-box optimization as an agent-driven, language-based reasoning process. We introduce Purely Agent-driven BLack-box Optimization (PABLO), a hierarchical agentic system that uses scientific LLMs pretrained on chemistry and biology literature to generate and iteratively refine biological candidates. On both the standard GuacaMol molecular design and antimicrobial peptide optimization tasks, PABLO achieves state-of-the-art performance, substantially improving sample efficiency and final objective values over established baselines. Compared to prior optimization methods that incorporate LLMs, PABLO achieves competitive token usage per run despite relying on LLMs throughout the optimization loop. Beyond raw performance, the agentic formulation offers key advantages for realistic design: it naturally incorporates semantic task descriptions, retrieval-augmented domain knowledge, and complex constraints. In follow-up in vitro validation, PABLO-optimized peptides showed strong activity against drug-resistant pathogens, underscoring the practical potential of PABLO for therapeutic discovery.", "date": "2026-01-29", "categories": []}
{"id": "2602.01839", "title": "DOGMA: Weaving Structural Information into Data-centric Single-cell Transcriptomics Analysis", "abstract": "Recently, data-centric AI methodology has been a dominant paradigm in single-cell transcriptomics analysis, which treats data representation rather than model complexity as the fundamental bottleneck. In the review of current studies, earlier sequence methods treat cells as independent entities and adapt prevalent ML models to analyze their directly inherited sequence data. Despite their simplicity and intuition, these methods overlook the latent intercellular relationships driven by the functional mechanisms of biological systems and the inherent quality issues of the raw sequencing data. Therefore, a series of structured methods has emerged. Although they employ various heuristic rules to capture intricate intercellular relationships and enhance the raw sequencing data, these methods often neglect biological prior knowledge. This omission incurs substantial overhead and yields suboptimal graph representations, hindering the utility of ML models.\n  To address these issues, we propose DOGMA, a data-centric framework designed for the structural reshaping and semantic enhancement of raw data through multi-level biological prior knowledge. Transcending reliance on purely data-driven heuristics, DOGMA provides a prior-guided graph construction pipeline that integrates statistical alignment with Cell Ontology and phylogenetic structure for biologically grounded cell-graph construction and robust cross-species alignment. Furthermore, Gene Ontology is utilized to bridge the feature-level semantic gap by incorporating functional priors. In complex multi-species and multi-organ benchmarks, DOGMA exhibits strong robustness in strict zero-shot cell-type evaluation and sample efficiency while using substantially lower GPU memory and inference time in downstream evaluation.", "date": "2026-02-02", "categories": []}
{"id": "2605.05370", "title": "SPADE: Faster Drug Discovery by Learning from Sparse Data", "abstract": "Drug discovery seeks molecules (ligands) that bind strongly and selectively to a target protein. However, fewer than 5% of candidate ligands pass the bar for even the early stages of drug discovery. Furthermore, we want methods that work for novel proteins for which we have no prior data. Starting from scratch, we have to iteratively select and test candidate ligands such that we find enough ligands of the desired quality in as few tests as possible. Our proposed algorithm, named SPADE, introduces a novel approach to ligand selection that requires only 40 tests on average to find 10 high-quality ligands. In one-vs-one comparisons, SPADE outperforms deep learning and Bayesian optimization methods on more proteins, achieving median improvements of 7%-32% in sample efficiency. SPADE is also 10x faster than its closest competitor at scoring candidate drugs. Dataset and code is available at https://anonymous.4open.science/r/SPADE_Fast_Drug_Discovery_by_Learning_from_Sparse_Data-F028/README.md", "date": "2026-05-06", "categories": []}
{"id": "2605.05913", "title": "Wisteria: A Unified Multi-Scale Feature Learning Framework for DNA Language Model", "abstract": "DNA language model aims to decipher the regulatory grammar and semantic of genomes by capturing long range dependencies in DNA sequences. Existing methods emphasize long range token interactions but often ignore the interplay between local motifs and global dependencies. In this paper, we propose Wisteria, a genomic language model that integrates multi scale feature learning within a unified framework for DNA sequence. Specifically, Wisteria augments the Mamba based architecture with gated dilated convolutions to capture local motifs and regulatory patterns, while gated multilayer perceptrons refine global dependencies. We further introduce a Fourier based attention mechanism to support frequency domain modeling, periodic extension and length generalization. Across four experimental settings with both short and long range dependencies, Wisteria demonstrates strong performance on downstream benchmarks against competitive DNA language model baselines. These results indicate that Wisteria effectively unifies local and global dependency modeling for multi scale genomic sequence analysis.", "date": "2026-05-07", "categories": []}
{"id": "2605.06552", "title": "Sequential Design of Genetic Circuits Under Uncertainty With Reinforcement Learning", "abstract": "The design of biological systems is hindered by uncertainty arising from both intrinsic stochasticity of biomolecular reactions and variability across laboratory or experimental conditions. In this work, we present a sequential framework to optimize genetic circuits under both forms of uncertainty. By employing simulator models based on differential equations or Markov jump processes alongside a reinforcement learning (RL) policy-based approach, our method suggests experiments that adapt to unknown laboratory conditions while accounting for inherent stochasticity. While previous Bayesian methods address uncertainty through iterative experiment-inference-optimization cycles, they typically require computationally expensive inference and optimization steps after each experimental round, leading to delays. To overcome this bottleneck, we propose an amortized approach trained up-front across a distribution of possible uncertain parameters. This strategy sidesteps the need for explicit parameter inference during the design cycle, enabling immediate, observation-based adaptation. We demonstrate our framework on models for heterologous gene expression and a repressilator circuit, showing that it efficiently handles both molecular noise and cross-laboratory variability.", "date": "2026-05-07", "categories": []}
{"id": "2409.09183", "title": "Quantum-inspired Reinforcement Learning for Synthesizable Drug Design", "abstract": "Synthesizable molecular design (also known as synthesizable molecular optimization) is a fundamental problem in drug discovery, and involves designing novel molecular structures to improve their properties according to drug-relevant oracle functions (i.e., objective) while ensuring synthetic feasibility. However, existing methods are mostly based on random search. To address this issue, in this paper, we introduce a novel approach using the reinforcement learning method with quantum-inspired simulated annealing policy neural network to navigate the vast discrete space of chemical structures intelligently. Specifically, we employ a deterministic REINFORCE algorithm using policy neural networks to output transitional probability to guide state transitions and local search using genetic algorithm to refine solutions to a local optimum within each iteration. Our methods are evaluated with the Practical Molecular Optimization (PMO) benchmark framework with a 10K query budget. We further showcase the competitive performance of our method by comparing it against the state-of-the-art genetic algorithms-based method.", "date": "2024-09-13", "categories": []}
{"id": "2603.26114", "title": "DPD-Cancer: Explainable Graph-Based Deep Learning for Small Molecule Anti-Cancer Activity Prediction", "abstract": "DPD-Cancer is a graph-attention deep learning framework for predicting small-molecule DPD-Cancer is a graph-attention deep learning framework for predicting small-molecule anti-cancer activity across the NCI-60 panel, trained and evaluated under a strict chemistry-aware data-partitioning scheme. On the hold-out test set, the classifier achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.87 (95% CI [0.86, 0.88]) and Area Under the Precision-Recall Curve (AUPRC) of 0.73 (95% CI [0.70, 0.76]); per-cell-line regression models for 73 cell lines produced a median Pearson's Correlation Coefficient (Pearson's R) of 0.64 and median Root Mean Squared Error (RMSE) of 0.67 for pGI50-value prediction. Benchmarks against pdCSM-Cancer, MLASM, and ACLPred under matched data conditions yielded consistently higher Matthew's Correlation Coefficient (MCC) scores, an occlusion-based attribution analysis confirmed that model explanations were quantitatively faithful to classifier decisions, and an applicability-domain analysis characterised reliability as a function of chemical distance. To facilitate widespread adoption, DPD-Cancer is available as a free, user-friendly web server for unrestricted use at https://biosig.lab.uq.edu.au/dpd_cancer/.", "date": "2026-03-27", "categories": []}
{"id": "2604.16851", "title": "Applications of deep generative models to DNA reaction kinetics and to cryogenic electron microscopy", "abstract": "This dissertation explores how deep generative models can advance the analysis of challenging biological problems by integrating domain knowledge with deep learning. It focuses on two areas: DNA reaction kinetics and cryogenic electron microscopy (cryo-EM). In the first part, we present ViDa, a biophysics-informed framework leveraging variational autoencoders (VAEs) and geometric scattering transforms to generate biophysically-plausible embeddings of DNA reaction kinetics simulations. These embeddings are reduced to a two-dimensional space to visualize DNA hybridization and toehold-mediated strand displacement reactions. ViDa preserves structure and clusters trajectory ensembles into reaction pathways, making simulation results more interpretable and revealing new mechanistic insights. In the second part, we address key challenges in cryo-EM density map interpretation and protein structure modeling. We provide a comprehensive review and benchmarking of deep learning methods for atomic model building, with improved evaluation metrics and practical guidance. We then present Struc2mapGAN, a generative adversarial network that synthesizes high-fidelity experimental-like cryo-EM density maps from protein structures. Finally, we present CryoSAMU, a structure-aware multimodal U-Net that enhances intermediate-resolution cryo-EM maps by integrating density features with structural embeddings from protein language models via cross-attention. Overall, these contributions demonstrate the potential of deep generative models to interpret DNA reaction mechanisms and advance cryo-EM density map analysis and protein structure modeling.", "date": "2026-04-18", "categories": []}
{"id": "2605.04119", "title": "Tree-Conditioned Edit Flows for Ancestral Sequence Reconstruction", "abstract": "Ancestral sequence reconstruction (ASR) aims to infer extinct protein sequences at internal nodes of a phylogenetic tree. Classical ASR methods are typically based on continuous-time Markov substitution models, but they treat sites largely independently and handle insertions and deletions only weakly or not at all. We introduce a tree-conditioned edit-flow model for variable-length ASR. Given two descendant sequences and their branch distances to a shared ancestor, the model reconstructs the ancestor through paired bidirectional edit trajectories constrained to agree on a common ancestral state. On a benchmark of experimentally evolved sequences with only context-independent substitutions, the model does not match the accuracy of the best classical method, yet still achieves reasonable performance despite being trained on natural sequences that include insertions, deletions, and substitutions. On a benchmark of natural homologous sequences with abundant insertions and deletions, the model most accurately localizes inferred evolutionary change.", "date": "2026-05-05", "categories": []}
{"id": "2605.04376", "title": "GraphPI: Efficient Protein Inference with Graph Neural Networks", "abstract": "The integration of deep learning approaches in biomedical research has been transformative, enabling breakthroughs in various applications. Despite these strides, its application in protein inference is impeded by the scarcity of extensively labeled datasets, a challenge compounded by the high costs and complexities of accurate protein annotation. In this study, we introduce GraphPI, a novel framework that treats protein inference as a node classification problem. We treat proteins as interconnected nodes within a protein-peptide-PSM graph, utilizing a Graph Neural Network-based architecture to elucidate their interrelations. To address label scarcity, we train the model on a set of unlabeled public protein datasets with pseudo-labels derived from an existing protein inference algorithm, enhanced by self-training to iteratively refine labels based on confidence scores. Contrary to prevalent methodologies necessitating dataset-specific training, our research illustrates that GraphPI, due to the well normalized nature of Percolator features, exhibits universal applicability without dataset-specific fine-tuning, a feature that not only mitigates the risk of overfitting but also enhances computational efficiency. Our empirical experiments reveal notable performance on various test datasets and deliver significantly reduced computation times compared to common protein inference algorithms.", "date": "2026-05-06", "categories": []}
{"id": "2605.04930", "title": "When Does Gene Regulatory Network Inference Break? A Controlled Diagnostic Study of Causal and Correlational Methods on Single-Cell Data", "abstract": "Despite theoretical advantages, causal methods for Gene Regulatory Network (GRN) inference from single-cell RNA-seq data consistently fail to match or outperform correlation-based baselines in many realistic benchmarks, a persistent puzzle which casts doubt on the value of causality for this task. We argue that existing benchmarks are insufficiently controlled to answer this question because they evaluate on real or semi-real data where multiple pathologies co-occur, confounding failure modes, and obscuring the specific conditions under which different inference methods excel or fail. To address this gap, we introduce a controlled diagnostic framework that isolates seven biologically motivated pathologies (dropout, latent confounders, cell-type mixing, feedback loops, network density, sample size, and pseudotime drift) and measure how six representative methods spanning three inference paradigms degrade as each pathology intensifies. Across 6,120 controlled experiments, we find that causal methods genuinely dominate in clean and structurally favorable regimes, but specific pathologies (notably dropout and latent confounders) selectively neutralize their advantages. We further introduce an error-type decomposition that reveals methods with similar aggregate accuracy commit qualitatively different errors. To probe whether single-pathology effects persist when multiple stressors co-occur, we perform an interaction sweep over the three most impactful pathologies and find that their joint effects are sub-additive, while also exposing density-conditional cross-overs invisible to single-dial analysis. Our findings offer a nuanced understanding of when and why different methods succeed or fail for GRN inference, providing actionable insights for method development and practical guidance for practitioners.", "date": "2026-05-06", "categories": []}
{"id": "2605.02937", "title": "Proteo-R1: Reasoning Foundation Models for De Novo Protein Design", "abstract": "Deep learning in \\emph{de novo} protein design has achieved atomic-level fidelity. However, existing models remain largely non-deliberative: they directly synthesize molecular geometries without explicitly reasoning about which residues or interactions are functionally essential. As a result, design decisions are entangled with continuous sampling dynamics, limiting interpretability, controllability, and systematic reuse of biochemical knowledge. We introduce \\textbf{Proteo-R1}, a reasoning-guided protein design framework that explicitly decouples \\emph{molecular understanding} from \\emph{geometric generation}. Proteo-R1 adopts a dual-expert architecture in which a multimodal large language model (MLLM) serves as an \\emph{understanding expert}, analyzing protein sequences, structures, and textual context to identify key functional residues that govern binding and specificity. These residue-level decisions are then passed as hard constraints to a separate diffusion-based \\emph{generation expert}, which performs conditional co-design while respecting the fixed interaction anchors. This factorization mirrors how human experts approach molecular engineering: first, reasoning about critical interactions, then optimizing geometry subject to those constraints. By operationalizing reasoning as explicit residue-level commitments rather than latent textual guidance, Proteo-R1 achieves stable, interpretable, and modular integration of LLM reasoning with state-of-the-art geometric generative models. Code, data, and demos are available at https://smiles724.github.io/r1/.", "date": "2026-05-01", "categories": []}
{"id": "2605.02949", "title": "Disease Is a Spectral Perturbation", "abstract": "We propose a novel method of understanding disease transformation from a healthy baseline with biomarker-level explainability. By modeling the biomarker covariance matrices of healthy controls and disease states, the perturbation can be individually characterized to accomplish mechanistic explanations of disease trajectories, both at a molecular level and for individual patients. Given a cohort of n patients each measured on p biomarkers, we define the biomarker \"Hamiltonian\" H = X^T X / n \\in R^{p \\times p}, where X \\in R^{n \\times p} is the covariant biomarker matrix. The eigenvectors of H define a set of normal modes of biomarker coordination, and the eigenvalues quantify the energy carried by each mode. In the healthy state, the reference Hamiltonian H_0 governs this structure where disease perturbs H_0 by an additive operator \\Delta H, thus shifting eigenvalues and rotating eigenvectors in proportion to the severity of pathological disruption. We formalize this framework, derive the spectral change given a disease perturbation, and demonstrate that the projection of a newly diagnosed patient's cumulative biomarker covariance structure onto disease-discriminant eigenmodes constitutes an optimal prognostic statistic for greater precision in disease prognosis. This work serves as a veritable white paper with application across a panoply of disease frameworks from cancer to neurodegenerative disorders.", "date": "2026-05-01", "categories": []}
{"id": "2605.02954", "title": "EFGPP: Exploratory framework for genotype-phenotype prediction", "abstract": "Predicting complex human traits from genetic data is challenging because different genetic, clinical, and molecular data sources often contain different parts of the signal. Here, we present EFGPP, a reproducible framework for generating, ranking, and combining multiple types of data for genotype-to-phenotype prediction. We applied EFGPP to migraine prediction using UK Biobank data from 733 individuals. The framework combined genotype-derived features, principal components, clinical and metabolomic covariates, and polygenic risk scores generated from migraine and depression GWAS using PLINK, PRSice-2, AnnoPred, and LDAK-GWAS. The best single data type achieved a test AUC of 0.644, while combining multiple data types improved performance to 0.688 using migraine-focused inputs and 0.663 using cross-trait depression-derived inputs. Genetic features alone did not outperform the covariates-only baseline, but genotype-derived features performed better than PRS alone, and depression-derived PRS showed useful predictive signal. Overall, EFGPP provides a practical proof-of-concept framework for prioritising and integrating heterogeneous genetic data sources for complex phenotype prediction.", "date": "2026-05-02", "categories": []}
{"id": "2605.03281", "title": "Donor-Aware scRNA-seq Benchmarks for IBD Classification", "abstract": "Donor-level disease classification from single-cell RNA sequencing (scRNA-seq) requires strict donor-aware cross-validation: naive pipelines that split cells randomly conflate training and test donors, inflating reported performance through pseudoreplication. We present a donor-aware benchmark evaluating three feature representations across two independent IBD cohorts: centered log-ratio (CLR) transformed cell-type composition, GatedStructuralCFN dependency embeddings, and scVI variational autoencoder latent embeddings. The cohorts are the SCP259 ulcerative colitis atlas (UC vs. Healthy, n=30 donors, 51 cell types) and the Kong 2023 Crohn's disease atlas (CD vs. Healthy, n=71 donors, 55-68 cell types across three intestinal regions).\n  Compartment-stratified CLR composition achieves AUROC 0.956 +/- 0.061 on SCP259; GatedStructuralCFN on the same features achieves 0.978 +/- 0.050. In the Kong cohort, CFN achieves its best performance in the colon region (0.960 +/- 0.055 after feature filtering), exceeding linear CLR (0.900 +/- 0.100), while terminal ileum classification is dominated by linear models (CatBoost CLR 0.967 +/- 0.075 vs. CFN 0.811 +/- 0.164). Cross-dataset transfer (CD->UC, four shared cell types) achieves AUC 0.833 with XGBoost CLR; the reverse direction performs at chance. CFN edge stability analysis shows that compartment-wise composition eliminates spurious unit-sum-induced instability present in global composition (Jaccard 0.026 vs. top-20 recurrence 1.0). CFN shows a consistent numerical advantage over linear models in the colon region of CD (AUROC 0.960 vs. 0.900), though no inter-method comparison reached statistical significance at n<=34 donors per region. Compartment-aware feature construction is critical for both classification performance and structural interpretability. Code: https://github.com/Jonathan-321/sfn-scrna-study", "date": "2026-05-05", "categories": []}
{"id": "2605.03360", "title": "A-CODE: Fully Atomic Protein Co-Design with Unified Multimodal Diffusion", "abstract": "We present A-CODE, a fully atomic unified one-stage protein co-design model that simultaneously refines discrete atom types and continuous atom coordinates. Unlike predominant two-stage methods that cascade structure design with amino acid-level sequence design, our approach is fully atomic within a unified multimodal diffusion framework, in which residue identities are inferred solely from atom-level predictions. Built upon the powerful all-atom architecture, A-CODE achieves superior designability for unconditional protein generation, outperforming all existing one-stage and two-stage design models. For binder design, A-CODE rivals and even outperforms existing state-of-the-art two-stage design models and, compared with the existing one-stage co-design model, achieves a drastic tenfold improvement in success rate on hard tasks. The inherent flexibility of our atomic formulation enables, for the first time, seamless adaptation to non-canonical amino acid (ncAA) modeling. Our fully atomic framework establishes a new, versatile foundation for all-atom generative modeling that can be naturally extended to complex biomolecular systems.", "date": "2026-05-05", "categories": []}
{"id": "2605.03578", "title": "Expanding functional protein sequence space using high entropy generative models", "abstract": "Boltzmann Machines trained on evolutionary sequence data have emerged as a powerful paradigm for the data-driven design of artificial proteins. However, the relationship between model architecture, specifically parameter density, and experimental performance remains poorly understood. Here, we investigate this relationship using the Chorismate Mutase enzyme family as a model system. We compare standard fully connected Boltzmann Machines for Direct Coupling Analysis (bmDCA) with sparse models generated via progressive edge activation (eaDCA) and edge decimation (edDCA). We identify a maximum-entropy model (meDCA) along the decimation trajectory that represents an optimal balance between constraint satisfaction and the flexibility of the probability distribution. We synthesized and tested artificial sequences from all models using an in vivo complementation assay, finding that all architectures, regardless of sparsity, generate functional enzymes with high success rates, even at significant divergence from natural sequences. Despite this functional equivalence, we demonstrate that the meDCA model samples a viable sequence space that is more than fifteen orders of magnitude larger than its low-entropy counterparts. Furthermore, comparative analyses reveal that high-entropy models systematically minimize overfitting and better capture the local neutral spaces surrounding natural proteins. These findings suggest that while various models satisfying coevolutionary statistics can generate functional sequences, high-entropy Boltzmann Machines provide a superior representation of the underlying evolutionary fitness landscape.", "date": "2026-05-05", "categories": []}
{"id": "2604.14241", "title": "Polyformer: a generative framework for thermodynamic modeling of polymeric molecules", "abstract": "The classic paradigm of structural biology is that the sequence of a biomolecule (protein, nucleic acid, lipid, etc) determines its conformation (shape) which determines its biological function. Protein folding programs like AlphaFold address this paradigm by predicting the single best conformation given a sequence that defines the molecule. However, biomolecules are not static structures, and their conformational ensemble determines their function. We present the Polyformer -- a generative framework for thermodynamic modeling of polymeric molecules. Given the sequence and temperature (or another thermodynamic variable), the Polyformer generates conformations faithful to the molecule's thermodynamic conformational ensemble. It is the first generative model that solves three problems simultaneously: how does a molecule fold, what is its conformational ensemble, and how does the conformational ensemble change as we change physical temperature. As a concrete test case, we apply Polyformer to protein domains with 50-111 residues and report good agreement of model predictions to Molecular Dynamics (MD) trajectories.", "date": "2026-04-15", "categories": []}
{"id": "2605.00925", "title": "Linking spatial biology and clinical histology via Haiku", "abstract": "Integrating molecular, morphological, and clinical data is essential for basic and translational biomedical research, yet systematic frameworks for jointly modeling these modalities remain limited. Here we present Haiku, a tri-modal contrastive learning model trained on multiplexed immunofluorescence (mIF). It comprises 26.7 million spatial proteomics patches from 3,218 tissue sections across 1,606 patients spanning 11 organ types, with matched hematoxylin and eosin (H&E) histology and clinical metadata aligned in a shared embedding space. Haiku enables three-way cross-modal retrieval, improves downstream classification and clinical prediction tasks over unimodal baselines, and supports zero-shot biomarker inference through fusion retrieval conditioned on clinical metadata-only text descriptions. Across tasks, Haiku outperforms competing approaches, achieving cross-modal retrieval (Recall@50 up to 0.611 versus near-zero baseline), survival prediction (C-index 0.737, +7.91% relative improvement), and zero-shot biomarker inference (mean Pearson correlation 0.718 across 52 biomarkers). Furthermore, we introduce a counterfactual prediction framework in which modifying only clinical metadata while fixing tissue morphology surfaces niche-specific molecular shifts associated with breast cancer stage progression and lung cancer survival outcomes. In a lung adenocarcinoma case study, the counterfactual analysis recovers niche-specific shifts characterized by increased CD8 and granzyme B, reduced PD-L1, and decreased Ki67, broadly consistent with patterns reported for favorable outcomes. We present these counterfactual results as exploratory, hypothesis-generating signals rather than mechanistic claims. These capabilities demonstrate that tri-modal alignment via Haiku enables integrative analysis of spatial biology, bridging molecular measurements with clinical context for biological exploration.", "date": "2026-04-30", "categories": []}
{"id": "2605.00930", "title": "CellxPert: Inference-Time MCMC Steering of a Multi-Omics Single-Cell Foundation Model for In-Silico Perturbation", "abstract": "In this work, we introduce CellxPert, a scalable multimodal foundation model that unifies single-cell and spatial multi-omics within a common representation space. CellxPert jointly encodes transcriptomic (scRNA-seq), chromatin-accessibility (ATAC-seq), and surface-proteomic (CITE-seq) measurements, while directly incorporating MERFISH and imaging mass-cytometry data as 2D or 3D spatial-visual layers. CellxPert facilitates four key downstream tasks out of the box: (i) cell-type annotation across a broad ontology of 154 largely overlapping identities -- the largest label space addressed to date and a stringent test of fine-grained discrimination, (ii) efficient fine-tuning using Low Rank Adaptation (LoRA), (iii) genome-wide transcriptomic response prediction to in-silico perturbations (ISP), and (iv) seamless multi-omic integration across various assays and platforms. Unlike current single-cell foundation models, which approximate gene perturbations by deleting or reordering tokenized gene expression ranks, CellxPert employs a Metropolis-Hastings sampler whose proposal kernel uses the model's masked conditional distributions to transition to new transcriptomic states conditioned on the perturbed genes. This Markov-chain procedure mitigates out-of-distribution artifacts introduced by abrupt token manipulation and produces trajectories that are biologically interpretable. Evaluations on PBMC68K, Replogle Perturb-seq, Systema, and BMMC benchmarks show that CellxPert surpasses classical and state-of-the-art baselines in cell-type annotation, perturbation response prediction, and multi-omic integration.", "date": "2026-04-30", "categories": []}
{"id": "2605.00948", "title": "Co-Generative De Novo Functional Protein Design", "abstract": "De novo functional protein design aims to generate protein sequences that realize specified biochemical functions without relying on evolutionary templates, enabling broad applications in biotechnology and medicine. Existing approaches adopt either direct function-to-sequence mapping or decoupled structure-sequence generation strategies but often fail to achieve functionality and foldability simultaneously. To address this, we propose CodeFP, a Co-generative protein language model for de novo Functional Protein design that simultaneously decodes sequence and structure tokens, thereby enabling superior simultaneous realization of functionality and foldability. CodeFP utilizes functional local structures to enrich functional semantic encodings, overcoming the suboptimal translation of flat encodings into structure tokens, while introducing auxiliary functional supervision to alleviate training ambiguity stemming from the one-to-many structure-to-token mapping. Extensive experiments show that CodeFP consistently achieves average improvements of 6.1% in functional consistency and 3.2% in foldability over the strongest baseline.", "date": "2026-05-01", "categories": []}
{"id": "2605.01513", "title": "Protein-Conditioned Multi-Objective Reinforcement Learning for Full-Length mRNA Design", "abstract": "Designing therapeutic messenger RNA (mRNA) requires creating full-length transcripts that carefully balance stability, translation efficiency, and immune safety. To address this challenge, we propose ProMORNA, a multi-objective generation framework that produces complete mRNA transcripts \\textit{de novo} directly from a target protein sequence. Our approach begins by training a BART-style encoder-decoder model on over 6 million natural protein-mRNA pairs. We then introduce Multi-Objective Group Relative Policy Optimization (MO-GRPO) to simultaneously optimize for various biological objectives in a unified way. As a case study, we evaluated ProMORNA on the widely used firefly luciferase target, excluding it from both our supervised training data and the prompt pool. The results indicate that ProMORNA improves the \\textit{in silico} Pareto frontier for predicted half-life and translation efficiency relative to standard supervised baselines. Additionally, it achieves higher predicted functional scores than a state-of-the-art baseline under the same evaluation pipeline. These computational findings demonstrate the feasibility of using multi-objective reinforcement learning for full-length mRNA design on unseen targets.", "date": "2026-05-02", "categories": []}
{"id": "2605.01945", "title": "PepSpecBench: A Unified Evaluation Benchmark for Peptide Tandem Mass Spectrometry Prediction", "abstract": "Tandem mass spectrometry provides a high-throughput framework for identifying and quantifying proteins in complex biological samples. In computational proteomics, predicting peptide MS/MS spectra is a critical task, enabling downstream applications such as large-scale peptide identification and quantification. While deep learning architectures have substantially improved prediction accuracy, three evaluation challenges obscure the true progress of the field. First, inconsistent data preprocessing and incompatible model output spaces hinder fair model comparison. Second, flawed data splitting strategies can permit hidden sequence leakage and inflate reported performance. Third, existing evaluations typically lack comprehensive cross-species benchmarking and systematic assessment of model robustness to influential experimental conditions. To address these challenges, we propose PepSpecBench, a unified benchmark for peptide MS/MS spectrum prediction. PepSpecBench standardizes data preprocessing across complementary public datasets, enforces a strict backbone-disjoint splitting strategy to eliminate sequence leakage, and evaluates diverse architectures within a shared fragment-ion representation space. It further introduces a comprehensive multi-species evaluation suite and physically grounded metadata perturbation probes to assess model robustness and instrument awareness. We uncover previously unrecognized performance discrepancies and robustness limitations across six representative models, providing actionable insights for future model design, evaluation and practical deployment.", "date": "2026-05-03", "categories": []}
{"id": "2512.01116", "title": "Structural Prognostic Event Modeling for Multimodal Cancer Survival Analysis", "abstract": "The integration of histology images and gene profiles has shown great promise for improving survival prediction in cancer. However, current approaches often struggle to model intra- and inter-modal interactions efficiently and effectively due to the high dimensionality and complexity of the inputs. A major challenge is capturing critical prognostic events that, though few, underlie the complexity of the observed inputs and largely determine patient outcomes. These events, manifested as high-level structural signals such as spatial histologic patterns or pathway co-activations, are typically sparse, patient-specific, and unannotated, making them inherently difficult to uncover. To address this, we propose SlotSPE, a slot-based framework for structural prognostic event modeling. Specifically, inspired by the principle of factorial coding, we compress each patient's multimodal inputs into compact, modality-specific sets of mutually distinctive slots using slot attention. By leveraging these slot representations as encodings for prognostic events, our framework enables both efficient and effective modeling of complex intra- and inter-modal interactions, while also facilitating seamless incorporation of biological priors that enhance prognostic relevance. Extensive experiments on ten cancer benchmarks show that SlotSPE outperforms existing methods in 8 out of 10 cohorts, achieving an overall improvement of 2.9%. It remains robust under missing genomic data and delivers markedly improved interpretability through structured event decomposition.", "date": "2025-11-30", "categories": []}
{"id": "2411.18796", "title": "Graph-Based Biomarker Discovery and Interpretation for Alzheimer's Disease", "abstract": "Early diagnosis and discovery of therapeutic drug targets are crucial objectives for effective management of Alzheimer's Disease (AD). Current approaches for AD diagnosis and treatment planning are based on radiological imaging and largely inaccessible for population-level screening due to prohibitive costs and limited availability. Recently, blood tests have shown promise in diagnosing AD and highlighting possible biomarkers that can be used as drug targets for AD management. Blood tests are significantly more accessible to disadvantaged populations, cost-effective, and minimally invasive. However, biomarker discovery in the context of AD diagnosis is complex as there exist important associations between various biomarkers. Here, we introduce BRAIN (Biomarker Representation, Analysis, and Interpretation Network), a novel machine learning (ML) framework to jointly optimize diagnostic accuracy and biomarker discovery processes to identify all relevant biomarkers that contribute to AD diagnosis. Using a holistic graph-based representation for biomarkers, we highlight their interdependencies and explain why different ML models identify different discriminative biomarkers. We apply BRAIN to a publicly available blood biomarker dataset, revealing three novel biomarker subnetworks whose interactions vary between the control and AD groups, offering a new paradigm for drug discovery and biomarker analysis for AD.", "date": "2024-11-27", "categories": []}
{"id": "2602.07408", "title": "Progressive Multi-Agent Reasoning for Biological Perturbation Prediction", "abstract": "Predicting gene regulation responses to biological perturbations requires reasoning about underlying biological causalities. While large language models (LLMs) show promise for such tasks, they are often overwhelmed by the entangled nature of high-dimensional perturbation results. Moreover, recent works have primarily focused on genetic perturbations in single-cell experiments, leaving bulk-cell chemical perturbations, which is central to drug discovery, largely unexplored. Motivated by this, we present LINCSQA, a novel benchmark for predicting target gene regulation under complex chemical perturbations in bulk-cell environments. We further propose PBio-Agent, a multi-agent framework that integrates difficulty-aware task sequencing with iterative knowledge refinement. Our key insight is that genes affected by the same perturbation share causal structure, allowing confidently predicted genes to contextualize more challenging cases. The framework employs specialized agents enriched with biological knowledge graphs, while a synthesis agent integrates outputs and specialized judges ensure logical coherence. PBio-Agent outperforms existing baselines on both LINCSQA and PerturbQA, enabling even smaller models to predict and explain complex biological processes without additional training.", "date": "2026-02-07", "categories": []}
{"id": "2604.25968", "title": "Mining Negative Sequential Patterns to Improve Viral Genomic Feature Representation and Classification", "abstract": "Viruses represent the most abundant biological entities on Earth and play a pivotal role in microbial ecosystems, yet, as prominent human pathogens, they are closely linked to human morbidity and mortality. Accurate identification of viral sequences from viral genome sequences is therefore essential, but existing genome-based classification models that largely relying on composition- or frequency-based subsequence features often suffer from limited interpretability and reduced accuracy, particularly on complex or imbalanced datasets. To address these limitations, we propose GeneNSPCla (Genomic Negative Sequential Pattern-based Classification), a novel viral classification framework based on Negative Sequential Patterns (NSPs) that extracts discriminative absence-based features from nucleotide sequences of RNA viral genomes. By transforming these NSPs into numerical feature vectors and integrating them into multiple supervised classifiers, GeneNSPCla effectively captures both presence and absence signals in viral sequences. Furthermore, we propose a negative pattern mining algorithm adapted for processing genomic data: GONPM+, which can discover longer and more biologically meaningful negative sequential patterns. The experimental results demonstrate that the average accuracy of GONPM+ in 8 classifiers has improved by 10.03% compared to the original negative pattern mining algorithm and by 24.75% compared to the positive pattern mining algorithm. These findings highlight the effectiveness of incorporating absence-based sequential information, providing a new and complementary perspective for viral genome analysis and classification.", "date": "2026-04-28", "categories": []}
{"id": "2507.14245", "title": "Curriculum-guided multimodal representation learning enables generalizable prediction of nanomaterial-protein interactions", "abstract": "Nanomaterial-protein interactions (NPI) are pivotal to realizing the therapeutic and diagnostic potential of nanomaterials. Although AI promises to accelerate mechanistic understanding and enable rational nanomaterial design, robust generalization to unseen nanomaterials or proteins remains unresolved. Here, we present CuMMI (curriculum-guided multimodal interaction model), a generalizable, explainable, and transferable model designed to infer NPI across complex biological settings. CuMMI leverages a self-constructed million-scale NPI dataset and adopts a multi-stage curriculum centered on human plasma, with progressively broader biofluid exposure to enhance data coverage and generalizability. By integrating protein sequence, structure, and a text-encoded experimental context of 37 features, CuMMI captures complementary material-specific, biochemical, and environmental information. Sample-level quality weights are assigned to ensure full utilization of available data while mitigating low-confidence and sparsely recorded entries. Ablation studies highlight the most influential tabular features, clarifying their contribution to the prediction. Through rigorous external validation across independence-preserving temporal, nanomaterial-held-out, and protein-held-out evaluations, our framework consistently achieves good performance (mean of five classification metrics exceeding 0.75), highlighting its robustness and generalizability to unseen data. Furthermore, fine-tuning on independent gold-nanoparticle data and a held-out protein subset further delivers better performance than training from scratch with substantially fewer samples. Together, our approach enables generalizable and transferable NPI prediction and may accelerate in vitro research and applications of nanomaterials.", "date": "2025-07-18", "categories": []}
{"id": "2604.25062", "title": "Learning biophysical models of gene regulation with probability flow matching", "abstract": "Cellular differentiation is governed by gene regulatory networks, the high-dimensional stochastic biochemical systems that determine the transcriptional landscape and mediate cellular responses to signals and perturbations. Although single-cell RNA sequencing provides quantitative snapshots of the transcriptome, current methods for inferring gene-regulatory dynamics often lack mechanistic interpretability and fail to generalize to unseen conditions. Here we introduce Probability Flow Matching (PFM), a scalable framework for learning biophysically consistent stochastic processes directly from time-resolved single-cell measurements. Applying PFM to three hematopoiesis datasets, we show that models with similar interpolation accuracy can encode fundamentally different dynamics, with only biophysically consistent formulations accurately capturing mechanisms of lineage transitions, fate specification, and gene perturbation responses. We further demonstrate that PFM accommodates unbalanced populations, enabling simultaneous inference of cellular proliferation and death dynamics. Together, these results establish PFM as a flexible, scalable framework for integrating mechanistic modeling with single-cell omics.", "date": "2026-04-27", "categories": []}
{"id": "2604.25244", "title": "Learning Structure, Energy, and Dynamics: A Survey of Artificial Intelligence for Protein Dynamics", "abstract": "Protein dynamics underlie many biological functions, yet remain difficult to characterize due to the high computational cost of molecular dynamics simulations and the scarcity of dynamic structural data. This survey reviews recent advances in artificial intelligence for protein dynamics from three perspectives: learning from structural ensembles and trajectories, learning from physical energy signals, and learning to accelerate molecular simulations. We summarize representative methods for conformation ensemble generation, trajectory generation, Boltzmann generators, physics-aware adaptation, machine learning potentials, coarse-grained modeling, and collective variable discovery. We further discuss available datasets and key open challenges, such as scalability, thermodynamic consistency, kinetic fidelity, and integration with experimental constraints.", "date": "2026-04-28", "categories": []}
{"id": "2409.10588", "title": "ADIOS: Antibody Development via Opponent Shaping", "abstract": "Anti-viral therapies are typically designed to target only the current strains of a virus, a myopic response. However, therapy-induced selective pressures drive the emergence of new viral strains, against which the original myopic therapies are no longer effective. This evolutionary response presents an opportunity: our therapies could both defend against and actively influence viral evolution. This motivates our method ADIOS: Antibody Development vIa Opponent Shaping. ADIOS is a meta-learning framework where the process of antibody therapy design, the outer loop, accounts for the virus's adaptive response, the inner loop. With ADIOS, antibodies are not only robust against potential future variants, they also influence, i.e., shape, which future variants emerge. In line with the opponent shaping literature, we refer to our optimised antibodies as shapers. To demonstrate the value of ADIOS, we build a viral evolution simulator using the Absolut! framework, in which shapers successfully target both current and future viral variants, outperforming myopic antibodies. Furthermore, we show that shapers modify the distribution over viral evolutionary trajectories to result in weaker variants. We believe that our ADIOS paradigm will facilitate the discovery of long-lived vaccines and antibody therapies while also generalising to other domains. Specifically, domains such as antimicrobial resistance, cancer treatment, and others with evolutionarily adaptive opponents. Our code is available at https://github.com/olakalisz/adios.", "date": "2024-09-16", "categories": []}
{"id": "2505.19763", "title": "AlphaFold's Bayesian Roots in Probability Kinematics", "abstract": "The seminal breakthrough of AlphaFold in protein structure prediction relied on a learned potential energy function parameterized by deep models, in contrast to its successors AlphaFold2 and AlphaFold3, which lack an explicit probabilistic interpretation. While AlphaFold's potential was originally justified by heuristic analogy to physical potentials of mean force, we show that it can instead be understood as a principled instance of probability kinematics (PK), also known as Jeffrey conditioning, a generalization of Bayesian updating. This reinterpretation reveals that AlphaFold is a generalized Bayesian model that explicitly defines a posterior distribution over structures, providing a deeper explanation of its success and a foundation for future model design. To demonstrate this framework with precision, we introduce a tractable synthetic model in which an angular random walk prior is updated with distance-based evidence via PK, directly mirroring AlphaFold's mechanism. This setting allows us to explore the probabilistic foundations of AlphaFold in a clear and interpretable way. Our work connects a landmark in protein structure prediction to a broader class of compositional deep generative models and points to new opportunities for principled probabilistic approaches.", "date": "2025-05-26", "categories": []}
{"id": "2506.21107", "title": "Doloris: Dual Conditional Diffusion Implicit Bridges with Sparsity Masking Strategy for Unpaired Single-Cell Perturbation Estimation", "abstract": "Estimating single-cell responses across various perturbations facilitates the identification of key genes and enhances drug screening, significantly boosting experimental efficiency. However, single-cell sequencing is a destructive process, making it impossible to capture the same cell's phenotype before and after perturbation. Consequently, data collected under perturbed and unperturbed conditions are inherently unpaired, creating a critical yet unresolved problem in single-cell perturbation modeling. Moreover, the high dimensionality and sparsity of single-cell expression make direct modeling prone to focusing on zeros and neglecting meaningful patterns. To address these problems, we propose a new paradigm for single-cell perturbation modeling. Specifically, we leverage dual diffusion models to learn the control and perturbed distributions separately, and implicitly align them through a shared Gaussian latent space, without requiring explicit cell pairing. Furthermore, we introduce a sparsity masking strategy in which the mask model learns to predict zero-expressed genes, allowing the diffusion model to focus on capturing meaningful patterns among expressed genes and thereby preserving diversity in high-dimensional sparse data. We introduce \\textbf{Doloris}, a generative framework that defines a new paradigm for modeling unpaired, high-dimensional, and sparse single-cell perturbation data. It leverages dual conditional diffusion models for separate learning of control and perturbed distributions, complemented by a sparsity masking strategy to enhance prediction of zero-valued genes. The results on publicly available datasets show that our model effectively captures the diversity of single-cell perturbations and achieves state-of-the-art performance. To facilitate reproducibility, we include the code in the supplementary materials.", "date": "2025-06-26", "categories": []}
{"id": "2509.19601", "title": "Learning Genetic Circuit Modules with Neural Networks: Full Version", "abstract": "In several applications, including in synthetic biology, one often has input/output data on a system composed of many modules, and although the modules' input/output functions and signals may be unknown, knowledge of the composition architecture can significantly reduce the amount of training data required to learn the system's input/output mapping. Learning the modules' input/output functions is also necessary for designing new systems from different composition architectures. Here, we propose a modular learning framework, which incorporates prior knowledge of the system's compositional structure to (a) identify the composing modules' input/output functions from the system's input/output data and (b) achieve this by using a reduced amount of data compared to what would be required without knowledge of the compositional structure. To achieve this, we introduce the notion of modular identifiability, which allows recovery of modules' input/output functions from a subset of the system's input/output data, and provide theoretical guarantees on a class of systems motivated by genetic circuits. We demonstrate the theory on computational studies showing that a neural network (NNET) that accounts for the compositional structure can learn the composing modules' input/output functions and predict the system's output on inputs outside of the training set distribution. By contrast, a neural network that is agnostic of the structure is unable to predict on inputs that fall outside of the training set distribution. By reducing the need for experimental data and allowing module identification, this framework offers the potential to ease the design of synthetic biological circuits and of multi-module systems more generally.", "date": "2025-09-23", "categories": []}
{"id": "2512.15930", "title": "Scalable Agentic Reasoning for Designing Biologics Targeting Intrinsically Disordered Proteins", "abstract": "Intrinsically disordered proteins (IDPs) represent crucial therapeutic targets due to their significant role in disease -- approximately 80\\% of cancer-related proteins contain long disordered regions -- but their lack of stable secondary/tertiary structures makes them \"undruggable\". While recent computational advances, such as diffusion models, can design high-affinity IDP binders, translating these to practical drug discovery requires autonomous systems capable of reasoning across complex conformational ensembles and orchestrating diverse computational tools at scale.To address this challenge, we designed and implemented StructBioReasoner, a scalable multi-agent system for designing biologics that can be used to target IDPs. StructBioReasoner employs a novel tournament-based reasoning framework where specialized agents compete to generate and refine therapeutic hypotheses, naturally distributing computational load for efficient exploration of the vast design space. Agents integrate domain knowledge with access to literature synthesis, AI-structure prediction, molecular simulations, and stability analysis, coordinating their execution on HPC infrastructure via an extensible federated agentic middleware, Academy. We benchmark StructBioReasoner across Der f 21 and NMNAT-2 and demonstrate that over 50\\% of 787 designed and validated candidates for Der f 21 outperformed the human-designed reference binders from literature, in terms of improved binding free energy. For the more challenging NMNAT-2 protein, we identified three binding modes from 97,066 binders, including the well-studied NMNAT2:p53 interface. Thus, StructBioReasoner lays the groundwork for agentic reasoning systems for IDP therapeutic discovery on Exascale platforms.", "date": "2025-12-17", "categories": []}
{"id": "2512.21786", "title": "VAMP-Net: An Interpretable Multi-Path Network of Genomic Permutation-Invariant Set Attention and Quality-Aware 1D-CNN for MTB Drug Resistance", "abstract": "Genomic prediction of drug resistance in Mycobacterium tuberculosis is often hindered by complex epistatic interactions and variable sequencing quality. We present the Interpretable Variant-Aware Multi-Path Network (VAMP-Net), a novel architecture addressing these challenges through a dual-pathway approach. Path-1 utilizes a Set Attention Transformer to model permutation-invariant variant sets and capture epistatic dependencies, while Path-2 employs a 1D-CNN to analyze VCF quality metrics for adaptive confidence scoring. Evaluated on four critical anti-TB drugs, VAMP-Net significantly outperforms baseline CNN and MLP models, achieving accuracies > 95% and AUCs around 0.97 for Rifampicin and Rifabutin. Feature attribution analysis via Integrated Gradients successfully recovered canonical targets (rpoB, embB, katG) and discovered high-impact novel loci. Functional enrichment confirmed these novel variants constitute non-random metabolic modules (p=0.00239) centered on cell-wall remodeling. Furthermore, systematic ablation of the Quality-Aware pathway demonstrates that the model performs a learned \"integrated audit,\" prioritizing the Fraction of Supporting Reads and relative confidence over raw depth to mitigate technical noise. This dual-layer interpretability, bridging genomic pathogenicity with technical reliability, establishes a new paradigm for robust, auditable, and clinically actionable resistance prediction, positioning VAMP-Net as an important tool for both diagnostic classification and mechanistic discovery in clinical genomics.", "date": "2025-12-25", "categories": []}
{"id": "2604.18467", "title": "An Integrated Deep-Learning Framework for Peptide-Protein Interaction Prediction and Target-Conditioned Peptide Generation with ConGA-PepPI and TC-PepGen", "abstract": "Motivation: Peptide-protein interactions (PepPIs) are central to cellular regulation and peptide therapeutics, but experimental characterization remains too slow for large-scale screening. Existing methods usually emphasize either interaction prediction or peptide generation, leaving candidate prioritization, residue-level interpretation, and target-conditioned expansion insufficiently integrated. Results: We present an integrated framework for early-stage peptide screening that combines a partner-aware prediction and localization model (ConGA-PepPI) with a target-conditioned generative model (TC-PepGen). ConGA-PepPI uses asymmetric encoding, bidirectional cross-attention, and progressive transfer from pair prediction to binding-site localization, while TC-PepGen preserves target information throughout autoregressive decoding via layerwise conditioning. In five-fold cross-validation, ConGA-PepPI achieved 0.839 accuracy and 0.921 AUROC, with binding-site AUPR values of 0.601 on the protein side and 0.950 on the peptide side, and remained competitive on external benchmarks. Under a controlled length-conditioned benchmark, 40.39% of TC-PepGen peptides exceeded native templates in AlphaFold 3 ipTM, and unconstrained generation retained evidence of target-conditioned signal.", "date": "2026-04-20", "categories": []}
{"id": "2604.23134", "title": "h-MINT: Modeling Pocket-Ligand Binding with Hierarchical Molecular Interaction Network", "abstract": "Accurate molecular representations are critical for drug discovery, and a central challenge lies in capturing the chemical environment of molecular fragments, as key interactions, such as H-bond and {\\pi} stacking, occur only under specific local conditions. Most existing approaches represent molecules as atom-level graphs; however, atom-level representations can hardly express higher-order chemical context (e.g., stereochemistry, lone pairs, conjugation). Fragment-based methods (e.g., principal subgraph, predefined functional groups) fail to preserve essential information such as chirality, aromaticity, and ionic states. This work addresses these limitations from two aspects. (i) OverlapBPE tokenization. We propose a novel data-driven molecule tokenization method. Unlike existing approaches, our method allows overlapping fragments, reflecting the inherently fuzzy boundaries of small-molecule substructures and, together with enriched chemical information at the token level, thereby preserving a more complete chemical context. (ii) h-MINT model. OverlapBPE induces many-to-many atom-fragment mappings, which necessitate a new hierarchical architecture. We therefore develop a hierarchical molecular interaction network capable of jointly modeling interactions at both atom and fragment levels. By supporting fragment overlaps, the model naturally accommodates the many-to-many atom-fragment mappings introduced by the OverlapBPE scheme. Extensive evaluation against state-of-the-art methods shows our method improves binding affinity prediction by 2-4% Pearson/Spearman correlation on PDBBind and LBA, enhances virtual screening by 1-3% in key metrics on DUD-E and LIT-PCBA, and achieves the best overall HTS performance on PubChem assays. Further analysis demonstrates that our method effectively captures interactive information while maintaining good generalization.", "date": "2026-04-25", "categories": []}
{"id": "2604.23307", "title": "CombiMOTS: Combinatorial Multi-Objective Tree Search for Dual-Target Molecule Generation", "abstract": "Dual-target molecule generation, which focuses on discovering compounds capable of interacting with two target proteins, has garnered significant attention due to its potential for improving therapeutic efficiency, safety and resistance mitigation. Existing approaches face two critical challenges. First, by simplifying the complex dual-target optimization problem to scalarized combinations of individual objectives, they fail to capture important trade-offs between target engagement and molecular properties. Second, they typically do not integrate synthetic planning into the generative process. This highlights a need for more appropriate objective function design and synthesis-aware methodologies tailored to the dual-target molecule generation task. In this work, we propose CombiMOTS, a Pareto Monte Carlo Tree Search (PMCTS) framework that generates dual-target molecules. CombiMOTS is designed to explore a synthesizable fragment space while employing vectorized optimization constraints to encapsulate target affinity and physicochemical properties. Extensive experiments on real-world databases demonstrate that CombiMOTS produces novel dual-target molecules with high docking scores, enhanced diversity, and balanced pharmacological characteristics, showcasing its potential as a powerful tool for dual-target drug discovery. The code and data is accessible through https://github.com/Tibogoss/CombiMOTS.", "date": "2026-04-25", "categories": []}
{"id": "2604.23368", "title": "TEMPO: Transformers for Temporal Disease Progression from Cross-Sectional Data", "abstract": "Event-Based Models (EBMs) infer biomarker progression from cross-sectional data but typically only as ordinal sequences and rely on rigid model assumptions. We propose \\textsc{Tempo}, a Transformer architecture that learns both ordinal and continuous event sequences through simulation-based supervised learning. \\textsc{Tempo} uses two Transformer modules: one treats biomarkers as tokens to infer event sequencing; the other treats patients as tokens, representing each by their per-biomarker abnormality profile, to infer patients' disease stages. On synthetic benchmarks, \\textsc{Tempo} reduces normalized Kendall's Tau distance by 52.89\\% and staging MAE by 25.33\\% compared to state-of-the-art SA-EBM, with larger reductions in high-dimensional settings (58.88\\% and 61.10\\%). Applied to ADNI, \\textsc{Tempo} recovers a biologically plausible Alzheimer's progression: early medial temporal atrophy, followed by amyloid accumulation and cognitive decline, and late-stage tau pathology with terminal acceleration of global neurodegeneration -- broadly consistent with established disease models. \\textsc{Tempo} also eliminates the need to derive custom inference algorithms and enables rapid empirical comparison of generative hypotheses.", "date": "2026-04-25", "categories": []}
{"id": "2604.23924", "title": "Agentic AI platforms for autonomous training and rule induction of human-human and virus-human protein-protein interactions", "abstract": "We instruct an AI agent to construct two separate agentic AI platforms: one for autonomous training of predictive ML models for human-human and virus-human PPI, and the other for inducing explicit general rules governing human-human and virus-human PPI. The first agentic AI platform for autonomous training of predictive ML models for PPI is designed to consist of five AI agents that handle autonomous data collection, data verification, feature embedding, model design, and training and validation on three-way protein-disjoint cross-fold datasets. For human-human and human-virus PPIs, the final three-way protein-disjoint ensemble achieves an accuracy of 87.3% and 86.5%, respectively. For cross-checking and interpretability purposes, the second agentic AI platform is designed to replace ML predictions with human-readable rules derived from protein embeddings, physicochemical autocovariance descriptors, compartment annotations, pathway-domain overlap, and graph contexts. For human-human PPI, it is defined by a two-rule induction, whereas human-virus is induced by a more complex set of weighted rules. The rules induced by the second agentic platform align with the SHAP-identified features from the predictive ML models built by the first agentic platform. Taken together, our work demonstrates the agentic AI's ability to orchestrate from data planning to execution, and from rule induction to explanation in ML, opening the door to various applications.", "date": "2026-04-27", "categories": []}
{"id": "2604.24371", "title": "PathMoG: A Pathway-Centric Modular Graph Neural Network for Multi-Omics Survival Prediction", "abstract": "Cancer survival prediction from multi-omics data remains challenging because prognostic signals are high-dimensional, heterogeneous, and distributed across interacting genes and pathways. We propose PathMoG, a pathway-centric modular graph neural network for multi-omics survival prediction. PathMoG reorganizes genome-scale inputs into 354 KEGG-informed pathway modules, introduces a Hierarchical Omics Modulation module to condition gene-expression representations on mutation, copy number variation, pathway, and clinical context, and uses dual-level attention to capture both intra-pathway driver signals and inter-pathway clinical relevance. We evaluated PathMoG on 5,650 patients across 10 TCGA cancer types and observed consistent improvements over representative survival baselines. The framework further provides gene-level, pathway-level, and patient-level interpretability, supporting biologically grounded and clinically relevant risk stratification.", "date": "2026-04-27", "categories": []}
{"id": "2604.24506", "title": "MIMIC: A Generative Multimodal Foundation Model for Biomolecules", "abstract": "Biological function emerges from coupled constraints across sequence, structure, regulation, evolution, and cellular context, yet most foundation models in biology are trained within one modality or for a fixed forward task. We present MIMIC, a generative multimodal foundation model trained on our newly curated and aligned dataset, LORE, linking nucleic acid, protein, evolutionary, structural, regulatory, and semantic/contextual modalities within partially observed biomolecular states. MIMIC uses a split-track encoder-decoder architecture to condition on arbitrary subsets of observed modalities and reconstruct or generate missing components of molecular state across the genome, transcriptome, and proteome. Multimodal conditioning consistently improves MIMIC's sequence reconstruction relative to sequence-only inputs, while its learned representations enable state-of-the-art performance on RNA and protein downstream tasks. MIMIC achieves state-of-the-art splicing prediction, and its joint generative formulation enables isoform-aware inference that further improves performance. Beyond prediction, the same generative framework supports constrained design. For RNA, MIMIC identifies corrective edits in a clinically relevant HBB splice-disrupting mutation without reverting it by using evolutionary and structural signals. For proteins, jointly conditioning on shape and surface chemistry of PD-L1 and hACE2 binding sites produces diverse, high-confidence sequences with strong in silico support for target binding. Finally, MIMIC uses experimental context as semantic conditioning to model assay-dependent RNA chemical probing, rather than treating context as a fixed output. Together, these results position MIMIC's aligned multimodal generative modeling as a strong foundation for unifying representation learning, conditional prediction, and constrained biomolecular design within a single model.", "date": "2026-04-27", "categories": []}
{"id": "2604.24634", "title": "Computational Design and Experimental Validation of Photoactive PARP1 Inhibitors", "abstract": "Light-activated drugs are a promising way to treat localized diseases for which existing treatments have severe side effects. However, their development is complicated by the set of photophysical and biological properties that must be simultaneously optimized. Here we used computational techniques to find a set of promising candidates for the photoactive inhibition of the poly(ADP-ribose) polymerase 1 (PARP1) cancer target. Using our recently developed methods based on atomistic simulation and machine learning (ML), we screened a set of 5 million hypothetical photoactive ligands. Our workflow used protein-ligand docking to identify candidates with differential PARP1 binding under light and dark conditions; ML force fields and quantum chemistry calculations to predict p$K_\\mathrm{a}$, absorption spectra, and thermal half-lives; graph-based surrogate models to screen additional compounds; excited-state nonadiabatic dynamics with ML force fields to estimate quantum yields; and free energy perturbation (FEP) to refine binding predictions. From these predictions, we prioritized a small set of synthetically feasible candidates expected to have red-shifted absorption spectra, thermal half-lives on the order of seconds to minutes, and isomer-dependent PARP1 binding under visible-light control. We synthesized 10 candidates and experimentally characterized their photobehavior and PARP1 inhibition constants. Among the validated compounds, \\textbf{1} showed a 15-fold increase in inhibition of PARP1 upon green-light irradiation at 519 nm (208.8 $\\pm$ 28.3 $\\mu$M vs 14.4 $\\pm$ 1.9 $\\mu$M). These results validate the computation-guided screening strategy for identifying red-shifted PARP1 photoinhibitors, while also underscoring current limitations such as rapid thermal relaxation in aqueous media.", "date": "2026-04-27", "categories": []}
{"id": "2512.09003", "title": "Digital Modeling of Spatial Pathway Activity from Histology Reveals Tumor Microenvironment Heterogeneity", "abstract": "Spatial transcriptomics (ST) enables simultaneous mapping of tissue morphology and spatially resolved gene expression, offering unique opportunities to study tumor microenvironment heterogeneity. Here, we introduce a computational framework that predicts spatial pathway activity directly from hematoxylin-and-eosin-stained histology images at microscale resolution 55 and 100 um. Using image features derived from a computational pathology foundation model, we found that TGFb signaling was the most accurately predicted pathway across three independent breast and lung cancer ST datasets. In 87-88% of reliably predicted cases, the resulting spatial TGFb activity maps reflected the expected contrast between tumor and adjacent non-tumor regions, consistent with the known role of TGFb in regulating interactions within the tumor microenvironment. Notably, linear and nonlinear predictive models performed similarly, suggesting that image features may relate to pathway activity in a predominantly linear fashion or that nonlinear structure is small relative to measurement noise. These findings demonstrate that features extracted from routine histopathology may recover spatially coherent and biologically interpretable pathway patterns, offering a scalable strategy for integrating image-based inference with ST information in tumor microenvironment studies.", "date": "2025-12-09", "categories": []}
{"id": "2603.12845", "title": "Multimodal Protein Language Models for Enzyme Kinetic Parameters: From Substrate Recognition to Conformational Adaptation", "abstract": "Predicting enzyme kinetic parameters quantifies how efficiently an enzyme catalyzes a specific substrate under defined biochemical conditions. Canonical parameters such as the turnover number ($k_\\text{cat}$), Michaelis constant ($K_\\text{m}$), and inhibition constant ($K_\\text{i}$) depend jointly on the enzyme sequence, the substrate chemistry, and the conformational adaptation of the active site during binding. Many learning pipelines simplify this process to a static compatibility problem between the enzyme and substrate, fusing their representations through shallow operations and regressing a single value. Such formulations overlook the staged nature of catalysis, which involves both substrate recognition and conformational adaptation. In this regard, we reformulate kinetic prediction as a staged multimodal conditional modeling problem and introduce the Enzyme-Reaction Bridging Adapter (ERBA), which injects cross-modal information via fine-tuning into Protein Language Models (PLMs) while preserving their biochemical priors. ERBA performs conditioning in two stages: Molecular Recognition Cross-Attention (MRCA) first injects substrate information into the enzyme representation to capture specificity; Geometry-aware Mixture-of-Experts (G-MoE) then integrates active-site structure and routes samples to pocket-specialized experts to reflect induced fit. To maintain semantic fidelity, Enzyme-Substrate Distribution Alignment (ESDA) enforces distributional consistency within the PLM manifold in a reproducing kernel Hilbert space. Experiments across three kinetic endpoints and multiple PLM backbones, ERBA delivers consistent gains and stronger out-of-distribution performance compared with sequence-only and shallow-fusion baselines, offering a biologically grounded route to scalable kinetic prediction and a foundation for adding cofactors, mutations, and time-resolved structural cues.", "date": "2026-03-13", "categories": []}
{"id": "2604.21573", "title": "CHRep: Cross-modal Histology Representation and Post-hoc Calibration for Spatial Gene Expression Prediction", "abstract": "Spatial transcriptomics (ST) enables spatially resolved gene profiling but remains expensive and low-throughput, limiting large-cohort studies and routine clinical use. Predicting spatial gene expression from routine hematoxylin and eosin (H&E) slides is a promising alternative, yet under realistic leave-one-slide-out evaluation, existing models often suffer from slide-level appearance shifts and regression-driven over-smoothing that suppress biologically meaningful variation. CHRep is a two-phase framework for robust histology-to-expression prediction. In the training phase, CHRep learns a structure-aware representation by jointly optimizing correlation-aware regression, symmetric image-expression alignment, and coordinate-induced spatial topology regularization. In the inference phase, cross-slide robustness is improved without backbone fine-tuning through a lightweight calibration module trained on the training slides, which combines a non-parametric estimate from a training gallery with a magnitude-regularized correction module. Unlike prior embedding-alignment or retrieval-based transfer methods that rely on a single prediction route, CHRep couples topology-preserving representation learning with post-hoc calibration, enabling stable neighborhood retrieval and controlled bias correction under slide-level shifts. Across the three cohorts, CHRep consistently improves gene-wise correlation under leave-one-slide-out evaluation, with the largest gains observed on Alex+10x. Relative to HAGE, the Pearson correlation coefficient on all considered genes [PCC(ACG)] increases by 4.0% on cSCC and 9.8% on HER2+. Relative to mclSTExp, PCC(ACG) further improves by 39.5% on Alex+10x, together with 9.7% and 9.0% reductions in mean squared error (MSE) and mean absolute error (MAE), respectively.", "date": "2026-04-23", "categories": []}
{"id": "2604.21690", "title": "Evaluating Post-hoc Explanations of the Transformer-based Genome Language Model DNABERT-2", "abstract": "Explaining deep neural network predictions on genome sequences enables biological insight and hypothesis generation-often of greater interest than predictive performance alone. While explanations of convolutional neural networks (CNNs) have been shown to capture relevant patterns in genome sequences, it is unclear whether this transfers to more expressive Transformer-based genome language models (gLMs). To answer this question, we adapt AttnLRP, an extension of layer-wise relevance propagation to the attention mechanism, and apply it to the state-of-the-art gLM DNABERT-2. Thereby, we propose strategies to transfer explanations from token and nucleotide level. We evaluate the adaption of AttnLRP on genomic datasets using multiple metrics. Further, we provide an extensive comparison between the explanations of DNABERT-2 and a baseline CNN. Our results demonstrate that AttnLRP yields reliable explanations corresponding to known biological patterns. Hence, like CNNs, gLMs can also help derive biological insights. This work contributes to the explainability of gLMs and addresses the comparability of relevance attributions across different architectures.", "date": "2026-04-23", "categories": []}
{"id": "2509.02060", "title": "Morphology-Aware Peptide Discovery via Masked Conditional Generative Modeling", "abstract": "Peptide self-assembly prediction offers a powerful bottom-up strategy for designing biocompatible, low-toxicity materials for large-scale synthesis in a broad range of biomedical and energy applications. However, screening the vast sequence space for categorization of aggregate morphology remains intractable. We introduce PepMorph, an end-to-end peptide discovery pipeline that generates novel sequences that are not only prone to aggregate but whose self-assembly is steered toward fibrillar or spherical morphologies by conditioning on isolated peptide descriptors that serve as morphology proxies. To this end, we compiled a new dataset by leveraging existing aggregation propensity datasets and extracting geometric and physicochemical descriptors. This dataset is then used to train a Transformer-based Conditional Variational Autoencoder with a masking mechanism, which generates novel peptides under arbitrary conditioning. After filtering to ensure design specifications and validation of generated sequences through coarse-grained molecular dynamics (CG-MD) simulations, PepMorph yielded 83% success rate under our CG-MD validation protocol and morphology criterion for the targeted class, showcasing its promise as a framework for application-driven peptide discovery.", "date": "2025-09-02", "categories": []}
{"id": "2604.20003", "title": "scpFormer: A Foundation Model for Unified Representation and Integration of the Single-Cell Proteomics", "abstract": "The integration of single-cell proteomic data is often hindered by the fragmented nature of targeted antibody panels. To address this limitation, we introduce scpFormer, a transformer-based foundation model designed for single-cell proteomics. Pre-trained on over 390 million cells, scpFormer replaces standard index-based tokenization with a continuous, sequence-anchored approach. By combining Evolutionary Scale Modeling (ESM) with value-aware expression embeddings, it dynamically maps variable panels into a shared semantic space without artificial discretization. We demonstrate that scpFormer generates global cell representations that perform competitively in large-scale batch integration and unsupervised clustering. Moreover, its open-vocabulary architecture facilitates in silico panel expansion, assisting in the reconstruction of biological manifolds in sparse clinical datasets. Finally, this learned protein co-expression logic is transferable to bulk-omics tasks, supporting applications like cancer drug response prediction. scpFormer provides a versatile, panel-agnostic framework to facilitate scalable biomarker discovery and precision oncology.", "date": "2026-04-21", "categories": []}
{"id": "2604.20263", "title": "AROMA: Augmented Reasoning Over a Multimodal Architecture for Virtual Cell Genetic Perturbation Modeling", "abstract": "Virtual cell modeling predicts molecular state changes under genetic perturbations in silico, which is essential for biological mechanism studies. However, existing approaches suffer from unconstrained reasoning, uninterpretable predictions, and retrieval signals that are weakly aligned with regulatory topology. To address these limitations, we propose AROMA, an Augmented Reasoning Over a Multimodal Architecture for virtual cell genetic perturbation modeling. AROMA integrates textual evidence, graph-topology information, and protein sequence features to model perturbation-target dependencies, and is trained with a two-stage optimization strategy to yield predictions that are both accurate and interpretable. We also construct two knowledge graphs and a perturbation reasoning dataset, PerturbReason, containing more than 498k samples, as reusable resources for the virtual cell domain. Experiments show that AROMA outperforms existing methods across multiple cell lines, and remains robust under zero-shot evaluation on an unseen cell line, as well as in knowledge-sparse, long-tail scenarios. Overall, AROMA demonstrates that combining knowledge-driven multimodal modeling with evidence retrieval provides a promising pathway toward more reliable and interpretable virtual cell perturbation prediction. Model weights are available at https://huggingface.co/blazerye/AROMA. Code is available at https://github.com/blazerye/AROMA.", "date": "2026-04-22", "categories": []}
{"id": "2604.04981", "title": "An Imbalanced Dataset with Multiple Feature Representations for Studying Quality Control of Next-Generation Sequencing", "abstract": "Next-generation sequencing (NGS) is a key technique for studying the DNA and RNA of organisms. However, identifying quality problems in NGS data across different experimental settings remains challenging. To develop automated quality-control tools, researchers require datasets with features that capture the characteristics of quality problems. Existing NGS repositories, however, offer only a limited number of quality-related features. To address this gap, we propose a dataset derived from 37,491 NGS samples with two types of quality-related feature representations. The first type consists of 34 features derived from quality control tools (QC-34 features). The second type has a variable number of features ranging from eight to 1,183. These features were derived from read counts in problematic genomic regions identified by the ENCODE blocklist (BL features). All features describe the same human and mouse samples from five genomic assays, allowing direct comparison of feature representations. The proposed dataset includes a binary quality label, derived from automated quality control and domain experts. Among all samples, $3.2\\%$ are of low quality. Supervised machine learning algorithms accurately predicted quality labels from the features, confirming the relevance of the provided feature representations. The proposed feature representations enable researchers to study how different feature types (QC-34 vs. BL features) and granularities (varying number of BL features) affect the detection of quality problems.", "date": "2026-04-04", "categories": []}
{"id": "2604.18621", "title": "Quantum AI for Cancer Diagnostic Biomarker Discovery", "abstract": "Quantum machine learning offers a promising new paradigm for computational biology by leveraging quantum mechanical principles to enhance cancer classification, biomarker discovery, and bioinformatics diagnostics. In this study, we apply QML to identify subtype specific biomarkers for lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), the two predominant forms of non-small cell lung cancer. Our methodology involves a two-phase process: in Phase 1, differential expression analysis and methylation analysis between tumor and normal samples allows us to identify LUAD-specific and LUSC-specific genes, revealing potential prognostic biomarkers for cancer subtypes. Phase 2 focuses on developing a quantum classifier capable of distinguishing between LUAD and LUSC tumors, as well as between tumor and normal samples. This classifier not only enhances diagnostic precision but also demonstrates the quantum advantage in processing large-scale multiomic datasets. Our results consistently demonstrated that Sample3, representing the combined gene set, achieved the highest overall predictive performance in all metrics. These results demonstrate that QML provides an effective and scalable approach for biomarker discovery and subtype specific cancer classification. GO enrichment analysis highlighted the significant involvement of genes in synaptic signaling, ion channel regulation, and neuronal development. In the quantum phase, KEGG analysis further identified enrichment in cancer-associated pathways, including neurotrophin, MAPK, Ras, and PI3KAkt signaling, with key genes such as NGFR, NTRK2, and NTF3 suggesting a central role in neurotrophinmediated oncogenic processes. Our findings highlight the growing potential of quantum computing to advance precision oncology and next-generation biomedical analytics.", "date": "2026-04-18", "categories": []}
{"id": "2604.19562", "title": "Structure-guided molecular design with contrastive 3D protein-ligand learning", "abstract": "Structure-based drug discovery faces the dual challenge of accurately capturing 3D protein-ligand interactions while navigating ultra-large chemical spaces to identify synthetically accessible candidates. In this work, we present a unified framework that addresses these challenges by combining contrastive 3D structure encoding with autoregressive molecular generation conditioned on commercial compound spaces. First, we introduce an SE(3)-equivariant transformer that encodes ligand and pocket structures into a shared embedding space via contrastive learning, achieving competitive results in zero-shot virtual screening. Second, we integrate these embeddings into a multimodal Chemical Language Model (MCLM). The model generates target-specific molecules conditioned on either pocket or ligand structures, with a learned dataset token that steers the output toward targeted chemical spaces, yielding candidates with favorable predicted binding properties across diverse targets.", "date": "2026-04-21", "categories": []}
{"id": "2506.03157", "title": "UniSim: A Unified Simulator for Time-Coarsened Dynamics of Biomolecules", "abstract": "Molecular Dynamics (MD) simulations are essential for understanding the atomic-level behavior of molecular systems, giving insights into their transitions and interactions. However, classical MD techniques are limited by the trade-off between accuracy and efficiency, while recent deep learning-based improvements have mostly focused on single-domain molecules, lacking transferability to unfamiliar molecular systems. Therefore, we propose \\textbf{Uni}fied \\textbf{Sim}ulator (UniSim), which leverages cross-domain knowledge to enhance the understanding of atomic interactions. First, we employ a multi-head pretraining approach to learn a unified atomic representation model from a large and diverse set of molecular data. Then, based on the stochastic interpolant framework, we learn the state transition patterns over long timesteps from MD trajectories, and introduce a force guidance module for rapidly adapting to different chemical environments. Our experiments demonstrate that UniSim achieves highly competitive performance across small molecules, peptides, and proteins.", "date": "2025-05-20", "categories": []}
{"id": "2604.16550", "title": "An Interpretable Framework Applying Protein Words to Predict Protein-Small Molecule Complementary Pairing Rules", "abstract": "Despite the high accuracy of 'black box' deep learning models, drug discovery still relies on protein-ligand interaction principles and heuristics. To improve interpretability of protein-small molecule binding predictions, we developed the PWRules framework, which applies binding affinity data to identify privileged small molecule fragments and subsequently defines complementary pairing rules between these fragments and protein words (semantic sequence units) through an interpretability module. The resulting word-fragment rules are then ranked by the PWScore function to prioritize active compounds. Evaluations on benchmark datasets show that PWScore achieves competitive performance comparable to the physics-based model (Glide) and the deep learning model (PSICHIC) and shows broad applicability for protein targets outside the training dataset, e.g., SARS-CoV-2 main protease. Notably, PWScore captures complementary interaction information, yielding superior enrichment performance when integrated with these established methods. Structural analysis of protein-ligand complexes indicates that learned word-fragment rules are significantly enriched near ligand-binding pockets, despite training without explicit structural guidance. By extracting and applying complementary pairing rules, PWRules provides an interpretable framework for drug discovery.", "date": "2026-04-17", "categories": []}
{"id": "2604.16685", "title": "Graph Transformer-Based Pathway Embedding for Cancer Prognosis", "abstract": "Accurate prediction of cancer progression remains a challenge due to the high heterogeneity of molecular omics data across patients. While biologically informed models have improved the interpretability of these predictions, a persistent limitation lies in how they encode individual genes to construct pathway representations. Existing hierarchical models typically derive gene features by directly mapping raw molecular inputs, whereas integration frameworks often rely on simple statistical aggregations of patient-level signals. These approaches often fail to explicitly learn a shared base representation for each gene, thereby limiting the expressiveness and biological accuracy of downstream pathway embeddings. To address this, we introduce PATH, a modulation-based, patient-conditioned gene embedding strategy. PATH represents a paradigm shift by starting from a shared base embedding for each gene, preserving a stable biological identity across the population, and then dynamically adapting it using patient-specific copy number variation (CNV) and mutation signals. This allows the model to capture subtle individual molecular variations while maintaining a consistent latent understanding of the gene itself. We integrate PATH into a graph transformer framework that models interactions among biologically connected pathways through pathway-guided attention. Across pancancer metastasis prediction, PATH achieves an F1 score of 0.8766, representing an 8.8 percent improvement over the current SOTA multi-omics benchmarks. Beyond superior predictive accuracy, our approach identifies biologically meaningful pathways and, crucially, reveals disease-state-specific pathway rewiring, offering new insights into the evolving pathway-pathway interactions that drive cancer progression.", "date": "2026-04-17", "categories": []}
{"id": "2604.16776", "title": "SAVE: A Generalizable Framework for Multi-Condition Single-Cell Generation with Gene Block Attention", "abstract": "Modeling single-cell gene expression across diverse biological and technical conditions is crucial for characterizing cellular states and simulating unseen scenarios. Existing methods often treat genes as independent tokens, overlooking their high-level biological relationships and leading to poor performance. We introduce SAVE, a unified generative framework based on conditional Transformers for multi-condition single-cell modeling. SAVE leverages a coarse-grained representation by grouping semantically related genes into blocks, capturing higher-order dependencies among gene modules. A Flow Matching mechanism and condition-masking strategy further enhance flexible simulation and enable generalization to unseen condition combinations. We evaluate SAVE on a range of benchmarks, including conditional generation, batch effect correction, and perturbation prediction. SAVE consistently outperforms state-of-the-art methods in generation fidelity and extrapolative generalization, especially in low-resource or combinatorially held-out settings. Overall, SAVE offers a scalable and generalizable solution for modeling complex single-cell data, with broad utility in virtual cell synthesis and biological interpretation. Our code is publicly available at https://github.com/fdu-wangfeilab/sc-save", "date": "2026-04-18", "categories": []}
{"id": "2604.18022", "title": "Boltzmann Machine Learning with a Parallel, Persistent Markov chain Monte Carlo method for Estimating Evolutionary Fields and Couplings from a Protein Multiple Sequence Alignment", "abstract": "The inverse Potts problem for estimating evolutionary single-site fields and pairwise couplings in homologous protein sequences from their single-site and pairwise amino acid frequencies observed in their multiple sequence alignment would be still one of useful methods in the studies of protein structure and evolution. Since the reproducibility of fields and couplings are the most important, the Boltzmann machine method is employed here, although it is computationally intensive. In order to reduce computational time required for the Boltzmann machine, parallel, persistent Markov chain Monte Carlo method is employed to estimate the single-site and pairwise marginal distributions in each learning step. Also, stochastic gradient descent methods are used to reduce computational time for each learning. Another problem is how to adjust the values of hyperparameters; there are two regularization parameters for evolutionary fields and couplings. The precision of contact residue pair prediction is often used to adjust the hyperparameters. However, it is not sensitive to these regularization parameters. Here, they are adjusted for the fields and couplings to satisfy a specific condition that is appropriate for protein conformations. This method has been applied to eight protein families.", "date": "2026-04-20", "categories": []}
{"id": "2604.18477", "title": "Multi-Scale Reversible Chaos Game Representation: A Unified Framework for Sequence Classification", "abstract": "Biological classification with interpretability remains a challenging task. For this, we introduce a novel encoding framework, Multi-Scale Reversible Chaos Game Representation (MS-RCGR), that transforms biological sequences into multi-resolution geometric representations with guaranteed reversibility. Unlike traditional sequence encoding methods, MS-RCGR employs rational arithmetic and hierarchical k-mer decomposition to generate scale-invariant features that preserve complete sequence information while enabling diverse analytical approaches. Our framework bridges three distinct paradigms for sequence analysis: (1) traditional machine learning using extracted geometric features, (2) computer vision models operating on CGR-generated images, and (3) hybrid approaches combining protein language model embeddings with CGR features. Through comprehensive experiments on synthetic DNA and protein datasets encompassing seven distinct sequence classes, we demonstrate that MS-RCGR features consistently enhance classification performance across all paradigms. Notably, our hybrid approach combining pre-trained language model embeddings (ESM2, ProtT5) with MS-RCGR features achieves superior performance compared to either method alone. The reversibility property of our encoding ensures no information loss during transformation, while multi-scale analysis captures patterns ranging from individual nucleotides to complex motif structures. Our results indicate that MS-RCGR provides a flexible, interpretable, and high-performing foundation for biological sequence analysis.", "date": "2026-04-20", "categories": []}
{"id": "2604.18559", "title": "ConforNets: Latents-Based Conformational Control in OpenFold3", "abstract": "Models from the AlphaFold (AF) family reliably predict one dominant conformation for most well-ordered proteins but struggle to capture biologically relevant alternate states. Several efforts have focused on eliciting greater conformational variability through ad hoc inference-time perturbations of AF models or their inputs. Despite their progress, these approaches remain inefficient and fail to consistently recover major conformational modes. Here, we investigate both the optimal location and manner-of-operation for perturbing latent representations in the AF3 architecture. We distill our findings in ConforNets: channel-wise affine transforms of the pre-Pairformer pair latents. Unlike previous methods, ConforNets globally modulate AF3 representations, making them reusable across proteins. On unsupervised generation of alternate states, ConforNets achieve state-of-the-art success rates on all existing multi-state benchmarks. On the novel supervised task of conformational transfer, ConforNets trained on one source protein can induce a conserved conformational change across a protein family. Collectively, these results introduce a mechanism for conformational control in AF3-based models.", "date": "2026-04-20", "categories": []}
{"id": "2509.21010", "title": "Bridging the phenotype-target gap for molecular generation via multi-objective reinforcement learning", "abstract": "The de novo generation of drug-like molecules capable of inducing desirable phenotypic changes is receiving increasing attention. However, previous methods predominantly rely on expression profiles to guide molecule generation, but overlook the perturbative effect of the molecules on cellular contexts. To overcome this limitation, we propose SmilesGEN, a novel generative model based on variational autoencoder (VAE) architecture to generate molecules with potential therapeutic effects. SmilesGEN integrates a pre-trained drug VAE (SmilesNet) with an expression profile VAE (ProfileNet), jointly modeling the interplay between drug perturbations and transcriptional responses in a common latent space. Specifically, ProfileNet is imposed to reconstruct pre-treatment expression profiles when eliminating drug-induced perturbations in the latent space, while SmilesNet is informed by desired expression profiles to generate drug-like molecules. Our empirical experiments demonstrate that SmilesGEN outperforms current state-of-the-art models in generating molecules with higher degree of validity, uniqueness, novelty, as well as higher Tanimoto similarity to known ligands targeting the relevant proteins. Moreover, we evaluate SmilesGEN for scaffold-based molecule optimization and generation of therapeutic agents, and confirmed its superior performance in generating molecules with higher similarity to approved drugs. SmilesGEN establishes a robust framework that leverages gene signatures to generate drug-like molecules that hold promising potential to induce desirable cellular phenotypic changes. The source code and datasets are available at: https://github.com/hliulab/SmilesGEN.", "date": "2025-09-25", "categories": []}
{"id": "2604.14334", "title": "Mamba-SSM with LLM Reasoning for Feature Selection: Faithfulness-Aware Biomarker Discovery", "abstract": "Gradient saliency from deep sequence models surfaces candidate biomarkers efficiently, but the resulting gene lists can be contaminated by tissue-composition confounders that degrade downstream classifiers. We study whether LLM chain-of-thought (CoT) reasoning can filter these confounders, and whether reasoning quality is associated with downstream performance. We train a Mamba SSM on TCGA-BRCA RNA-seq and extract the top-50 genes by gradient saliency; DeepSeek-R1 evaluates every candidate with structured CoT to produce a final 17-gene set. On the held-out test split, the raw 50-gene saliency set (no LLM) performs worse than a 5,000-gene variance baseline (AUC 0.832 vs. 0.903), while the LLM-filtered set surpasses it (AUC 0.927), using 294x fewer features. A faithfulness audit (COSMIC CGC, OncoKB, PAM50) shows that 6 of 17 selected genes (35.3%) are validated BRCA biomarkers, while 10 of 16 known BRCA genes present in the input were missed - including FOXA1. This divergence between downstream performance and reasoning faithfulness suggests selective faithfulness in this setting: targeted confounder removal can improve predictive performance without comprehensive recall.", "date": "2026-04-15", "categories": []}
{"id": "2502.21274", "title": "BAnG: Bidirectional Anchored Generation for Conditional RNA Design", "abstract": "Designing RNA molecules that interact with specific proteins is a critical challenge in experimental and computational biology. Existing computational approaches require a substantial amount of previously known interacting RNA sequences for each specific protein or a detailed knowledge of RNA structure, restricting their utility in practice. To address this limitation, we develop RNA-BAnG, a deep learning-based model designed to generate RNA sequences for protein interactions without these requirements. Central to our approach is a novel generative method, Bidirectional Anchored Generation (BAnG), which leverages the observation that protein-binding RNA sequences often contain functional binding motifs embedded within broader sequence contexts. We first validate our method on generic synthetic tasks involving similar localized motifs to those appearing in RNAs, demonstrating its benefits over existing generative approaches. We then evaluate our model on biological sequences, showing its effectiveness for conditional RNA sequence design given a binding protein.", "date": "2025-02-28", "categories": []}
{"id": "2603.22564", "title": "MIOFlow 2.0: A unified framework for inferring cellular stochastic dynamics from single cell and spatial transcriptomics data", "abstract": "Understanding cellular trajectories via time-resolved single-cell transcriptomics is vital for studying development, regeneration, and disease. A key challenge is inferring continuous trajectories from discrete snapshots. Biological complexity stems from stochastic cell fate decisions, temporal proliferation changes, and spatial environmental influences. Current methods often use deterministic interpolations treating cells in isolation, failing to capture the probabilistic branching, population shifts, and niche-dependent signaling driving real biological processes.\n  We introduce Manifold Interpolating Optimal-Transport Flow (MIOFlow) 2.0. This framework learns biologically informed cellular trajectories by integrating manifold learning, optimal transport, and neural differential equations. It models three core processes: (1) stochasticity and branching via Neural Stochastic Differential Equations; (2) non-conservative population changes using a learned growth-rate model initialized with unbalanced optimal transport; and (3) environmental influence through a joint latent space unifying gene expression with spatial features like local cell type composition and signaling.\n  By operating in a PHATE-distance matching autoencoder latent space, MIOFlow 2.0 ensures trajectories respect the data's intrinsic geometry. Empirical comparisons show expressive trajectory learning via neural differential equations outperforms existing generative models, including simulation-free flow matching. Validated on synthetic datasets, embryoid body differentiation, and spatially resolved axolotl brain regeneration, MIOFlow 2.0 improves trajectory accuracy and reveals hidden drivers of cellular transitions, like specific signaling niches. MIOFlow 2.0 thus bridges single-cell and spatial transcriptomics to uncover tissue-scale trajectories.", "date": "2026-03-23", "categories": []}
{"id": "2604.14796", "title": "PUFFIN: Protein Unit Discovery with Functional Supervision", "abstract": "Proteins carry out biological functions through the coordinated action of groups of residues organized into structural arrangements. These arrangements, which we refer to as protein units, exist at an intermediate scale, being larger than individual residues yet smaller than entire proteins. A deeper understanding of protein function can be achieved by identifying these units and their associations with function. However, existing approaches either focus on residue-level signals, rely on curated annotations, or segment protein structures without incorporating functional information, thereby limiting interpretable analysis of structure-function relationships. We introduce PUFFIN, a data-driven framework for discovering protein units by jointly learning structural partitioning and functional supervision. PUFFIN represents proteins as residue-level structure graphs and applies a graph neural network with a structure-aware pooling mechanism that partitions each protein into multi-residue units, with functional supervision that shapes the partition. We show that the learned units are structurally coherent, exhibit organized associations with molecular function, and show meaningful correspondence with curated InterPro annotations. Together, these results demonstrate that PUFFIN provides an interpretable framework for analyzing structure-function relationships using learned protein units and their statistical function associations. We made our source code available at https://github.com/boun-tabi-lifelu/puffin.", "date": "2026-04-16", "categories": []}
{"id": "2604.14838", "title": "Intermediate Layers Encode Optimal Biological Representations in Single-Cell Foundation Models", "abstract": "Current single-cell foundation model benchmarks universally extract final layer embeddings, assuming these represent optimal feature spaces. We systematically evaluate layer-wise representations from scFoundation (100M parameters) and Tahoe-X1 (1.3B parameters) across trajectory inference and perturbation response prediction. Our analysis reveals that optimal layers are task-dependent (trajectory peaks at 60% depth, 31% above final layers) and context-dependent (perturbation optima shift 0-96% across T cell activation states). Notably, first-layer embeddings outperform all deeper layers in quiescent cells, challenging assumptions about hierarchical feature abstraction. These findings demonstrate that \"where\" to extract features matters as much as \"what\" the model learns, necessitating systematic layer evaluation tailored to biological task and cellular context rather than defaulting to final-layer embeddings.", "date": "2026-04-16", "categories": []}
{"id": "2505.00598", "title": "Fast and Low-Cost Genomic Foundation Models via Outlier Removal", "abstract": "To address the challenge of scarce computational resources in genomic modeling, we introduce GERM, a genomic foundation model with strong compression performance and fast adaptability. GERM improves upon models like DNABERT-2 by eliminating outliers that hinder low-rank adaptation and post-training quantization, enhancing both efficiency and robustness. We replace the vanilla attention layer with an outlier-free mechanism inspired by associative memory models. By removing outliers during both pre-training and fine-tuning, this approach accelerates adaptation, reduces computational costs, and enhances quantization robustness within acceptable loss margins. Additionally, we propose GERM-T, a strategy that employs small-step continual learning within the outlier-free framework, leveraging original checkpoints to avoid retraining from scratch. Empirically, GERM improves fine-tuning performance by 37.98% and quantization by 64.34% over the baseline model. It also reduces average kurtosis by 92.14% and maximum infinity norm by 82.77%. Compared to leading methods, GERM consistently delivers superior performance, offering a practical solution for genomic modeling in resource-constrained settings. Code is available at https://github.com/MAGICS-LAB/GERM.", "date": "2025-05-01", "categories": []}
{"id": "2604.13256", "title": "Counterfactual Peptide Editing for Causal TCR--pMHC Binding Inference", "abstract": "Neural models for TCR-pMHC binding prediction are susceptible to shortcut learning: they exploit spurious correlations in training data -- such as peptide length bias or V-gene co-occurrence -- rather than the physical binding interface. This renders predictions brittle under family-held-out and distance-aware evaluation, where such shortcuts do not transfer. We introduce \\emph{Counterfactual Invariant Prediction} (CIP), a training framework that generates biologically constrained counterfactual peptide edits and enforces invariance to edits at non-anchor positions while amplifying sensitivity at MHC anchor residues. CIP augments the base classifier with two auxiliary objectives: (1) an invariance loss penalizing prediction changes under conservative non-anchor substitutions, and (2) a contrastive loss encouraging large prediction changes under anchor-position disruptions. Evaluated on a curated VDJdb-IEDB benchmark under family-held-out, distance-aware, and random splits, CIP achieves AUROC 0.831 and counterfactual consistency (CFC) 0.724 under the challenging family-held-out protocol -- a 39.7\\% reduction in shortcut index relative to the unconstrained baseline. Ablations confirm that anchor-aware edit generation is the dominant driver of OOD gains, providing a practical recipe for causally-grounded TCR specificity modeling.", "date": "2026-04-14", "categories": []}
{"id": "2604.13471", "title": "Computational framework for multistep metabolic pathway design", "abstract": "In silico tools are important for generating novel hypotheses and exploring alternatives in de novo metabolic pathway design. However, while many computational frameworks have been proposed for retrobiosynthesis, few successful examples of algorithm-guided xenobiotic biochemical retrosynthesis have been reported in the literature. Deep learning has improved the quality of synthesis and retrosynthesis in organic chemistry applications. Inspired by this progress, we explored combining deep learning of biochemical transformations with the traditional retrobiosynthetic workflow to improve in silico synthetic metabolic pathway designs. To develop our computational biosynthetic pathway design framework, we assembled metabolic reaction and enzymatic template data from public databases. A data augmentation procedure, adapted from literature, was carried out to enrich the assembled reaction dataset with artificial metabolic reactions generated by enzymatic reaction templates. Two neural network-based pathway ranking models were trained as binary classifiers to distinguish assembled reactions from artificial counterparts; each model output a scalar quantifying the plausibility of a 1-step or 2-step pathway. Combining these two models with enzymatic templates, we built a multistep retrobiosynthesis pipeline and validated it by reproducing some natural and non-natural pathways computationally.", "date": "2026-04-15", "categories": []}
{"id": "2604.13980", "title": "BOAT: Navigating the Sea of In Silico Predictors for Antibody Design via Multi-Objective Bayesian Optimization", "abstract": "Antibody lead optimization is inherently a multi-objective challenge in drug discovery. Achieving a balance between different drug-like properties is crucial for the development of viable candidates, and this search becomes exponentially challenging as desired properties grow. The ever-growing zoo of sophisticated in silico tools for predicting antibody properties calls for an efficient joint optimization procedure to overcome resource-intensive sequential filtering pipelines. We present BOAT, a versatile Bayesian optimization framework for multi-property antibody engineering. Our `plug-and-play' framework couples uncertainty-aware surrogate modeling with a genetic algorithm to jointly optimize various predicted antibody traits while enabling efficient exploration of sequence space. Through systematic benchmarking against genetic algorithms and newer generative learning approaches, we demonstrate competitive performance with state-of-the-art methods for multi-objective protein optimization. We identify clear regimes where surrogate-driven optimization outperforms expensive generative approaches and establish practical limits imposed by sequence dimensionality and oracle costs.", "date": "2026-04-15", "categories": []}
{"id": "2509.02648", "title": "Optimizing Prognostic Biomarker Discovery in Pancreatic Cancer Through Hybrid Ensemble Feature Selection and Multi-Omics Data", "abstract": "Prediction of patient survival using high-dimensional multi-omics data requires systematic feature selection methods that ensure predictive performance, sparsity, and reliability for prognostic biomarker discovery. We developed a hybrid ensemble feature selection (hEFS) approach that combines data subsampling with multiple prognostic models, integrating both embedded and wrapper-based strategies for survival prediction. Omics features are ranked using a voting-theory-inspired aggregation mechanism across models and subsamples, while the optimal number of features is selected via a Pareto front, balancing predictive accuracy and model sparsity without any user-defined thresholds. When applied to multi-omics datasets from three pancreatic cancer cohorts, hEFS identifies significantly fewer and more stable biomarkers compared to the conventional, late-fusion CoxLasso models, while maintaining comparable discrimination performance. Implemented within the open-source mlr3fselect R package, hEFS offers a robust, interpretable, and clinically valuable tool for prognostic modelling and biomarker discovery in high-dimensional survival settings.", "date": "2025-09-02", "categories": []}
{"id": "2511.21211", "title": "Robust gene prioritization for Dietary Restriction via Fast-mRMR Feature Selection techniques", "abstract": "Gene prioritization (identifying genes potentially associated with a biological process) is increasingly tackled with Artificial Intelligence. However, existing methods struggle with the high dimensionality and incomplete labelling of biomedical data. This work proposes a more robust and efficient pipeline that leverages Fast-mRMR Feature Selection to retain only relevant, non-redundant features for classifiers, building simpler, more interpretable and more efficient models. Experiments in our domain of interest, prioritizing genes related to Dietary Restriction (DR), show significant improvements over existing methods and enables us to integrate heterogeneous biological feature sets for better performance, a strategy that previously degraded performance due to noise accumulation. This work focuses on DR given the availability of curated data and expert knowledge for validation, yet this pipeline would be applicable to other biological processes, proving that feature selection is critical for reliable gene prioritization in high-dimensional omics.", "date": "2025-11-26", "categories": []}
{"id": "2603.24626", "title": "A Large-Scale Comparative Analysis of Imputation Methods for Single-Cell RNA Sequencing Data", "abstract": "Background: Single-cell RNA sequencing (scRNA-seq) enables gene expression profiling at cellular resolution but is inherently affected by sparsity caused by dropout events, where expressed genes are recorded as zeros due to technical limitations. These artifacts distort gene expression distributions and compromise downstream analyses. Numerous imputation methods have been proposed to recover latent transcriptional signals. These methods range from traditional statistical models to deep learning (DL)-based methods. However, their comparative performance remains unclear, as existing benchmarks evaluate only a limited subset of methods, datasets, and downstream analyses. Results: We present a comprehensive benchmark of 15 scRNA-seq imputation methods spanning 7 methodological categories, including traditional and DL-based methods. Methods are evaluated across 30 datasets from 10 experimental protocols on 6 downstream analyses. Results show that traditional methods, such as model-based, smoothing-based, and low-rank matrix-based methods, generally outperform DL-based methods, including diffusion-based, GAN-based, GNN-based, and autoencoder-based methods. In addition, strong performance in numerical gene expression recovery does not necessarily translate into improved biological interpretability in downstream analyses, including cell clustering, differential expression analysis, marker gene analysis, trajectory analysis, and cell type annotation. Furthermore, method performance varies substantially across datasets, protocols, and downstream analyses, with no single method consistently outperforming others. Conclusions: Our findings provide practical guidance for selecting imputation methods tailored to specific analytical objectives and underscore the importance of task-specific evaluation when assessing imputation performance in scRNA-seq data analysis.", "date": "2026-03-25", "categories": []}
{"id": "2604.12026", "title": "TriFit: Trimodal Fusion with Protein Dynamics for Mutation Fitness Prediction", "abstract": "Predicting the functional impact of single amino acid substitutions (SAVs) is central to understanding genetic disease and engineering therapeutic proteins. While protein language models and structure-based methods have achieved strong performance on this task, they systematically neglect protein dynamics; residue flexibility, correlated motions, and allosteric coupling are well-established determinants of mutational tolerance in structural biology, yet have not been incorporated into supervised variant effect predictors. We present TriFit, a multimodal framework that integrates sequence, structure, and protein dynamics through a four-expert Mixture-of-Experts (MoE) fusion module with trimodal cross-modal contrastive learning. Sequence embeddings are extracted via masked marginal scoring with ESM-2 (650M); structural embeddings from AlphaFold2-predicted C-alpha geometries; and dynamics embeddings from Gaussian Network Model (GNM) B-factors, mode shapes, and residue-residue cross-correlations. The MoE router adaptively weights modality combinations conditioned on the input, enabling protein-specific fusion without fixed modality assumptions. On the ProteinGym substitution benchmark (217 DMS assays, 696k SAVs), TriFit achieves AUROC 0.897 +/- 0.0002, outperforming all supervised baselines including Kermut (0.864) and ProteinNPT (0.844), and the best zero-shot model ESM3 (0.769). Ablation studies confirm that dynamics provides the largest marginal contribution over pairwise modality combinations, and TriFit achieves well-calibrated probabilistic outputs (ECE = 0.044) without post-hoc correction.", "date": "2026-04-13", "categories": []}
{"id": "2502.18026", "title": "ExPath: Targeted Pathway Inference for Biological Knowledge Bases via Graph Learning and Explanation", "abstract": "Retrieving targeted pathways in biological knowledge bases, particularly when incorporating wet-lab experimental data, remains a challenging task and often requires downstream analyses and specialized expertise. In this paper, we frame this challenge as a solvable graph learning and explaining task and propose a novel subgraph inference framework, ExPAth, that explicitly integrates experimental data to classify various graphs (bio-networks) in biological databases. The links (representing pathways) that contribute more to classification can be considered as targeted pathways. Our framework can seamlessly integrate biological foundation models to encode the experimental molecular data. We propose ML-oriented biological evaluations and a new metric. The experiments involving 301 bio-networks evaluations demonstrate that pathways inferred by ExPath are biologically meaningful, achieving up to 4.5x higher Fidelity+ (necessity) and 14x lower Fidelity- (sufficiency) than explainer baselines, while preserving signaling chains up to 4x longer.", "date": "2025-02-25", "categories": []}
{"id": "2510.07286", "title": "Evolutionary Profiles for Protein Fitness Prediction", "abstract": "Predicting the fitness impact of mutations is central to protein engineering but constrained by limited assays relative to the size of sequence space. Protein language models (pLMs) trained with masked language modeling (MLM) exhibit strong zero-shot fitness prediction; we provide a unifying view by interpreting natural evolution as implicit reward maximization and MLM as inverse reinforcement learning (IRL), in which extant sequences act as expert demonstrations and pLM log-odds serve as fitness estimates. Building on this perspective, we introduce EvoIF, a lightweight model that integrates two complementary sources of evolutionary signal: (i) within-family profiles from retrieved homologs and (ii) cross-family structural-evolutionary constraints distilled from inverse folding logits. EvoIF fuses sequence-structure representations with these profiles via a compact transition block, yielding calibrated probabilities for log-odds scoring. On ProteinGym (217 mutational assays; >2.5M mutants), EvoIF and its MSA-enabled variant achieve state-of-the-art or competitive performance while using only 0.15% of the training data and fewer parameters than recent large models. Ablations confirm that within-family and cross-family profiles are complementary, improving robustness across function types, MSA depths, taxa, and mutation depths. The codes will be made publicly available.", "date": "2025-10-08", "categories": []}
{"id": "2601.14346", "title": "DiSPA: Differential Substructure-Pathway Attention for Drug Response Prediction", "abstract": "Accurate prediction of drug response in precision medicine requires models that capture how specific chemical substructures interact with cellular pathway states. However, most existing deep learning approaches treat chemical and transcriptomic modalities independently or combine them only at late stages, limiting their ability to model fine-grained, context-dependent mechanisms of drug action. In addition, vanilla attention mechanisms are often sensitive to noise and sparsity in high-dimensional biological networks, hindering both generalization and interpretability. We present DiSPA (Differential Substructure-Pathway Attention), a framework that models bidirectional interactions between chemical substructures and pathway-level gene expression. DiSPA introduces differential cross-attention to suppress spurious associations while enhancing context-relevant interactions. On the GDSC benchmark, DiSPA achieves state-of-the-art performance, with strong improvements in the disjoint setting. These gains are consistent across random and drug-blind splits, suggesting improved robustness. Analyses of attention patterns indicate more selective and concentrated interactions compared to standard cross-attention. Exploratory evaluation shows that differential attention better prioritizes predefined target-related pathways, although this does not constitute mechanistic validation. DiSPA also shows promising generalization on external datasets (CTRP) and cross-dataset settings, although further validation is needed. It further enables zero-shot application to spatial transcriptomics, providing exploratory insights into region-specific drug sensitivity patterns without ground-truth validation.", "date": "2026-01-20", "categories": []}
{"id": "2602.17330", "title": "SubQuad: Near-Quadratic-Free Structure Inference with Distribution-Balanced Objectives in Adaptive Receptor framework", "abstract": "Comparative analysis of adaptive immune repertoires at population scale is hampered by two practical bottlenecks: the near-quadratic cost of pairwise affinity evaluations and dataset imbalances that obscure clinically important minority clonotypes. We introduce SubQuad, an end-to-end pipeline that addresses these challenges by combining antigen-aware, near-subquadratic retrieval with GPU-accelerated affinity kernels, learned multimodal fusion, and fairness-constrained clustering. The system employs compact MinHash prefiltering to sharply reduce candidate comparisons, a differentiable gating module that adaptively weights complementary alignment and embedding channels on a per-pair basis, and an automated calibration routine that enforces proportional representation of rare antigen-specific subgroups. On large viral and tumor repertoires SubQuad achieves measured gains in throughput and peak memory usage while preserving or improving recall@k, cluster purity, and subgroup equity. By co-designing indexing, similarity fusion, and equity-aware objectives, SubQuad offers a scalable, bias-aware platform for repertoire mining and downstream translational tasks such as vaccine target prioritization and biomarker discovery.", "date": "2026-02-19", "categories": []}
{"id": "2602.17709", "title": "UBio-MolFM: A Universal Molecular Foundation Model for Bio-Systems", "abstract": "All-atom molecular simulation serves as a quintessential ``computational microscope'' for understanding the machinery of life, yet it remains fundamentally limited by the trade-off between quantum-mechanical (QM) accuracy and biological scale. We present UBio-MolFM, a universal foundation model framework specifically engineered to bridge this gap. UBio-MolFM introduces three synergistic innovations: (1) UBio-Mol26, a large bio-specific dataset constructed via a multi-fidelity ``Two-Pronged Strategy'' that combines systematic bottom-up enumeration with top-down sampling of native protein environments (up to 1,200 atoms); (2) E2Former-V2, a linear-scaling equivariant transformer that integrates Equivariant Axis-Aligned Sparsification (EAAS) and Long-Short Range (LSR) modeling to capture non-local physics with up to ~4x higher inference throughput in our large-system benchmarks; and (3) a Three-Stage Curriculum Learning protocol that transitions from energy initialization to energy-force consistency, with force-focused supervision to mitigate energy offsets. Rigorous benchmarking across microscopic forces and macroscopic observables -- including liquid water structure, ionic solvation, and peptide folding -- demonstrates that UBio-MolFM achieves ab initio-level fidelity on large, out-of-distribution biomolecular systems (up to ~1,500 atoms) and realistic MD observables. By reconciling scalability with quantum precision, UBio-MolFM provides a robust, ready-to-use tool for the next generation of computational biology.", "date": "2026-02-13", "categories": []}
{"id": "2604.10531", "title": "PepBenchmark: A Standardized Benchmark for Peptide Machine Learning", "abstract": "Peptide therapeutics are widely regarded as the \"third generation\" of drugs, yet progress in peptide Machine Learning (ML) are hindered by the absence of standardized benchmarks. Here we present PepBenchmark, which unifies datasets, preprocessing, and evaluation protocols for peptide drug discovery. PepBenchmark comprises three components: (1) PepBenchData, a well-curated collection comprising 29 canonical-peptide and 6 non-canonical-peptide datasets across 7 groups, systematically covering key aspects of peptide drug development, representing, to the best of our knowledge, the most comprehensive AI-ready dataset resource to date; (2) PepBenchPipeline, a standardized preprocessing pipeline that ensures consistent dataset cleaning, construction, splitting, and feature transformation, mitigating quality issues common in ad hoc pipelines; and (3) PepBenchLeaderboard, a unified evaluation protocol and leaderboard with strong baselines across 4 major methodological families: Fingerprint-based, GNN-based, PLM-based, and SMILES-based models. Together, PepBenchmark provides the first standardized and comparable foundation for peptide drug discovery, facilitating methodological advances and translation into real-world applications. The data and code are publicly available at https://github.com/ZGCI-AI4S-Pep/PepBenchmark/.", "date": "2026-04-12", "categories": []}
{"id": "2604.10970", "title": "Using Deep Learning Models Pretrained by Self-Supervised Learning for Protein Localization", "abstract": "Background: Task-specific microscopy datasets are often small, making it difficult to train deep learning models that learn robust features. While self-supervised learning (SSL) has shown promise through pretraining on large, domain-specific datasets, generalizability across datasets with differing staining protocols and channel configurations remains underexplored. We investigated the generalizability of SSL models pretrained on ImageNet-1k and HPA FOV, evaluating their embeddings on OpenCell with and without fine-tuning, two channel-mismatch strategies, and varying fine-tuning data fractions. We additionally analyzed single-cell embeddings on a labeled OpenCell subset.\n  Result: DINO-based ViT backbones pretrained on HPA FOV or ImageNet-1k transfer well to OpenCell even without fine-tuning. The HPA FOV-pretrained model achieved the highest zero-shot performance (macro $F_1$ 0.822 $\\pm$ 0.007). Fine-tuning further improved performance to 0.860 $\\pm$ 0.013. At the single-cell level, the HPA single-cell-pretrained model achieved the highest k-nearest neighbor performance across all neighborhood sizes (macro $F_1$ $\\geq$ 0.796).\n  Conclusion: SSL methods like DINO, pretrained on large domain-relevant datasets, enable effective use of deep learning features for fine-tuning on small, task-specific microscopy datasets.", "date": "2026-04-13", "categories": []}
{"id": "2604.11272", "title": "AbLWR:A Context-Aware Listwise Ranking Framework for Antibody-Antigen Binding Affinity Prediction via Positive-Unlabeled Learning", "abstract": "Accurate prediction of antibody-antigen binding affinity is fundamental to therapeutic design, yet remains constrained by severe label sparsity and the complexity of antigenic variations. In this paper, we propose AbLWR (Antibody-antigen binding affinity List-Wise Ranking), a novel framework that reformulates the conventional affinity regression task as a listwise ranking problem. To mitigate label sparsity, AbLWR incorporates a PU (Positive-Unlabeled) learning mechanism leveraging a dual-level contrastive objective and meta-optimized label refinement to learn robust representations. Furthermore, we address antigenic variation by employing a homologous antigen sampling strategy where Multi-Head Self-Attention (MHSA) explicitly models inter-sample relationships within training lists to capture subtle affinity nuances. Extensive experiments demonstrate that AbLWR significantly outperforms state-of-the-art baselines, improving the Precision@1 (P@1) by over 10$\\%$ in randomized cross-validation experiments. Notably, case studies on Influenza and IL-33 validate its practical utility, demonstrating robust ranking consistency in distinguishing subtle viral mutations and efficiently prioritizing top-tier candidates for wet-lab screening.", "date": "2026-04-13", "categories": []}
{"id": "2602.10603", "title": "dnaHNet: A Scalable and Hierarchical Foundation Model for Genomic Sequence Learning", "abstract": "Genomic foundation models have the potential to decode DNA syntax, yet face a fundamental tradeoff in their input representation. Standard fixed-vocabulary tokenizers fragment biologically meaningful motifs such as codons and regulatory elements, while nucleotide-level models preserve biological coherence but incur prohibitive computational costs for long contexts. We introduce dnaHNet, a state-of-the-art tokenizer-free autoregressive model that segments and models genomic sequences end-to-end. Using a differentiable dynamic chunking mechanism, dnaHNet compresses raw nucleotides into latent tokens adaptively, balancing compression with predictive accuracy. Pretrained on prokaryotic genomes, dnaHNet outperforms leading architectures including StripedHyena2 in scaling and efficiency. This recursive chunking yields quadratic FLOP reductions, enabling $>3 \\times$ inference speedup over Transformers. On zero-shot tasks, dnaHNet achieves superior performance in predicting protein variant fitness and gene essentiality, while automatically discovering hierarchical biological structures without supervision. These results establish dnaHNet as a scalable, interpretable framework for next-generation genomic modeling.", "date": "2026-02-11", "categories": []}
{"id": "2604.08698", "title": "EvoLen: Evolution-Guided Tokenization for DNA Language Model", "abstract": "Tokens serve as the basic units of representation in DNA language models (DNALMs), yet their design remains underexplored. Unlike natural language, DNA lacks inherent token boundaries or predefined compositional rules, making tokenization a fundamental modeling decision rather than a naturally specified one. While existing approaches like byte-pair encoding (BPE) excel at capturing token structures that reflect human-generated linguistic regularities, DNA is organized by biological function and evolutionary constraint rather than linguistic convention. We argue that DNA tokenization should prioritize functional sequence patterns like regulatory motifs-short, recurring segments under evolutionary constraint and typically preserved across species. We incorporate evolutionary information directly into the tokenization process through EvoLen, a tokenizer that combines evolutionary stratification with length-aware decoding to better preserve motif-scale functional sequence units. EvoLen uses cross-species evolutionary signals to group DNA sequences, trains separate BPE tokenizers on each group, merges the resulting vocabularies via a rule prioritizing preserved patterns, and applies length-aware decoding with dynamic programming. Through controlled experiments, EvoLen improves the preservation of functional sequence patterns, differentiation across genomic contexts, and alignment with evolutionary constraint, while matching or outperforming standard BPE across diverse DNALM benchmarks. These results demonstrate that tokenization introduces a critical inductive bias and that incorporating evolutionary information yields more biologically meaningful and interpretable sequence representations.", "date": "2026-04-09", "categories": []}
{"id": "2604.09369", "title": "Biologically-Grounded Multi-Encoder Architectures as Developability Oracles for Antibody Design", "abstract": "Generative models can now propose thousands of \\emph{de novo} antibody sequences, yet translating these designs into viable therapeutics remains constrained by the cost of biophysical characterization. Here we present CrossAbSense, a framework of property-specific neural oracles that combine frozen protein language model encoders with configurable attention decoders, identified through a systematic hyperparameter campaign totaling over 200 runs per property. On the GDPa1 benchmark of 242 therapeutic IgGs, our oracles achieve notable improvements of 12--20\\% over established baselines on three of five developability assays and competitive performance on the remaining two. The central finding is that optimal decoder architectures \\emph{invert} our initial biological hypotheses: self-attention alone suffices for aggregation-related properties (hydrophobic interaction chromatography, polyreactivity), where the relevant sequence signatures -- such as CDR-H3 hydrophobic patches -- are already fully resolved within single-chain embeddings by the high-capacity 6B encoder. Bidirectional cross-attention, by contrast, is required for expression yield and thermal stability -- properties that inherently depend on the compatibility between heavy and light chains. Learned chain fusion weights independently confirm heavy-chain dominance in aggregation ($w_H = 0.62$) versus balanced contributions for stability ($w_H = 0.51$). We demonstrate practical utility by deploying CrossAbSense on 100 IgLM-generated antibody designs, illustrating a path toward substantial reduction in experimental screening costs.", "date": "2026-04-10", "categories": []}
{"id": "2504.06316", "title": "GraphGDel: Constructing and Learning Graph Representations of Genome-Scale Metabolic Models for Growth-Coupled Gene Deletion Prediction", "abstract": "In genome-scale constraint-based metabolic models, gene deletion strategies are essential for achieving growth-coupled production, where cell growth and target metabolite synthesis occur simultaneously. Despite the inherently networked nature of genome-scale metabolic models, existing computational approaches rely primarily on sequential data and lack graph representations that capture their complex relationships, as both well-defined graph constructions and learning frameworks capable of exploiting them remain largely unexplored. To address this gap, we present a twofold solution. First, we introduce a systematic pipeline for constructing graph representations from constraint-based metabolic models. Second, we develop a deep learning framework that integrates these graph representations with gene and metabolite sequence data to predict growth-coupled gene deletion strategies. Across three metabolic models, our approach consistently outperforms established baselines, with improvements in overall accuracy of 14.04%, 16.26%, and 13.18% over a deep feedforward neural network baseline, 6.17%, 4.96%, and 5.31% over a sequence-learning baseline, and 5.10%, 4.36%, and 4.70% over a topology-aware graph aggregation baseline on the same metabolite graph, respectively. The source code and example datasets are available at: https://github.com/MetNetComp/GraphGDel.", "date": "2025-04-08", "categories": []}
{"id": "2505.00017", "title": "ReCellTy: Domain-Specific Knowledge Graph Retrieval-Augmented LLMs Reasoning Workflow for Single-Cell Annotation", "abstract": "With the rapid development of large language models (LLMs), their application to cell type annotation has drawn increasing attention. However, general-purpose LLMs often face limitations in this specific task due to the lack of guidance from external domain knowledge. To enable more accurate and fully automated cell type annotation, we develop a globally connected knowledge graph comprising 18850 biological information nodes, including cell types, gene markers, features, and other related entities, along with 48,944 edges connecting these nodes, which is used by LLMs to retrieve entities associated with differential genes for cell reconstruction. Additionally, a multi-task reasoning workflow is designed to optimise the annotation process. Compared to general-purpose LLMs, our method improves human evaluation scores by up to 0.21 and semantic similarity by 6.1% across multiple tissue types, while more closely aligning with the cognitive logic of manual annotation. Meanwhile, it narrows the performance gap between large and small LLMs in cell type annotation, offering a paradigm for structured knowledge integration and reasoning in bioinformatics.", "date": "2025-04-24", "categories": []}
{"id": "2603.11703", "title": "EvoFlows: Evolutionary Edit-Based Flow-Matching for Protein Engineering", "abstract": "We introduce EvoFlows, a variable-length protein sequence-to-sequence modeling approach designed for protein engineering. Existing protein language models are poorly suited for optimization tasks: autoregressive models require full sequence generation, masked language and discrete diffusion models rely on pre-specified mutation locations, and no existing methods naturally support insertions and deletions relative to a template sequence. EvoFlows learns mutational trajectories between evolutionarily related protein sequences via edit flows, allowing it to perform a controllable number of mutations (insertions, deletions, and substitutions) on a template sequence, predicting not only _which_ mutation to perform, but also _where_ it should occur. Through extensive _in silico_ evaluation on diverse protein families from UniRef and OAS, we show that EvoFlows generates variants that remain consistent with natural protein families while exploring farther from template sequences than leading baselines.", "date": "2026-03-12", "categories": []}
{"id": "2604.07276", "title": "Making Room for AI: Multi-GPU Molecular Dynamics with Deep Potentials in GROMACS", "abstract": "GROMACS is a de-facto standard for classical Molecular Dynamics (MD). The rise of AI-driven interatomic potentials that pursue near-quantum accuracy at MD throughput now poses a significant challenge: embedding neural-network inference into multi-GPU simulations retaining high-performance. In this work, we integrate the MLIP framework DeePMD-kit into GROMACS, enabling domain-decomposed, GPU-accelerated inference across multi-node systems. We extend the GROMACS NNPot interface with a DeePMD backend, and we introduce a domain decomposition layer decoupled from the main simulation. The inference is executed concurrently on all processes, with two MPI collectives used each step to broadcast coordinates and to aggregate and redistribute forces. We train an in-house DPA-1 model (1.6 M parameters) on a dataset of solvated protein fragments. We validate the implementation on a small protein system, then we benchmark the GROMACS-DeePMD integration with a 15,668 atom protein on NVIDIA A100 and AMD MI250x GPUs up to 32 devices. Strong-scaling efficiency reaches 66% at 16 devices and 40% at 32; weak-scaling efficiency is 80% to 16 devices and reaches 48% (MI250x) and 40% (A100) at 32 devices. Profiling with the ROCm System profiler shows that >90% of the wall time is spent in DeePMD inference, while MPI collectives contribute <10%, primarily since they act as a global synchronization point. The principal bottlenecks are the irreducible ghost-atom cost set by the cutoff radius, confirmed by a simple throughput model, and load imbalance across ranks. These results demonstrate that production MD with near ab initio fidelity is feasible at scale in GROMACS.", "date": "2026-04-08", "categories": []}
{"id": "2604.04403", "title": "MolDA: Molecular Understanding and Generation via Large Language Diffusion Model", "abstract": "Large Language Models (LLMs) have significantly advanced molecular discovery, but existing multimodal molecular architectures fundamentally rely on autoregressive (AR) backbones. This strict left-to-right inductive bias is sub-optimal for generating chemically valid molecules, as it struggles to account for non-local global constraints (e.g., ring closures) and often accumulates structural errors during sequential generation. To address these limitations, we propose MolDA (Molecular language model with masked Diffusion with mAsking), a novel multimodal framework that replaces the conventional AR backbone with a discrete Large Language Diffusion Model. MolDA extracts comprehensive structural representations using a hybrid graph encoder, which captures both local and global topologies, and aligns them into the language token space via a Q-Former. Furthermore, we mathematically reformulate Molecular Structure Preference Optimization specifically for the masked diffusion. Through bidirectional iterative denoising, MolDA ensures global structural coherence, chemical validity, and robust reasoning across molecule generation, captioning, and property prediction.", "date": "2026-04-06", "categories": []}
{"id": "2604.05181", "title": "General Multimodal Protein Design Enables DNA-Encoding of Chemistry", "abstract": "Evolution is an extraordinary engine for enzymatic diversity, yet the chemistry it has explored remains a narrow slice of what DNA can encode. Deep generative models can design new proteins that bind ligands, but none have created enzymes without pre-specifying catalytic residues. We introduce DISCO (DIffusion for Sequence-structure CO-design), a multimodal model that co-designs protein sequence and 3D structure around arbitrary biomolecules, as well as inference-time scaling methods that optimize objectives across both modalities. Conditioned solely on reactive intermediates, DISCO designs diverse heme enzymes with novel active-site geometries. These enzymes catalyze new-to-nature carbene-transfer reactions, including alkene cyclopropanation, spirocyclopropanation, B-H, and C(sp$^3$)-H insertions, with high activities exceeding those of engineered enzymes. Random mutagenesis of a selected design further confirmed that enzyme activity can be improved through directed evolution. By providing a scalable route to evolvable enzymes, DISCO broadens the potential scope of genetically encodable transformations. Code is available at https://github.com/DISCO-design/DISCO.", "date": "2026-04-06", "categories": []}
{"id": "2604.05775", "title": "PhageBench: Can LLMs Understand Raw Bacteriophage Genomes?", "abstract": "Bacteriophages, often referred to as the dark matter of the biosphere, play a critical role in regulating microbial ecosystems and in antibiotic alternatives. Thus, accurate interpretation of their genomes holds significant scientific and practical value. While general-purpose Large Language Models (LLMs) excel at understanding biological texts, their ability to directly interpret raw nucleotide sequences and perform biological reasoning remains underexplored. To address this, we introduce PhageBench, the first benchmark designed to evaluate phage genome understanding by mirroring the workflow of bioinformatics experts. The dataset contains 5,600 high-quality samples covering five core tasks across three stages: Screening, Quality Control, and Phenotype Annotation. Our evaluation of eight LLMs reveals that general-purpose reasoning models significantly outperform random baselines in phage contig identification and host prediction, demonstrating promising potential for genomic understanding. However, they exhibit significant limitations in complex reasoning tasks involving long-range dependencies and fine-grained functional localization. These findings highlight the necessity of developing next-generation models with enhanced reasoning capabilities for biological sequences.", "date": "2026-04-07", "categories": []}
{"id": "2604.03630", "title": "A Multimodal Foundation Model of Spatial Transcriptomics and Histology for Biological Discovery and Clinical Prediction", "abstract": "Spatial transcriptomics (ST) enables gene expression mapping within anatomical context but remains costly and low-throughput. Hematoxylin and eosin (H\\&E) staining offers rich morphology yet lacks molecular resolution. We present \\textbf{\\ours} (\\textbf{S}patial \\textbf{T}ranscriptomics and hist\\textbf{O}logy \\textbf{R}epresentation \\textbf{M}odel), a foundation model trained on 1.2 million spatially resolved transcriptomic profiles with matched histology across 18 organs. Using a hierarchical architecture integrating morphological features, gene expression, and spatial context, STORM bridges imaging and omics through robust molecular--morphological representations. STORM enhances spatial domain discovery, producing biologically coherent tissue maps, and outperforms existing methods in predicting spatial gene expression from H\\&E images across 11 tumor types. The model is platform-agnostic, performing consistently across Visium, Xenium, Visium HD, and CosMx. Applied to 23 independent cohorts comprising 7,245 patients, STORM significantly improves immunotherapy response prediction and prognostication over established biomarkers, providing a scalable framework for spatially informed discovery and clinical precision medicine.", "date": "2026-04-04", "categories": []}
{"id": "2604.04677", "title": "Towards protein folding pathways by reconstructing protein residue networks with a policy-driven model", "abstract": "A method that reconstructs protein residue networks using suitable node selection and edge recovery policies produced numerical observations that correlate strongly (Pearson's correlation coefficient < -0.83) with published folding rates for 52 two-state folders and 21 multi-state folders; correlations are also strong at the fold-family level. These results were obtained serendipitously with the ND model, which was introduced previously, but is here extended with policies that dictate actions according to feature states. This result points to the importance of both the starting search point and the prevailing condition (random seed) for the quick success of policy search by a simple hill-climber. The two conditions, suitable policies and random seed, which (evidenced by the strong correlation statistic) setup a conducive environment for modelling protein folding within ND, could be compared to appropriate physiological conditions required by proteins to fold naturally. Of interest is an examination of the sequence of restored edges for potential as plausible protein folding pathways. Towards this end, trajectory data is collected for analysis and further model evaluation and development.", "date": "2026-04-06", "categories": []}
{"id": "2510.17569", "title": "Towards best practices in low-dimensional semi-supervised latent Bayesian optimization for the design of antimicrobial peptides", "abstract": "Generative deep learning techniques have demonstrated an impressive capacity for tackling biomolecular design problems in recent years. Despite their high performance, however, they still suffer from a lack of interpretability and rigorous quantification of associated search spaces, which are necessary to unlock their full potential for scientific inquiry beyond efficient design. An area in which they are of particular interest is in the design of antimicrobial peptides, which are a promising class of therapeutics to treat bacterial infections. Discovering and designing such peptides is difficult because of the vast number of possible sequences and comparatively small amount of experimental information. In this work, we perform a theoretical investigation of latent Bayesian optimization for searching through peptide sequence spaces, with a focus on antimicrobial peptides. We investigate (1) whether searching through a dimensionally-reduced variant of the latent design space may facilitate optimization, (2) how organizing latent spaces by differing amounts of more and less relevant information may improve the efficiency of arriving at an optimal peptide design, and (3) the interpretability of the spaces. We find that employing a dimensionally-reduced version of the latent space is more interpretable and can be advantageous, while the use of less-relevant but more easily-computable physicochemical properties is advantageous to latent space organization in certain contexts and the use of more-relevant but sparser properties associated with the latent Bayesian objective function is advantageous in others. This work lays crucial groundwork for biophysically-motivated peptide design procedures, with an especial focus on antimicrobial peptides.", "date": "2025-10-20", "categories": []}
{"id": "2604.01949", "title": "annbatch unlocks terabyte-scale training of biological data in anndata", "abstract": "The scale of biological datasets now routinely exceeds system memory, making data access rather than model computation the primary bottleneck in training machine-learning models. This bottleneck is particularly acute in biology, where widely used community data formats must support heterogeneous metadata, sparse and dense assays, and downstream analysis within established computational ecosystems. Here we present annbatch, a mini-batch loader native to anndata that enables out-of-core training directly on disk-backed datasets. Across single-cell transcriptomics, microscopy and whole-genome sequencing benchmarks, annbatch increases loading throughput by up to an order of magnitude and shortens training from days to hours, while remaining fully compatible with the scverse ecosystem. Annbatch establishes a practical data-loading infrastructure for scalable biological AI, allowing increasingly large and diverse datasets to be used without abandoning standard biological data formats. Github: https://github.com/scverse/annbatch", "date": "2026-04-02", "categories": []}
{"id": "2510.16082", "title": "BIOGEN: Evidence-Grounded Multi-Agent Reasoning Framework for Transcriptomic Interpretation in Antimicrobial Resistance", "abstract": "Interpreting gene clusters from RNA sequencing (RNA-seq) remains challenging, especially in antimicrobial resistance studies where mechanistic insight is important for hypothesis generation. Existing pathway enrichment methods can summarize co-expressed modules, but they often provide limited cluster-specific explanations and weak connections to supporting literature. We present BIOGEN, an evidence-grounded multi-agent framework for post hoc interpretation of RNA-seq transcriptional modules. BIOGEN combines biomedical retrieval, structured reasoning, and multi-critic verification to generate traceable cluster-level explanations with explicit evidence and confidence labels. On a primary Salmonella enterica dataset, BIOGEN achieved strong biological grounding, including BERTScore 0.689, Semantic Alignment Score 0.715, KEGG Functional Similarity 0.342, and a hallucination rate of 0.000, compared with 0.100 for an LLM-only baseline. Across four additional bacterial RNA-seq datasets, BIOGEN also maintained zero hallucination under the same fixed pipeline. In comparisons with representative open-source agentic AI baselines, BIOGEN was the only framework that consistently preserved zero hallucination across all five datasets. These findings suggest that retrieval alone is not enough for reliable biological interpretation, and that evidence-grounded orchestration is important for transparent and source-traceable transcriptomic reasoning.", "date": "2025-10-17", "categories": []}
{"id": "2510.00512", "title": "Adaptive Data-Knowledge Alignment in Genetic Perturbation Prediction", "abstract": "The transcriptional response to genetic perturbation reveals fundamental insights into complex cellular systems. While current approaches have made progress in predicting genetic perturbation responses, they provide limited biological understanding and cannot systematically refine existing knowledge. Overcoming these limitations requires an end-to-end integration of data-driven learning and existing knowledge. However, this integration is challenging due to inconsistencies between data and knowledge bases, such as noise, misannotation, and incompleteness. To address this challenge, we propose ALIGNED (Adaptive aLignment for Inconsistent Genetic kNowledgE and Data), a neuro-symbolic framework based on the Abductive Learning (ABL) paradigm. This end-to-end framework aligns neural and symbolic components and performs systematic knowledge refinement. We introduce a balanced consistency metric to evaluate the predictions' consistency against both data and knowledge. Our results show that ALIGNED outperforms state-of-the-art methods by achieving the highest balanced consistency, while also re-discovering biologically meaningful knowledge. Our work advances beyond existing methods to enable both the transparency and the evolution of mechanistic biological understanding.", "date": "2025-10-01", "categories": []}
{"id": "2603.27048", "title": "MOOZY: A Patient-First Foundation Model for Computational Pathology", "abstract": "Computational pathology needs whole-slide image (WSI) foundation models that transfer across diverse clinical tasks, yet current approaches remain largely slide-centric, often depend on private data and expensive paired-report supervision, and do not explicitly model relationships among multiple slides from the same patient. We present MOOZY, a patient-first pathology foundation model in which the patient case, not the individual slide, is the core unit of representation. MOOZY explicitly models dependencies across all slides from the same patient via a case transformer during pretraining, combining multi-stage open self-supervision with scaled low-cost task supervision. In Stage 1, we pretrain a vision-only slide encoder on 77,134 public slide feature grids using masked self-distillation. In Stage 2, we align these representations with clinical semantics using a case transformer and multi-task supervision over 333 tasks from 56 public datasets, including 205 classification and 128 survival tasks across four endpoints. Across eight held-out tasks with five-fold frozen-feature probe evaluation, MOOZY achieves best or tied-best performance on most metrics and improves macro averages over TITAN by +7.37%, +5.50%, and +7.83% and over PRISM by +8.83%, +10.70%, and +9.78% for weighted F1, weighted ROC-AUC, and balanced accuracy, respectively. MOOZY is also parameter efficient with 85.77M parameters, 14x smaller than GigaPath. These results demonstrate that open, reproducible patient-level pretraining yields transferable embeddings, providing a practical path toward scalable patient-first histopathology foundation models.", "date": "2026-03-27", "categories": []}
{"id": "2604.00058", "title": "GenoBERT: A Language Model for Accurate Genotype Imputation", "abstract": "Genotype imputation enables dense variant coverage for genome-wide association and risk-prediction studies, yet conventional reference-panel methods remain limited by ancestry bias and reduced rare-variant accuracy. We present Genotype Bidirectional Encoder Representations from Transformers (GenoBERT), a transformer-based, reference-free framework that tokenizes phased genotypes and uses a self-attention mechanism to capture both short- and long-range linkage disequilibrium (LD) dependencies. Benchmarking on two independent datasets including the Louisiana Osteoporosis Study (LOS) and the 1000 Genomes Project (1KGP) across ancestry groups and multiple genotype missingness levels (5-50%) shows that GenoBERT achieves the highest overall accuracy compared to four baseline methods (Beagle5.4, SCDA, BiU-Net, and STICI). At practical sparsity levels (up to 25% missing), GenoBERT attains high overall imputation accuracy ($r^2 approx 0.98$) across datasets, and maintains robust performance ($r^2 > 0.90$) even at 50% missingness. Experimental results across different ancestries confirm consistent gains across datasets, with resilience to small sample sizes and weak LD. A 128-SNP (single-nucleotide polymorphism) context window (approximately 100 Kb) is validated through LD-decay analyses as sufficient to capture local correlation structures. By eliminating reference-panel dependence while preserving high accuracy, GenoBERT provides a scalable and robust solution for genotype imputation and a foundation for downstream genomic modeling.", "date": "2026-03-31", "categories": []}
{"id": "2604.00065", "title": "Genetic algorithms for multi-omic feature selection: a comparative study in cancer survival analysis", "abstract": "Multi-omic datasets offer opportunities for improved biomarker discovery in cancer research, but their high dimensionality and limited sample sizes make identifying compact and effective biomarker panels challenging. Feature selection in large-scale omics can be efficiently addressed by combining machine learning with genetic algorithms, which naturally support multi-objective optimization of predictive accuracy and biomarker set size. However, genetic algorithms remain relatively underexplored for multi-omic feature selection, where most approaches concatenate all layers into a single feature space. To address this limitation, we introduce Sweeping*, a multi-view, multi-objective algorithm alternating between single- and multi-view optimization. It employs a nested single-view multi-objective optimizer, and for this study we use the genetic algorithm NSGA3-CHS. It first identifies informative biomarkers within each layer, then jointly evaluates cross-layer interactions; these multi-omic solutions guide the next single-view search. Through repeated sweeps, the algorithm progressively identifies compact biomarker panels capturing cross-modal complementary signals. We benchmark five Sweeping* strategies, including hierarchical and concatenation-based variants, using survival prediction on three TCGA cohorts. Each strategy jointly optimizes predictive accuracy and set size, measured via the concordance index and root-leanness. Overall performance and estimation error are assessed through cross hypervolume and Pareto delta under 5-fold cross-validation. Our results show that Sweeping* can improve the accuracy-complexity trade-off when sufficient survival signal is present and that integrating omic layers can enhance survival prediction beyond clinical-only models, although benefits remain cohort-dependent.", "date": "2026-03-31", "categories": []}
{"id": "2604.00580", "title": "Representation choice shapes the interpretation of protein conformational dynamics", "abstract": "Molecular dynamics simulations provide detailed trajectories at the atomic level, but extracting interpretable and robust insights from these high-dimensional data remains challenging. In practice, analyses typically rely on a single representation. Here, we show that representation choice is not neutral: it fundamentally shapes the conformational organization, similarity relationships, and apparent transitions inferred from identical simulation data.\n  To complement existing representations, we introduce Orientation features, a geometrically grounded, rotation-aware encoding of protein backbone. We compare it against common descriptions across three dynamical regimes: fast-folding proteins, large-scale domain motions, and protein-protein association. Across these systems, we find that different representations emphasize complementary aspects of conformational space, and that no single representation provides a complete picture of the underlying dynamics.\n  To facilitate systematic comparison, we developed ManiProt, a library for efficient computation and analysis of multiple protein representations. Our results motivate a comparative, representation-aware framework for the interpretation of molecular dynamics simulations.", "date": "2026-04-01", "categories": []}
{"id": "2604.00739", "title": "BioCOMPASS: Integrating Biomarkers into Transformer-Based Immunotherapy Response Prediction", "abstract": "Datasets used in immunotherapy response prediction are typically small in size, as well as diverse in cancer type, drug administered, and sequencer used. Models often drop in performance when tested on patient cohorts that are not included in the training process. Recent work has shown that transformer-based models along with self-supervised learning show better generalisation performance than threshold-based biomarkers, but is still suboptimal. We present BioCOMPASS, an extension of a transformer-based model called COMPASS, that integrates biomarkers and treatment information to further improve its generalisability. Instead of feeding biomarker data as input, we built loss components to align them with the model's intermediate representations. We found that components such as treatment gating and pathway consistency loss improved generalisability when evaluated with Leave-one-cohort-out, Leave-one-cancer-type-out and Leave-one-treatment-out strategies. Results show that building components that exploit biomarker and treatment information can help in generalisability of immunotherapy response prediction. Careful curation of additional components that leverage complementary clinical information and domain knowledge represents a promising direction for future research.", "date": "2026-04-01", "categories": []}
{"id": "2604.00911", "title": "Event Embedding of Protein Networks : Compositional Learning of Biological Function", "abstract": "In this work, we study whether enforcing strict compositional structure in sequence embeddings yields meaningful geometric organization when applied to protein-protein interaction networks. Using Event2Vec, an additive sequence embedding model, we train 64-dimensional representations on random walks from the human STRING interactome, and compare against a DeepWalk baseline based on Word2Vec, trained on the same walks. We find that compositional structure substantially improves pathway coherence (30.2$\\times$ vs 2.9$\\times$ above random), functional analogy accuracy (mean similarity 0.966 vs 0.650), and hierarchical pathway organization, while geometric properties such as norm--degree anticorrelation are shared with or exceeded by the non-compositional baseline. These results indicate that enforced compositionality specifically benefits relational and compositional reasoning tasks in biological networks.", "date": "2026-04-01", "categories": []}
{"id": "2509.20702", "title": "Incorporating LLM Embeddings for Variation Across the Human Genome", "abstract": "Recent advances in large language model (LLM) embeddings have enabled powerful representations for biological data, but most applications to date focus on gene-level information. We present one of the first systematic frameworks to generate genetic variant-level embeddings across the entire human genome. Using curated annotations from FAVOR, ClinVar, and the GWAS Catalog, we construct functional text descriptions for 8.9 billion possible variants and generated embeddings at three scales: 1.5 million HapMap3/MEGA variants, 90 million imputed UK Biobank (UKB) variants, and 9 billion all possible variants. Embeddings were produced using general purpose models including both OpenAI's text-embedding-3-large and the open-source Qwen3-Embedding-0.6B models. Baseline quality control experiments demonstrate high predictive accuracy for variant-level properties, validating the embeddings as structured representations of genomic variation. We further apply them to real-world embedding-augmented genetic risk predictions that demonstrate the performance of using LLM embeddings in polygenic risk score (PRS) style predictions over the UK Biobank cohort data. These resources, publicly available on Hugging Face, provide a foundation for advancing large-scale genomic discovery and precision medicine.", "date": "2025-09-25", "categories": []}
{"id": "2603.25240", "title": "Lingshu-Cell: A generative cellular world model for transcriptome modeling toward virtual cells", "abstract": "Modeling cellular states and predicting their responses to perturbations are central challenges in computational biology and the development of virtual cells. Existing foundation models for single-cell transcriptomics provide powerful static representations, but they do not explicitly model the distribution of cellular states for generative simulation. Here, we introduce Lingshu-Cell, a masked discrete diffusion model that learns transcriptomic state distributions and supports conditional simulation under perturbation. By operating directly in a discrete token space that is compatible with the sparse, non-sequential nature of single-cell transcriptomic data, Lingshu-Cell captures complex transcriptome-wide expression dependencies across approximately 18,000 genes without relying on prior gene selection, such as filtering by high variability or ranking by expression level. Across diverse tissues and species, Lingshu-Cell accurately reproduces transcriptomic distributions, marker-gene expression patterns and cell-subtype proportions, demonstrating its ability to capture complex cellular heterogeneity. Moreover, by jointly embedding cell type or donor identity with perturbation, Lingshu-Cell can predict whole-transcriptome expression changes for novel combinations of identity and perturbation. It achieves leading performance on the Virtual Cell Challenge H1 genetic perturbation benchmark and in predicting cytokine-induced responses in human PBMCs. Together, these results establish Lingshu-Cell as a flexible cellular world model for in silico simulation of cell states and perturbation responses, laying the foundation for a new paradigm in biological discovery and perturbation screening.", "date": "2026-03-26", "categories": []}
{"id": "2603.26790", "title": "Elucidating the Design Space of Flow Matching for Cellular Microscopy", "abstract": "Flow-matching generative models are increasingly used to simulate cell responses to biological perturbations. However, the design space for building such models is large and underexplored. We systematically analyse the design space of flow matching models for cell-microscopy images, finding that many popular techniques are unnecessary and can even hurt performance. We develop a simple, stable, and scalable recipe which we use to train our foundation model. We scale our model to two orders of magnitude larger than prior methods, achieving a two-fold FID and ten-fold KID improvement over prior methods. We then fine-tune our model with pre-trained molecular embeddings to achieve state-of-the-art performance simulating responses to unseen molecules.\n  Code is available at https://github.com/valence-labs/microscopy-flow-matching", "date": "2026-03-25", "categories": []}
{"id": "2603.26827", "title": "Central-to-Local Adaptive Generative Diffusion Framework for Improving Gene Expression Prediction in Data-Limited Spatial Transcriptomics", "abstract": "Spatial Transcriptomics (ST) provides spatially resolved gene expression profiles within intact tissue architecture, enabling molecular analysis in histological context. However, the high cost, limited throughput, and restricted data sharing of ST experiments result in severe data scarcity, constraining the development of robust computational models. To address this limitation, we present a Central-to-Local adaptive generative diffusion framework for ST (C2L-ST) that integrates large-scale morphological priors with limited molecular guidance. A global central model is first pretrained on extensive histopathology datasets to learn transferable morphological representations, and institution-specific local models are then adapted through lightweight gene-conditioned modulation using a small number of paired image-gene spots. This strategy enables the synthesis of realistic and molecularly consistent histology patches under data-limited conditions. The generated images exhibit high visual and structural fidelity, reproduce cellular composition, and show strong embedding overlap with real data across multiple organs, reflecting both realism and diversity. When incorporated into downstream training, synthetic image-gene pairs improve gene expression prediction accuracy and spatial coherence, achieving performance comparable to real data while requiring only a fraction of sampled spots. C2L-ST provides a scalable and data-efficient framework for molecular-level data augmentation, offering a domain-adaptive and generalizable approach for integrating histology and transcriptomics in spatial biology and related fields.", "date": "2026-03-27", "categories": []}
{"id": "2603.26994", "title": "ImmSET: Sequence-Based Predictor of TCR-pMHC Specificity at Scale", "abstract": "T cells are a critical component of the adaptive immune system, playing a role in infectious disease, autoimmunity, and cancer. T cell function is mediated by the T cell receptor (TCR) protein, a highly diverse receptor targeting specific peptides presented by the major histocompatibility complex (pMHCs). Predicting the specificity of TCRs for their cognate pMHCs is central to understanding adaptive immunity and enabling personalized therapies. However, accurate prediction of this protein-protein interaction remains challenging due to the extreme diversity of both TCRs and pMHCs. Here, we present ImmSET (Immune Synapse Encoding Transformer), a novel sequence-based architecture designed to model interactions among sets of variable-length biological sequences. We train this model across a range of dataset sizes and compositions and study the resulting models' generalization to pMHC targets. We describe a failure mode in prior sequence-based approaches that inflates previously reported performance on this task and show that ImmSET remains robust under stricter evaluation. In systematically testing the scaling behavior of ImmSET with training data, we show that performance scales consistently with data volume across multiple data types and compares favorably with the pre-trained protein language model ESM2 fine-tuned on the same datasets. Finally, we demonstrate that ImmSET can outperform AlphaFold2 and AlphaFold3-based pipelines on TCR-pMHC specificity prediction when provided sufficient training data. This work establishes ImmSET as a scalable modeling paradigm for multi-sequence interaction problems, demonstrated in the TCR-pMHC setting but generalizable to other biological domains where high-throughput sequence-driven reasoning complements structure prediction and experimental mapping.", "date": "2026-03-27", "categories": []}
{"id": "2603.27145", "title": "Pan-Cancer Mapping of the Tumor Immune Landscape through Metagene Clustering and Predictive Modeling", "abstract": "As immunotherapies become standard cancer treatments, it is increasingly important to identify a patient's immune profile, which encompasses the activity of immune cells within the tumor microenvironment and the presence of specific biomarkers. However, we lack mechanistic explanations drivers of immune phenotypes. Despite advances in immune profiling with high-throughput sequencing, the mechanisms driving them remain unclear. This study aimed to identify novel, robust immune-related gene clusters (metagenes) and evaluate their prognostic significance and functional relevance across various pan-cancer types using a comprehensive computational pipeline. We acquired pan-cancer bulk RNA-seq and established immune subtypes from The Cancer Genome Atlas (TCGA). Using expression-based filtering and clustering of genes with ANOVA and Gaussian Mixture Model (GMM), we identified 48 unique metagenes. These metagenes achieved 87% accuracy in predicting the established subtypes. SHAP analysis revealed the most predictive metagenes per subtype, while functional enrichment analysis identified their associated pathways. Genes were ranked by differential expression between high- and low-expression groups. The metagenes revealed insights, including co-expression of immune activation and regulatory factors, links between cell cycle regulation and immune evasion, and dynamic microenvironment remodeling signatures. Kaplan-Meier survival analysis and multivariate Cox Regression revealed that many metagenes had prognostic value for overall survival. Overall, the metagenes represent coordinated biological programs across diverse cancer types, providing a foundation for developing robust, broadly applicable immuno-oncology biomarkers that extend beyond single-gene markers. They demonstrate prognostic value across cancer types and hold potential to guide immunotherapy treatment decisions.", "date": "2026-03-28", "categories": []}
{"id": "2603.27303", "title": "Self-evolving AI agents for protein discovery and directed evolution", "abstract": "Protein scientific discovery is bottlenecked by the manual orchestration of information and algorithms, while general agents are insufficient in complex domain projects. VenusFactory2 provides an autonomous framework that shifts from static tool usage to dynamic workflow synthesis via a self-evolving multi-agent infrastructure to address protein-related demands. It outperforms a set of well-known agents on the VenusAgentEval benchmark, and autonomously organizes the discovery and optimization of proteins from a single natural language prompt.", "date": "2026-03-28", "categories": []}
{"id": "2603.27526", "title": "Q-BIOLAT: Binary Latent Protein Fitness Landscapes for QUBO-Based Optimization", "abstract": "Protein fitness optimization is inherently a discrete combinatorial problem, yet most learning-based approaches rely on continuous representations and are primarily evaluated through predictive accuracy. We introduce Q-BIOLAT, a framework for modeling and optimizing protein fitness landscapes in compact binary latent spaces. Starting from pretrained protein language model embeddings, we construct binary latent representations and learn a quadratic unconstrained binary optimization (QUBO) surrogate that captures unary and pairwise interactions.\n  Beyond its formulation, Q-BIOLAT provides a representation-centric perspective on protein fitness modeling. We show that representations with similar predictive performance can induce fundamentally different optimization landscapes. In particular, learned autoencoder-based representations collapse after binarization, producing degenerate latent spaces that fail to support combinatorial search, whereas simple structured representations such as PCA yield high-entropy, decodable, and optimization-friendly latent spaces.\n  Across multiple datasets and data regimes, we demonstrate that classical combinatorial optimization methods, including simulated annealing, genetic algorithms, and greedy hill climbing, are highly effective in structured binary latent spaces. By expressing the objective in QUBO form, our approach connects modern machine learning with discrete and quantum-inspired optimization.\n  Our implementation and dataset are publicly available at: https://github.com/HySonLab/Q-BIOLAT-Extended", "date": "2026-03-29", "categories": []}
{"id": "2603.27950", "title": "Scaling Atomistic Protein Binder Design with Generative Pretraining and Test-Time Compute", "abstract": "Protein interaction modeling is central to protein design, which has been transformed by machine learning with applications in drug discovery and beyond. In this landscape, structure-based de novo binder design is cast as either conditional generative modeling or sequence optimization via structure predictors (\"hallucination\"). We argue that this is a false dichotomy and propose Proteina-Complexa, a novel fully atomistic binder generation method unifying both paradigms. We extend recent flow-based latent protein generation architectures and leverage the domain-domain interactions of monomeric computationally predicted protein structures to construct Teddymer, a new large-scale dataset of synthetic binder-target pairs for pretraining. Combined with high-quality experimental multimers, this enables training a strong base model. We then perform inference-time optimization with this generative prior, unifying the strengths of previously distinct generative and hallucination methods. Proteina-Complexa sets a new state of the art in computational binder design benchmarks: it delivers markedly higher in-silico success rates than existing generative approaches, and our novel test-time optimization strategies greatly outperform previous hallucination methods under normalized compute budgets. We also demonstrate interface hydrogen bond optimization, fold class-guided binder generation, and extensions to small molecule targets and enzyme design tasks, again surpassing prior methods. Code, models and new data will be publicly released.", "date": "2026-03-30", "categories": []}
{"id": "2509.24779", "title": "MarS-FM: Generative Modeling of Molecular Dynamics via Markov State Models", "abstract": "Molecular Dynamics (MD) is a powerful computational microscope for probing protein functions. However, the need for fine-grained integration and the long timescales of biomolecular events make MD computationally expensive. To address this, several generative models have been proposed to generate surrogate trajectories at lower cost. Yet, these models typically learn a fixed-lag transition density, causing the training signal to be dominated by frequent but uninformative transitions. We introduce a new class of generative models, MSM Emulators, which instead learn to sample transitions across discrete states defined by an underlying Markov State Model (MSM). We instantiate this class with Markov Space Flow Matching (MarS-FM), whose sampling offers more than two orders of magnitude speedup compared to implicit- or explicit-solvent MD simulations. We benchmark Mars-FM ability to reproduce MD statistics through structural observables such as RMSD, radius of gyration, and secondary structure content. Our evaluation spans protein domains (up to 500 residues) with significant chemical and structural diversity, including unfolding events, and enforces strict sequence dissimilarity between training and test sets to assess generalization. Across all metrics, MarS-FM outperforms existing methods, often by a substantial margin.", "date": "2025-09-29", "categories": []}
{"id": "2603.25855", "title": "Incorporating contextual information into KGWAS for interpretable GWAS discovery", "abstract": "Genome-Wide Association Studies (GWAS) identify associations between genetic variants and disease; however, moving beyond associations to causal mechanisms is critical for therapeutic target prioritization. The recently proposed Knowledge Graph GWAS (KGWAS) framework addresses this challenge by linking genetic variants to downstream gene-gene interactions via a knowledge graph (KG), thereby improving detection power and providing mechanistic insights. However, the original KGWAS implementation relies on a large general-purpose KG, which can introduce spurious correlations. We hypothesize that cell-type specific KGs from disease-relevant cell types will better support disease mechanism discovery. Here, we show that the general-purpose KG in KGWAS can be substantially pruned with no loss of statistical power on downstream tasks, and that performance further improves by incorporating gene-gene relationships derived from perturb-seq data. Importantly, using a sparse, context-specific KG from direct perturb-seq evidence yields more consistent and biologically robust disease-critical networks.", "date": "2026-03-26", "categories": []}
{"id": "2603.25880", "title": "Spectral Coherence Index: A Model-Free Metric for Protein Structural Ensemble Quality Assessment", "abstract": "Protein structural ensembles from NMR spectroscopy capture biologically important conformational heterogeneity, but it remains difficult to determine whether observed variation reflects coordinated motion or noise-like artifacts. We evaluate the Spectral Coherence Index (SCI), a model-free, rotation-invariant summary derived from the participation-ratio effective rank of the inter-model pairwise distance-variance matrix. Under grouped primary analysis of a Main110 cohort of 110 NMR ensembles (30--403 residues; 10--30 models per entry), SCI separated experimental ensembles from matched synthetic incoherent controls with AUC-ROC $= 0.973$ and Cliff's $\\delta = -0.945$. Relative to an internal 27-protein pilot, discrimination softened modestly, showing that pilot-era thresholds do not transfer perfectly to a larger, more heterogeneous cohort: the primary operating point $\\tau = 0.811$ yielded 95.5\\% sensitivity and 89.1\\% specificity. PDB-level sensitivity remained nearly unchanged (AUC $= 0.972$), and an independent 11-protein holdout reached AUC $= 0.983$. Across 5-fold grouped stratified cross-validation and leave-one-function-class-out testing, SCI remained strong (AUC $= 0.968$ and $0.971$), although $\\sigma_{R_g}$ was the stronger single-feature discriminator and a QC-augmented multifeature model generalized best (AUC $= 0.989$ and $0.990$). Residue-level validation linked SCI-derived contributions to experimental RMSF across 110 proteins and showed broad concordance with GNM-based flexibility patterns. Rescue analyses showed that Main110 softening arose mainly from size and ensemble normalization rather than from loss of spectral signal. Together, these results establish SCI as an interpretable, bounded coherence summary that is most useful when embedded in a multimetric QC workflow for heterogeneous protein ensembles.", "date": "2026-03-26", "categories": []}
{"id": "2603.26378", "title": "Generative Modeling in Protein Design: Neural Representations, Conditional Generation, and Evaluation Standards", "abstract": "Generative modeling has become a central paradigm in protein research, extending machine learning beyond structure prediction toward sequence design, backbone generation, inverse folding, and biomolecular interaction modeling. However, the literature remains fragmented across representations, model classes, and task formulations, making it difficult to compare methods or identify appropriate evaluation standards. This survey provides a systematic synthesis of generative AI in protein research, organized around (i) foundational representations spanning sequence, geometric, and multimodal encodings; (ii) generative architectures including $\\mathrm{SE}(3)$-equivariant diffusion, flow matching, and hybrid predictor-generator systems; and (iii) task settings from structure prediction and de novo design to protein-ligand and protein-protein interactions. Beyond cataloging methods, we compare assumptions, conditioning mechanisms, and controllability, and we synthesize evaluation best practices that emphasize leakage-aware splits, physical validity checks, and function-oriented benchmarks. We conclude with critical open challenges: modeling conformational dynamics and intrinsically disordered regions, scaling to large assemblies while maintaining efficiency, and developing robust safety frameworks for dual-use biosecurity risks. By unifying architectural advances with practical evaluation standards and responsible development considerations, this survey aims to accelerate the transition from predictive modeling to reliable, function-driven protein engineering.", "date": "2026-03-27", "categories": []}
{"id": "2506.14861", "title": "BMFM-RNA: whole-cell expression decoding improves transcriptomic foundation models", "abstract": "Transcriptomic foundation models pretrained with masked language modeling can achieve low pretraining loss yet produce poor cell representations for downstream tasks. We introduce whole-cell expression decoding (WCED), where models reconstruct the entire gene vocabulary from a single CLS token embedding, even with limited inputs, creating a maximally informative bottleneck. WCED consistently outperforms MLM on all downstream metrics despite higher reconstruction error during training. Gene-level error tracking reveals that both methods preferentially learn genes whose expression co-varies with stable transcriptional programs rather than those driven by transient factors. We further add hierarchical cross-entropy loss that exploits Cell Ontology structure for zero-shot annotation at multiple granularity levels. Models trained with these objectives achieve best overall performance across CZI benchmarks, on zero-shot batch integration and linear probing cell-type annotation. Methods are implemented in biomed-multi-omic ( https://github.com/BiomedSciAI/biomed-multi-omic ), an open-source framework for transcriptomic foundation model development.", "date": "2025-06-17", "categories": []}
{"id": "2511.01464", "title": "Split-Flows: Measure Transport and Information Loss Across Molecular Resolutions", "abstract": "By reducing resolution, coarse-grained models greatly accelerate molecular simulations, unlocking access to long-timescale phenomena, though at the expense of microscopic information. Recovering this fine-grained detail is essential for tasks that depend on atomistic accuracy, making backmapping a central challenge in molecular modeling. We introduce split-flows, a novel flow-based approach that reinterprets backmapping as a continuous-time measure transport across resolutions. Unlike existing generative strategies, split-flows establish a direct probabilistic link between resolutions, enabling expressive conditional sampling of atomistic structures and -- for the first time -- a tractable route to computing mapping entropies, an information-theoretic measure of the irreducible detail lost in coarse-graining. We demonstrate these capabilities on diverse molecular systems, including chignolin, a lipid bilayer, and alanine dipeptide, highlighting split-flows as a principled framework for accurate backmapping and systematic evaluation of coarse-grained models.", "date": "2025-11-03", "categories": []}
{"id": "2512.05245", "title": "STAR-GO: Improving Protein Function Prediction by Learning to Hierarchically Integrate Ontology-Informed Semantic Embeddings", "abstract": "Accurate prediction of protein function is essential for elucidating molecular mechanisms and advancing biological and therapeutic discovery. Yet experimental annotation lags far behind the rapid growth of protein sequence data. Computational approaches address this gap by associating proteins with Gene Ontology (GO) terms, which encode functional knowledge through hierarchical relations and textual definitions. However, existing models often emphasize one modality over the other, limiting their ability to generalize, particularly to unseen or newly introduced GO terms that frequently arise as the ontology evolves, and making the previously trained models outdated. We present STAR-GO, a Transformer-based framework that jointly models the semantic and structural characteristics of GO terms to enhance zero-shot protein function prediction. STAR-GO integrates textual definitions with ontology graph structure to learn unified GO representations, which are processed in hierarchical order to propagate information from general to specific terms. These representations are then aligned with protein sequence embeddings to capture sequence-function relationships. STAR-GO achieves state-of-the-art performance and superior zero-shot generalization, demonstrating the utility of integrating semantics and structure for robust and adaptable protein function prediction. Code is available at https://github.com/boun-tabi-lifelu/stargo.", "date": "2025-12-04", "categories": []}
{"id": "2603.23361", "title": "Central Dogma Transformer III: Interpretable AI Across DNA, RNA, and Protein", "abstract": "Biological AI models increasingly predict complex cellular responses, yet their learned representations remain disconnected from the\n  molecular processes they aim to capture. We present CDT-III, which extends mechanism-oriented AI across the full central dogma: DNA, RNA, and\n  protein. Its two-stage Virtual Cell Embedder architecture mirrors the spatial compartmentalization of the cell: VCE-N models transcription in\n  the nucleus and VCE-C models translation in the cytosol. On five held-out genes, CDT-III achieves per-gene RNA r=0.843 and protein r=0.969.\n  Adding protein prediction improves RNA performance (r=0.804 to 0.843), demonstrating that downstream tasks regularize upstream\n  representations. Protein supervision sharpens DNA-level interpretability, increasing CTCF enrichment by 30%. Analysis of experimentally\n  measured mRNA and protein responses reveals that the majority of genes with observable mRNA changes show opposite protein-level changes (66.7%\n  at |log2FC|>0.01, rising to 87.5% at |log2FC|>0.02), exposing a fundamental limitation of RNA-only perturbation models. Despite this\n  pervasive direction discordance, CDT-III correctly predicts both mRNA and protein responses. Applied to in silico CD52 knockdown approximating\n  Alemtuzumab, the model predicts 29/29 protein changes correctly and rediscovers 5 of 7 known clinical side effects without clinical data.\n  Gradient-based side effect profiling requires only unperturbed baseline data (r=0.939), enabling screening of all 2,361 genes without new\n  experiments.", "date": "2026-03-24", "categories": []}
{"id": "2511.04854", "title": "SigmaDock: Untwisting Molecular Docking With Fragment-Based SE(3) Diffusion", "abstract": "Determining the binding pose of a ligand to a protein, known as molecular docking, is a fundamental task in drug discovery. Generative approaches promise faster, improved, and more diverse pose sampling than physics-based methods, but are often hindered by chemically implausible outputs, poor generalisability, and high computational cost. To address these challenges, we introduce a novel fragmentation scheme, leveraging inductive biases from structural chemistry, to decompose ligands into rigid-body fragments. Building on this decomposition, we present SigmaDock, an SE(3) Riemannian diffusion model that generates poses by learning to reassemble these rigid bodies within the binding pocket. By operating at the level of fragments in SE(3), SigmaDock exploits well-established geometric priors while avoiding overly complex diffusion processes and unstable training dynamics. Experimentally, we show SigmaDock achieves state-of-the-art performance, reaching Top-1 success rates (RMSD<2 & PB-valid) above 79.9% on the PoseBusters set, compared to 12.7-30.8% reported by recent deep learning approaches, whilst demonstrating consistent generalisation to unseen proteins. SigmaDock is the first deep learning approach to surpass classical physics-based docking under the PB train-test split, marking a significant leap forward in the reliability and feasibility of deep learning for molecular modelling.", "date": "2025-11-06", "categories": []}
{"id": "2603.23583", "title": "ZeroFold: Protein-RNA Binding Affinity Predictions from Pre-Structural Embeddings", "abstract": "The accurate prediction of protein-RNA binding affinity remains an unsolved problem in structural biology, limiting opportunities in understanding gene regulation and designing RNA-targeting therapeutics. A central obstacle is the structural flexibility of RNA, as, unlike proteins, RNA molecules exist as dynamic conformational ensembles. Thus, committing to a single predicted structure discards information relevant to binding. Here, we show that this obstacle can be addressed by extracting pre-structural embeddings, which are intermediate representations from a biomolecular foundation model captured before the structure decoding step. Pre-structural embeddings implicitly encode conformational ensemble information without requiring predicted structures. We build ZeroFold, a transformer-based model that combines pre-structural embeddings from Boltz-2 for both protein and RNA molecules through a cross-modal attention mechanism to predict binding affinity directly from sequence. To support training and evaluation, we construct PRADB, a curated dataset of 2,621 unique protein-RNA pairs with experimentally measured affinities drawn from four complementary databases. On a held-out test set constructed with 40% sequence identity thresholds, ZeroFold achieves a Spearman correlation of 0.65, a value approaching the ceiling imposed by experimental measurement noise. Under progressively fairer evaluation conditions that control for training-set overlap, ZeroFold compares favourably with respect to leading structure-based and leading sequence-based predictors, with the performance gap widening as sequence similarity to competitor training data is reduced. These results illustrate how pre-structural embeddings offer a representation strategy for flexible biomolecules, opening a route to affinity prediction for protein-RNA pairs for which no structural data exist.", "date": "2026-03-24", "categories": []}
{"id": "2603.24265", "title": "DeepDTF: Dual-Branch Transformer Fusion for Multi-Omics Anticancer Drug Response Prediction", "abstract": "Cancer drug response varies widely across tumors due to multi-layer molecular heterogeneity, motivating computational decision support for precision oncology. Despite recent progress in deep CDR models, robust alignment between high-dimensional multi-omics and chemically structured drugs remains challenging due to cross-modal misalignment and limited inductive bias. We present DeepDTF, an end-to-end dual-branch Transformer fusion framework for joint log(IC50) regression and drug sensitivity classification. The cell-line branch uses modality-specific encoders for multi-omics profiles with Transformer blocks to capture long-range dependencies, while the drug branch represents compounds as molecular graphs and encodes them with a GNN-Transformer to integrate local topology with global context. Omics and drug representations are fused by a Transformer-based module that models cross-modal interactions and mitigates feature misalignment. On public pharmacogenomic benchmarks under 5-fold cold-start cell-line evaluation, DeepDTF consistently outperforms strong baselines across omics settings, achieving up to RMSE=1.248, R^2=0.875, and AUC=0.987 with full multi-omics inputs, while reducing classification error (1-ACC) by 9.5%. Beyond accuracy, DeepDTF provides biologically grounded explanations via SHAP-based gene attributions and pathway enrichment with pre-ranked GSEA.", "date": "2026-03-25", "categories": []}
{"id": "2412.05430", "title": "DART-Eval: A Comprehensive DNA Language Model Evaluation Benchmark on Regulatory DNA", "abstract": "Recent advances in self-supervised models for natural language, vision, and protein sequences have inspired the development of large genomic DNA language models (DNALMs). These models aim to learn generalizable representations of diverse DNA elements, potentially enabling various genomic prediction, interpretation and design tasks. Despite their potential, existing benchmarks do not adequately assess the capabilities of DNALMs on key downstream applications involving an important class of non-coding DNA elements critical for regulating gene activity. In this study, we introduce DART-Eval, a suite of representative benchmarks specifically focused on regulatory DNA to evaluate model performance across zero-shot, probed, and fine-tuned scenarios against contemporary ab initio models as baselines. Our benchmarks target biologically meaningful downstream tasks such as functional sequence feature discovery, predicting cell-type specific regulatory activity, and counterfactual prediction of the impacts of genetic variants. We find that current DNALMs exhibit inconsistent performance and do not offer compelling gains over alternative baseline models for most tasks, while requiring significantly more computational resources. We discuss potentially promising modeling, data curation, and evaluation strategies for the next generation of DNALMs. Our code is available at https://github.com/kundajelab/DART-Eval.", "date": "2024-12-06", "categories": []}
{"id": "2504.16956", "title": "GeneMamba: An Efficient and Effective Foundation Model on Single Cell Data", "abstract": "Single-cell RNA sequencing (scRNA-seq) enables high-resolution analysis of cellular heterogeneity, but its complexity, which is marked by high dimensionality, sparsity, and batch effects, which poses major computational challenges. Transformer-based models have made significant advances in this domain but are often limited by their quadratic complexity and suboptimal handling of long-range dependencies. In this work, we introduce GeneMamba, a scalable and efficient foundation model for single-cell transcriptomics built on state space modeling. Leveraging the Bi-Mamba architecture, GeneMamba captures bidirectional gene context with linear-time complexity, offering substantial computational gains over transformer baselines. The model is pretrained on nearly 30 million cells and incorporates biologically informed objectives, including pathway-aware contrastive loss and rank-based gene encoding. We evaluate GeneMamba across diverse tasks, including multi-batch integration, cell type annotation, and gene-gene correlation, demonstrating strong performance, interpretability, and robustness. These results position GeneMamba as a practical and powerful alternative to transformer-based methods, advancing the development of biologically grounded, scalable tools for large-scale single-cell data analysis.", "date": "2025-04-22", "categories": []}
{"id": "2603.22330", "title": "Fair splits flip the leaderboard: CHANRG reveals limited generalization in RNA secondary-structure prediction", "abstract": "Accurate prediction of RNA secondary structure underpins transcriptome annotation, mechanistic analysis of non-coding RNAs, and RNA therapeutic design. Recent gains from deep learning and RNA foundation models are difficult to interpret because current benchmarks may overestimate generalization across RNA families. We present the Comprehensive Hierarchical Annotation of Non-coding RNA Groups (CHANRG), a benchmark of 170{,}083 structurally non-redundant RNAs curated from more than 10 million sequences in Rfam~15.0 using structure-aware deduplication, genome-aware split design and multiscale structural evaluation. Across 29 predictors, foundation-model methods achieved the highest held-out accuracy but lost most of that advantage out of distribution, whereas structured decoders and direct neural predictors remained markedly more robust. This gap persisted after controlling for sequence length and reflected both loss of structural coverage and incorrect higher-order wiring. Together, CHANRG and a padding-free, symmetry-aware evaluation stack provide a stricter and batch-invariant framework for developing RNA structure predictors with demonstrable out-of-distribution robustness.", "date": "2026-03-20", "categories": []}
{"id": "2603.22369", "title": "SynLeaF: A Dual-Stage Multimodal Fusion Framework for Synthetic Lethality Prediction Across Pan- and Single-Cancer Contexts", "abstract": "Accurate prediction of synthetic lethality (SL) is important for guiding the development of cancer drugs and therapies. SL prediction faces significant challenges in the effective fusion of heterogeneous multi-source data. Existing multimodal methods often suffer from \"modality laziness\" due to disparate convergence speeds, which hinders the exploitation of complementary information. This is also one reason why most existing SL prediction models cannot perform well on both pan-cancer and single-cancer SL pair prediction. In this study, we propose SynLeaF, a dual-stage multimodal fusion framework for SL prediction across pan- and single-cancer contexts. The framework employs a VAE-based cross-encoder with a product of experts mechanism to fuse four omics data types (gene expression, mutation, methylation, and CNV), while simultaneously utilizing a relational graph convolutional network to capture structured gene representations from biomedical knowledge graphs. To mitigate modality laziness, SynLeaF introduces a dual-stage training mechanism employing featurelevel knowledge distillation with adaptive uni-modal teacher and ensemble strategies. In extensive experiments across eight specific cancer types and a pancancer dataset, SynLeaF achieves superior performance in 17 out of 19 scenarios. Ablation studies and gradient analyses further validate the critical contributions of the proposed fusion and distillation mechanisms to model robustness and generalization. To facilitate community use, a web server is available at https://synleaf.bioinformatics-lilab.cn.", "date": "2026-03-23", "categories": []}
{"id": "2405.14108", "title": "Assessing the potential of deep learning for protein-ligand docking", "abstract": "The effects of ligand binding on protein structures and their in vivo functions carry numerous implications for modern biomedical research and biotechnology development efforts such as drug discovery. Although several deep learning (DL) methods and benchmarks designed for protein-ligand docking have recently been introduced, to date no prior works have systematically studied the behavior of the latest docking and structure prediction methods within the broadly applicable context of (1) using predicted (apo) protein structures for docking (e.g., for applicability to new proteins); (2) binding multiple (cofactor) ligands concurrently to a given target protein (e.g., for enzyme design); and (3) having no prior knowledge of binding pockets (e.g., for generalization to unknown pockets). To enable a deeper understanding of docking methods' real-world utility, we introduce PoseBench, the first comprehensive benchmark for broadly applicable protein-ligand docking. PoseBench enables researchers to rigorously and systematically evaluate DL methods for apo-to-holo protein-ligand docking and protein-ligand structure prediction using both primary ligand and multi-ligand benchmark datasets, the latter of which we introduce for the first time to the DL community. Empirically, using PoseBench, we find that (1) DL co-folding methods generally outperform comparable conventional and DL docking baseline algorithms, yet popular methods such as AlphaFold 3 are still challenged by prediction targets with novel binding poses; (2) certain DL co-folding methods are highly sensitive to their input multiple sequence alignments, while others are not; and (3) DL methods struggle to strike a balance between structural accuracy and chemical specificity when predicting novel or multi-ligand protein targets. Code, data, tutorials, and benchmark results are available at https://github.com/BioinfoMachineLearning/PoseBench.", "date": "2024-05-23", "categories": []}
{"id": "2509.19988", "title": "BioBO: Biology-informed Bayesian Optimization for Perturbation Design", "abstract": "Efficient design of genomic perturbation experiments is crucial for accelerating drug discovery and therapeutic target identification, yet exhaustive perturbation of the human genome remains infeasible due to the vast search space of potential genetic interactions and experimental constraints. Bayesian optimization (BO) has emerged as a powerful framework for selecting informative interventions, but existing approaches often fail to exploit domain-specific biological prior knowledge. We propose Biology-Informed Bayesian Optimization (BioBO), a method that integrates Bayesian optimization with multimodal gene embeddings and enrichment analysis, a widely used tool for gene prioritization in biology, to enhance surrogate modeling and acquisition strategies. BioBO combines biologically grounded priors with acquisition functions in a principled framework, which biases the search toward promising genes while maintaining the ability to explore uncertain regions. Through experiments on established public benchmarks and datasets, we demonstrate that BioBO improves labeling efficiency by 25-40%, and consistently outperforms conventional BO by identifying top-performing perturbations more effectively. Moreover, by incorporating enrichment analysis, BioBO yields pathway-level explanations for selected perturbations, offering mechanistic interpretability that links designs to biologically coherent regulatory circuits.", "date": "2025-09-24", "categories": []}
{"id": "2512.03497", "title": "Cell-cell Communication Inference and Analysis: Biological Mechanisms, Computational Approaches, and Future Opportunities", "abstract": "In multicellular organisms, cells coordinate their activities through cell-cell communication (CCC), which is crucial for development, tissue homeostasis, and disease progression. Recent advances in single-cell and spatial omics technologies provide unprecedented opportunities to systematically infer and analyze CCC from these omics data, either by integrating prior knowledge of ligand-receptor interactions (LRIs) or through de novo approaches. A variety of computational methods have been developed, focusing on methodological innovations, accurate modeling of complex signaling mechanisms, and investigation of broader biological questions. These advances have greatly enhanced our ability to analyze CCC and generate biological hypotheses. Here, we introduce the biological mechanisms and modeling strategies of CCC, and provide a focused overview of more than 140 computational methods for inferring CCC from single-cell and spatial transcriptomic data, emphasizing the diversity in methodological frameworks and biological questions. Finally, we discuss the current challenges and future opportunities in this rapidly evolving field, and summarize available methods in an interactive online resource (https://cellchat.whu.edu.cn) to facilitate more efficient method comparison and selection.", "date": "2025-12-03", "categories": []}
{"id": "2603.20825", "title": "Cross-Granularity Representations for Biological Sequences: Insights from ESM and BiGCARP", "abstract": "Recent advances in general-purpose foundation models have stimulated the development of large biological sequence models. While natural language shows symbolic granularity (characters, words, sentences), biological sequences exhibit hierarchical granularity whose levels (nucleotides, amino acids, protein domains, genes) further encode biologically functional information. In this paper, we investigate the integration of cross-granularity knowledge from models through a case study of BiGCARP, a Pfam domain-level model for biosynthetic gene clusters, and ESM, an amino acid-level protein language model. Using representation analysis tools and a set of probe tasks, we first explain why a straightforward cross-model embedding initialization fails to improve downstream performance in BiGCARP, and show that deeper-layer embeddings capture a more contextual and faithful representation of the model's learned knowledge. Furthermore, we demonstrate that representations at different granularities encode complementary biological knowledge, and that combining them yields measurable performance gains in intermediate-level prediction tasks. Our findings highlight cross-granularity integration as a promising strategy for improving both the performance and interpretability of biological foundation models.", "date": "2026-03-21", "categories": []}
{"id": "2603.20955", "title": "Beyond Expression Similarity: Contrastive Learning Recovers Functional Gene Associations from Protein Interaction Structure", "abstract": "The Predictive Associative Memory (PAM) framework posits that useful relationships often connect items that co-occur in shared contexts rather than items that appear similar in embedding space. A contrastive MLP trained on co-occurrence annotations--Contrastive Association Learning (CAL)--has improved multi-hop passage retrieval and discovered narrative function at corpus scale in text. We test whether this principle transfers to molecular biology, where protein-protein interactions provide functional associations distinct from gene expression similarity. Four experiments across two biological domains map the operating envelope. On gene perturbation data (Replogle K562 CRISPRi, 2,285 genes), CAL trained on STRING protein interactions achieves cross-boundary AUC of 0.908 where expression similarity scores 0.518. A second gene dataset (DepMap, 17,725 genes) confirms the result after negative sampling correction, reaching cross-boundary AUC of 0.947. Two drug sensitivity experiments produce informative negatives that sharpen boundary conditions. Three cross-domain findings emerge: (1) inductive transfer succeeds in biology--a node-disjoint split with unseen genes yields AUC 0.826 (Delta +0.127)--where it fails in text (+/-0.10), suggesting physically grounded associations are more transferable than contingent co-occurrences; (2) CAL scores anti-correlate with interaction degree (Spearman r = -0.590), with gains concentrating on understudied genes with focused interaction profiles; (3) tighter association quality outperforms larger but noisier training sets, reversing the text pattern. Results are stable across training seeds (SD < 0.001) and cross-boundary threshold choices.", "date": "2026-03-21", "categories": []}
{"id": "2603.19636", "title": "RiboSphere: Learning Unified and Efficient Representations of RNA Structures", "abstract": "Accurate RNA structure modeling remains difficult because RNA backbones are highly flexible, non-canonical interactions are prevalent, and experimentally determined 3D structures are comparatively scarce. We introduce \\emph{RiboSphere}, a framework that learns \\emph{discrete} geometric representations of RNA by combining vector quantization with flow matching. Our design is motivated by the modular organization of RNA architecture: complex folds are composed from recurring structural motifs. RiboSphere uses a geometric transformer encoder to produce SE(3)-invariant (rotation/translation-invariant) features, which are discretized with finite scalar quantization (FSQ) into a finite vocabulary of latent codes. Conditioned on these discrete codes, a flow-matching decoder reconstructs atomic coordinates, enabling high-fidelity structure generation. We find that the learned code indices are enriched for specific RNA motifs, suggesting that the model captures motif-level compositional structure rather than acting as a purely compressive bottleneck. Across benchmarks, RiboSphere achieves strong performance in structure reconstruction (RMSD 1.25\\,\\AA, TM-score 0.84), and its pretrained discrete representations transfer effectively to inverse folding and RNA--ligand binding prediction, with robust generalization in data-scarce regimes.", "date": "2026-03-20", "categories": []}
{"id": "2603.20115", "title": "Conditioning Protein Generation via Hopfield Pattern Multiplicity", "abstract": "Protein sequence generation via stochastic attention produces plausible family members from small alignments without training, but treats all stored sequences equally and cannot direct generation toward a functional subset of interest. We show that a single scalar parameter, added as a bias to the sampler's attention logits, continuously shifts generation from the full family toward a user-specified subset, with no retraining and no change to the model architecture. A practitioner supplies a small set of sequences (for example, hits from a binding screen) and a multiplicity ratio that controls how strongly generation favors them. The method is agnostic to what the subset represents: binding, stability, specificity, or any other property. We find that the conditioning is exact at the level of the sampler's internal representation, but that the decoded sequence phenotype can fall short because the dimensionality reduction used to encode sequences does not always preserve the residue-level variation that defines the functional split. We term this discrepancy the calibration gap and show that it is predicted by a simple geometric measure of how well the encoding separates the functional subset from the rest of the family. Experiments on five Pfam families (Kunitz, SH3, WW, Homeobox, and Forkhead domains) confirm the monotonic relationship between separation and gap across a fourfold range of geometries. Applied to omega-conotoxin peptides targeting a calcium channel involved in pain signaling, curated seeding from 23 characterized binders produces over a thousand candidates that preserve the primary pharmacophore and all experimentally identified binding determinants. These results show that stochastic attention enables practitioners to expand a handful of experimentally characterized sequences into diverse candidate libraries without retraining a generative model.", "date": "2026-03-20", "categories": []}
{"id": "2603.20132", "title": "Revisiting Gene Ontology Knowledge Discovery with Hierarchical Feature Selection and Virtual Study Group of AI Agents", "abstract": "Large language models have achieved great success in multiple challenging tasks, and their capacity can be further boosted by the emerging agentic AI techniques. This new computing paradigm has already started revolutionising the traditional scientific discovery pipelines. In this work, we propose a novel agentic AI-based knowledge discovery-oriented virtual study group that aims to extract meaningful ageing-related biological knowledge considering highly ageing-related Gene Ontology terms that are selected by hierarchical feature selection methods. We investigate the performance of the proposed agentic AI framework by considering four different model organisms' ageing-related Gene Ontology terms and validate the biological findings by reviewing existing research articles. It is found that the majority of the AI agent-generated scientific claims can be supported by existing literatures and the proposed internal mechanisms of the virtual study group also play an important role in the designed agentic AI-based knowledge discovery framework.", "date": "2026-03-20", "categories": []}
{"id": "2603.18505", "title": "From Snapshots to Symphonies: The Evolution of Protein Prediction from Static Structures to Generative Dynamics and Multimodal Interactions", "abstract": "The protein folding problem has been fundamentally transformed by artificial intelligence, evolving from static structure prediction toward the modeling of dynamic conformational ensembles and complex biomolecular interactions. This review systematically examines the paradigm shift in AI driven protein science across five interconnected dimensions: unified multimodal representations that integrate sequences, geometries, and textual knowledge; refinement of static prediction through MSA free architectures and all atom complex modeling; generative frameworks, including diffusion models and flow matching, that capture conformational distributions consistent with thermodynamic ensembles; prediction of heterogeneous interactions spanning protein ligand, protein nucleic acid, and protein protein complexes; and functional inference of fitness landscapes, mutational effects, and text guided property prediction. We critically analyze current bottlenecks, including data distribution biases, limited mechanistic interpretability, and the disconnect between geometric metrics and biophysical reality, while identifying future directions toward physically consistent generative models, multimodal foundation architectures, and experimental closed loop systems. This methodological transformation marks artificial intelligence's transition from a structural analysis tool into a universal simulator capable of understanding and ultimately rewriting the dynamic language of life.", "date": "2026-03-19", "categories": []}
{"id": "2603.18571", "title": "CAPSUL: A Comprehensive Human Protein Benchmark for Subcellular Localization", "abstract": "Subcellular localization is a crucial biological task for drug target identification and function annotation. Although it has been biologically realized that subcellular localization is closely associated with protein structure, no existing dataset offers comprehensive 3D structural information with detailed subcellular localization annotations, thus severely hindering the application of promising structure-based models on this task. To address this gap, we introduce a new benchmark called $\\mathbf{CAPSUL}$, a $\\mathbf{C}$omprehensive hum$\\mathbf{A}$n $\\mathbf{P}$rotein benchmark for $\\mathbf{SU}$bcellular $\\mathbf{L}$ocalization. It features a dataset that integrates diverse 3D structural representations with fine-grained subcellular localization annotations carefully curated by domain experts. We evaluate this benchmark using a variety of state-of-the-art sequence-based and structure-based models, showcasing the importance of involving structural features in this task. Furthermore, we explore reweighting and single-label classification strategies to facilitate future investigation on structure-based methods for this task. Lastly, we showcase the powerful interpretability of structure-based methods through a case study on the Golgi apparatus, where we discover a decisive localization pattern $\\alpha$-helix from attention mechanisms, demonstrating the potential for bridging the gap with intuitive biological interpretability and paving the way for data-driven discoveries in cell biology.", "date": "2026-03-19", "categories": []}
{"id": "2603.17247", "title": "Binary Latent Protein Fitness Landscapes for Quantum Annealing Optimization", "abstract": "We propose Q-BIOLAT, a framework for modeling and optimizing protein fitness landscapes in binary latent spaces. Starting from protein sequences, we leverage pretrained protein language models to obtain continuous embeddings, which are then transformed into compact binary latent representations. In this space, protein fitness is approximated using a quadratic unconstrained binary optimization (QUBO) model, enabling efficient combinatorial search via classical heuristics such as simulated annealing and genetic algorithms.\n  On the ProteinGym benchmark, we demonstrate that Q-BIOLAT captures meaningful structure in protein fitness landscapes and enables the identification of high-fitness variants. Despite using a simple binarization scheme, our method consistently retrieves sequences whose nearest neighbors lie within the top fraction of the training fitness distribution, particularly under the strongest configurations. We further show that different optimization strategies exhibit distinct behaviors, with evolutionary search performing better in higher-dimensional latent spaces and local search remaining competitive in preserving realistic sequences.\n  Beyond its empirical performance, Q-BIOLAT provides a natural bridge between protein representation learning and combinatorial optimization. By formulating protein fitness as a QUBO problem, our framework is directly compatible with emerging quantum annealing hardware, opening new directions for quantum-assisted protein engineering.\n  Our implementation is publicly available at: https://github.com/HySonLab/Q-BIOLAT", "date": "2026-03-18", "categories": []}
{"id": "2603.17633", "title": "Atomic Trajectory Modeling with State Space Models for Biomolecular Dynamics", "abstract": "Understanding the dynamic behavior of biomolecules is fundamental to elucidating biological function and facilitating drug discovery. While Molecular Dynamics (MD) simulations provide a rigorous physical basis for studying these dynamics, they remain computationally expensive for long timescales. Conversely, recent deep generative models accelerate conformation generation but are typically either failing to model temporal relationship or built only for monomeric proteins. To bridge this gap, we introduce ATMOS, a novel generative framework based on State Space Models (SSM) designed to generate atom-level MD trajectories for biomolecular systems. ATMOS integrates a Pairformer-based state transition mechanism to capture long-range temporal dependencies, with a diffusion-based module to decode trajectory frames in an autoregressive manner. ATMOS is trained across crystal structures from PDB and conformation trajectory from large-scale MD simulation datasets including mdCATH and MISATO. We demonstrate that ATMOS achieves state-of-the-art performance in generating conformation trajectories for both protein monomers and complex protein-ligand systems. By enabling efficient inference of atomic trajectory of motions, this work establishes a promising foundation for modeling biomolecular dynamics.", "date": "2026-03-18", "categories": []}
{"id": "2405.08979", "title": "drGT: Attention-Guided Gene Assessment of Drug Response Utilizing a Drug-Cell-Gene Heterogeneous Network", "abstract": "For translational impact, both accurate drug response prediction and biological plausibility of predictive features are needed. We present drGT, a heterogeneous graph deep learning model over drugs, genes, and cell lines that couples prediction with mechanism-oriented interpretability via attention coefficients (ACs).\n  We assess both predictive generalization (random, unseen-drug, unseen-cell, and zero-shot splits) and biological plausibility (use of text-mined PubMed gene-drug co-mentions and comparison to a structure-based DTI predictor) on GDSC, NCI60, and CTRP datasets. Across benchmarks, drGT consistently delivers top regression performance while maintaining competitive classification accuracy for drug sensitivity. Under random 5-fold cross-validation, drGT attains an AUROC of up to 0.945 (3rd overall) and an $R^2$ up to 0.690, outperforming all baselines on regression. In leave-one-out tests for unseen cell lines and drugs, drGT achieves AUROCs of 0.706 and 0.844, and $R^2$ values of 0.692 and 0.022, the only model yielding positive $R^2$ for unseen drugs. In zero-shot prediction, drGT achieves an AUROC of 0.786 and a regression $R^2$ of 0.334, both representing the highest scores among all models. For interpretability, AC-derived drug-gene links recover known biology: among 976 drugs with known DTIs, 36.9% of predicted links match established DTIs, and 63.7% are supported by either PubMed abstracts or a structure-based predictive model. Enrichment analyses of AC-prioritized genes reveal drug-perturbed biological processes, providing pathway-level explanations.\n  drGT advances predictive generalization and mechanism-centered interpretability, offering state-of-the-art regression accuracy and literature-supported biological hypotheses that demonstrate the use of graph learning from heterogeneous input data for biological discovery. Code: https://github.com/sciluna/drGT", "date": "2024-05-14", "categories": []}
{"id": "2603.11872", "title": "ELISA: An Interpretable Hybrid Generative AI Agent for Expression-Grounded Discovery in Single-Cell Genomics", "abstract": "Translating single-cell RNA sequencing (scRNA-seq) data into mechanistic biological hypotheses remains a critical bottleneck, as agentic AI systems lack direct access to transcriptomic representations while expression foundation models remain opaque to natural language. Here we introduce ELISA (Embedding-Linked Interactive Single-cell Agent), an interpretable framework that unifies scGPT expression embeddings with BioBERT-based semantic retrieval and LLM-mediated interpretation for interactive single-cell discovery. An automatic query classifier routes inputs to gene marker scoring, semantic matching, or reciprocal rank fusion pipelines depending on whether the query is a gene signature, natural language concept, or mixture of both. Integrated analytical modules perform pathway activity scoringacross 60+ gene sets, ligand--receptor interaction prediction using 280+ curated pairs, condition-aware comparative analysis, and cell-type proportion estimation all operating directly on embedded data without access to the original count matrix. Benchmarked across six diverse scRNA-seq datasets spanning inflammatory lung disease, pediatric and adult cancers, organoid models, healthy tissue, and neurodevelopment, ELISA significantly outperforms CellWhisperer in cell type retrieval (combined permutation test, $p < 0.001$), with particularly large gains on gene-signature queries (Cohen's $d = 5.98$ for MRR). ELISA replicates published biological findings (mean composite score 0.90) with near-perfect pathway alignment and theme coverage (0.98 each), and generates candidate hypotheses through grounded LLM reasoning, bridging the gap between transcriptomic data exploration and biological discovery. Code available at: https://github.com/omaruno/ELISA-An-AI-Agent-for-Expression-Grounded-Discovery-in-Single-Cell-Genomics.git (If you use ELISA in your research, please cite this work).", "date": "2026-03-12", "categories": []}
{"id": "2603.16562", "title": "Understanding Cell Fate Decisions with Temporal Attention", "abstract": "Understanding non-genetic determinants of cell fate is critical for developing and improving cancer therapies, as genetically identical cells can exhibit divergent outcomes under the same treatment conditions. In this work, we present a deep learning approach for cell fate prediction from raw long-term live-cell recordings of cancer cell populations under chemotherapeutic treatment. Our Transformer model is trained to predict cell fate directly from raw image sequences, without relying on predefined morphological or molecular features. Beyond classification, we introduce a comprehensive explainability framework for interpreting the temporal and morphological cues guiding the model's predictions. We demonstrate that prediction of cell outcomes is possible based on the video only, our model achieves balanced accuracy of 0.94 and an F1-score of 0.93. Attention and masking experiments further indicate that the signal predictive of the cell fate is not uniquely located in the final frames of a cell trajectory, as reliable predictions are possible up to 10 h before the event. Our analysis reveals distinct temporal distribution of predictive information in the mitotic and apoptotic sequences, as well as the role of cell morphology and p53 signaling in determining cell outcomes. Together, these findings demonstrate that attention-based temporal models enable accurate cell fate prediction while providing biologically interpretable insights into non-genetic determinants of cellular decision-making. The code is available at https://github.com/bozeklab/Cell-Fate-Prediction.", "date": "2026-03-17", "categories": []}
{"id": "2508.13201", "title": "Benchmarking LLM-based agents for single-cell omics analysis", "abstract": "Background: The surge in single-cell omics data exposes limitations in traditional, manually defined analysis workflows. AI agents offer a paradigm shift, enabling adaptive planning, executable code generation, traceable decisions, and real-time knowledge fusion. However, the lack of a comprehensive benchmark critically hinders progress.\n  Results: We introduce a novel benchmarking evaluation system to rigorously assess agent capabilities in single-cell omics analysis. This system comprises: a unified platform compatible with diverse agent frameworks and LLMs; multidimensional metrics assessing cognitive program synthesis, collaboration, execution efficiency, bioinformatics knowledge integration, and task completion quality; and 50 diverse real-world single-cell omics analysis tasks spanning multi-omics, species, and sequencing technologies. Our evaluation reveals that Grok3-beta achieves state-of-the-art performance among tested agent frameworks. Multi-agent frameworks significantly enhance collaboration and execution efficiency over single-agent approaches through specialized role division. Attribution analyses of agent capabilities identify that high-quality code generation is crucial for task success, and self-reflection has the most significant overall impact, followed by retrieval-augmented generation (RAG) and planning.\n  Conclusions: This work highlights persistent challenges in code generation, long-context handling, and context-aware knowledge retrieval, providing a critical empirical foundation and best practices for developing robust AI agents in computational biology.", "date": "2025-08-16", "categories": []}
{"id": "2510.01618", "title": "Automated Genomic Interpretation via Concept Bottleneck Models for Medical Robotics", "abstract": "We propose an automated genomic interpretation module that transforms raw DNA sequences into actionable, interpretable decisions suitable for integration into medical automation and robotic systems. Our framework combines Chaos Game Representation (CGR) with a Concept Bottleneck Model (CBM), enforcing predictions to flow through biologically meaningful concepts such as GC content, CpG density, and k mer motifs. To enhance reliability, we incorporate concept fidelity supervision, prior consistency alignment, KL distribution matching, and uncertainty calibration. Beyond accurate classification of HIV subtypes across both in-house and LANL datasets, our module delivers interpretable evidence that can be directly validated against biological priors. A cost aware recommendation layer further translates predictive outputs into decision policies that balance accuracy, calibration, and clinical utility, reducing unnecessary retests and improving efficiency. Extensive experiments demonstrate that the proposed system achieves state of the art classification performance, superior concept prediction fidelity, and more favorable cost benefit trade-offs compared to existing baselines. By bridging the gap between interpretable genomic modeling and automated decision-making, this work establishes a reliable foundation for robotic and clinical automation in genomic medicine.", "date": "2025-10-02", "categories": []}
{"id": "2602.08751", "title": "Central Dogma Transformer II: An AI Microscope for Understanding Cellular Regulatory Mechanisms", "abstract": "Current biological AI models lack interpretability -- their internal representations do not correspond to biological relationships that researchers can\n  examine. Understanding gene regulation requires models whose learned structure can be directly interrogated to generate experimentally testable\n  hypotheses. CDT-II mirrors the central dogma in its architecture -- DNA self-attention, RNA self-attention, and cross-attention for transcriptional\n  control -- requiring only genomic embeddings and raw per-cell expression. Applied to K562 CRISPRi data with five genes held out entirely, CDT-II predicts\n  perturbation effects (per-gene mean r = 0.84), recovers the GFI1B regulatory network (6.6-fold enrichment, P = 3.5 x 10^{-17}), and shows that\n  cross-attention focuses on ENCODE regulatory elements including CTCF sites (mean 7.67x across 28 targets, P < 0.001). Gradient-based attribution\n  accurately predicts downstream consequences of perturbing therapeutic targets (mean r = 0.82). Applied to TFRC, the target of the anti-TfR1 antibody\n  PPMX-T003, gradient analysis identifies genes involved in erythrocyte structure, iron-dependent DNA synthesis, and oxidative stress -- pathways that align\n  with anemia and reticulocyte decrease reported in Phase 1 trials and ferroptosis demonstrated in preclinical studies, without any clinical data as input,\n  establishing CDT-II as an AI microscope that reveals clinically relevant regulatory structure from perturbation experiments alone.", "date": "2026-02-09", "categories": []}
{"id": "2603.05110", "title": "BLINK: Behavioral Latent Modeling of NK Cell Cytotoxicity", "abstract": "Machine learning models of cellular interaction dynamics hold promise for understanding cell behavior. Natural killer (NK) cell cytotoxicity is a prominent example of such interaction dynamics and is commonly studied using time-resolved multi-channel fluorescence microscopy. Although tumor cell death events can be annotated at single frames, NK cytotoxic outcome emerges over time from cellular interactions and cannot be reliably inferred from frame-wise classification alone. We introduce BLINK, a trajectory-based recurrent state-space model that serves as a cell world model for NK-tumor interactions. BLINK learns latent interaction dynamics from partially observed NK-tumor interaction sequences and predicts apoptosis increments that accumulate into cytotoxic outcomes. Experiments on long-term time-lapse NK-tumor recordings show improved cytotoxic outcome detection and enable forecasting of future outcomes, together with an interpretable latent representation that organizes NK trajectories into coherent behavioral modes and temporally structured interaction phases. BLINK provides a unified framework for quantitative evaluation and structured modeling of NK cytotoxic behavior at the single-cell level.", "date": "2026-03-05", "categories": []}
{"id": "2603.13342", "title": "MS2MetGAN: Latent-space adversarial training for metabolite-spectrum matching in MS/MS database search", "abstract": "Database search is a widely used approach for identifying metabolites from tandem mass spectra (MS/MS). In this strategy, an experimental spectrum is matched against a user-specified database of candidate metabolites, and candidates are ranked such that true metabolite-spectrum matches receive the highest scores. Machine-learning methods have been widely incorporated into database-search-based identification tools and have substantially improved performance. To further improve identification accuracy, we propose a new framework for generating negative training samples. The framework first uses autoencoders to learn latent representations of metabolite structures and MS/MS spectra, thereby recasting metabolite-spectrum matching as matching between latent vectors. It then uses a GAN to generate latent vectors of decoy metabolites and constructs decoy metabolite-spectrum matches as negative samples for training. Experimental results show that our tool, MS2MetGAN, achieves better overall performance than existing metabolite identification methods.", "date": "2026-03-07", "categories": []}
{"id": "2603.13401", "title": "MAD: Microenvironment-Aware Distillation -- A Pretraining Strategy for Virtual Spatial Omics from Microscopy", "abstract": "Bridging microscopy and omics would allow us to read molecular states from images-at single-cell resolution and tissue scale-without the cost and throughput limits of omics technologies. Self-supervised pretraining offers a scalable approach with minimal labels, yet how to encode single-cell identity within tissue environments-and the extent of biological information such models can capture-remains an open question. Here, we introduce MAD (microenvironment-aware distillation), a pretraining strategy that learns cell-centric embeddings by jointly self-distilling the morphology view and the microenvironment view of the same indexed cell into a unified embedding space. Across diverse tissues and imaging modalities, MAD achieves state-of-the-art prediction performance on downstream tasks including cell subtyping, transcriptomic prediction, and bioinformatic inference. MAD even outperforms foundation models with a similar number of model parameters that have been trained on substantially larger datasets. These results demonstrate that MAD's dual-view joint self-distillation effectively captures the complexity and diversity of cells within tissues. Together, this establishes MAD as a general tool for representation learning in microscopy, enabling virtual spatial omics and biological insights from vast microscopy datasets.", "date": "2026-03-11", "categories": []}
{"id": "2603.13787", "title": "Advancing Cancer Prognosis with Hierarchical Fusion of Genomic, Proteomic and Pathology Imaging Data from a Systems Biology Perspective", "abstract": "To enhance the precision of cancer prognosis, recent research has increasingly focused on multimodal survival methods by integrating genomic data and histology images. However, current approaches overlook the fact that the proteome serves as an intermediate layer bridging genomic alterations and histopathological features while providing complementary biological information essential for survival prediction. This biological reality exposes another architectural limitation: existing integrative analysis studies fuse these heterogeneous data sources in a flat manner that fails to capture their inherent biological hierarchy. To address these limitations, we propose HFGPI, a hierarchical fusion framework that models the biological progression from genes to proteins to histology images from a systems biology perspective. Specifically, we introduce Molecular Tokenizer, a molecular encoding strategy that integrates identity embeddings with expression profiles to construct biologically informed representations for genes and proteins. We then develop Gene-Regulated Protein Fusion (GRPF), which employs graph-aware cross-attention with structure-preserving alignment to explicitly model gene-protein regulatory relationships and generate gene-regulated protein representations. Additionally, we propose Protein-Guided Hypergraph Learning (PGHL), which establishes associations between proteins and image patches, leveraging hypergraph convolution to capture higher-order protein-morphology relationships. The final features are progressively fused across hierarchical layers to achieve precise survival outcome prediction. Extensive experiments on five benchmark datasets demonstrate the superiority of HFGPI over state-of-the-art methods.", "date": "2026-03-14", "categories": []}
{"id": "2603.14717", "title": "Training-Free Generation of Protein Sequences from Small Family Alignments via Stochastic Attention", "abstract": "Most protein families have fewer than 100 known members, a regime where deep generative models overfit or collapse. We propose stochastic attention (SA), a training-free sampler that treats the modern Hopfield energy over a protein alignment as a Boltzmann distribution and draws samples via Langevin dynamics. The score function is a closed-form softmax attention operation requiring no training, no pretraining data, and no GPU, with cost linear in alignment size. Across eight Pfam families, SA generates sequences with low amino acid compositional divergence, substantial novelty, and structural plausibility confirmed by ESMFold and AlphaFold2. Generated sequences fold more faithfully to canonical family structures than natural members in six of eight families. Against profile HMMs, EvoDiff, and the MSA Transformer, which produce sequences that drift far outside the family, SA maintains 51 to 66 percent identity while remaining novel, in seconds on a laptop. The critical temperature governing generation is predicted from PCA dimensionality alone, enabling fully automatic operation. Controls confirm SA encodes correlated substitution patterns, not just per-position amino acid frequencies.", "date": "2026-03-16", "categories": []}
{"id": "2603.14797", "title": "Multi-Task Genetic Algorithm with Multi-Granularity Encoding for Protein-Nucleotide Binding Site Prediction", "abstract": "Accurate identification of protein-nucleotide binding sites is fundamental to deciphering molecular mechanisms and accelerating drug discovery. However, current computational methods often struggle with suboptimal performance due to inadequate feature representation and rigid fusion mechanisms, which hinder the effective exploitation of cross-task information synergy. To bridge this gap, we propose MTGA-MGE, a framework that integrates a Multi-Task Genetic Algorithm with Multi-Granularity Encoding to enhance binding site prediction. Specifically, we develop a Multi-Granularity Encoding (MGE) network that synergizes multi-scale convolutions and self-attention mechanisms to distill discriminative signals from high-dimensional, redundant biological data. To overcome the constraints of static fusion, a genetic algorithm is employed to adaptively evolve task-specific fusion strategies, thereby effectively improving model generalization. Furthermore, to catalyze collaborative learning, we introduce an External-Neighborhood Mechanism (ENM) that leverages biological similarities to facilitate targeted information exchange across tasks. Extensive evaluations on fifteen nucleotide datasets demonstrate that MTGA-MGE not only establishes a new state-of-the-art in data-abundant, high-resource scenarios but also maintains a robust competitive edge in rare, low-resource regimes, presenting a highly adaptive scheme for decoding complex protein-ligand interactions in the post-genomic era.", "date": "2026-03-16", "categories": []}
{"id": "2603.14806", "title": "Fold-CP: A Context Parallelism Framework for Biomolecular Modeling", "abstract": "Understanding cellular machinery requires atomic-scale reconstruction of large biomolecular assemblies. However, predicting the structures of these systems has been constrained by hardware memory requirements of models like AlphaFold 3, imposing a practical ceiling of a few thousand residues that can be processed on a single GPU. Here we present NVIDIA BioNeMo Fold-CP, a context parallelism framework that overcomes this barrier by distributing the inference and training pipelines of co-folding models across multiple GPUs. We use the Boltz models as open source reference architectures and implement custom multidimensional primitives that efficiently parallelize both the dense triangular updates and the irregular, data-dependent pattern of window-batched local attention. Our approach achieves efficient memory scaling; for an N-token input distributed across P GPUs, per-device memory scales as $O(N^2/P)$, enabling the structure prediction of assemblies exceeding 30,000 residues on 64 NVIDIA B300 GPUs. We demonstrate the scientific utility of this approach through successful developer use cases: Fold-CP enabled the scoring of over 90% of Comprehensive Resource of Mammalian protein complexes (CORUM) database, as well as folding of disease-relevant PI4KA lipid kinase complex bound to an intrinsically disordered region without cropping. By providing a scalable pathway for modeling massive systems with full global context, Fold-CP represents a significant step toward the realization of a virtual cell.", "date": "2026-03-16", "categories": []}
{"id": "2603.14870", "title": "IgPose: A Generative Data-Augmented Pipeline for Robust Immunoglobulin-Antigen Binding Prediction", "abstract": "Predicting immunoglobulin-antigen (Ig-Ag) binding remains a significant challenge due to the paucity of experimentally-resolved complexes and the limited accuracy of de novo Ig structure prediction. We introduce IgPose, a generalizable framework for Ig-Ag pose identification and scoring, built on a generative data-augmentation pipeline. To mitigate data scarcity, we constructed the Structural Immunoglobulin Decoy Database (SIDD), a comprehensive repository of high-fidelity synthetic decoys. IgPose integrates equivariant graph neural networks, ESM-2 embeddings, and gated recurrent units to synergistically capture both geometric and evolutionary features. We implemented interface-focused k-hop sampling with biologically guided pooling to enhance generalization across diverse interfaces. The framework comprises two sub-networks--IgPoseClassifier for binding pose discrimination and IgPoseScore for DockQ score estimation--and achieves robust performance on curated internal test sets and the CASP-16 benchmark compared to physics and deep learning baselines. IgPose serves as a versatile computational tool for high-throughput antibody discovery pipelines by providing accurate pose filtering and ranking. IgPose is available on GitHub (https://github.com/arontier/igpose).", "date": "2026-03-16", "categories": []}
{"id": "2603.14897", "title": "BiTro: Bidirectional Transfer Learning Enhances Bulk and Spatial Transcriptomics Prediction in Cancer Pathological Images", "abstract": "Cancer pathological analysis requires modeling tumor heterogeneity across multiple modalities, primarily through transcriptomics and whole slide imaging (WSI), along with their spatial relations. On one hand, bulk transcriptomics and WSI images are largely available but lack spatial mapping; on the other hand, spatial transcriptomics (ST) data can offer high spatial resolution, yet facing challenges of high cost, low sequencing depth, and limited sample sizes. Therefore, the data foundation of either side is flawed and has its limit in accurately finding the mapping between the two modalities. To this end, we propose BiTro, a bidirectional transfer learning framework that can enhance bulk and spatial transcriptomics prediction from pathological images. Our contributions are twofold. First, we design a universal and transferable model architecture that works for both bulk+WSI and ST data. A major highlight is that we model WSI images on the cellular level to better capture cells' visual features, morphological phenotypes, and their spatial relations; to map cells' features to their transcriptomics measured in bulk or ST, we adopt multiple instance learning. Second, by using LoRA, our model can be efficiently transferred between bulk and ST data to exploit their complementary information. To test our framework, we conducted comprehensive experiments on five cancer datasets. Results demonstrate that 1) our base model can achieve better or competitive performance compared to existing models on bulk or spatial transcriptomics prediction, and 2) transfer learning can further improve the base model's performance.", "date": "2026-03-16", "categories": []}
{"id": "2603.15006", "title": "Empowering Chemical Structures with Biological Insights for Scalable Phenotypic Virtual Screening", "abstract": "Motivation: The scalable identification of bioactive compounds is essential for contemporary drug discovery. This process faces a key trade-off: structural screening offers scalability but lacks biological context, whereas high-content phenotypic profiling provides deep biological insights but is resource-intensive. The primary challenge is to extract robust biological signals from noisy data and encode them into representations that do not require biological data at inference. Results: This study presents DECODE (DEcomposing Cellular Observations of Drug Effects), a framework that bridges this gap by empowering chemical representations with intrinsic biological semantics to enable structure-based in silico biological profiling. DECODE leverages limited paired transcriptomic and morphological data as supervisory signals during training, enabling the extraction of a measurement-invariant biological fingerprint from chemical structures and explicit filtering of experimental noise. Our evaluations demonstrate that DECODE retrieves functionally similar drugs in zero-shot settings with over 20% relative improvement over chemical baselines in mechanism-of-action (MOA) prediction. Furthermore, the framework achieves a 6-fold increase in hit rates for novel anti-cancer agents during external validation. Availability and implementation: The codes and datasets of DECODE are available at https://github.com/lian-xiao/DECODE.", "date": "2026-03-16", "categories": []}
{"id": "2603.15047", "title": "CrossADR: enhancing adverse drug reactions prediction for combination pharmacotherapy with cross-layer feature integration and cross-level associative learning", "abstract": "Combination pharmacotherapy offers substantial therapeutic advantages but also poses substantial risks of adverse drug reactions (ADRs). The accurate prediction of ADRs with interpretable computational methods is crucial for clinical safety management, drug development, and precision medicine. However, managing ADRs remains a challenge due to the vast search space of drug combinations and the complexity of physiological responses. Current graph-based architectures often struggle to effectively integrate multi-scale biological information and frequently rely on fixed association matrices, which limits their ability to capture dynamic organ-level dependencies and generalize across diverse datasets. Here we propose CrossADR, a hierarchical framework for organ-level ADR prediction through cross-layer feature integration and cross-level associative learning. It incorporates a gated-residual-flow graph neural network to fuse multi-scale molecular features and utilizes a learnable ADR embedding space to dynamically capture latent biological correlations across 15 organ systems. Systematic evaluation on the newly constructed CrossADR-Dataset-covering 1,376 drugs and 946,000 unique combinations-demonstrates that CrossADR consistently achieves state-of-the-art performance across 80 distinct experimental scenarios and provides high-resolution insights into drug-related protein protein interactions and pathways. Overall, CrossADR represents a robust tool for cross-scale biomedical information integration, cross-layer feature integration as well as cross-level associative learning, and can be effectively utilized to prevent ADRs in clinical decision-making.", "date": "2026-03-16", "categories": []}
{"id": "2603.15198", "title": "Geometric framework for biological evolution", "abstract": "We develop a generally covariant description of evolutionary dynamics that operates consistently in both genotype and phenotype spaces. We show that the maximum entropy principle yields a fundamental identification between the inverse metric tensor and the covariance matrix, revealing the Lande equation as a covariant gradient ascent equation. This demonstrates that evolution can be modeled as a learning process on the fitness landscape, with the specific learning algorithm determined by the functional relation between the metric tensor and the noise covariance arising from microscopic dynamics. While the metric (or the inverse genotypic covariance matrix) has been extensively characterized empirically, the noise covariance and its associated observable (the covariance of evolutionary changes) have never been directly measured. This poses the experimental challenge of determining the functional form relating metric to noise covariance.", "date": "2026-03-16", "categories": []}
{"id": "2409.03658", "title": "A DNN Biophysics Model with Topological and Electrostatic Features", "abstract": "In this project, we present a deep neural network (DNN)-based biophysics model that uses multi-scale and uniform topological and electrostatic features to predict protein properties, such as Coulomb energies or solvation energies. The topological features are generated using element-specific persistent homology (ESPH) on a selection of heavy atoms or carbon atoms. The electrostatic features are generated using a novel Cartesian treecode, which adds underlying electrostatic interactions to further improve the model prediction. These features are uniform in number for proteins of varying sizes; therefore, the widely available protein structure databases can be used to train the network. These features are also multi-scale, allowing users to balance resolution and computational cost. The optimal model trained on more than 17,000 proteins for predicting Coulomb energy achieves MSE of approximately 0.024, MAPE of 0.073 and $R^2$ of 0.976. Meanwhile, the optimal model trained on more than 4,000 proteins for predicting solvation energy achieves MSE of approximately 0.064, MAPE of 0.081, and $R^2$ of 0.926, showing the efficiency and fidelity of these features in representing the protein structure and force field. The feature generation algorithms also have the potential to serve as general tools for assisting machine learning based prediction of protein properties and functions.", "date": "2024-09-05", "categories": []}
{"id": "2603.12293", "title": "Multi-objective Genetic Programming with Multi-view Multi-level Feature for Enhanced Protein Secondary Structure Prediction", "abstract": "Predicting protein secondary structure is essential for understanding protein function and advancing drug discovery. However, the intricate sequence-structure relationship poses significant challenges for accurate modeling. To address these, we propose MOGP-MMF, a multi-objective genetic programming framework that reformulates PSSP as an automated optimization task focused on feature selection and fusion. Specifically, MOGP-MMF introduces a multi-view multi-level representation strategy that integrates evolutionary, semantic, and newly introduced structural views to capture the comprehensive protein folding logic. Leveraging an enriched operator set, the framework evolves both linear and nonlinear fusion functions, effectively capturing high-order feature interactions while reducing fusion complexity. To resolve the accuracy-complexity trade-off, an improved multi-objective GP algorithm is developed, incorporating a knowledge transfer mechanism that utilizes prior evolutionary experience to guide the population toward global optima. Extensive experiments across seven benchmark datasets demonstrate that MOGP-MMF surpasses state-of-the-art methods, particularly in Q8 accuracy and structural integrity. Furthermore, MOGP-MMF generates a diverse set of non-dominated solutions, offering flexible model selection schemes for various practical application scenarios. The source code is available on GitHub: https://github.com/qian-ann/MOGP-MMF/tree/main.", "date": "2026-03-11", "categories": []}
{"id": "2603.12694", "title": "RXNRECer Enables Fine-grained Enzymatic Function Annotation through Active Learning and Protein Language Models", "abstract": "A key challenge in enzyme annotation is identifying the biochemical reactions catalyzed by proteins. Most existing methods rely on Enzyme Commission (EC) numbers as intermediaries: they first predict an EC number and then retrieve the associated reactions. This indirect strategy introduces ambiguity due to the complex many-to-many mappings among proteins, EC numbers, and reactions, and is further complicated by frequent updates to EC numbers and inconsistencies across databases. To address these challenges, we present RXNRECer, a transformer-based ensemble framework that directly predicts enzyme-catalyzed reactions without relying on EC numbers. It integrates protein language modeling and active learning to capture both high-level sequence semantics and fine-grained transformation patterns. Evaluations on curated cross-validation and temporal test sets demonstrate consistent improvements over six EC-based baselines, with gains of 16.54% in F1 score and 15.43% in accuracy. Beyond accuracy gains, the framework offers clear advantages for downstream applications, including scalable proteome-wide reaction annotation, enhanced specificity in refining generic reaction schemas, systematic annotation of previously uncurated proteins, and reliable identification of enzyme promiscuity. By incorporating large language models, it also provides interpretable rationales for predictions. These capabilities make RXNRECer a robust and versatile solution for EC-free, fine-grained enzyme function prediction, with potential applications across multiple areas of enzyme research and industrial applications.", "date": "2026-03-13", "categories": []}
{"id": "2603.13051", "title": "Causal Cellular Context Transfer Learning (C3TL): An Efficient Architecture for Prediction of Unseen Perturbation Effects", "abstract": "Predicting the effects of chemical and genetic perturbations on quantitative cell states is a central challenge in computational biology, molecular medicine and drug discovery. Recent work has leveraged large-scale single-cell data and massive foundation models to address this task. However, such computational resources and extensive datasets are not always accessible in academic or clinical settings, hence limiting utility. Here we propose a lightweight framework for perturbation effect prediction that exploits the structured nature of biological interventions and specific inductive biases/invariances. Our approach leverages available information concerning perturbation effects to allow generalization to novel contexts and requires only widely-available bulk molecular data. Extensive testing, comparing predictions of context-specific perturbation effects against real, large-scale interventional experiments, demonstrates accurate prediction in new contexts. The proposed approach is competitive with SOTA foundation models but requires simpler data, much smaller model sizes and less time. Focusing on robust bulk signals and efficient architectures, we show that accurate prediction of perturbation effects is possible without proprietary hardware or very large models, hence opening up ways to leverage causal learning approaches in biomedicine generally.", "date": "2026-03-13", "categories": []}
{"id": "2602.21550", "title": "Extending Sequence Length is Not All You Need: Effective Integration of Multimodal Signals for Gene Expression Prediction", "abstract": "Gene expression prediction, which predicts mRNA expression levels from DNA sequences, presents significant challenges. Previous works often focus on extending input sequence length to locate distal enhancers, which may influence target genes from hundreds of kilobases away. Our work first reveals that for current models, long sequence modeling can decrease performance. Even carefully designed algorithms only mitigate the performance degradation caused by long sequences. Instead, we find that proximal multimodal epigenomic signals near target genes prove more essential. Hence we focus on how to better integrate these signals, which has been overlooked. We find that different signal types serve distinct biological roles, with some directly marking active regulatory elements while others reflect background chromatin patterns that may introduce confounding effects. Simple concatenation may lead models to develop spurious associations with these background patterns. To address this challenge, we propose Prism, a framework that learns multiple combinations of high-dimensional epigenomic features to represent distinct background chromatin states and uses backdoor adjustment to mitigate confounding effects. Our experimental results demonstrate that proper modeling of multimodal epigenomic signals achieves state-of-the-art performance using only short sequences for gene expression prediction.", "date": "2026-02-25", "categories": []}
{"id": "2603.11061", "title": "Hybrid Quantum-Classical Encoding for Accurate Residue-Level pKa Prediction", "abstract": "Accurate prediction of residue-level pKa values is essential for understanding protein function, stability, and reactivity. While existing resources such as DeepKaDB and CpHMD-derived datasets provide valuable training data, their descriptors remain primarily classical and often struggle to generalize across diverse biochemical environments. We introduce a reproducible hybrid quantum-classical framework that enriches residue-level representations with a Gaussian kernel-based quantum-inspired feature mapping. These quantum-enhanced descriptors are combined with normalized structural features to form a unified hybrid encoding processed by a Deep Quantum Neural Network (DQNN). This architecture captures nonlinear relationships in residue microenvironments that are not accessible to classical models. Benchmarking across multiple curated descriptor sets demonstrates that the DQNN achieves improved cross-context generalization relative to classical baselines. External evaluation on the PKAD-R experimental benchmark and an A$\\beta$40 case study further highlights the robustness and transferability of the quantum-inspired representation. By integrating quantum-inspired feature transformations with classical biochemical descriptors, this work establishes a scalable and experimentally transferable approach for residue-level pKa prediction and broader applications in protein electrostatics.", "date": "2026-03-09", "categories": []}
{"id": "2603.11940", "title": "Exhaustive Circuit Mapping of a Single-Cell Foundation Model Reveals Massive Redundancy, Heavy-Tailed Hub Architecture, and Layer-Dependent Differentiation Control", "abstract": "Mechanistic interpretability of biological foundation models has relied on selective feature sampling, pairwise interaction testing, and observational trajectory analysis. Each of these can introduce systematic bias. Here we present three experiments that address these limitations through exhaustive circuit tracing, higher order combinatorial ablation, and causal trajectory steering in Geneformer, a transformer based single cell foundation model. First, exhaustive tracing of all 4065 active sparse autoencoder features at layer 5 yields 1393850 significant downstream edges, a 27 fold expansion over selective sampling. This reveals a heavy tailed hub distribution in which 1.8 percent of features account for disproportionate connectivity and 40 percent of the top 20 hubs lack biological annotation. These results indicate systematic annotation bias in prior selective analyses. Second, three way combinatorial ablation across 8 feature triplets shows that redundancy deepens monotonically with interaction order, with a three way ratio of 0.59 versus a pairwise ratio of 0.74, and with zero synergy. This confirms that the model architecture is subadditive at all tested orders. Third, trajectory guided feature steering establishes a causal link between layer position and differentiation directionality. Late layer features at L17 consistently push cell states toward maturity, with fraction positive equal to 1.0. Early and mid layer features at L0 and L11 mostly push away from maturity, with fraction positive ranging from 0.00 to 0.58. Together these results move from correlation toward causal evidence for layer dependent control of cell state.", "date": "2026-03-12", "categories": []}
{"id": "2603.12073", "title": "A Multi-Label Temporal Convolutional Framework for Transcription Factor Binding Characterization", "abstract": "Transcription factors (TFs) regulate gene expression through complex and co-operative mechanisms. While many TFs act together, the logic underlying TFs binding and their interactions is not fully understood yet. Most current approaches for TF binding site prediction focus on individual TFs and binary classification tasks, without a full analysis of the possible interactions among various TFs. In this paper we investigate DNA TF binding site recognition as a multi-label classification problem, achieving reliable predictions for multiple TFs on DNA sequences retrieved in public repositories. Our deep learning models are based on Temporal Convolutional Networks (TCNs), which are able to predict multiple TF binding profiles, capturing correlations among TFs andtheir cooperative regulatory mechanisms. Our results suggest that multi-label learning leading to reliable predictive performances can reveal biologically meaningful motifs and co-binding patterns consistent with known TF interactions, while also suggesting novel relationships and cooperation among TFs.", "date": "2026-03-12", "categories": []}
{"id": "2512.09566", "title": "Toward Closed-loop Molecular Discovery via Language Model, Property Alignment and Strategic Search", "abstract": "Drug discovery is a time-consuming and expensive process, with traditional high-throughput and docking-based virtual screening hampered by low success rates and limited scalability. Recent advances in generative modelling, including autoregressive, diffusion, and flow-based approaches, have enabled de novo ligand design beyond the limits of enumerative screening. Yet these models often suffer from inadequate generalization, limited interpretability, and an overemphasis on binding affinity at the expense of key pharmacological properties, thereby restricting their translational utility. Here we present Trio, a molecular generation framework integrating fragment-based molecular language modeling, reinforcement learning, and Monte Carlo tree search, for effective and interpretable closed-loop targeted molecular design. Through the three key components, Trio enables context-aware fragment assembly, enforces physicochemical and synthetic feasibility, and guides a balanced search between the exploration of novel chemotypes and the exploitation of promising intermediates within protein binding pockets. Experimental results show that Trio reliably achieves chemically valid and pharmacologically enhanced ligands, outperforming state-of-the-art approaches with improved binding affinity (+7.85%), drug-likeness (+11.10%) and synthetic accessibility (+12.05%), while expanding molecular diversity more than fourfold. By combining generalization, plausibility, and interpretability, Trio establishes a closed-loop generative paradigm that redefines how chemical space can be navigated, offering a transformative foundation for the next era of AI-driven drug discovery.", "date": "2025-12-10", "categories": []}
{"id": "2603.10261", "title": "Discovery of a Hematopoietic Manifold in scGPT Yields a Method for Extracting Performant Algorithms from Biological Foundation Model Internals", "abstract": "We report the discovery and extraction of a compact hematopoietic algorithm from the single-cell foundation model scGPT, to our knowledge the first biologically useful, competitive algorithm extracted from a foundation model via mechanistic interpretability. We show that scGPT internally encodes a compact hematopoietic manifold with significant developmental branch structure, validated on a strict non-overlap Tabula Sapiens external panel and confirmed via frozen-head zero-shot transfer to an independent multi-donor immune panel. To isolate this geometry, we introduce a general three-stage extraction method consisting of direct operator export from frozen attention weights, a lightweight learned adaptor, and a task-specific readout, producing a standalone algorithm without target-dataset retraining. In 88-split donor-holdout benchmarks against scVI, Palantir, DPT, CellTypist, PCA, and raw-expression baselines, the extracted algorithm achieves the strongest pseudotime-depth ordering and leads on key subtype endpoints (CD4/CD8 AUROC 0.867, mono/macro AUROC 0.951). Compared to standard probing of frozen scGPT embeddings with a 3-layer MLP, the extracted head is BH-significantly better on 6/8 classification endpoints while completing a full 12-split evaluation campaign 34.5x faster with approximately 1000x fewer trainable parameters. The exported operator compresses from three pooled attention heads to a single head without statistically significant loss, and further to a rank-64 surrogate. Mechanistic interpretability of the compact operator reveals a concentrated four-factor core explaining 66.2% of ablation impact, with factors resolving into explicit T/lymphoid, B/plasma, granulocytic, and monocyte/macrophage gene programs. A supplementary second-manifold validation (intercellular communication geometry) confirms that the extraction method generalizes beyond hematopoiesis.", "date": "2026-03-10", "categories": []}
{"id": "2603.10873", "title": "SNPgen: Phenotype-Supervised Genotype Representation and Synthetic Data Generation via Latent Diffusion", "abstract": "Polygenic risk scores and other genomic analyses require large individual-level genotype datasets, yet strict data access restrictions impede sharing. Synthetic genotype generation offers a privacy-preserving alternative, but most existing methods operate unconditionally, producing samples without phenotype alignment, or rely on unsupervised compression, creating a gap between statistical fidelity and downstream task utility. We present SNPgen, a two-stage conditional latent diffusion framework for generating phenotype-supervised synthetic genotypes. SNPgen combines GWAS-guided variant selection (1,024-2,048 trait-associated SNPs) with a variational autoencoder for genotype compression and a latent diffusion model conditioned on binary disease labels via classifier-free guidance. Evaluated on 458,724 UK Biobank individuals across four complex diseases (coronary artery disease, breast cancer, type 1 and type 2 diabetes), models trained on synthetic data matched real-data predictive performance in a train-on-synthetic, test-on-real protocol, approaching genome-wide PRS methods that use $2$-$6\\times$ more variants. Privacy analysis confirmed zero identical matches, near-random membership inference (AUC $\\approx 0.50$), preserved linkage disequilibrium structure, and high allele frequency correlation ($r \\geq 0.95$) with source data. A controlled simulation with known causal effects verified faithful recovery of the imposed genetic association structure.", "date": "2026-03-11", "categories": []}
{"id": "2603.10885", "title": "Continuous Diffusion Transformers for Designing Synthetic Regulatory Elements", "abstract": "We present a parameter-efficient Diffusion Transformer (DiT) for generating 200bp cell-type-specific regulatory DNA sequences. By replacing the U-Net backbone of DNA-Diffusion with a transformer denoiser equipped with a 2D CNN input encoder, our model matches the U-Net's best validation loss in 13 epochs (60$\\times$ fewer) and converges 39% lower, while reducing memorization from 5.3% to 1.7% of generated sequences aligning to training data via BLAT. Ablations show the CNN encoder is essential: without it, validation loss increases 70% regardless of positional embedding choice. We further apply DDPO finetuning using Enformer as a reward model, achieving a 38$\\times$ improvement in predicted regulatory activity. Cross-validation against DRAKES on an independent prediction task confirms that improvements reflect genuine regulatory signal rather than reward model overfitting.", "date": "2026-03-11", "categories": []}
{"id": "2501.17901", "title": "Molecular Fingerprints Are Strong Models for Peptide Function Prediction", "abstract": "Understanding peptide properties is often assumed to require modeling long-range molecular interactions, motivating the use of complex graph neural networks and pretrained transformers. Yet, whether such long-range dependencies are essential remains unclear. We investigate if simple, domain-specific molecular fingerprints can capture peptide function without these assumptions. Atomic-level representation aims to provide richer information than purely sequence-based models and better efficiency than structural ones. Across 132 datasets, including LRGB and five other peptide benchmarks, models using count-based ECFP, Topological Torsion, and RDKit fingerprints with LightGBM achieve state-of-the-art accuracy. Despite encoding only short-range molecular features, these models outperform GNNs and transformer-based approaches. Control experiments with sequence shuffling and amino acid counts confirm that fingerprints, though inherently local, suffice for robust peptide property prediction. Our results challenge the presumed necessity of long-range interaction modeling and highlight molecular fingerprints as efficient, interpretable, and computationally lightweight alternatives for peptide prediction.", "date": "2025-01-29", "categories": []}
{"id": "2603.06748", "title": "Property-driven Protein Inverse Folding With Multi-Objective Preference Alignment", "abstract": "Protein sequence design must balance designability, defined as the ability to recover a target backbone, with multiple, often competing, developability properties such as solubility, thermostability, and expression. Existing approaches address these properties through post hoc mutation, inference-time biasing, or retraining on property-specific subsets, yet they are target dependent and demand substantial domain expertise or careful hyperparameter tuning. In this paper, we introduce ProtAlign, a multi-objective preference alignment framework that fine-tunes pretrained inverse folding models to satisfy diverse developability objectives while preserving structural fidelity. ProtAlign employs a semi-online Direct Preference Optimization strategy with a flexible preference margin to mitigate conflicts among competing objectives and constructs preference pairs using in silico property predictors. Applied to the widely used ProteinMPNN backbone, the resulting model MoMPNN enhances developability without compromising designability across tasks including sequence design for CATH 4.3 crystal structures, de novo generated backbones, and real-world binder design scenarios, making it an appealing framework for practical protein sequence design.", "date": "2026-03-06", "categories": []}
{"id": "2505.23354", "title": "Representing local protein environments with machine learning force fields", "abstract": "The local structure of a protein strongly impacts its function and interactions with other molecules. Therefore, a concise, informative representation of a local protein environment is essential for modeling and designing proteins and biomolecular interactions. However, these environments' extensive structural and chemical variability makes them challenging to model, and such representations remain under-explored. In this work, we propose a novel representation for a local protein environment derived from the intermediate features of atomistic foundation models (AFMs). We demonstrate that this embedding effectively captures both local structure (e.g., secondary motifs), and chemical features (e.g., amino-acid identity and protonation state). We further show that the AFM-derived representation space exhibits meaningful structure, enabling the construction of data-driven priors over the distribution of biomolecular environments. Finally, in the context of biomolecular NMR spectroscopy, we demonstrate that the proposed representations enable a first-of-its-kind physics-informed chemical shift predictor that achieves state-of-the-art accuracy. Our results demonstrate the surprising effectiveness of atomistic foundation models and their emergent representations for protein modeling beyond traditional molecular simulations. We believe this will open new lines of work in constructing effective functional representations for protein environments.", "date": "2025-05-29", "categories": []}
{"id": "2512.00126", "title": "RadDiff: Retrieval-Augmented Denoising Diffusion for Protein Inverse Folding", "abstract": "Protein inverse folding, the design of an amino acid sequence based on a target protein structure, is a fundamental problem of computational protein engineering. Existing methods either generate sequences without leveraging external knowledge or relying on protein language models~(PLMs). The former omits the knowledge stored in natural protein data, while the latter is parameter-inefficient and inflexible to adapt to ever-growing protein data. To overcome the above drawbacks, in this paper we propose a novel method, called $\\underline{\\text{r}}$etrieval-$\\underline{\\text{a}}$ugmented $\\underline{\\text{d}}$enoising $\\underline{\\text{diff}}$usion~($\\mbox{RadDiff}$), for protein inverse folding. In RadDiff, a novel retrieval-augmentation mechanism is designed to capture the up-to-date protein knowledge. We further design a knowledge-aware diffusion model that integrates this protein knowledge into the diffusion process via a lightweight module. Experimental results on the CATH, TS50, and PDB2022 datasets show that $\\mbox{RadDiff}$ consistently outperforms existing methods, improving sequence recovery rate by up to 19\\%. Experimental results also demonstrate that RadDiff generates highly foldable sequences and scales effectively with database size.", "date": "2025-11-28", "categories": []}
{"id": "2603.01396", "title": "HarmonyCell: Automating Single-Cell Perturbation Modeling under Semantic and Distribution Shifts", "abstract": "Single-cell perturbation studies face dual heterogeneity bottlenecks: (i) semantic heterogeneity--identical biological concepts encoded under incompatible metadata schemas across datasets; and (ii) statistical heterogeneity--distribution shifts from biological variation demanding dataset-specific inductive biases. We propose HarmonyCell, an end-to-end agent framework resolving each challenge through a dedicated mechanism: an LLM-driven Semantic Unifier autonomously maps disparate metadata into a canonical interface without manual intervention; and an adaptive Monte Carlo Tree Search engine operates over a hierarchical action space to synthesize architectures with optimal statistical inductive biases for distribution shifts. Evaluated across diverse perturbation tasks under both semantic and distribution shifts, HarmonyCell achieves a 95% valid execution rate on heterogeneous input datasets (versus 0% for general agents) while matching or even exceeding expert-designed baselines in rigorous out-of-distribution evaluations. This dual-track orchestration enables scalable automatic virtual cell modeling without dataset-specific engineering.", "date": "2026-03-02", "categories": []}
{"id": "2603.06618", "title": "Distilling and Adapting: A Topology-Aware Framework for Zero-Shot Interaction Prediction in Multiplex Biological Networks", "abstract": "Multiplex Biological Networks (MBNs), which represent multiple interaction types between entities, are crucial for understanding complex biological systems. Yet, existing methods often inadequately model multiplexity, struggle to integrate structural and sequence information, and face difficulties in zero-shot prediction for unseen entities with no prior neighbourhood information. To address these limitations, we propose a novel framework for zero-shot interaction prediction in MBNs by leveraging context-aware representation learning and knowledge distillation. Our approach leverages domain-specific foundation models to generate enriched embeddings, introduces a topology-aware graph tokenizer to capture multiplexity and higher-order connectivity, and employs contrastive learning to align embeddings across modalities. A teacher-student distillation strategy further enables robust zero-shot generalization. Experimental results demonstrate that our framework outperforms state-of-the-art methods in interaction prediction for MBNs, providing a powerful tool for exploring various biological interactions and advancing personalized therapeutics.", "date": "2026-02-20", "categories": []}
{"id": "2603.06950", "title": "How Private Are DNA Embeddings? Inverting Foundation Model Representations of Genomic Sequences", "abstract": "DNA foundation models have become transformative tools in bioinformatics and healthcare applications. Trained on vast genomic datasets, these models can be used to generate sequence embeddings, dense vector representations that capture complex genomic information. These embeddings are increasingly being shared via Embeddings-as-a-Service (EaaS) frameworks to facilitate downstream tasks, while supposedly protecting the privacy of the underlying raw sequences. However, as this practice becomes more prevalent, the security of these representations is being called into question. This study evaluates the resilience of DNA foundation models to model inversion attacks, whereby adversaries attempt to reconstruct sensitive training data from model outputs. In our study, the model's output for reconstructing the DNA sequence is a zero-shot embedding, which is then fed to a decoder. We evaluated the privacy of three DNA foundation models: DNABERT-2, Evo 2, and Nucleotide Transformer v2 (NTv2). Our results show that per-token embeddings allow near-perfect sequence reconstruction across all models. For mean-pooled embeddings, reconstruction quality degrades as sequence length increases, though it remains substantially above random baselines. Evo 2 and NTv2 prove to be most vulnerable, especially for shorter sequences with reconstruction similarities > 90%, while DNABERT-2's BPE tokenization provides the greatest resilience. We found that the correlation between embedding similarity and sequence similarity was a key predictor of reconstruction success. Our findings emphasize the urgent need for privacy-aware design in genomic foundation models prior to their widespread deployment in EaaS settings. Training code, model weights and evaluation pipeline are released on: https://github.com/not-a-feature/DNA-Embedding-Inversion.", "date": "2026-03-06", "categories": []}
{"id": "2603.07233", "title": "Retrieval-Augmented Generation for Predicting Cellular Responses to Gene Perturbation", "abstract": "Predicting how cells respond to genetic perturbations is fundamental to understanding gene function, disease mechanisms, and therapeutic development. While recent deep learning approaches have shown promise in modeling single-cell perturbation responses, they struggle to generalize across cell types and perturbation contexts due to limited contextual information during generation. We introduce PT-RAG (Perturbation-aware Two-stage Retrieval-Augmented Generation), a novel framework that extends Retrieval-Augmented Generation beyond traditional language-model applications to cellular biology. Unlike standard RAG systems designed for text retrieval with pre-trained LLMs, perturbation retrieval lacks established similarity metrics and requires learning what constitutes relevant context, making differentiable retrieval essential. PT-RAG addresses this through a two-stage pipeline: first, retrieving candidate perturbations $K$ using GenePT embeddings, then adaptively refining the selection through Gumbel-Softmax discrete sampling conditioned on both the cell state and the input perturbation. This cell-type-aware differentiable retrieval enables end-to-end optimization of the retrieval objective jointly with generation. On the Replogle-Nadig single-gene perturbation dataset, we demonstrate that PT-RAG outperforms both STATE and vanilla RAG under identical experimental conditions, with the strongest gains in distributional similarity metrics ($W_1$, $W_2$). Notably, vanilla RAG's dramatic failure is itself a key finding: it demonstrates that differentiable, cell-type-aware retrieval is essential in this domain, and that naive retrieval can actively harm performance. Our results establish retrieval-augmented generation as a promising paradigm for modelling cellular responses to gene perturbation. The code to reproduce our experiments is available at https://github.com/difra100/PT-RAG_ICLR.", "date": "2026-03-07", "categories": []}
{"id": "2603.07472", "title": "Contact-Guided 3D Genome Structure Generation of E. coli via Diffusion Transformers", "abstract": "In this study, we present a conditional diffusion-transformer framework for generating ensembles of three-dimensional Escherichia coli genome conformations guided by Hi-C contact maps. Instead of producing a single deterministic structure, we formulate genome reconstruction as a conditional generative modeling problem that samples heterogeneous conformations whose ensemble-averaged contacts are consistent with the input Hi-C data. A synthetic dataset is constructed using coarse-grained molecular dynamics simulations to generate chromatin ensembles and corresponding Hi-C maps under circular topology. Our models operate in a latent diffusion setting with a variational autoencoder that preserves per-bin alignment and supports replication-aware representations. Hi-C information is injected through a transformer-based encoder and cross-attention, enforcing a physically interpretable one-way constraint from Hi-C to structure. The model is trained using a flow-matching objective for stable optimization. On held-out ensembles, generated structures reproduce the input Hi-C distance-decay and structural correlation metrics while maintaining substantial conformational diversity, demonstrating the effectiveness of diffusion-based generative modeling for ensemble-level 3D genome reconstruction.", "date": "2026-03-08", "categories": []}
{"id": "2603.07710", "title": "Reverse Distillation: Consistently Scaling Protein Language Model Representations", "abstract": "Unlike the predictable scaling laws in natural language processing and computer vision, protein language models (PLMs) scale poorly: for many tasks, models within the same family plateau or even decrease in performance, with mid-sized models often outperforming the largest in the family. We introduce Reverse Distillation, a principled framework that decomposes large PLM representations into orthogonal subspaces guided by smaller models of the same family. The resulting embeddings have a nested, Matryoshka-style structure: the first k dimensions of a larger model's embedding are exactly the representation from the smaller model. This ensures that larger reverse-distilled models consistently outperform smaller ones. A motivating intuition is that smaller models, constrained by capacity, preferentially encode broadly-shared protein features. Reverse distillation isolates these shared features and orthogonally extracts additional contributions from larger models, preventing interference between the two. On ProteinGym benchmarks, reverse-distilled ESM-2 variants outperform their respective baselines at the same embedding dimensionality, with the reverse-distilled 15 billion parameter model achieving the strongest performance. Our framework is generalizable to any model family where scaling challenges persist. Code and trained models are available at https://github.com/rohitsinghlab/plm_reverse_distillation.", "date": "2026-03-08", "categories": []}
{"id": "2405.16861", "title": "BInD: Bond and Interaction-generating Diffusion Model for Multi-objective Structure-based Drug Design", "abstract": "Recent remarkable advancements in geometric deep generative models, coupled with accumulated structural data, enable structure-based drug design (SBDD) using only target protein information. However, existing models often struggle to balance multiple objectives, excelling only in specific tasks. BInD, a diffusion model with knowledge-based guidance, is introduced to address this limitation by co-generating molecules and their interactions with a target protein. This approach ensures balanced consideration of key objectives, including target-specific interactions, molecular properties, and local geometry. Comprehensive evaluations demonstrate that BInD achieves robust performance across all objectives, matching or surpassing state-of-the-art methods. Additionally, an NCI-driven molecule design and optimization method is proposed, enabling the enhancement of target binding and specificity by elaborating the adequate interaction patterns.", "date": "2024-05-27", "categories": []}
{"id": "2510.04377", "title": "TCR-EML: Explainable Model Layers for TCR-pMHC Prediction", "abstract": "T cell receptor (TCR) recognition of peptide-MHC (pMHC) complexes is a central component of adaptive immunity, with implications for vaccine design, cancer immunotherapy, and autoimmune disease. While recent advances in machine learning have improved prediction of TCR-pMHC binding, the most effective approaches are black-box transformer models that cannot provide a rationale for predictions. Post-hoc explanation methods can provide insight with respect to the input but do not explicitly model biochemical mechanisms (e.g. known binding regions), as in TCR-pMHC binding. ``Explain-by-design'' models (i.e., with architectural components that can be examined directly after training) have been explored in other domains, but have not been used for TCR-pMHC binding. We propose explainable model layers (TCR-EML) that can be incorporated into protein-language model backbones for TCR-pMHC modeling. Our approach uses prototype layers for amino acid residue contacts drawn from known TCR-pMHC binding mechanisms, enabling high-quality explanations for predicted TCR-pMHC binding. Experiments of our proposed method on large-scale datasets demonstrate competitive predictive accuracy and generalization, and evaluation on the TCR-XAI benchmark demonstrates improved explainability compared with existing approaches.", "date": "2025-10-05", "categories": []}
{"id": "2601.13327", "title": "PepEDiff: Zero-Shot Peptide Binder Design via Protein Embedding Diffusion", "abstract": "We present PepEDiff, a novel peptide binder generator that designs binding sequences given a target receptor protein sequence and its pocket residues. Peptide binder generation is critical in therapeutic and biochemical applications, yet many existing methods rely heavily on intermediate structure prediction, adding complexity and limiting sequence diversity. Our approach departs from this paradigm by generating binder sequences directly in a continuous latent space derived from a pretrained protein embedding model, without relying on predicted structures, thereby improving structural and sequence diversity. To encourage the model to capture binding-relevant features rather than memorizing known sequences, we perform latent-space exploration and diffusion-based sampling, enabling the generation of peptides beyond the limited distribution of known binders. This zero-shot generative strategy leverages the global protein embedding manifold as a semantic prior, allowing the model to propose novel peptide sequences in previously unseen regions of the protein space. We evaluate PepEDiff on TIGIT, a challenging target with a large, flat protein-protein interaction interface that lacks a druggable pocket. Despite its simplicity, our method outperforms state-of-the-art approaches across benchmark tests and in the TIGIT case study, demonstrating its potential as a general, structure-free framework for zero-shot peptide binder design. The code for this research is available at GitHub: https://github.com/LabJunBMI/PepEDiff-An-Peptide-binder-Embedding-Diffusion-Model", "date": "2026-01-19", "categories": []}
{"id": "2602.10152", "title": "Validating Interpretability in siRNA Efficacy Prediction: A Perturbation-Based, Dataset-Aware Protocol", "abstract": "Saliency maps are increasingly used as design guidance in siRNA efficacy prediction, yet attribution methods are rarely validated before motivating sequence edits. We introduce a pre-synthesis gate: a protocol for counterfactual sensitivity faithfulness that tests whether mutating high-saliency positions changes model output more than composition-matched controls. Cross-dataset transfer reveals two failure modes that would otherwise go undetected: faithful-but-wrong (saliency valid, predictions fail) and inverted saliency (top-saliency edits less impactful than random). Strikingly, models trained on mRNA-level assays collapse on a luciferase reporter dataset, demonstrating that protocol shifts can silently invalidate deployment. Across four benchmarks, 19/20 fold instances pass; the single failure shows inverted saliency. A biology-informed regularizer (BioPrior) strengthens saliency faithfulness with modest, dataset-dependent predictive trade-offs. Our results establish saliency validation as essential pre-deployment practice for explanation-guided therapeutic design. Code is available at https://github.com/shadi97kh/BioPrior.", "date": "2026-02-09", "categories": []}
{"id": "2602.22289", "title": "What Topological and Geometric Structure Do Biological Foundation Models Learn? Evidence from 141 Hypotheses", "abstract": "When biological foundation models such as scGPT and Geneformer process single-cell gene expression, what geometric and topological structure forms in their internal representations? Is that structure biologically meaningful or a training artifact, and how confident should we be in such claims? We address these questions through autonomous large-scale hypothesis screening: an AI-driven executor-brainstormer loop that proposed, tested, and refined 141 geometric and topological hypotheses across 52 iterations, covering persistent homology, manifold distances, cross-model alignment, community structure, and directed topology, all with explicit null controls and disjoint gene-pool splits.\n  Three principal findings emerge. First, the models learn genuine geometric structure. Gene embedding neighborhoods exhibit non-trivial topology, with persistent homology significant in 11 of 12 transformer layers at p < 0.05 in the weakest domain and 12 of 12 in the other two. A multi-level distance hierarchy shows that manifold-aware metrics outperform Euclidean distance for identifying regulatory gene pairs, and graph community partitions track known transcription factor target relationships. Second, this structure is shared across independently trained models. CCA alignment between scGPT and Geneformer yields canonical correlation of 0.80 and gene retrieval accuracy of 72 percent, yet none of 19 tested methods reliably recover gene-level correspondences. The models agree on the global shape of gene space but not on precise gene placement. Third, the structure is more localized than it first appears. Under stringent null controls applied across all null families, robust signal concentrates in immune tissue, while lung and external lung signals weaken substantially.", "date": "2026-02-25", "categories": []}
{"id": "2603.02406", "title": "Rigidity-Aware Geometric Pretraining for Protein Design and Conformational Ensembles", "abstract": "Generative models have recently advanced $\\textit{de novo}$ protein design by learning the statistical regularities of natural structures. However, current approaches face three key limitations: (1) Existing methods cannot jointly learn protein geometry and design tasks, where pretraining can be a solution; (2) Current pretraining methods mostly rely on local, non-rigid atomic representations for property prediction downstream tasks, limiting global geometric understanding for protein generation tasks; and (3) Existing approaches have yet to effectively model the rich dynamic and conformational information of protein structures. To overcome these issues, we introduce $\\textbf{RigidSSL}$ ($\\textit{Rigidity-Aware Self-Supervised Learning}$), a geometric pretraining framework that front-loads geometry learning prior to generative finetuning. Phase I (RigidSSL-Perturb) learns geometric priors from 432K structures from the AlphaFold Protein Structure Database with simulated perturbations. Phase II (RigidSSL-MD) refines these representations on 1.3K molecular dynamics trajectories to capture physically realistic transitions. Underpinning both phases is a bi-directional, rigidity-aware flow matching objective that jointly optimizes translational and rotational dynamics to maximize mutual information between conformations. Empirically, RigidSSL variants improve designability by up to 43% while enhancing novelty and diversity in unconditional generation. Furthermore, RigidSSL-Perturb improves the success rate by 5.8% in zero-shot motif scaffolding and RigidSSL-MD captures more biophysically realistic conformational ensembles in G protein-coupled receptor modeling.", "date": "2026-03-02", "categories": []}
{"id": "2603.05532", "title": "On the Reliability of AI Methods in Drug Discovery: Evaluation of Boltz-2 for Structure and Binding Affinity Prediction", "abstract": "Despite continuing hype about the role of AI in drug discovery, no \"AI-discovered drugs\" have so far received regulatory approval. Here we assess one of the latest AI based tools in this domain. The ability to rapidly predict protein-ligand structures and binding affinities is pivotal for accelerating drug discovery. Boltz-2, a recently developed biomolecular foundation model, aims to bridge the gap between AI efficiency and physics-based precision through a joint \"co-folding\" approach. In this study, we provide an extensive evaluation of Boltz-2 using two large-scale datasets: 16,780 compounds for 3CLPro and 21,702 compounds for TNKS2. We compare Boltz-2 predicted structures with traditional docking and binding affinities with binding free energies derived from the physics-based ESMACS protocol. Structural analysis reveals significant global RMSD variations, indicating that Boltz-2 predicts multiple protein conformations and ligand binding positions rather than a single converged pose. Energetic evaluations exhibit only weak to moderate correlations across the global datasets. Furthermore, a focused analysis of the top 100 compounds yields no significant correlation between the Boltz-2 predictions and the binding free energies from fine-grained ESMACS, alongside observed saturation difference in ligand structures. Our results show that while Boltz-2 offers substantial speed for initial screening, it lacks the energetic resolution required for lead identification. These findings highlight the necessity of employing physics-based methods for the reliability and refinement of AI-derived models.", "date": "2026-03-02", "categories": []}
{"id": "2602.24007", "title": "Inference-time optimization for experiment-grounded protein ensemble generation", "abstract": "Protein function relies on dynamic conformational ensembles, yet current generative models like AlphaFold3 often fail to produce ensembles that match experimental data. Recent experiment-guided generators attempt to address this by steering the reverse diffusion process. However, these methods are limited by fixed sampling horizons and sensitivity to initialization, often yielding thermodynamically implausible results. We introduce a general inference-time optimization framework to solve these challenges. First, we optimize over latent representations to maximize ensemble log-likelihood, rather than perturbing structures post hoc. This approach eliminates dependence on diffusion length, removes initialization bias, and easily incorporates external constraints. Second, we present novel sampling schemes for drawing Boltzmann-weighted ensembles. By combining structural priors from AlphaFold3 with force-field-based priors, we sample from their product distribution while balancing experimental likelihoods. Our results show that this framework consistently outperforms state-of-the-art guidance, improving diversity, physical energy, and agreement with data in X-ray crystallography and NMR, often fitting the experimental data better than deposited PDB structures. Finally, inference-time optimization experiments maximizing ipTM scores reveal that perturbing AlphaFold3 embeddings can artificially inflate model confidence. This exposes a vulnerability in current design metrics, whose mitigation could offer a pathway to reduce false discovery rates in binder engineering.", "date": "2026-02-27", "categories": []}
{"id": "2603.04480", "title": "AbAffinity: A Large Language Model for Predicting Antibody Binding Affinity against SARS-CoV-2", "abstract": "Machine learning-based antibody design is emerging as one of the most promising approaches to combat infectious diseases, due to significant advancements in the field of artificial intelligence and an exponential surge in experimental antibody data (in particular related to COVID-19). The ability of an antibody to bind to an antigens (called binding affinity) is one of the the most critical properties in designing neutralizing antibodies. In this study we introduce Ab-Affinity, a new large language model that can accurately predict the binding affinity of antibodies against a target peptide, e.g., the SARS-CoV-2 spike protein. Code and model are available at https://github.com/ucrbioinfo/AbAffinity.", "date": "2026-03-04", "categories": []}
{"id": "2603.01752", "title": "Causal Circuit Tracing Reveals Distinct Computational Architectures in Single-Cell Foundation Models: Inhibitory Dominance, Biological Coherence, and Cross-Model Convergence", "abstract": "Motivation: Sparse autoencoders (SAEs) decompose foundation model activations into interpretable features, but causal feature-to-feature interactions across network depth remain unknown for biological foundation models.\n  Results: We introduce causal circuit tracing by ablating SAE features and measuring downstream responses, and apply it to Geneformer V2-316M and scGPT whole-human across four conditions (96,892 edges, 80,191 forward passes). Both models show approximately 53 percent biological coherence and 65 to 89 percent inhibitory dominance, invariant to architecture and cell type. scGPT produces stronger effects (mean absolute d = 1.40 vs. 1.05) with more balanced dynamics. Cross-model consensus yields 1,142 conserved domain pairs (10.6x enrichment, p < 0.001). Disease-associated domains are 3.59x more likely to be consensus. Gene-level CRISPRi validation shows 56.4 percent directional accuracy, confirming co-expression rather than causal encoding.", "date": "2026-03-02", "categories": []}
{"id": "2510.08946", "title": "Physically Valid Biomolecular Interaction Modeling with Gauss-Seidel Projection", "abstract": "Biomolecular interaction modeling has been substantially advanced by foundation models, yet they often produce all-atom structures that violate basic steric feasibility. We address this limitation by enforcing physical validity as a strict constraint during both training and inference with a uniffed module. At its core is a differentiable projection that maps the provisional atom coordinates from the diffusion model to the nearest physically valid conffguration. This projection is achieved using a Gauss-Seidel scheme, which exploits the locality and sparsity of the constraints to ensure stable and fast convergence at scale. By implicit differentiation to obtain gradients, our module integrates seamlessly into existing frameworks for end-to-end ffnetuning. With our Gauss-Seidel projection module in place, two denoising steps are sufffcient to produce biomolecular complexes that are both physically valid and structurally accurate. Across six benchmarks, our 2-step model achieves the same structural accuracy as state-of-the-art 200-step diffusion baselines, delivering approximately 10 times faster wall-clock speed while guaranteeing physical validity. The code is available at https://github.com/chensiyuan030105/ProteinGS.git.", "date": "2025-10-10", "categories": []}
{"id": "2512.20833", "title": "CHAMMI-75: Pre-training multi-channel models with heterogeneous microscopy images", "abstract": "Quantifying cell morphology using images and machine learning has proven to be a powerful tool to study the response of cells to treatments. However, models used to quantify cellular morphology are typically trained with a single microscopy imaging type. This results in specialized models that cannot be reused across biological studies because the technical specifications do not match (e.g., different number of channels). Here, we present CHAMMI-75, an open access dataset of heterogeneous, multi-channel microscopy images from 75 diverse biological studies. We curated this resource from publicly available sources to investigate cellular morphology models that are channel-adaptive and can process any microscopy image type. Our experiments show that training with CHAMMI-75 can improve performance in multi-channel bioimaging tasks primarily because of its high diversity in microscopy modalities. This work paves the way to create the next generation of cellular morphology models for biological studies.", "date": "2025-12-23", "categories": []}
{"id": "2603.02273", "title": "Graph Attention Based Prioritization of Disease Responsible Genes from Multimodal Alzheimer's Network", "abstract": "Prioritizing disease-associated genes is central to understanding the molecular mechanisms of complex disorders such as Alzheimer's disease (AD). Traditional network-based approaches rely on static centrality measures and often fail to capture cross-modal biological heterogeneity. We propose NETRA (Node Evaluation through Transformer-based Representation and Attention), a multimodal graph transformer framework that replaces heuristic centrality metrics with attention-driven relevance scoring. Using AD as a case study, gene regulatory networks are independently constructed from microarray, single-cell RNA-seq, and single-nucleus RNA-seq data. Random-walk sequences derived from these networks are used to train a BERT-based model for learning global gene embeddings, while modality-specific gene expression profiles are compressed using variational autoencoders. These representations are integrated with auxiliary biological networks, including protein-protein interactions, Gene Ontology semantic similarity, and diffusion-based gene similarity, into a unified multimodal graph. A graph transformer assigns NETRA scores that quantify gene relevance in a disease-specific and context-aware manner. Gene set enrichment analysis shows that NETRA achieves a normalized enrichment score of about 3.9 for the Alzheimer's disease pathway, substantially outperforming classical centrality measures and diffusion models. Top-ranked genes enrich multiple neurodegenerative pathways, recover a known late-onset AD susceptibility locus at chr12q13, and reveal conserved cross-disease gene modules. The framework preserves biologically realistic heavy-tailed network topology and is readily extensible to other complex disorders.", "date": "2026-03-01", "categories": []}
{"id": "2603.02753", "title": "Deep learning-guided evolutionary optimization for protein design", "abstract": "Designing novel proteins with desired characteristics remains a significant challenge due to the large sequence space and the complexity of sequence-function relationships. Efficient exploration of this space to identify sequences that meet specific design criteria is crucial for advancing therapeutics and biotechnology. Here, we present BoGA (Bayesian Optimization Genetic Algorithm), a framework that combines evolutionary search with Bayesian optimization to efficiently navigate the sequence space. By integrating a genetic algorithm as a stochastic proposal generator within a surrogate modeling loop, BoGA prioritizes candidates based on prior evaluations and surrogate model predictions, enabling data-efficient optimization. We demonstrate the utility of BoGA through benchmarking on sequence and structure design tasks, followed by its application in designing peptide binders against pneumolysin, a key virulence factor of \\textit{Streptococcus pneumoniae}. BoGA accelerates the discovery of high-confidence binders, demonstrating the potential for efficient protein design across diverse objectives. The algorithm is implemented within the BoPep suite and is available under an MIT license at \\href{https://github.com/ErikHartman/bopep}{GitHub}.", "date": "2026-03-03", "categories": []}
{"id": "2603.02952", "title": "Sparse autoencoders reveal organized biological knowledge but minimal regulatory logic in single-cell foundation models: a comparative atlas of Geneformer and scGPT", "abstract": "Background: Single-cell foundation models such as Geneformer and scGPT encode rich biological information, but whether this includes causal regulatory logic rather than statistical co-expression remains unclear. Sparse autoencoders (SAEs) can resolve superposition in neural networks by decomposing dense activations into interpretable features, yet they have not been systematically applied to biological foundation models.\n  Results: We trained TopK SAEs on residual stream activations from all layers of Geneformer V2-316M (18 layers, d=1152) and scGPT whole-human (12 layers, d=512), producing atlases of 82525 and 24527 features, respectively. Both atlases confirm massive superposition, with 99.8 percent of features invisible to SVD. Systematic characterization reveals rich biological organization: 29 to 59 percent of features annotate to Gene Ontology, KEGG, Reactome, STRING, or TRRUST, with U-shaped layer profiles reflecting hierarchical abstraction. Features organize into co-activation modules (141 in Geneformer, 76 in scGPT), exhibit causal specificity (median 2.36x), and form cross-layer information highways (63 to 99.8 percent). When tested against genome-scale CRISPRi perturbation data, only 3 of 48 transcription factors (6.2 percent) show regulatory-target-specific feature responses. A multi-tissue control yields marginal improvement (10.4 percent, 5 of 48 TFs), establishing model representations as the bottleneck.\n  Conclusions: These models have internalized organized biological knowledge, including pathway membership, protein interactions, functional modules, and hierarchical abstraction, yet they encode minimal causal regulatory logic. We release both feature atlases as interactive web platforms enabling exploration of more than 107000 features across 30 layers of two leading single-cell foundation models.", "date": "2026-03-03", "categories": []}
{"id": "2405.08205", "title": "Generative Enzyme Design Guided by Functionally Important Sites and Small-Molecule Substrates", "abstract": "Enzymes are genetically encoded biocatalysts capable of accelerating chemical reactions. How can we automatically design functional enzymes? In this paper, we propose EnzyGen, an approach to learn a unified model to design enzymes across all functional families. Our key idea is to generate an enzyme's amino acid sequence and their three-dimensional (3D) coordinates based on functionally important sites and substrates corresponding to a desired catalytic function. These sites are automatically mined from enzyme databases. EnzyGen consists of a novel interleaving network of attention and neighborhood equivariant layers, which captures both long-range correlation in an entire protein sequence and local influence from nearest amino acids in 3D space. To learn the generative model, we devise a joint training objective, including a sequence generation loss, a position prediction loss and an enzyme-substrate interaction loss. We further construct EnzyBench, a dataset with 3157 enzyme families, covering all available enzymes within the protein data bank (PDB). Experimental results show that our EnzyGen consistently achieves the best performance across all 323 testing families, surpassing the best baseline by 10.79% in terms of substrate binding affinity. These findings demonstrate EnzyGen's superior capability in designing well-folded and effective enzymes binding to specific substrates with high affinities.", "date": "2024-05-13", "categories": []}
{"id": "2406.16227", "title": "VICatMix: variational Bayesian clustering and variable selection for discrete biomedical data", "abstract": "Effective clustering of biomedical data is crucial in precision medicine, enabling accurate stratifiction of patients or samples. However, the growth in availability of high-dimensional categorical data, including `omics data, necessitates computationally efficient clustering algorithms. We present VICatMix, a variational Bayesian finite mixture model designed for the clustering of categorical data. The use of variational inference (VI) in its training allows the model to outperform competitors in term of efficiency, while maintaining high accuracy. VICatMix furthermore performs variable selection, enhancing its performance on high-dimensional, noisy data. The proposed model incorporates summarisation and model averaging to mitigate poor local optima in VI, allowing for improved estimation of the true number of clusters simultaneously with feature saliency. We demonstrate the performance of VICatMix with both simulated and real-world data, including applications to datasets from The Cancer Genome Atlas (TCGA), showing its use in cancer subtyping and driver gene discovery. We demonstrate VICatMix's utility in integrative cluster analysis with different `omics datasets, enabling the discovery of novel subtypes.\n  \\textbf{Availability:} VICatMix is freely available as an R package, incorporating C++ for faster computation, at https://github.com/j-ackierao/VICatMix.", "date": "2024-06-23", "categories": []}
{"id": "2411.02109", "title": "One protein is all you need", "abstract": "Generalization beyond training data remains a central challenge in machine learning for biology. A common way to enhance generalization is self-supervised pre-training on large datasets. However, aiming to perform well on all possible proteins can limit a model's capacity to excel on any specific one, whereas experimentalists typically need accurate predictions for individual proteins they study, often not covered in training data. To address this limitation, we propose a method that enables self-supervised customization of protein language models to one target protein at a time, on the fly, and without assuming any additional data. We show that our Protein Test-Time Training (ProteinTTT) method consistently enhances generalization across different models, their sizes, and datasets. ProteinTTT improves structure prediction for challenging targets, achieves new state-of-the-art results on protein fitness prediction, and enhances function prediction on two tasks. Through two challenging case studies, we also show that customization via ProteinTTT achieves more accurate antibody-antigen loop modeling and enhances 19% of structures in the Big Fantastic Virus Database, delivering improved predictions where general-purpose AlphaFold2 and ESMFold struggle.", "date": "2024-11-04", "categories": []}
{"id": "2506.02052", "title": "General Protein Pretraining or Domain-Specific Designs? Benchmarking Protein Modeling on Realistic Applications", "abstract": "Recently, extensive deep learning architectures and pretraining strategies have been explored to support downstream protein applications. Additionally, domain-specific models incorporating biological knowledge have been developed to enhance performance in specialized tasks. In this work, we introduce $\\textbf{Protap}$, a comprehensive benchmark that systematically compares backbone architectures, pretraining strategies, and domain-specific models across diverse and realistic downstream protein applications. Specifically, Protap covers five applications: three general tasks and two novel specialized tasks, i.e., enzyme-catalyzed protein cleavage site prediction and targeted protein degradation, which are industrially relevant yet missing from existing benchmarks. For each application, Protap compares various domain-specific models and general architectures under multiple pretraining settings. Our empirical studies imply that: (i) Though large-scale pretraining encoders achieve great results, they often underperform supervised encoders trained on small downstream training sets. (ii) Incorporating structural information during downstream fine-tuning can match or even outperform protein language models pretrained on large-scale sequence corpora. (iii) Domain-specific biological priors can enhance performance on specialized downstream tasks. Code and datasets are publicly available at https://github.com/Trust-App-AI-Lab/protap.", "date": "2025-06-01", "categories": []}
{"id": "2506.07459", "title": "ProteinZero: Self-Improving Protein Generation via Online Reinforcement Learning", "abstract": "Protein generative models have shown remarkable promise in protein design, yet their success rates remain constrained by reliance on curated sequence-structure datasets and by misalignment between supervised objectives and real design goals. We present ProteinZero, an online reinforcement learning framework for inverse folding models that enables scalable, automated, and continuous self-improvement with computationally efficient feedback. ProteinZero employs a reward pipeline that combines structural guidance from ESMFold with a novel self-derived ddG predictor, providing stable multi-objective signals while avoiding the prohibitive cost of physics-based methods. To ensure robustness in online RL, we further introduce a novel embedding-level diversity regularizer that mitigates mode collapse and promotes functionally meaningful sequence variation. Within a general RL formulation balancing multi-reward optimization, KL-divergence from a reference model, and diversity regularization, ProteinZero achieves robust improvements across designability, stability, recovery, and diversity. On the CATH-4.3 benchmark, it consistently outperforms state-of-the-art baselines including ProteinMPNN, ESM-IF, and InstructPLM, reducing design failure rates by 36-48% and achieving success rates above 90% across diverse folds. Importantly, a complete RL run can be executed on a single 8 X GPU node within three days, including reward computation and data generation. These results indicate that efficient online RL fine-tuning can complement supervised pretraining by allowing protein generative models to evolve continuously from their own outputs and optimize multiple design objectives without labeled data, opening new possibilities for exploring the vast protein design space. Full source code and model checkpoints will be released upon publication.", "date": "2025-06-09", "categories": []}
{"id": "2506.09332", "title": "InstructPro: Natural Language Guided Ligand-Binding Protein Design", "abstract": "The de novo design of ligand-binding proteins with tailored functions is essential for advancing biotechnology and molecular medicine, yet existing AI approaches are limited by scarce protein-ligand complex data. To circumvent this data bottleneck, we leverage the abundant natural language descriptions characterizing protein-ligand interactions. Here, we introduce InstructPro, a family of generative models that design proteins following the guidance of natural language instructions and ligand formulas. InstructPro produces protein sequences consistent with specified function descriptions and ligand targets. To enable training and evaluation, we develop InstructProBench, a large-scale dataset of 9.6 million (function description, ligand, protein) triples. We train two model variants -- InstructPro-1B and InstructPro-3B -- that substantially outperform strong baselines. InstructPro-1B achieves an AlphaFold3 ipTM of 0.918 and a binding affinity of -8.764 on seen ligands, while maintaining robust performance in a zero-shot setting with scores of 0.869 and -6.713, respectively. These results are accompanied by novelty scores of 70.1% and 68.8%, underscoring the model's ability to generalize beyond the training set. Furthermore, the model yields a superior binding free energy of -20.9 kcal/mol and an average of 5.82 intermolecular hydrogen bonds, validating its proficiency in designing high-affinity ligand-binding proteins. Notably, scaling to InstructPro-3B further improves the zero-shot ipTM to 0.882, binding affinity to -6.797, and binding free energy to -25.8 kcal/mol, demonstrating clear performance gains associated with increased model capacity. These findings highlight the power of natural language-guided generative models to mitigate the data bottlenecks in traditional structure-based methods, significantly broadening the scope of de novo protein design.", "date": "2025-06-11", "categories": []}
{"id": "2511.11758", "title": "Protein Structure Tokenization via Geometric Byte Pair Encoding", "abstract": "Protein structure is central to biological function, and enabling multimodal protein models requires joint reasoning over sequence, structure, and function. A key barrier is the lack of principled protein structure tokenizers (PSTs): existing approaches fix token size or rely on continuous vector codebooks, limiting interpretability, multi-scale control, and transfer across architectures. We introduce GeoBPE, a geometry-grounded PST that transforms continuous, noisy, multi-scale backbone conformations into discrete ``sentences'' of geometry while enforcing global constraints. Analogous to byte-pair encoding, GeoBPE generates a hierarchical vocabulary of geometric primitives by iteratively (i) clustering Geo-Pair occurrences with k-medoids to yield a resolution-controllable vocabulary; (ii) quantizing each Geo-Pair to its closest medoid prototype; and (iii) reducing drift through differentiable inverse kinematics that optimizes boundary glue angles under an $\\mathrm{SE}(3)$ end-frame loss. GeoBPE offers compression ($>$10x reduction in bits-per-residue at similar distortion rate), data efficiency ($>$10x less training data), and generalization (maintains test/train distortion ratio of $1.0-1.1$). It is architecture-agnostic: (a) its hierarchical vocabulary provides a strong inductive bias for coarsening residue-level embeddings from large PLMs into motif- and protein-level representations, consistently outperforming leading PSTs across $12$ tasks and $24$ test splits; (b) paired with a transformer, GeoBPE supports unconditional backbone generation via language modeling; and (c) tokens align with CATH functional families and support expert-interpretable case studies, offering functional meaning absent in prior PSTs. Code is available at https://github.com/shiningsunnyday/PT-BPE/.", "date": "2025-11-13", "categories": []}
{"id": "2603.00143", "title": "GrapHist: Graph Self-Supervised Learning for Histopathology", "abstract": "Self-supervised vision models have achieved notable success in digital pathology. However, their domain-agnostic transformer architectures are not originally designed to account for fundamental biological elements of histopathology images, namely cells and their complex interactions. In this work, we hypothesize that a biologically-informed modeling of tissues as cell graphs offers a more efficient representation learning. Thus, we introduce GrapHist, a novel graph-based self-supervised learning framework for histopathology, which learns generalizable and structurally-informed embeddings that enable diverse downstream tasks. GrapHist integrates masked autoencoders and heterophilic graph neural networks that are explicitly designed to capture the heterogeneity of tumor microenvironments. We pre-train GrapHist on a large collection of 11 million cell graphs derived from breast tissues and evaluate its transferability across in- and out-of-domain benchmarks. Our results show that GrapHist achieves competitive performance compared to its vision-based counterparts in slide-, region-, and cell-level tasks, while requiring four times fewer parameters. It also drastically outperforms fully-supervised graph models on cancer subtyping tasks. Finally, we also release five graph-based digital pathology datasets used in our study at https://huggingface.co/ogutsevda/datasets , establishing the first large-scale graph benchmark in this field. Our code is available at https://github.com/ogutsevda/graphist .", "date": "2026-02-24", "categories": []}
{"id": "2603.01780", "title": "D3LM: A Discrete DNA Diffusion Language Model for Bidirectional DNA Understanding and Generation", "abstract": "Early DNA foundation models adopted BERT-style training, achieving good performance on DNA understanding tasks but lacking generative capabilities. Recent autoregressive models enable DNA generation, but employ left-to-right causal modeling that is suboptimal for DNA where regulatory relationships are inherently bidirectional. We present D3LM (\\textbf{D}iscrete \\textbf{D}NA \\textbf{D}iffusion \\textbf{L}anguage \\textbf{M}odel), which unifies bidirectional representation learning and DNA generation through masked diffusion. D3LM directly adopts the Nucleotide Transformer (NT) v2 architecture but reformulates the training objective as masked diffusion in discrete DNA space, enabling both bidirectional understanding and generation capabilities within a single model. Compared to NT v2 of the same size, D3LM achieves improved performance on understanding tasks. Notably, on regulatory element generation, D3LM achieves an SFID of 10.92, closely approaching real DNA sequences (7.85) and substantially outperforming the previous best result of 29.16 from autoregressive models. Our work suggests diffusion language models as a promising paradigm for unified DNA foundation models. We further present the first systematic study of masked diffusion models in the DNA domain, investigating practical design choices such as tokenization schemes and sampling strategies, thereby providing empirical insights and a solid foundation for future research. D3LM has been released at https://huggingface.co/collections/Hengchang-Liu/d3lm.", "date": "2026-03-02", "categories": []}
{"id": "2512.00306", "title": "VCWorld: A Biological World Model for Virtual Cell Simulation", "abstract": "Virtual cell modeling aims to predict cellular responses to perturbations. Existing virtual cell models rely heavily on large-scale single-cell datasets, learning explicit mappings between gene expression and perturbations. Although recent models attempt to incorporate multi-source biological information, their generalization remains constrained by data quality, coverage, and batch effects. More critically, these models often function as black boxes, offering predictions without interpretability or consistency with biological principles, which undermines their credibility in scientific research. To address these challenges, we present VCWorld, a cell-level white-box simulator that integrates structured biological knowledge with the iterative reasoning capabilities of large language models to instantiate a biological world model. VCWorld operates in a data-efficient manner to reproduce perturbation-induced signaling cascades and generates interpretable, stepwise predictions alongside explicit mechanistic hypotheses. In drug perturbation benchmarks, VCWorld achieves state-of-the-art predictive performance, and the inferred mechanistic pathways are consistent with publicly available biological evidence.", "date": "2025-11-29", "categories": []}
{"id": "2601.17582", "title": "GenAI-Net: A Generative AI Framework for Automated Biomolecular Network Design", "abstract": "Biomolecular networks underpin emerging technologies in synthetic biology-from robust biomanufacturing and metabolic engineering to smart therapeutics and cell-based diagnostics-and also provide a mechanistic language for understanding complex dynamics in natural and ecological systems. Yet designing chemical reaction networks (CRNs) that implement a desired dynamical function remains largely manual: while a proposed network can be checked by simulation, the reverse problem of discovering a network from a behavioral specification is difficult, requiring substantial human insight to navigate a vast space of topologies and kinetic parameters with nonlinear and possibly stochastic dynamics. Here we introduce GenAI-Net, a generative AI framework that automates CRN design by coupling an agent that proposes reactions to simulation-based evaluation defined by a user-specified objective. GenAI-Net efficiently produces novel, topologically diverse solutions across multiple design tasks, including dose responses, complex logic gates, classifiers, oscillators, and robust perfect adaptation in deterministic and stochastic settings (including noise reduction). By turning specifications into families of circuit candidates and reusable motifs, GenAI-Net provides a general route to programmable biomolecular circuit design and accelerates the translation from desired function to implementable mechanisms.", "date": "2026-01-24", "categories": []}
{"id": "2602.07588", "title": "Unified Biomolecular Trajectory Generation via Pretrained Variational Bridge", "abstract": "Molecular Dynamics (MD) simulations provide a fundamental tool for characterizing molecular behavior at full atomic resolution, but their applicability is severely constrained by the computational cost. To address this, a surge of deep generative models has recently emerged to learn dynamics at coarsened timesteps for efficient trajectory generation, yet they either generalize poorly across systems or, due to limited molecular diversity of trajectory data, fail to fully exploit structural information to improve generative fidelity. Here, we present the Pretrained Variational Bridge (PVB) in an encoder-decoder fashion, which maps the initial structure into a noised latent space and transports it toward stage-specific targets through augmented bridge matching. This unifies training on both single-structure and paired trajectory data, enabling consistent use of cross-domain structural knowledge across training stages. Moreover, for protein-ligand complexes, we further introduce a reinforcement learning-based optimization via adjoint matching that speeds progression toward the holo state, which supports efficient post-optimization of docking poses. Experiments on proteins and protein-ligand complexes demonstrate that PVB faithfully reproduces thermodynamic and kinetic observables from MD while delivering stable and efficient generative dynamics.", "date": "2026-02-07", "categories": []}
{"id": "2410.19704", "title": "Multi-view biomedical foundation models for molecule-target and property prediction", "abstract": "Quality molecular representations are key to foundation model development in bio-medical research. Previous efforts have typically focused on a single representation or molecular view, which may have strengths or weaknesses on a given task. We develop Multi-view Molecular Embedding with Late Fusion (MMELON), an approach that integrates graph, image and text views in a foundation model setting and may be readily extended to additional representations. Single-view foundation models are each pre-trained on a dataset of up to 200M molecules. The multi-view model performs robustly, matching the performance of the highest-ranked single-view. It is validated on over 120 tasks, including molecular solubility, ADME properties, and activity against G Protein-Coupled receptors (GPCRs). We identify 33 GPCRs that are related to Alzheimer's disease and employ the multi-view model to select strong binders from a compound screen. Predictions are validated through structure-based modeling and identification of key binding motifs.", "date": "2024-10-25", "categories": []}
{"id": "2507.16801", "title": "Decoding Translation-Related Functional Sequences in 5'UTRs Using Interpretable Deep Learning Models", "abstract": "Understanding how 5' untranslated regions (5'UTRs) regulate mRNA translation is critical for controlling protein expression and designing effective therapeutic mRNAs. While recent deep learning models have shown promise in predicting translational efficiency from 5'UTR sequences, most are constrained by fixed input lengths and limited interpretability. We introduce UTR-STCNet, a Transformer-based architecture for flexible and biologically grounded modeling of variable-length 5'UTRs. UTR-STCNet integrates a Saliency-Aware Token Clustering (SATC) module that iteratively aggregates nucleotide tokens into multi-scale, semantically meaningful units based on saliency scores. A Saliency-Guided Transformer (SGT) block then captures both local and distal regulatory dependencies using a lightweight attention mechanism. This combined architecture achieves efficient and interpretable modeling without input truncation or increased computational cost. Evaluated across three benchmark datasets, UTR-STCNet consistently outperforms state-of-the-art baselines in predicting mean ribosome load (MRL), a key proxy for translational efficiency. Moreover, the model recovers known functional elements such as upstream AUGs and Kozak motifs, highlighting its potential for mechanistic insight into translation regulation.", "date": "2025-07-22", "categories": []}
{"id": "2508.04724", "title": "Understanding protein function with a multimodal retrieval-augmented foundation model", "abstract": "Protein language models (PLMs) learn probability distributions over natural protein sequences. By learning from hundreds of millions of natural protein sequences, protein understanding and design capabilities emerge. Recent works have shown that scaling these models improves structure prediction, but does not seem to improve mutation understanding and representation quality for protein function prediction. We introduce PoET-2, a multimodal, retrieval-augmented protein foundation model that incorporates in-context learning of family-specific evolutionary constraints with optional structure conditioning to learn generative distributions over protein sequences. PoET-2 uses a hierarchical transformer encoder that is equivariant to sequence context ordering and a dual decoder architecture with both causal and masked language modeling objectives, allowing PoET-2 to operate in both fully generative and bidirectional representation learning modes. PoET-2 achieves state-of-the-art performance on zero-shot variant effect prediction, excelling at scoring variants with multiple mutations and challenging indel mutations. In supervised settings, PoET-2 embeddings outperform previous methods for learning sequence-function relationships, especially with small datasets. This work highlights the benefits of combining retrieval augmentation with multimodal, family-centric modeling for advancing protein foundation models.", "date": "2025-08-05", "categories": []}
{"id": "2509.15429", "title": "Random Matrix Theory-guided sparse PCA for single-cell RNA-seq data", "abstract": "Single-cell RNA-seq provides detailed molecular snapshots of individual cells but is notoriously noisy. Variability stems from biological differences and technical factors, such as amplification bias and limited RNA capture efficiency, making it challenging to adapt computational pipelines to heterogeneous datasets or evolving technologies. As a result, most studies still rely on principal component analysis (PCA) for dimensionality reduction, valued for its interpretability and robustness, in spite of its known bias in high dimensions. Here, we improve upon PCA with a Random Matrix Theory (RMT)-based approach that guides the inference of sparse principal components using existing sparse PCA algorithms. We first introduce a novel biwhitening algorithm which self-consistently estimates the magnitude of transcriptomic noise affecting each gene in individual cells, without assuming a specific noise distribution. This enables the use of an RMT-based criterion to automatically select the sparsity level, rendering sparse PCA nearly parameter-free. Our mathematically grounded approach retains the interpretability of PCA while enabling robust, hands-off inference of sparse principal components. Across seven single-cell RNA-seq technologies and four sparse PCA algorithms, we show that this method systematically improves the reconstruction of the principal subspace and consistently outperforms PCA-, autoencoder-, and diffusion-based methods in cell-type classification tasks.", "date": "2025-09-18", "categories": []}
{"id": "2602.22236", "title": "CrossLLM-Mamba: Multimodal State Space Fusion of LLMs for RNA Interaction Prediction", "abstract": "Accurate prediction of RNA-associated interactions is essential for understanding cellular regulation and advancing drug discovery. While Biological Large Language Models (BioLLMs) such as ESM-2 and RiNALMo provide powerful sequence representations, existing methods rely on static fusion strategies that fail to capture the dynamic, context-dependent nature of molecular binding. We introduce CrossLLM-Mamba, a novel framework that reformulates interaction prediction as a state-space alignment problem. By leveraging bidirectional Mamba encoders, our approach enables deep ``crosstalk'' between modality-specific embeddings through hidden state propagation, modeling interactions as dynamic sequence transitions rather than static feature overlaps. The framework maintains linear computational complexity, making it scalable to high-dimensional BioLLM embeddings. We further incorporate Gaussian noise injection and Focal Loss to enhance robustness against hard-negative samples. Comprehensive experiments across three interaction categories, RNA-protein, RNA-small molecule, and RNA-RNA demonstrate that CrossLLM-Mamba achieves state-of-the-art performance. On the RPI1460 benchmark, our model attains an MCC of 0.892, surpassing the previous best by 5.2\\%. For binding affinity prediction, we achieve Pearson correlations exceeding 0.95 on riboswitch and repeat RNA subtypes. These results establish state-space modeling as a powerful paradigm for multi-modal biological interaction prediction.", "date": "2026-02-23", "categories": []}
{"id": "2602.22239", "title": "VAE-MS: An Asymmetric Variational Autoencoder for Mutational Signature Extraction", "abstract": "Mutational signature analysis has emerged as a powerful method for uncovering the underlying biological processes driving cancer development. However, the signature extraction process, typically performed using non-negative matrix factorization (NMF), often lacks reliability and clinical applicability. To address these limitations, several solutions have been introduced, including the use of neural networks to achieve more accurate estimates and probabilistic methods to better capture natural variation in the data. In this work, we introduce a Variational Autoencoder for Mutational Signatures (VAE-MS), a novel model that leverages both an asymmetric architecture and probabilistic methods for the extraction of mutational signatures. VAE-MS is compared to with three state-of-the-art models for mutational signature extraction: SigProfilerExtractor, the NMF-based gold standard; MUSE-XAE, an autoencoder that employs an asymmetric design without probabilistic components; and SigneR, a Bayesian NMF model, to illustrate the strength in combining a nonlinear extraction with a probabilistic model. In the ability to reconstruct input data and generalize to unseen data, models with probabilistic components (VAE-MS, SigneR) dramatically outperformed models without (SigProfilerExtractor, MUSE-XAE). The NMF-baed models (SigneR, SigProfilerExtractor) had the most accurate reconstructions in simulated data, while VAE-MS reconstructed more accurately on real cancer data. Upon evaluating the ability to extract signatures consistently, no model exhibited a clear advantage over the others. Software for VAE-MS is available at https://github.com/CLINDA-AAU/VAE-MS.", "date": "2026-02-24", "categories": []}
{"id": "2602.22247", "title": "Multi-Dimensional Spectral Geometry of Biological Knowledge in Single-Cell Transformer Representations", "abstract": "Single-cell foundation models such as scGPT learn high-dimensional gene representations, but what biological knowledge these representations encode remains unclear. We systematically decode the geometric structure of scGPT internal representations through 63 iterations of automated hypothesis screening (183 hypotheses tested), revealing that the model organizes genes into a structured biological coordinate system rather than an opaque feature space.\n  The dominant spectral axis separates genes by subcellular localization, with secreted proteins at one pole and cytosolic proteins at the other. Intermediate transformer layers transiently encode mitochondrial and ER compartments in a sequence that mirrors the cellular secretory pathway. Orthogonal axes encode protein-protein interaction networks with graded fidelity to experimentally measured interaction strength (Spearman rho = 1.000 across n = 5 STRING confidence quintiles, p = 0.017).\n  In a compact six-dimensional spectral subspace, the model distinguishes transcription factors from their target genes (AUROC = 0.744, all 12 layers significant). Early layers preserve which specific genes regulate which targets, while deeper layers compress this into a coarser regulator versus regulated distinction. Repression edges are geometrically more prominent than activation edges, and B-cell master regulators BATF and BACH2 show convergence toward the B-cell identity anchor PAX5 across transformer depth. Cell-type marker genes cluster with high fidelity (AUROC = 0.851). Residual-stream geometry encodes biological structure complementary to attention patterns. These results indicate that biological transformers learn an interpretable internal model of cellular organization, with implications for regulatory network inference, drug target prioritization, and model auditing.", "date": "2026-02-24", "categories": []}
{"id": "2602.22850", "title": "MEDNA-DFM: A Dual-View FiLM-MoE Model for Explainable DNA Methylation Prediction", "abstract": "Accurate computational identification of DNA methylation is essential for understanding epigenetic regulation. Although deep learning excels in this binary classification task, its \"black-box\" nature impedes biological insight. We address this by introducing a high-performance model MEDNA-DFM, alongside mechanism-inspired signal purification algorithms. Our investigation demonstrates that MEDNA-DFM effectively captures conserved methylation patterns, achieving robust distinction across diverse species. Validation on external independent datasets confirms that the model's generalization is driven by conserved intrinsic motifs (e.g., GC content) rather than phylogenetic proximity. Furthermore, applying our developed algorithms extracted motifs with significantly higher reliability than prior studies. Finally, empirical evidence from a Drosophila 6mA case study prompted us to propose a \"sequence-structure synergy\" hypothesis, suggesting that the GAGG core motif and an upstream A-tract element function cooperatively. We further validated this hypothesis via in silico mutagenesis, confirming that the ablation of either or both elements significantly degrades the model's recognition capabilities. This work provides a powerful tool for methylation prediction and demonstrates how explainable deep learning can drive both methodological innovation and the generation of biological hypotheses.", "date": "2026-02-26", "categories": []}
{"id": "2510.01988", "title": "PepCompass: Navigating peptide embedding spaces using Riemannian Geometry", "abstract": "Antimicrobial peptide discovery is challenged by the astronomical size of peptide space and the relative scarcity of active peptides. Generative models provide continuous latent \"maps\" of peptide space, but conventionally ignore decoder-induced geometry and rely on flat Euclidean metrics, rendering exploration and optimization distorted and inefficient. Prior manifold-based remedies assume fixed intrinsic dimensionality, which critically fails in practice for peptide data. Here, we introduce PepCompass, a geometry-aware framework for peptide exploration and optimization. At its core, we define a Union of $\\kappa$-Stable Riemannian Manifolds $\\mathbb{M}^{\\kappa}$, a family of decoder-induced manifolds that captures local geometry while ensuring computational stability. We propose two local exploration methods: Second-Order Riemannian Brownian Efficient Sampling, which provides a convergent second-order approximation to Riemannian Brownian motion, and Mutation Enumeration in Tangent Space, which reinterprets tangent directions as discrete amino-acid substitutions. Combining these yields Local Enumeration Bayesian Optimization (LE-BO), an efficient algorithm for local activity optimization. Finally, we introduce Potential-minimizing Geodesic Search (PoGS), which interpolates between prototype embeddings along property-enriched geodesics, biasing discovery toward seeds, i.e. peptides with favorable activity. In-vitro validation confirms the effectiveness of PepCompass: PoGS yields four novel seeds, and subsequent optimization with LE-BO discovers 25 highly active peptides with broad-spectrum activity, including against resistant bacterial strains. These results demonstrate that geometry-informed exploration provides a powerful new paradigm for antimicrobial peptide design.", "date": "2025-10-02", "categories": []}
{"id": "2602.22018", "title": "Disease Progression and Subtype Modeling for Combined Discrete and Continuous Input Data", "abstract": "Disease progression modeling provides a robust framework to identify long-term disease trajectories from short-term biomarker data. It is a valuable tool to gain a deeper understanding of diseases with a long disease trajectory, such as Alzheimer's disease. A key limitation of most disease progression models is that they are specific to a single data type (e.g., continuous data), thereby limiting their applicability to heterogeneous, real-world datasets. To address this limitation, we propose the Mixed Events model, a novel disease progression model that handles both discrete and continuous data types. This model is implemented within the Subtype and Stage Inference (SuStaIn) framework, resulting in Mixed-SuStaIn, enabling subtype and progression modeling. We demonstrate the effectiveness of Mixed-SuStaIn through simulation experiments and real-world data from the Alzheimer's Disease Neuroimaging Initiative, showing that it performs well on mixed datasets. The code is available at: https://github.com/ucl-pond/pySuStaIn.", "date": "2026-02-25", "categories": []}
{"id": "2509.15796", "title": "Monte Carlo Tree Diffusion with Multiple Experts for Protein Design", "abstract": "The goal of protein design is to generate amino acid sequences that fold into functional structures with desired properties. Prior methods combining autoregressive language models with Monte Carlo Tree Search (MCTS) struggle with long-range dependencies and suffer from an impractically large search space. We propose MCTD-ME, Monte Carlo Tree Diffusion with Multiple Experts, which integrates masked diffusion models with tree search to enable multi-token planning and efficient exploration under the guidance of multiple experts. Unlike autoregressive planners, MCTD-ME uses biophysical-fidelity-enhanced diffusion denoising as the rollout engine, jointly revising multiple positions and scaling to large sequence spaces. It further leverages experts of varying capacities to enrich exploration, guided by a pLDDT-based masking schedule that targets low-confidence regions while preserving reliable residues. We propose a novel multi-expert selection rule ( PH-UCT-ME) extends Shannon-entropy-based UCT to expert ensembles with mutual information. MCTD-ME achieves superior performance on the CAMEO and PDB benchmarks, excelling in protein design tasks such as inverse folding, folding, and conditional design challenges like motif scaffolding on lead optimization tasks. Our framework is model-agnostic, plug-and-play, and extensible to denovo protein engineering and beyond.", "date": "2025-09-19", "categories": []}
{"id": "2602.02620", "title": "CryoLVM: Self-supervised Learning from Cryo-EM Density Maps with Large Vision Models", "abstract": "Cryo-electron microscopy (cryo-EM) has revolutionized structural biology by enabling near-atomic-level visualization of biomolecular assemblies. However, the exponential growth in cryo-EM data throughput and complexity, coupled with diverse downstream analytical tasks, necessitates unified computational frameworks that transcend current task-specific deep learning approaches with limited scalability and generalizability. We present CryoLVM, a foundation model that learns rich structural representations from experimental density maps with resolved structures by leveraging the Joint-Embedding Predictive Architecture (JEPA) integrated with SCUNet-based backbone, which can be rapidly adapted to various downstream tasks. We further introduce a novel histogram-based distribution alignment loss that accelerates convergence and enhances fine-tuning performance. We demonstrate CryoLVM's effectiveness across three critical cryo-EM tasks: density map sharpening, density map super-resolution, and missing wedge restoration. Our method consistently outperforms state-of-the-art baselines across multiple density map quality metrics, confirming its potential as a versatile model for a wide spectrum of cryo-EM applications.", "date": "2026-02-02", "categories": []}
{"id": "2602.20198", "title": "KEMP-PIP: A Feature-Fusion Based Approach for Pro-inflammatory Peptide Prediction", "abstract": "Pro-inflammatory peptides (PIPs) play critical roles in immune signaling and inflammation but are difficult to identify experimentally due to costly and time-consuming assays. To address this challenge, we present KEMP-PIP, a hybrid machine learning framework that integrates deep protein embeddings with handcrafted descriptors for robust PIP prediction. Our approach combines contextual embeddings from pretrained ESM protein language models with multi-scale k-mer frequencies, physicochemical descriptors, and modlAMP sequence features. Feature pruning and class-weighted logistic regression manage high dimensionality and class imbalance, while ensemble averaging with an optimized decision threshold enhances the sensitivity--specificity balance. Through systematic ablation studies, we demonstrate that integrating complementary feature sets consistently improves predictive performance. On the standard benchmark dataset, KEMP-PIP achieves an MCC of 0.505, accuracy of 0.752, and AUC of 0.762, outperforming ProIn-fuse, MultiFeatVotPIP, and StackPIP. Relative to StackPIP, these results represent improvements of 9.5% in MCC and 4.8% in both accuracy and AUC. The KEMP-PIP web server is freely available at https://nilsparrow1920-kemp-pip.hf.space/ and the full implementation at https://github.com/S18-Niloy/KEMP-PIP.", "date": "2026-02-22", "categories": []}
{"id": "2602.20209", "title": "Regressor-guided Diffusion Model for De Novo Peptide Sequencing with Explicit Mass Control", "abstract": "The discovery of novel proteins relies on sensitive protein identification, for which de novo peptide sequencing (DNPS) from mass spectra is a crucial approach. While deep learning has advanced DNPS, existing models inadequately enforce the fundamental mass consistency constraint, that a predicted peptide's mass must match the experimental measured precursor mass. Previous DNPS methods often treat this critical information as a simple input feature or use it in post-processing, leading to numerous implausible predictions that do not adhere to this fundamental physical property. To address this limitation, we introduce DiffuNovo, a novel regressor-guided diffusion model for de novo peptide sequencing that provides explicit peptide-level mass control. Our approach integrates the mass constraint at two critical stages: during training, a novel peptide-level mass loss guides model optimization, while at inference, regressor-based guidance from gradient-based updates in the latent space steers the generation to compel the predicted peptide adheres to the mass constraint. Comprehensive evaluations on established benchmarks demonstrate that DiffuNovo surpasses state-of-the-art methods in DNPS accuracy. Additionally, as the first DNPS model to employ a diffusion model as its core backbone, DiffuNovo leverages the powerful controllability of diffusion architecture and achieves a significant reduction in mass error, thereby producing much more physically plausible peptides. These innovations represent a substantial advancement toward robust and broadly applicable DNPS. The source code is available in the supplementary material.", "date": "2026-02-23", "categories": []}
{"id": "2507.07390", "title": "Learning Collective Variables from BioEmu with Time-Lagged Generation", "abstract": "Molecular dynamics is crucial for understanding molecular systems but its applicability is often limited by the vast timescales of rare events like protein folding. Enhanced sampling techniques overcome this by accelerating the simulation along key reaction pathways, which are defined by collective variables (CVs). However, identifying effective CVs that capture the slow, macroscopic dynamics of a system remains a major bottleneck. This work proposes a novel framework coined BioEmu-CV that learns these essential CVs automatically from BioEmu, a recently proposed foundation model for generating protein equilibrium samples. In particular, we re-purpose BioEmu to learn time-lagged generation conditioned on the learned CV, i.e., predict the distribution of molecular states after a certain amount of time. This training process promotes the CV to encode only the slow, long-term information while disregarding fast, random fluctuations. We validate our learned CV on fast-folding proteins with two key applications: (1) estimating free energy differences using on-the-fly probability enhanced sampling and (2) sampling transition paths with steered molecular dynamics. Our empirical study also serves as a new systematic and comprehensive benchmark for MLCVs on fast-folding proteins larger than Alanine Dipeptide.", "date": "2025-07-10", "categories": []}
{"id": "2602.18476", "title": "BioLM-Score: Language-Prior Conditioned Probabilistic Geometric Potentials for Protein-Ligand Scoring", "abstract": "Protein-ligand scoring is a central component of structure-based drug design, underpinning molecular docking, virtual screening, and pose optimization. Conventional physics-based energy functions are often computationally expensive, limiting their utility in large-scale screening. In contrast, deep learning-based scoring models offer improved computational efficiency but frequently suffer from limited cross-target generalization and poor interpretability, which restrict their practical applicability. Here we present BioLM-Score, a simple yet generalizable protein-ligand scoring model that couples geometric modeling with representation learning. Specifically, it employs modality-specific and structure-aware encoders for proteins and ligands, each augmented with biomolecular language models to enrich structural and chemical representations. Subsequently, these representations are integrated through a mixture density network to predict multimodal interatomic distance distributions, from which statistically grounded likelihood-based scores are derived. Evaluations on the CASF-2016 benchmark demonstrate that BioLM-Score achieves significant improvements across docking, scoring, ranking, and screening tasks. Moreover, the proposed scoring function serves as an effective optimization objective for guiding docking protocols and conformational search. In summary, BioLM-Score provides a principled and practical alternative to existing scoring functions, combining efficiency, generalization, and interpretability for structure-based drug discovery.", "date": "2026-02-09", "categories": []}
{"id": "2602.18915", "title": "AAVGen: Precision Engineering of Adeno-associated Viral Capsids for Renal Selective Targeting", "abstract": "Adeno-associated viruses (AAVs) are promising vectors for gene therapy, but their native serotypes face limitations in tissue tropism, immune evasion, and production efficiency. Engineering capsids to overcome these hurdles is challenging due to the vast sequence space and the difficulty of simultaneously optimizing multiple functional properties. The complexity also adds when it comes to the kidney, which presents unique anatomical barriers and cellular targets that require precise and efficient vector engineering. Here, we present AAVGen, a generative artificial intelligence framework for de novo design of AAV capsids with enhanced multi-trait profiles. AAVGen integrates a protein language model (PLM) with supervised fine-tuning (SFT) and a reinforcement learning technique termed Group Sequence Policy Optimization (GSPO). The model is guided by a composite reward signal derived from three ESM-2-based regression predictors, each trained to predict a key property: production fitness, kidney tropism, and thermostability. Our results demonstrate that AAVGen produces a diverse library of novel VP1 protein sequences. In silico validations revealed that the majority of the generated variants have superior performance across all three employed indices, indicating successful multi-objective optimization. Furthermore, structural analysis via AlphaFold3 confirms that the generated sequences preserve the canonical capsid folding despite sequence diversification. AAVGen establishes a foundation for data-driven viral vector engineering, accelerating the development of next-generation AAV vectors with tailored functional characteristics.", "date": "2026-02-21", "categories": []}
{"id": "2602.19444", "title": "PIS: A Physics-Informed System for Accurate State Partitioning of $A\\beta_{42}$ Protein Trajectories", "abstract": "Understanding the conformational evolution of $\\beta$-amyloid ($A\\beta$), particularly the $A\\beta_{42}$ isoform, is fundamental to elucidating the pathogenic mechanisms underlying Alzheimer's disease. However, existing end-to-end deep learning models often struggle to capture subtle state transitions in protein trajectories due to a lack of explicit physical constraints. In this work, we introduce PIS, a Physics-Informed System designed for robust metastable state partitioning. By integrating pre-computed physical priors, such as the radius of gyration and solvent-accessible surface area, into the extraction of topological features, our model achieves superior performance on the $A\\beta_{42}$ dataset. Furthermore, PIS provides an interactive platform that features dynamic monitoring of physical characteristics and multi-dimensional result validation. This system offers biological researchers a powerful set of analytical tools with physically grounded interpretability. A demonstration video of PIS is available on https://youtu.be/AJHGzUtRCg0.", "date": "2026-02-23", "categories": []}
{"id": "2602.20100", "title": "Transcending the Annotation Bottleneck: AI-Powered Discovery in Biology and Medicine", "abstract": "The dependence on expert annotation has long constituted the primary rate-limiting step in the application of artificial intelligence to biomedicine. While supervised learning drove the initial wave of clinical algorithms, a paradigm shift towards unsupervised and self-supervised learning (SSL) is currently unlocking the latent potential of biobank-scale datasets. By learning directly from the intrinsic structure of data - whether pixels in a magnetic resonance image (MRI), voxels in a volumetric scan, or tokens in a genomic sequence - these methods facilitate the discovery of novel phenotypes, the linkage of morphology to genetics, and the detection of anomalies without human bias. This article synthesises seminal and recent advances in \"learning without labels,\" highlighting how unsupervised frameworks can derive heritable cardiac traits, predict spatial gene expression in histology, and detect pathologies with performance that rivals or exceeds supervised counterparts.", "date": "2026-02-23", "categories": []}
{"id": "2602.17747", "title": "AgriVariant: Variant Effect Prediction using DeepChem-Variant for Precision Breeding in Rice", "abstract": "Predicting functional consequences of genetic variants in crop genes remains a critical bottleneck for precision breeding programs. We present AgriVariant, an end-to-end pipeline for variant-effect prediction in rice (Oryza sativa) that addresses the lack of crop-specific variant-interpretation tools and can be extended to any crop species with available reference genomes and gene annotations. Our approach integrates deep learning-based variant calling (DeepChem-Variant) with custom plant genomics annotation using RAP-DB gene models and database-independent deleteriousness scoring that combines the Grantham distance and the BLOSUM62 substitution matrix. We validate the pipeline through targeted mutations in stress-response genes (OsDREB2a, OsDREB1F, SKC1), demonstrating correct classification of stop-gained, missense, and synonymous variants with appropriate HIGH / MODERATE / LOW impact assignments. An exhaustive mutagenesis study of OsMT-3a analyzed all 1,509 possible single-nucleotide variants in 10 days, identifying 353 high-impact, 447 medium-impact, and 709 low-impact variants - an analysis that would have required 2-4 years using traditional wet-lab approaches. This computational framework enables breeders to prioritize variants for experimental validation across diverse crop species, reducing screening costs and accelerating development of climate-resilient crop varieties.", "date": "2026-02-19", "categories": []}
{"id": "2602.16327", "title": "Guide-Guard: Off-Target Predicting in CRISPR Applications", "abstract": "With the introduction of cyber-physical genome sequencing and editing technologies, such as CRISPR, researchers can more easily access tools to investigate and create remedies for a variety of topics in genetics and health science (e.g. agriculture and medicine). As the field advances and grows, new concerns present themselves in the ability to predict the off-target behavior. In this work, we explore the underlying biological and chemical model from a data driven perspective. Additionally, we present a machine learning based solution named \\textit{Guide-Guard} to predict the behavior of the system given a gRNA in the CRISPR gene-editing process with 84\\% accuracy. This solution is able to be trained on multiple different genes at the same time while retaining accuracy.", "date": "2026-02-18", "categories": []}
{"id": "2602.16643", "title": "Factorization Machine with Quadratic-Optimization Annealing for RNA Inverse Folding and Evaluation of Binary-Integer Encoding and Nucleotide Assignment", "abstract": "The RNA inverse folding problem aims to identify nucleotide sequences that preferentially adopt a given target secondary structure. While various heuristic and machine learning-based approaches have been proposed, many require a large number of sequence evaluations, which limits their applicability when experimental validation is costly. We propose a method to solve the problem using a factorization machine with quadratic-optimization annealing (FMQA). FMQA is a discrete black-box optimization method reported to obtain high-quality solutions with a limited number of evaluations. Applying FMQA to the problem requires converting nucleotides into binary variables. However, the influence of integer-to-nucleotide assignments and binary-integer encoding on the performance of FMQA has not been thoroughly investigated, even though such choices determine the structure of the surrogate model and the search landscape, and thus can directly affect solution quality. Therefore, this study aims both to establish a novel FMQA framework for RNA inverse folding and to analyze the effects of these assignments and encoding methods. We evaluated all 24 possible assignments of the four nucleotides to the ordered integers (0-3), in combination with four binary-integer encoding methods. Our results demonstrated that one-hot and domain-wall encodings outperform binary and unary encodings in terms of the normalized ensemble defect value. In domain-wall encoding, nucleotides assigned to the boundary integers (0 and 3) appeared with higher frequency. In the RNA inverse folding problem, assigning guanine and cytosine to these boundary integers promoted their enrichment in stem regions, which led to more thermodynamically stable secondary structures than those obtained with one-hot encoding.", "date": "2026-02-18", "categories": []}
{"id": "2602.16696", "title": "Parameter-free representations outperform single-cell foundation models on downstream benchmarks", "abstract": "Single-cell RNA sequencing (scRNA-seq) data exhibit strong and reproducible statistical structure. This has motivated the development of large-scale foundation models, such as TranscriptFormer, that use transformer-based architectures to learn a generative model for gene expression by embedding genes into a latent vector space. These embeddings have been used to obtain state-of-the-art (SOTA) performance on downstream tasks such as cell-type classification, disease-state prediction, and cross-species learning. Here, we ask whether similar performance can be achieved without utilizing computationally intensive deep learning-based representations. Using simple, interpretable pipelines that rely on careful normalization and linear methods, we obtain SOTA or near SOTA performance across multiple benchmarks commonly used to evaluate single-cell foundation models, including outperforming foundation models on out-of-distribution tasks involving novel cell types and organisms absent from the training data. Our findings highlight the need for rigorous benchmarking and suggest that the biology of cell identity can be captured by simple linear representations of single cell gene expression data.", "date": "2026-02-18", "categories": []}
{"id": "2504.18367", "title": "A Novel 4-D Dataset Paradigm for Studying Complete Ligand-Protein Dissociation Dynamics", "abstract": "The kinetics and dynamics of drug-protein binding and dissociation are crucial to understanding drug absorption and metabolism. Despite advances in artificial intelligence (AI) tools for drug-protein interaction studies, existing training datasets remain limited to static structures or quasi-static conformations. This paper proposes a novel computational approach for rapidly generating drug-protein dissociation trajectories and presents the inaugural dynamically time-resolved 4-D (t, x, y, z) trajectory database DD-13M. This dataset captures over 26,000 complete dissociation processes for 565 ligand-protein complexes, providing nearly 13 million frames of all-atom simulation trajectories. A deep equivariant generative model, UnbindingFlow, was trained using the DD-13M dataset. This model has the capacity to produce dissociation trajectories for novel targets whilst accurately predicting their rate constants (koff). DD-13M introduces a new type of training dataset for AI models, establishing a de novo paradigm for studying the dynamics of drug-protein interactions.", "date": "2025-04-25", "categories": []}
{"id": "2602.13346", "title": "CellMaster: Collaborative Cell Type Annotation in Single-Cell Analysis", "abstract": "Single-cell RNA-seq (scRNA-seq) enables atlas-scale profiling of complex tissues, revealing rare lineages and transient states. Yet, assigning biologically valid cell identities remains a bottleneck because markers are tissue- and state-dependent, and novel states lack references. We present CellMaster, an AI agent that mimics expert practice for zero-shot cell-type annotation. Unlike existing automated tools, CellMaster leverages LLM-encoded knowledge (e.g., GPT-4o) to perform on-the-fly annotation with interpretable rationales, without pre-training or fixed marker databases. Across 9 datasets spanning 8 tissues, CellMaster improved accuracy by 7.1% over best-performing baselines (including CellTypist and scTab) in automatic mode. With human-in-the-loop refinement, this advantage increased to 18.6%, with a 22.1% gain on subtype populations. The system demonstrates particular strength in rare and novel cell states where baselines often fail. Source code and the web application are available at \\href{https://github.com/AnonymousGym/CellMaster}{https://github.com/AnonymousGym/CellMaster}.", "date": "2026-02-12", "categories": []}
{"id": "2602.13791", "title": "MechPert: Mechanistic Consensus as an Inductive Bias for Unseen Perturbation Prediction", "abstract": "Predicting transcriptional responses to unseen genetic perturbations is essential for understanding gene regulation and prioritizing large-scale perturbation experiments. Existing approaches either rely on static, potentially incomplete knowledge graphs, or prompt language models for functionally similar genes, retrieving associations shaped by symmetric co-occurrence in scientific text rather than directed regulatory logic. We introduce MechPert, a lightweight framework that encourages LLM agents to generate directed regulatory hypotheses rather than relying solely on functional similarity. Multiple agents independently propose candidate regulators with associated confidence scores; these are aggregated through a consensus mechanism that filters spurious associations, producing weighted neighborhoods for downstream prediction. We evaluate MechPert on Perturb-seq benchmarks across four human cell lines. For perturbation prediction in low-data regimes ($N=50$ observed perturbations), MechPert improves Pearson correlation by up to 10.5\\% over similarity-based baselines. For experimental design, MechPert-selected anchor genes outperform standard network centrality heuristics by up to 46\\% in well-characterized cell lines.", "date": "2026-02-14", "categories": []}
{"id": "2602.14177", "title": "Towards Spatial Transcriptomics-driven Pathology Foundation Models", "abstract": "Spatial transcriptomics (ST) provides spatially resolved measurements of gene expression, enabling characterization of the molecular landscape of human tissue beyond histological assessment as well as localized readouts that can be aligned with morphology. Concurrently, the success of multimodal foundation models that integrate vision with complementary modalities suggests that morphomolecular coupling between local expression and morphology can be systematically used to improve histological representations themselves. We introduce Spatial Expression-Aligned Learning (SEAL), a vision-omics self-supervised learning framework that infuses localized molecular information into pathology vision encoders. Rather than training new encoders from scratch, SEAL is designed as a parameter-efficient vision-omics finetuning method that can be flexibly applied to widely used pathology foundation models. We instantiate SEAL by training on over 700,000 paired gene expression spot-tissue region examples spanning tumor and normal samples from 14 organs. Tested across 38 slide-level and 15 patch-level downstream tasks, SEAL provides a drop-in replacement for pathology foundation models that consistently improves performance over widely used vision-only and ST prediction baselines on slide-level molecular status, pathway activity, and treatment response prediction, as well as patch-level gene expression prediction tasks. Additionally, SEAL encoders exhibit robust domain generalization on out-of-distribution evaluations and enable new cross-modal capabilities such as gene-to-image retrieval. Our work proposes a general framework for ST-guided finetuning of pathology foundation models, showing that augmenting existing models with localized molecular supervision is an effective and practical step for improving visual representations and expanding their cross-modal utility.", "date": "2026-02-15", "categories": []}
{"id": "2602.14926", "title": "MAC-AMP: A Closed-Loop Multi-Agent Collaboration System for Multi-Objective Antimicrobial Peptide Design", "abstract": "To address the global health threat of antimicrobial resistance, antimicrobial peptides (AMP) are being explored for their potent and promising ability to fight resistant pathogens. While artificial intelligence (AI) is being employed to advance AMP discovery and design, most AMP design models struggle to balance key goals like activity, toxicity, and novelty, using rigid or unclear scoring methods that make results hard to interpret and optimize. As the capabilities of Large Language Models (LLM) advance and evolve swiftly, we turn to AI multi-agent collaboration based on such models (multi-agent LLMs), which show rapidly rising potential in complex scientific design scenarios. Based on this, we introduce MAC-AMP, a closed-loop multi-agent collaboration (MAC) system for multi-objective AMP design. The system implements a fully autonomous simulated peer review-adaptive reinforcement learning framework that requires only a task description and example dataset to design novel AMPs. The novelty of our work lies in introducing a closed-loop multi-agent system for AMP design, with cross-domain transferability, that supports multi-objective optimization while remaining explainable rather than a 'black box'. Experiments show that MAC-AMP outperforms other AMP generative models by effectively optimizing AMP generation for multiple key molecular properties, demonstrating exceptional results in antibacterial activity, AMP likeliness, toxicity compliance, and structural reliability.", "date": "2026-02-16", "categories": []}
{"id": "2502.15110", "title": "Variational phylogenetic inference with products over bipartitions", "abstract": "Bayesian phylogenetics is vital for understanding evolutionary dynamics, and requires accurate and efficient approximation of posterior distributions over trees. In this work, we develop a variational Bayesian approach for ultrametric phylogenetic trees. We present a novel variational family based on coalescent times of a single-linkage clustering and derive a closed-form density for the resulting distribution over trees. Unlike existing methods for ultrametric trees, our method performs inference over all of tree space, it does not require any Markov chain Monte Carlo subroutines, and our variational family is differentiable. Through experiments on benchmark genomic datasets and an application to the viral RNA of SARS-CoV-2, we demonstrate that our method achieves competitive accuracy while requiring significantly fewer gradient evaluations than existing state-of-the-art techniques.", "date": "2025-02-21", "categories": []}
{"id": "2602.10168", "title": "EVA: Towards a universal model of the immune system", "abstract": "The effective application of foundation models to translational research in immune-mediated diseases requires multimodal patient-level representations that can capture complex phenotypes emerging from multicellular interactions. Yet most current biological foundation models focus only on single-cell resolution and are evaluated on technical metrics often disconnected from actual drug development tasks and challenges. Here, we introduce EVA, the first cross-species, multimodal foundation model of immunology and inflammation, a therapeutic area where shared pathogenic mechanisms create unique opportunities for transfer learning. EVA harmonizes transcriptomics data across species, platforms, and resolutions, and integrates histology data to produce rich, unified patient representations. We establish clear scaling laws, demonstrating that increasing model size and compute translates to improvements in both pretraining and downstream tasks performance. We introduce a comprehensive evaluation suite of 39 tasks spanning the drug development pipeline: zero-shot target efficacy and gene function prediction for discovery, cross-species or cross-diseases molecular perturbations for preclinical development, and patient stratification with treatment response prediction or disease activity prediction for clinical trials applications. We benchmark EVA against several state-of-the-art biological foundation models and baselines on these tasks, and demonstrate state-of-the-art results on each task category. Using mechanistic interpretability, we further identify biological meaningful features, revealing intertwined representations across species and technologies. We release an open version of EVA for transcriptomics to accelerate research on immune-mediated diseases.", "date": "2026-02-10", "categories": []}
{"id": "2602.12470", "title": "Designing RNAs with Language Models", "abstract": "RNA design, the task of finding a sequence that folds into a target secondary structure, has broad biological and biomedical impact but remains computationally challenging due to the exponentially large sequence space and exponentially many competing folds. Traditional approaches treat it as an optimization problem, relying on per-instance heuristics or constraint-based search. We instead reframe RNA design as conditional sequence generation and introduce a reusable neural approximator, instantiated as an autoregressive language model (LM), that maps target structures directly to sequences. We first train our model in a supervised setting on random-induced structure-sequence pairs, and then use reinforcement learning (RL) to optimize end-to-end metrics. We also propose methods to select a small subset for RL that greatly improves RL efficiency and quality. Across four datasets, our approach outperforms state-of-the-art systems on key metrics such as Boltzmann probability while being 1.7x faster, establishing conditional LM generation as a scalable, task-agnostic alternative to per-instance optimization for RNA design. Our code and data are available at https://github.com/KuNyaa/RNA-Design-LM.", "date": "2026-02-12", "categories": []}
{"id": "2602.12651", "title": "Uncovering spatial tissue domains and cell types in spatial omics through cross-scale profiling of cellular and genomic interactions", "abstract": "Cellular identity and function are linked to both their intrinsic genomic makeup and extrinsic spatial context within the tissue microenvironment. Spatial transcriptomics (ST) offers an unprecedented opportunity to study this, providing in situ gene expression profiles at single-cell resolution and illuminating the spatial and functional organization of cells within tissues. However, a significant hurdle remains: ST data is inherently noisy, large, and structurally complex. This complexity makes it intractable for existing computational methods to effectively capture the interplay between spatial interactions and intrinsic genomic relationships, thus limiting our ability to discern critical biological patterns. Here, we present CellScape, a deep learning framework designed to overcome these limitations for high-performance ST data analysis and pattern discovery. CellScape jointly models cellular interactions in tissue space and genomic relationships among cells, producing comprehensive representations that seamlessly integrate spatial signals with underlying gene regulatory mechanisms. This technique uncovers biologically informative patterns that improve spatial domain segmentation and supports comprehensive spatial cellular analyses across diverse transcriptomics datasets, offering an accurate and versatile framework for deep analysis and interpretation of ST data.w", "date": "2026-02-13", "categories": []}
{"id": "2507.10136", "title": "A PBN-RL-XAI Framework for Discovering a \"Hit-and-Run\" Therapeutic Strategy in Melanoma", "abstract": "Innate resistance to anti-PD-1 immunotherapy remains a major clinical challenge in metastatic melanoma, with the underlying molecular networks being poorly understood. To address this, we constructed a dynamic Probabilistic Boolean Network model using transcriptomic data from patient tumor biopsies to elucidate the regulatory logic governing therapy response. We then employed a reinforcement learning agent to systematically discover optimal, multi-step therapeutic interventions and used explainable artificial intelligence to mechanistically interpret the agent's control policy. The analysis revealed that a precisely timed, 4-step temporary inhibition of the lysyl oxidase like 2 protein (LOXL2) was the most effective strategy. Our explainable analysis showed that this ''hit-and-run\" intervention is sufficient to erase the molecular signature driving resistance, allowing the network to self-correct without requiring sustained intervention. This study presents a novel, time-dependent therapeutic hypothesis for overcoming immunotherapy resistance and provides a powerful computational framework for identifying non-obvious intervention protocols in complex biological systems.", "date": "2025-07-14", "categories": []}
{"id": "2602.02128", "title": "Scalable Spatio-Temporal SE(3) Diffusion for Long-Horizon Protein Dynamics", "abstract": "Molecular dynamics (MD) simulations remain the gold standard for studying protein dynamics, but their computational cost limits access to biologically relevant timescales. Recent generative models have shown promise in accelerating simulations, yet they struggle with long-horizon generation due to architectural constraints, error accumulation, and inadequate modeling of spatio-temporal dynamics. We present STAR-MD (Spatio-Temporal Autoregressive Rollout for Molecular Dynamics), a scalable SE(3)-equivariant diffusion model that generates physically plausible protein trajectories over microsecond timescales. Our key innovation is a causal diffusion transformer with joint spatio-temporal attention that efficiently captures complex space-time dependencies while avoiding the memory bottlenecks of existing methods. On the standard ATLAS benchmark, STAR-MD achieves state-of-the-art performance across all metrics--substantially improving conformational coverage, structural validity, and dynamic fidelity compared to previous methods. STAR-MD successfully extrapolates to generate stable microsecond-scale trajectories where baseline methods fail catastrophically, maintaining high structural quality throughout the extended rollout. Our comprehensive evaluation reveals severe limitations in current models for long-horizon generation, while demonstrating that STAR-MD's joint spatio-temporal modeling enables robust dynamics simulation at biologically relevant timescales, paving the way for accelerated exploration of protein function.", "date": "2026-02-02", "categories": []}
{"id": "2602.09067", "title": "AntigenLM: Structure-Aware DNA Language Modeling for Influenza", "abstract": "Language models have advanced sequence analysis, yet DNA foundation models often lag behind task-specific methods for unclear reasons. We present AntigenLM, a generative DNA language model pretrained on influenza genomes with intact, aligned functional units. This structure-aware pretraining enables AntigenLM to capture evolutionary constraints and generalize across tasks. Fine-tuned on time-series hemagglutinin (HA) and neuraminidase (NA) sequences, AntigenLM accurately forecasts future antigenic variants across regions and subtypes, including those unseen during training, outperforming phylogenetic and evolution-based models. It also achieves near-perfect subtype classification. Ablation studies show that disrupting genomic structure through fragmentation or shuffling severely degrades performance, revealing the importance of preserving functional-unit integrity in DNA language modeling. AntigenLM thus provides both a powerful framework for antigen evolution prediction and a general principle for building biologically grounded DNA foundation models.", "date": "2026-02-09", "categories": []}
{"id": "2602.10156", "title": "STRAND: Sequence-Conditioned Transport for Single-Cell Perturbations", "abstract": "Predicting how genetic perturbations change cellular state is a core problem for building controllable models of gene regulation. Perturbations targeting the same gene can produce different transcriptional responses depending on their genomic locus, including different transcription start sites and regulatory elements. Gene-level perturbation models collapse these distinct interventions into the same representation.\n  We introduce STRAND, a generative model that predicts single-cell transcriptional responses by conditioning on regulatory DNA sequence. STRAND represents a perturbation by encoding the sequence at its genomic locus and uses this representation to parameterize a conditional transport process from control to perturbed cell states. Representing perturbations by sequence, rather than by a fixed set of gene identifiers, supports zero-shot inference at loci not seen during training and expands inference-time genomic coverage from ~1.5% for gene-level single-cell foundation models to ~95% of the genome.\n  We evaluate STRAND on CRISPR perturbation datasets in K562, Jurkat, and RPE1 cells. STRAND improves discrimination scores by up to 33% in low-sample regimes, achieves the best average rank on unseen gene perturbation benchmarks, and improves transfer to novel cell lines by up to 0.14 in Pearson correlation. Ablations isolate the gains to sequence conditioning and transport, and case studies show that STRAND resolves functionally alternative transcription start sites missed by gene-level models.", "date": "2026-02-10", "categories": []}
{"id": "2602.10984", "title": "Sample Efficient Generative Molecular Optimization with Joint Self-Improvement", "abstract": "Generative molecular optimization aims to design molecules with properties surpassing those of existing compounds. However, such candidates are rare and expensive to evaluate, yielding sample efficiency essential. Additionally, surrogate models introduced to predict molecule evaluations, suffer from distribution shift as optimization drives candidates increasingly out-of-distribution. To address these challenges, we introduce Joint Self-Improvement, which benefits from (i) a joint generative-predictive model and (ii) a self-improving sampling scheme. The former aligns the generator with the surrogate, alleviating distribution shift, while the latter biases the generative part of the joint model using the predictive one to efficiently generate optimized molecules at inference-time. Experiments across offline and online molecular optimization benchmarks demonstrate that Joint Self-Improvement outperforms state-of-the-art methods under limited evaluation budgets.", "date": "2026-02-11", "categories": []}
{"id": "2406.16821", "title": "General Binding Affinity Guidance for Diffusion Models in Structure-Based Drug Design", "abstract": "Structure-based drug design (SBDD) aims to generate ligands that bind strongly and specifically to target protein pockets. Recent diffusion models have advanced SBDD by capturing the distributions of atomic positions and types, yet they often underemphasize binding affinity control during generation. To address this limitation, we introduce \\textbf{\\textnormal{\\textbf{BADGER}}}, a general \\textbf{binding-affinity guidance framework for diffusion models in SBDD}. \\textnormal{\\textbf{BADGER} }incorporates binding affinity awareness through two complementary strategies: (1) \\textit{classifier guidance}, which applies gradient-based affinity signals during sampling in a plug-and-play fashion, and (2) \\textit{classifier-free guidance}, which integrates affinity conditioning directly into diffusion model training. Together, these approaches enable controllable ligand generation guided by binding affinity. \\textnormal{\\textbf{BADGER} } can be added to any diffusion model and achieves up to a \\textbf{60\\% improvement in ligand--protein binding affinity} of sampled molecules over prior methods. Furthermore, we extend the framework to \\textbf{multi-constraint diffusion guidance}, jointly optimizing for binding affinity, drug-likeness (QED), and synthetic accessibility (SA) to design realistic and synthesizable drug candidates.", "date": "2024-06-24", "categories": []}
{"id": "2507.02883", "title": "DISPROTBENCH: Uncovering the Functional Limits of Protein Structure Prediction Models in Intrinsically Disordered Regions", "abstract": "Intrinsically disordered regions (IDRs) play central roles in cellular function, yet remain poorly evaluated by existing protein structure prediction benchmarks. Current evaluations largely focus on well-folded domains, overlooking three fundamental challenges in realistic biological settings: the structural complexity of proteins, the resulting low availability of reliable ground truth, and prediction uncertainty that can propagate into high-risk downstream failures, such as in drug discovery, protein-protein interaction modeling, and functional annotation. We present DisProtBench, an IDR-centric benchmark that explicitly incorporates prediction uncertainty into the evaluation of protein structure prediction models (PSPMs). To address structural complexity and ground-truth scarcity, we curate and unify a large-scale, multi-modal dataset spanning disease-relevant IDRs, GPCR-ligand interactions, and multimeric protein complexes. To assess predictive uncertainty, we introduce Functional Uncertainty Sensitivity (FUS), a novel prediction uncertainty-stratified metric that quantifies downstream task performance under prediction uncertainty. Using this benchmark, we conduct a systematic evaluation of state-of-the-art PSPMs and reveal clear, task-dependent failure modes. Protein-protein interaction prediction degrades sharply in IDRs, while structure-based drug discovery remains comparatively robust. These effects are largely invisible to standard global accuracy metrics, which overestimate functional reliability under prediction uncertainty. We have open-sourced our benchmark and the codebase at https://github.com/Susan571/DisProtBench.", "date": "2025-06-18", "categories": []}
{"id": "2602.09036", "title": "Predicting Gene Disease Associations in Type 2 Diabetes Using Machine Learning on Single-Cell RNA-Seq Data", "abstract": "Diabetes is a chronic metabolic disorder characterized by elevated blood glucose levels due to impaired insulin production or function. Two main forms are recognized: type 1 diabetes (T1D), which involves autoimmune destruction of insulin-producing \\b{eta}-cells, and type 2 diabetes (T2D), which arises from insulin resistance and progressive \\b{eta}-cell dysfunction. Understanding the molecular mechanisms underlying these diseases is essential for the development of improved therapeutic strategies, particularly those targeting \\b{eta}-cell dysfunction. To investigate these mechanisms in a controlled and biologically interpretable setting, mouse models have played a central role in diabetes research. Owing to their genetic and physiological similarity to humans, together with the ability to precisely manipulate their genome, mice enable detailed investigation of disease progression and gene function. In particular, mouse models have provided critical insights into \\b{eta}-cell development, cellular heterogeneity, and functional failure under diabetic conditions. Building on these experimental advances, this study applies machine learning methods to single-cell transcriptomic data from mouse pancreatic islets. Specifically, we evaluate two supervised approaches identified in the literature; Extra Trees Classifier (ETC) and Partial Least Squares Discriminant Analysis (PLS-DA), to assess their ability to identify T2D-associated gene expression signatures at single-cell resolution. Model performance is evaluated using standard classification metrics, with an emphasis on interpretability and biological relevance", "date": "2026-01-30", "categories": []}
{"id": "2602.09063", "title": "scBench: Evaluating AI Agents on Single-Cell RNA-seq Analysis", "abstract": "As single-cell RNA sequencing datasets grow in adoption, scale, and complexity, data analysis remains a bottleneck for many research groups. Although frontier AI agents have improved dramatically at software engineering and general data analysis, it remains unclear whether they can extract biological insight from messy, real-world single-cell datasets. We introduce scBench, a benchmark of 394 verifiable problems derived from practical scRNA-seq workflows spanning six sequencing platforms and seven task categories. Each problem provides a snapshot of experimental data immediately prior to an analysis step and a deterministic grader that evaluates recovery of a key biological result. Benchmark data on eight frontier models shows that accuracy ranges from 29-53%, with strong model-task and model-platform interactions. Platform choice affects accuracy as much as model choice, with 40+ percentage point drops on less-documented technologies. scBench complements SpatialBench to cover the two dominant single-cell modalities, serving both as a measurement tool and a diagnostic lens for developing agents that can analyze real scRNA-seq datasets faithfully and reproducibly.", "date": "2026-02-09", "categories": []}
{"id": "2601.21560", "title": "HistoPrism: Unlocking Functional Pathway Analysis from Pan-Cancer Histology via Gene Expression Prediction", "abstract": "Predicting spatial gene expression from H&E histology offers a scalable and clinically accessible alternative to sequencing, but realizing clinical impact requires models that generalize across cancer types and capture biologically coherent signals. Prior work is often limited to per-cancer settings and variance-based evaluation, leaving functional relevance underexplored. We introduce HistoPrism, an efficient transformer-based architecture for pan-cancer prediction of gene expression from histology. To evaluate biological meaning, we introduce a pathway-level benchmark, shifting assessment from isolated gene-level variance to coherent functional pathways. HistoPrism not only surpasses prior state-of-the-art models on highly variable genes , but also more importantly, achieves substantial gains on pathway-level prediction, demonstrating its ability to recover biologically coherent transcriptomic patterns. With strong pan-cancer generalization and improved efficiency, HistoPrism establishes a new standard for clinically relevant transcriptomic modeling from routinely available histology.", "date": "2026-01-29", "categories": []}
{"id": "2602.06020", "title": "Mechanisms of AI Protein Folding in ESMFold", "abstract": "How do protein structure prediction models fold proteins? We investigate this question by tracing how ESMFold folds a beta hairpin, a prevalent structural motif. Through counterfactual interventions on model latents, we identify two computational stages in the folding trunk. In the first stage, early blocks initialize pairwise biochemical signals: residue identities and associated biochemical features such as charge flow from sequence representations into pairwise representations. In the second stage, late blocks develop pairwise spatial features: distance and contact information accumulate in the pairwise representation. We demonstrate that the mechanisms underlying structural decisions of ESMFold can be localized, traced through interpretable representations, and manipulated with strong causal effects.", "date": "2026-02-05", "categories": []}
{"id": "2602.07002", "title": "MolLIBRA: Genetic Molecular Optimization with Multi-Fingerprint Surrogates and Text-Molecule Aligned Critic", "abstract": "We study sample-efficient molecular optimization under a limited budget of oracle evaluations. We propose MolLIBRA (MultimOdaLity and Language Integrated Bayesian and evolutionaRy optimizAtion), a genetic algorithm based framework that pre-ranks candidate molecules using multiple critics before oracle calls: (i) an ensemble of Gaussian process (GP) surrogates defined over multiple molecular fingerprints and (ii) a pretrained text-molecule aligned encoder CLAMP. The GP ensemble enables adaptive selection of task-appropriate fingerprints, while CLAMP provides a zero-shot scoring signal from task descriptions by measuring the similarity between molecular and text embeddings. On the Practical Molecular Optimization (PMO) benchmark with a budget of 1,000 evaluations (PMO-1K), MolLIBRA-L, our variant with a language-model-based candidate generator, attains the best Top-10 AUC on 14/22 tasks and the highest overall sum of Top-10 AUC across tasks among prior methods.", "date": "2026-01-29", "categories": []}
{"id": "2602.07084", "title": "AbFlow : End-to-end Paratope-Centric Antibody Design by Interaction Enhanced Flow Matching", "abstract": "Antigen-antibody binding is a critical process in the immune response. Although recent progress has advanced antibody design, current methods lack a generative framework for end-to-end modeling of full-atom antibody structures and struggle to fully exploit antigen-specific geometric information for optimizing local binding interfaces and global structures. To overcome these limitations, we introduce AbFlow, a flow-matching framework that leverages optimal transport to design full-atom antibodies end-to-end. AbFlow incorporates an extended velocity field network featuring an equivariant Surface Multi-channel Encoder, which uses surface-level antigen interaction data to refine the antibody structure, particularly the CDR-H3 region. Extensive experiments in paratoep-centric antibody design, multi-CDRs and full-atom antibody design, binding affinity optimization, and complex structure prediction show that AbFlow produces superior antigen-antibody complexes, especially at the contact interface, and markedly improves the binding affinity of generated antibodies.", "date": "2026-02-06", "categories": []}
{"id": "2602.07103", "title": "scDFM: Distributional Flow Matching Model for Robust Single-Cell Perturbation Prediction", "abstract": "A central goal in systems biology and drug discovery is to predict the transcriptional response of cells to perturbations. This task is challenging due to the noisy and sparse nature of single-cell measurements, as well as the fact that perturbations often induce population-level shifts rather than changes in individual cells. Existing deep learning methods typically assume cell-level correspondences, limiting their ability to capture such global effects. We present scDFM, a generative framework based on conditional flow matching that models the full distribution of perturbed cells conditioned on control states. By incorporating a maximum mean discrepancy (MMD) objective, our method aligns perturbed and control populations beyond cell-level correspondences. To further improve robustness to sparsity and noise, we introduce the Perturbation-Aware Differential Transformer (PAD-Transformer), a backbone architecture that leverages gene interaction graphs and differential attention to capture context-specific expression changes. Across multiple genetic and drug perturbation benchmarks, scDFM consistently outperforms prior methods, demonstrating strong generalization in both unseen and combinatorial settings. In the combinatorial setting, it reduces mean squared error by 19.6% relative to the strongest baseline. These results highlight the importance of distribution-level generative modeling for robust in silico perturbation prediction. The code is available at https://github.com/AI4Science-WestlakeU/scDFM", "date": "2026-02-06", "categories": []}
{"id": "2602.07735", "title": "TerraBind: Fast and Accurate Binding Affinity Prediction through Coarse Structural Representations", "abstract": "We present TerraBind, a foundation model for protein-ligand structure and binding affinity prediction that achieves 26-fold faster inference than state-of-the-art methods while improving affinity prediction accuracy by $\\sim$20\\%. Current deep learning approaches to structure-based drug design rely on expensive all-atom diffusion to generate 3D coordinates, creating inference bottlenecks that render large-scale compound screening computationally intractable. We challenge this paradigm with a critical hypothesis: full all-atom resolution is unnecessary for accurate small molecule pose and binding affinity prediction. TerraBind tests this hypothesis through a coarse pocket-level representation (protein C$_\\beta$ atoms and ligand heavy atoms only) within a multimodal architecture combining COATI-3 molecular encodings and ESM-2 protein embeddings that learns rich structural representations, which are used in a diffusion-free optimization module for pose generation and a binding affinity likelihood prediction module. On structure prediction benchmarks (FoldBench, PoseBusters, Runs N' Poses), TerraBind matches diffusion-based baselines in ligand pose accuracy. Crucially, TerraBind outperforms Boltz-2 by $\\sim$20\\% in Pearson correlation for binding affinity prediction on both a public benchmark (CASP16) and a diverse proprietary dataset (18 biochemical/cell assays). We show that the affinity prediction module also provides well-calibrated affinity uncertainty estimates, addressing a critical gap in reliable compound prioritization for drug discovery. Furthermore, this module enables a continual learning framework and a hedged batch selection strategy that, in simulated drug discovery cycles, achieves 6$\\times$ greater affinity improvement of selected molecules over greedy-based approaches.", "date": "2026-02-08", "categories": []}
{"id": "2510.14586", "title": "Matcha: Multi-Stage Riemannian Flow Matching for Accurate and Physically Valid Molecular Docking", "abstract": "Accurate prediction of protein-ligand binding poses is crucial for structure-based drug design, yet existing methods struggle to balance speed, accuracy, and physical plausibility. We introduce Matcha, a novel molecular docking pipeline that combines multi-stage flow matching with physically-aware post-processing. Our approach consists of three sequential stages applied consecutively to progressively refine docking predictions, each implemented as a flow matching model operating on appropriate geometric spaces ($\\mathbb{R}^3$, $\\mathrm{SO}(3)$, and $\\mathrm{SO}(2)$). We enhance the prediction quality through GNINA energy minimization and apply unsupervised physical validity filters to eliminate unrealistic poses. Compared to various approaches, Matcha demonstrates superior physical plausibility across all considered benchmarks. Moreover, our method works approximately 31 times faster than modern large-scale co-folding models. The model weights and inference code to reproduce our results are available at https://github.com/LigandPro/Matcha.", "date": "2025-10-16", "categories": []}
{"id": "2510.25814", "title": "Optimizing Mirror-Image Peptide Sequence Design for Data Storage via Peptide Bond Cleavage Prediction", "abstract": "Traditional non-biological storage media, such as hard drives, face limitations in both storage density and lifespan due to the rapid growth of data in the big data era. Mirror-image peptides composed of D-amino acids have emerged as a promising biological storage medium due to their high storage density, structural stability, and long lifespan. The sequencing of mirror-image peptides relies on \\textit{de-novo} technology. However, its accuracy is limited by the scarcity of tandem mass spectrometry datasets and the challenges that current algorithms encounter when processing these peptides directly. This study is the first to propose improving sequencing accuracy indirectly by optimizing the design of mirror-image peptide sequences. In this work, we introduce DBond, a deep neural network based model that integrates sequence features, precursor ion properties, and mass spectrometry environmental factors for the prediction of mirror-image peptide bond cleavage. In this process, sequences with a high peptide bond cleavage ratio, which are easy to sequence, are selected. The main contributions of this study are as follows. First, we constructed MiPD513, a tandem mass spectrometry dataset containing 513 mirror-image peptides. Second, we developed the peptide bond cleavage labeling algorithm (PBCLA), which generated approximately 12.5 million labeled data based on MiPD513. Third, we proposed a dual prediction strategy that combines multi-label and single-label classification. On an independent test set, the single-label classification strategy outperformed other methods in both single and multiple peptide bond cleavage prediction tasks, offering a strong foundation for sequence optimization.", "date": "2025-10-29", "categories": []}
{"id": "2602.05527", "title": "Generalization of Self-Supervised Vision Transformers for Protein Localization Across Microscopy Domains", "abstract": "Task-specific microscopy datasets are often too small to train deep learning models that learn robust feature representations. Self-supervised learning (SSL) can mitigate this by pretraining on large unlabeled datasets, but it remains unclear how well such representations transfer across microscopy domains with different staining protocols and channel configurations. We investigate the cross-domain transferability of DINO-pretrained Vision Transformers for protein localization on the OpenCell dataset. We generate image embeddings using three DINO backbones pretrained on ImageNet-1k, the Human Protein Atlas (HPA), and OpenCell, and evaluate them by training a supervised classification head on OpenCell labels. All pretrained models transfer well, with the microscopy-specific HPA-pretrained model achieving the best performance (mean macro $F_1$-score = 0.8221 $\\pm$ 0.0062), slightly outperforming a DINO model trained directly on OpenCell (0.8057 $\\pm$ 0.0090). These results highlight the value of large-scale pretraining and indicate that domain-relevant SSL representations can generalize effectively to related but distinct microscopy datasets, enabling strong downstream performance even when task-specific labeled data are limited.", "date": "2026-02-05", "categories": []}
{"id": "2602.06706", "title": "SaDiT: Efficient Protein Backbone Design via Latent Structural Tokenization and Diffusion Transformers", "abstract": "Generative models for de novo protein backbone design have achieved remarkable success in creating novel protein structures. However, these diffusion-based approaches remain computationally intensive and slower than desired for large-scale structural exploration. While recent efforts like Proteina have introduced flow-matching to improve sampling efficiency, the potential of tokenization for structural compression and acceleration remains largely unexplored in the protein domain. In this work, we present SaDiT, a novel framework that accelerates protein backbone generation by integrating SaProt Tokenization with a Diffusion Transformer (DiT) architecture. SaDiT leverages a discrete latent space to represent protein geometry, significantly reducing the complexity of the generation process while maintaining theoretical SE(3) equivalence. To further enhance efficiency, we introduce an IPA Token Cache mechanism that optimizes the Invariant Point Attention (IPA) layers by reusing computed token states during iterative sampling. Experimental results demonstrate that SaDiT outperforms state-of-the-art models, including RFDiffusion and Proteina, in both computational speed and structural viability. We evaluate our model across unconditional backbone generation and fold-class conditional generation tasks, where SaDiT shows superior ability to capture complex topological features with high designability.", "date": "2026-02-06", "categories": []}
{"id": "2508.02276", "title": "CellForge: Agentic Design of Virtual Cell Models", "abstract": "Virtual cell modeling aims to predict cellular responses to diverse perturbations but faces challenges from biological complexity, multimodal data heterogeneity, and the need for interdisciplinary expertise. We introduce CellForge, a multi-agent framework that autonomously designs and synthesizes neural network architectures tailored to specific single-cell datasets and perturbation tasks. Given raw multi-omics data and task descriptions, CellForge discovers candidate architectures through collaborative reasoning among specialized agents, then generates executable implementations. Our core contribution is the framework itself: showing that multi-agent collaboration mechanisms - rather than manual human design or single-LLM prompting - can autonomously produce executable, high-quality computational methods. This approach goes beyond conventional hyperparameter tuning by enabling entirely new architectural components such as trajectory-aware encoders and perturbation diffusion modules to emerge from agentic deliberation. We evaluate CellForge on six datasets spanning gene knockouts, drug treatments, and cytokine stimulations across multiple modalities (scRNA-seq, scATAC-seq, CITE-seq). The results demonstrate that the models generated by CellForge are highly competitive with established baselines, while revealing systematic patterns of architectural innovation. CellForge highlights the scientific value of multi-agent frameworks: collaboration among specialized agents enables genuine methodological innovation and executable solutions that single agents or human experts cannot achieve. This represents a paradigm shift toward autonomous scientific method development in computational biology. Code is available at https://github.com/gersteinlab/CellForge.", "date": "2025-08-04", "categories": []}
{"id": "2602.05479", "title": "Phi-Former: A Pairwise Hierarchical Approach for Compound-Protein Interactions Prediction", "abstract": "Drug discovery remains time-consuming, labor-intensive, and expensive, often requiring years and substantial investment per drug candidate. Predicting compound-protein interactions (CPIs) is a critical component in this process, enabling the identification of molecular interactions between drug candidates and target proteins. Recent deep learning methods have successfully modeled CPIs at the atomic level, achieving improved efficiency and accuracy over traditional energy-based approaches. However, these models do not always align with chemical realities, as molecular fragments (motifs or functional groups) typically serve as the primary units of biological recognition and binding. In this paper, we propose Phi-former, a pairwise hierarchical interaction representation learning method that addresses this gap by incorporating the biological role of motifs in CPIs. Phi-former represents compounds and proteins hierarchically and employs a pairwise pre-training framework to model interactions systematically across atom-atom, motif-motif, and atom-motif levels, reflecting how biological systems recognize molecular partners. We design intra-level and inter-level learning pipelines that make different interaction levels mutually beneficial. Experimental results demonstrate that Phi-former achieves superior performance on CPI-related tasks. A case study shows that our method accurately identifies specific atoms or motifs activated in CPIs, providing interpretable model explanations. These insights may guide rational drug design and support precision medicine applications.", "date": "2026-02-05", "categories": []}
{"id": "2602.05811", "title": "STProtein: predicting spatial protein expression from multi-omics data", "abstract": "The integration of spatial multi-omics data from single tissues is crucial for advancing biological research. However, a significant data imbalance impedes progress: while spatial transcriptomics data is relatively abundant, spatial proteomics data remains scarce due to technical limitations and high costs. To overcome this challenge we propose STProtein, a novel framework leveraging graph neural networks with multi-task learning strategy. STProtein is designed to accurately predict unknown spatial protein expression using more accessible spatial multi-omics data, such as spatial transcriptomics. We believe that STProtein can effectively addresses the scarcity of spatial proteomics, accelerating the integration of spatial multi-omics and potentially catalyzing transformative breakthroughs in life sciences. This tool enables scientists to accelerate discovery by identifying complex and previously hidden spatial patterns of proteins within tissues, uncovering novel relationships between different marker genes, and exploring the biological \"Dark Matter\".", "date": "2026-02-05", "categories": []}
{"id": "2504.02148", "title": "OmniCellTOSG: The First Cell Text-Omic Signaling Graphs Dataset for Graph Language Foundation Modeling", "abstract": "With the rapid growth of large-scale single-cell omic datasets, omic foundation models (FMs) have emerged as powerful tools for advancing research in life sciences and precision medicine. However, most existing omic FMs rely primarily on numerical transcriptomic data by sorting genes as sequences, while lacking explicit integration of biomedical prior knowledge and signaling interactions that are critical for scientific discovery. Here, we introduce the Text-Omic Signaling Graph (TOSG), a novel data structure that unifies human-interpretable biomedical textual knowledge, quantitative omic data, and signaling network information. Using this framework, we construct OmniCellTOSG, a large-scale resource comprising approximately half million meta-cell TOSGs derived from around 80 million single-cell and single-nucleus RNA-seq profiles across organs and diseases. We further develop CellTOSG-FM, a multimodal graph language FM, to jointly analyze textual, omic and signaling network context. Across diverse downstream tasks, CellTOSG-FM outperforms existing omic FMs, and provides interpretable insights into disease-associated targets and signaling pathways.", "date": "2025-04-02", "categories": []}
{"id": "2512.03312", "title": "Unlocking hidden biomolecular conformational landscapes in diffusion models at inference time", "abstract": "The function of biomolecules such as proteins depends on their ability to interconvert between a wide range of structures or \"conformations.\" Researchers have endeavored for decades to develop computational methods to predict the distribution of conformations, which is far harder to determine experimentally than a static folded structure. We present ConforMix, an inference-time algorithm that enhances sampling of conformational distributions using a combination of classifier guidance, filtering, and free energy estimation. Our approach upgrades diffusion models -- whether trained for static structure prediction or conformational generation -- to enable more efficient discovery of conformational variability without requiring prior knowledge of major degrees of freedom. ConforMix is orthogonal to improvements in model pretraining and would benefit even a hypothetical model that perfectly reproduced the Boltzmann distribution. Remarkably, when applied to a diffusion model trained for static structure prediction, ConforMix captures structural changes including domain motion, cryptic pocket flexibility, and transporter cycling, while avoiding unphysical states. Case studies of biologically critical proteins demonstrate the scalability, accuracy, and utility of this method.", "date": "2025-12-02", "categories": []}
{"id": "2602.00157", "title": "ProDCARL: Reinforcement Learning-Aligned Diffusion Models for De Novo Antimicrobial Peptide Design", "abstract": "Antimicrobial resistance threatens healthcare sustainability and motivates low-cost computational discovery of antimicrobial peptides (AMPs). De novo peptide generation must optimize antimicrobial activity and safety through low predicted toxicity, but likelihood-trained generators do not enforce these goals explicitly. We introduce ProDCARL, a reinforcement-learning alignment framework that couples a diffusion-based protein generator (EvoDiff OA-DM 38M) with sequence property predictors for AMP activity and peptide toxicity. We fine-tune the diffusion prior on AMP sequences to obtain a domain-aware generator. Top-k policy-gradient updates use classifier-derived rewards plus entropy regularization and early stopping to preserve diversity and reduce reward hacking. In silico experiments show ProDCARL increases the mean predicted AMP score from 0.081 after fine-tuning to 0.178. The joint high-quality hit rate reaches 6.3\\% with pAMP $>$0.7 and pTox $<$0.3. ProDCARL maintains high diversity, with $1-$mean pairwise identity equal to 0.929. Qualitative analyses with AlphaFold3 and ProtBERT embeddings suggest candidates show plausible AMP-like structural and semantic characteristics. ProDCARL serves as a candidate generator that narrows experimental search space, and experimental validation remains future work.", "date": "2026-01-29", "categories": []}
{"id": "2602.03902", "title": "All-Atom GPCR-Ligand Simulation via Residual Isometric Latent Flow", "abstract": "G-protein-coupled receptors (GPCRs), primary targets for over one-third of approved therapeutics, rely on intricate conformational transitions to transduce signals. While Molecular Dynamics (MD) is essential for elucidating this transduction process, particularly within ligand-bound complexes, conventional all-atom MD simulation is computationally prohibitive. In this paper, we introduce GPCRLMD, a deep generative framework for efficient all-atom GPCR-ligand simulation.GPCRLMD employs a Harmonic-Prior Variational Autoencoder (HP-VAE) to first map the complex into a regularized isometric latent space, preserving geometric topology via physics-informed constraints. Within this latent space, a Residual Latent Flow samples evolution trajectories, which are subsequently decoded back to atomic coordinates. By capturing temporal dynamics via relative displacements anchored to the initial structure, this residual mechanism effectively decouples static topology from dynamic fluctuations. Experimental results demonstrate that GPCRLMD achieves state-of-the-art performance in GPCR-ligand dynamics simulation, faithfully reproducing thermodynamic observables and critical ligand-receptor interactions.", "date": "2026-02-03", "categories": []}
{"id": "2602.04637", "title": "RIGA-Fold: A General Framework for Protein Inverse Folding via Recurrent Interaction and Geometric Awareness", "abstract": "Protein inverse folding, the task of predicting amino acid sequences for desired structures, is pivotal for de novo protein design. However, existing GNN-based methods typically suffer from restricted receptive fields that miss long-range dependencies and a \"single-pass\" inference paradigm that leads to error accumulation. To address these bottlenecks, we propose RIGA-Fold, a framework that synergizes Recurrent Interaction with Geometric Awareness. At the micro-level, we introduce a Geometric Attention Update (GAU) module where edge features explicitly serve as attention keys, ensuring strictly SE(3)-invariant local encoding. At the macro-level, we design an attention-based Global Context Bridge that acts as a soft gating mechanism to dynamically inject global topological information. Furthermore, to bridge the gap between structural and sequence modalities, we introduce an enhanced variant, RIGA-Fold*, which integrates trainable geometric features with frozen evolutionary priors from ESM-2 and ESM-IF via a dual-stream architecture. Finally, a biologically inspired ``predict-recycle-refine'' strategy is implemented to iteratively denoise sequence distributions. Extensive experiments on CATH 4.2, TS50, and TS500 benchmarks demonstrate that our geometric framework is highly competitive, while RIGA-Fold* significantly outperforms state-of-the-art baselines in both sequence recovery and structural consistency.", "date": "2026-02-04", "categories": []}
{"id": "2602.04696", "title": "Beyond Learning on Molecules by Weakly Supervising on Molecules", "abstract": "Molecular representations are inherently task-dependent, yet most pre-trained molecular encoders are not. Task conditioning promises representations that reorganize based on task descriptions, but existing approaches rely on expensive labeled data. We show that weak supervision on programmatically derived molecular motifs is sufficient. Our Adaptive Chemical Embedding Model (ACE-Mol) learns from hundreds of motifs paired with natural language descriptors that are cheap to compute, trivial to scale. Conventional encoders slowly search the embedding space for task-relevant structure, whereas ACE-Mol immediately aligns its representations with the task. ACE-Mol achieves state-of-the-art performance across molecular property prediction benchmarks with interpretable, chemically meaningful representations.", "date": "2026-02-04", "categories": []}
{"id": "2505.13197", "title": "Inferring stochastic dynamics with growth from cross-sectional data", "abstract": "Time-resolved single-cell omics data offers high-throughput, genome-wide measurements of cellular states, which are instrumental to reverse-engineer the processes underpinning cell fate. Such technologies are inherently destructive, allowing only cross-sectional measurements of the underlying stochastic dynamical system. Furthermore, cells may divide or die in addition to changing their molecular state. Collectively these present a major challenge to inferring realistic biophysical models. We present a novel approach, unbalanced probability flow inference, that addresses this challenge for biological processes modelled as stochastic dynamics with growth. By leveraging a Lagrangian formulation of the Fokker-Planck equation, our method accurately disentangles drift from intrinsic noise and growth. We showcase the applicability of our approach through evaluation on a range of simulated and real single-cell RNA-seq datasets. Comparing to several existing methods, we find our method achieves higher accuracy while enjoying a simple two-step training scheme.", "date": "2025-05-19", "categories": []}
{"id": "2602.00197", "title": "Rank-and-Reason: Multi-Agent Collaboration Accelerates Zero-Shot Protein Mutation Prediction", "abstract": "Zero-shot mutation prediction is vital for low-resource protein engineering, yet existing protein language models (PLMs) often yield statistically confident results that ignore fundamental biophysical constraints. Currently, selecting candidates for wet-lab validation relies on manual expert auditing of PLM outputs, a process that is inefficient, subjective, and highly dependent on domain expertise. To address this, we propose Rank-and-Reason (VenusRAR), a two-stage agentic framework to automate this workflow and maximize expected wet-lab fitness. In the Rank-Stage, a Computational Expert and Virtual Biologist aggregate a context-aware multi-modal ensemble, establishing a new Spearman correlation record of 0.551 (vs. 0.518) on ProteinGym. In the Reason-Stage, an agentic Expert Panel employs chain-of-thought reasoning to audit candidates against geometric and structural constraints, improving the Top-5 Hit Rate by up to 367% on ProteinGym-DMS99. The wet-lab validation on Cas12i3 nuclease further confirms the framework's efficacy, achieving a 46.7% positive rate and identifying two novel mutants with 4.23-fold and 5.05-fold activity improvements. Code and datasets are released on GitHub (https://github.com/ai4protein/VenusRAR/).", "date": "2026-01-30", "categories": []}
{"id": "2602.03477", "title": "ScDiVa: Masked Discrete Diffusion for Joint Modeling of Single-Cell Identity and Expression", "abstract": "Single-cell RNA-seq profiles are high-dimensional, sparse, and unordered, causing autoregressive generation to impose an artificial ordering bias and suffer from error accumulation. To address this, we propose scDiVa, a masked discrete diffusion foundation model that aligns generation with the dropout-like corruption process by defining a continuous-time forward masking mechanism in token space. ScDiVa features a bidirectional denoiser that jointly models discrete gene identities and continuous values, utilizing entropy-normalized serialization and a latent anchor token to maximize information efficiency and preserve global cell identity. The model is trained via depth-invariant time sampling and a dual denoising objective to simulate varying sparsity levels while ensuring precise recovery of both identity and magnitude. Pre-trained on 59 million cells, scDiVa achieves strong transfer performance across major benchmarks, including batch integration, cell type annotation, and perturbation response prediction. These results suggest that masked discrete diffusion serves as a biologically coherent and effective alternative to autoregression.", "date": "2026-02-03", "categories": []}
{"id": "2302.13268", "title": "Revolutionizing Genomics with Reinforcement Learning Techniques", "abstract": "In recent years, Reinforcement Learning (RL) has emerged as a powerful tool for solving a wide range of problems, including decision-making and genomics. The exponential growth of raw genomic data over the past two decades has exceeded the capacity of manual analysis, leading to a growing interest in automatic data analysis and processing. RL algorithms are capable of learning from experience with minimal human supervision, making them well-suited for genomic data analysis and interpretation. One of the key benefits of using RL is the reduced cost associated with collecting labeled training data, which is required for supervised learning. While there have been numerous studies examining the applications of Machine Learning (ML) in genomics, this survey focuses exclusively on the use of RL in various genomics research fields, including gene regulatory networks (GRNs), genome assembly, and sequence alignment. We present a comprehensive technical overview of existing studies on the application of RL in genomics, highlighting the strengths and limitations of these approaches. We then discuss potential research directions that are worthy of future exploration, including the development of more sophisticated reward functions as RL heavily depends on the accuracy of the reward function, the integration of RL with other machine learning techniques, and the application of RL to new and emerging areas in genomics research. Finally, we present our findings and conclude by summarizing the current state of the field and the future outlook for RL in genomics.", "date": "2023-02-26", "categories": []}
{"id": "2502.06914", "title": "UniZyme: A Unified Protein Cleavage Site Predictor Enhanced with Enzyme Active-Site Knowledge", "abstract": "Enzyme-catalyzed protein cleavage is essential for many biological functions. Accurate prediction of cleavage sites can facilitate various applications such as drug development, enzyme design, and a deeper understanding of biological mechanisms. However, most existing models are restricted to an individual enzyme, which neglects shared knowledge of enzymes and fails to generalize to novel enzymes. Thus, we introduce a unified protein cleavage site predictor named UniZyme, which can generalize across diverse enzymes. To enhance the enzyme encoding for the protein cleavage site prediction, UniZyme employs a novel biochemically-informed model architecture along with active-site knowledge of proteolytic enzymes. Extensive experiments demonstrate that UniZyme achieves high accuracy in predicting cleavage sites across a range of proteolytic enzymes, including unseen enzymes. The code is available in https://github.com/Ao-LiChen/UniZyme", "date": "2025-02-10", "categories": []}
{"id": "2502.07272", "title": "GENERator: A Long-Context Generative Genomic Foundation Model", "abstract": "The rapid advancement of DNA sequencing has produced vast genomic datasets, yet interpreting and engineering genomic function remain fundamental challenges. Recent large language models have opened new avenues for genomic analysis, but existing approaches are often limited by restricted training scope, constrained generative capability, or prohibitive computational cost. We introduce GENErator, a generative genomic foundation model for long-context DNA modeling, with a context length of 98k nucleotides, pre-trained on 386 billion nucleotides of eukaryotic DNA. Without task-specific fine-tuning, GENERator exhibits strong intrinsic capabilities: unsupervised embedding analyses reveal phylogenetically coherent structure, and sequence recovery benchmarks demonstrate generative accuracy comparable to or exceeding state-of-the-art models with substantially improved computational efficiency. In a zero-shot setting, GENERator achieves competitive variant effect prediction performance relative to alignment-based methods, while remaining fully alignment-free and broadly applicable across species. With task-specific fine-tuning, the model attains leading performance on established genomic benchmarks. We further demonstrate practical generative applications. GENERator can generate protein-coding DNA sequences that translate into structurally plausible proteins and, through a prompt-guided design framework, design cis-regulatory elements with targeted activity profiles, including synthetic super-enhancers validated by high-throughput UMI-STARR-seq assays. Together, these results establish GENERator as an efficient and biologically grounded framework for genomic interpretation and programmable sequence design. Code and supplementary resources are available at https://github.com/GenerTeam/GENERator.", "date": "2025-02-11", "categories": []}
{"id": "2512.13635", "title": "SCR2-ST: Combine Single Cell with Spatial Transcriptomics for Efficient Active Sampling via Reinforcement Learning", "abstract": "Spatial transcriptomics (ST) is an emerging technology that enables researchers to investigate the molecular relationships underlying tissue morphology. However, acquiring ST data remains prohibitively expensive, and traditional fixed-grid sampling strategies lead to redundant measurements of morphologically similar or biologically uninformative regions, thus resulting in scarce data that constrain current methods. The well-established single-cell sequencing field, however, could provide rich biological data as an effective auxiliary source to mitigate this limitation. To bridge these gaps, we introduce SCR2-ST, a unified framework that leverages single-cell prior knowledge to guide efficient data acquisition and accurate expression prediction. SCR2-ST integrates a single-cell guided reinforcement learning-based (SCRL) active sampling and a hybrid regression-retrieval prediction network SCR2Net. SCRL combines single-cell foundation model embeddings with spatial density information to construct biologically grounded reward signals, enabling selective acquisition of informative tissue regions under constrained sequencing budgets. SCR2Net then leverages the actively sampled data through a hybrid architecture combining regression-based modeling with retrieval-augmented inference, where a majority cell-type filtering mechanism suppresses noisy matches and retrieved expression profiles serve as soft labels for auxiliary supervision. We evaluated SCR2-ST on three public ST datasets, demonstrating SOTA performance in both sampling efficiency and prediction accuracy, particularly under low-budget scenarios. Code is publicly available at: https://github.com/hrlblab/SCR2ST", "date": "2025-12-15", "categories": []}
{"id": "2601.19149", "title": "GPCR-Filter: a deep learning framework for efficient and precise GPCR modulator discovery", "abstract": "G protein-coupled receptors (GPCRs) govern diverse physiological processes and are central to modern pharmacology. Yet discovering GPCR modulators remains challenging because receptor activation often arises from complex allosteric effects rather than direct binding affinity, and conventional assays are slow, costly, and not optimized for capturing these dynamics. Here we present GPCR-Filter, a deep learning framework specifically developed for GPCR modulator discovery. We assembled a high-quality dataset of over 90,000 experimentally validated GPCR-ligand pairs, providing a robust foundation for training and evaluation. GPCR-Filter integrates the ESM-3 protein language model for high-fidelity GPCR sequence representations with graph neural networks that encode ligand structures, coupled through an attention-based fusion mechanism that learns receptor-ligand functional relationships. Across multiple evaluation settings, GPCR-Filter consistently outperforms state-of-the-art compound-protein interaction models and exhibits strong generalization to unseen receptors and ligands. Notably, the model successfully identified micromolar-level agonists of the 5-HT\\textsubscript{1A} receptor with distinct chemical frameworks. These results establish GPCR-Filter as a scalable and effective computational approach for GPCR modulator discovery, advancing AI-assisted drug development for complex signaling systems.", "date": "2026-01-27", "categories": []}
{"id": "2602.00019", "title": "AutoBinder Agent: An MCP-Based Agent for End-to-End Protein Binder Design", "abstract": "Modern AI technologies for drug discovery are distributed across heterogeneous platforms-including web applications, desktop environments, and code libraries-leading to fragmented workflows, inconsistent interfaces, and high integration overhead. We present an agentic end-to-end drug design framework that leverages a Large Language Model (LLM) in conjunction with the Model Context Protocol (MCP) to dynamically coordinate access to biochemical databases, modular toolchains, and task-specific AI models. The system integrates four state-of-the-art components: MaSIF (MaSIF-site and MaSIF-seed-search) for geometric deep learning-based identification of protein-protein interaction (PPI) sites, Rosetta for grafting protein fragments onto protein backbones to form mini proteins, ProteinMPNN for amino acid sequences redesign, and AlphaFold3 for near-experimental accuracy in complex structure prediction. Starting from a target structure, the framework supports de novo binder generation via surface analysis, scaffold grafting and pose construction, sequence optimization, and structure prediction. Additionally, by replacing rigid, script-based workflows with a protocol-driven, LLM-coordinated architecture, the framework improves reproducibility, reduces manual overhead, and ensures extensibility, portability, and auditability across the entire drug design process.", "date": "2026-01-16", "categories": []}
{"id": "2602.00782", "title": "Controlling Repetition in Protein Language Models", "abstract": "Protein language models (PLMs) have enabled advances in structure prediction and de novo protein design, yet they frequently collapse into pathological repetition during generation. Unlike in text, where repetition merely reduces readability, in proteins it undermines structural confidence and functional viability. To unify this problem, we present the first systematic study of repetition in PLMs. We first propose quantitative metrics to characterize motif-level and homopolymer repetition and then demonstrate their negative impact on folding reliability. To address this challenge, we propose UCCS (Utility-Controlled Contrastive Steering), which steers protein generation with a constrained dataset. Instead of naively contrasting high- vs. low-repetition sequences, we construct contrastive sets that maximize differences in repetition while tightly controlling for structural utility. This disentanglement yields steering vectors that specifically target repetition without degrading foldability. Injected at inference, these vectors consistently reduce repetition without retraining or heuristic decoding. Experiments with ESM-3 and ProtGPT2 in CATH, UniRef50, and SCOP show that our method outperforms decoding penalties and other baselines, substantially lowering repetition while preserving AlphaFold confidence scores. Our results establish repetition control as a central challenge for PLMs and highlight dataset-guided steering as a principled approach for reliable protein generation.", "date": "2026-01-31", "categories": []}
{"id": "2602.01772", "title": "DIA-CLIP: a universal representation learning framework for zero-shot DIA proteomics", "abstract": "Data-independent acquisition mass spectrometry (DIA-MS) has established itself as a cornerstone of proteomic profiling and large-scale systems biology, offering unparalleled depth and reproducibility. Current DIA analysis frameworks, however, require semi-supervised training within each run for peptide-spectrum match (PSM) re-scoring. This approach is prone to overfitting and lacks generalizability across diverse species and experimental conditions. Here, we present DIA-CLIP, a pre-trained model shifting the DIA analysis paradigm from semi-supervised training to universal cross-modal representation learning. By integrating dual-encoder contrastive learning framework with encoder-decoder architecture, DIA-CLIP establishes a unified cross-modal representation for peptides and corresponding spectral features, achieving high-precision, zero-shot PSM inference. Extensive evaluations across diverse benchmarks demonstrate that DIA-CLIP consistently outperforms state-of-the-art tools, yielding up to a 45% increase in protein identification while achieving a 12% reduction in entrapment identifications. Moreover, DIA-CLIP holds immense potential for diverse practical applications, such as single-cell and spatial proteomics, where its enhanced identification depth facilitates the discovery of novel biomarkers and the elucidates of intricate cellular mechanisms.", "date": "2026-02-02", "categories": []}
{"id": "2602.02425", "title": "Repurposing Protein Language Models for Latent Flow-Based Fitness Optimization", "abstract": "Protein fitness optimization is challenged by a vast combinatorial landscape where high-fitness variants are extremely sparse. Many current methods either underperform or require computationally expensive gradient-based sampling. We present CHASE, a framework that repurposes the evolutionary knowledge of pretrained protein language models by compressing their embeddings into a compact latent space. By training a conditional flow-matching model with classifier-free guidance, we enable the direct generation of high-fitness variants without predictor-based guidance during the ODE sampling steps. CHASE achieves state-of-the-art performance on AAV and GFP protein design benchmarks. Finally, we show that bootstrapping with synthetic data can further enhance performance in data-constrained settings.", "date": "2026-02-02", "categories": []}
{"id": "2511.09529", "title": "SiDGen: Structure-informed Diffusion for Generative modeling of Ligands for Proteins", "abstract": "Structure-based drug design (SBDD) faces a fundamental scaling fidelity dilemma: rich pocket-aware conditioning captures interaction geometry but can be costly, often scales quadratically ($O(L^2)$) or worse with protein length ($L$), while efficient sequence-only conditioning can miss key interaction structure. We propose SiDGen, a structure-informed discrete diffusion framework that resolves this trade-off through a Topological Information Bottleneck (TIB). SiDGen leverages a learned, soft assignment mechanism to compress residue-level protein representations into a compact bottleneck enabling downstream pairwise computations on the coarse grid ($O(L^2/s^2)$). This design reduces memory and computational cost without compromising generative accuracy. Our approach achieves state-of-the-art performance on CrossDocked2020 and DUD-E benchmarks while significantly reducing pairwise-tensor memory. SiDGen bridges the gap between sequence-based efficiency and pocket-aware conditioning, offering a scalable path for high-throughput structure-based discovery.", "date": "2025-11-12", "categories": []}
{"id": "2601.22203", "title": "Beyond Conditional Computation: Retrieval-Augmented Genomic Foundation Models with Gengram", "abstract": "Current genomic foundation models (GFMs) rely on extensive neural computation to implicitly approximate conserved biological motifs from single-nucleotide inputs. We propose Gengram, a conditional memory module that introduces an explicit and highly efficient lookup primitive for multi-base motifs via a genomic-specific hashing scheme, establishing genomic \"syntax\". Integrated into the backbone of state-of-the-art GFMs, Gengram achieves substantial gains (up to 14%) across several functional genomics tasks. The module demonstrates robust architectural generalization, while further inspection of Gengram's latent space reveals the emergence of meaningful representations that align closely with fundamental biological knowledge. By establishing structured motif memory as a modeling primitive, Gengram simultaneously boosts empirical performance and mechanistic interpretability, providing a scalable and biology-aligned pathway for the next generation of GFMs. The code is available at https://github.com/zhejianglab/Genos, and the model checkpoint is available at https://huggingface.co/ZhejiangLab/Gengram.", "date": "2026-01-29", "categories": []}
{"id": "2601.22408", "title": "Minimal-Action Discrete Schr\\\"odinger Bridge Matching for Peptide Sequence Design", "abstract": "Generative modeling of peptide sequences requires navigating a discrete and highly constrained space in which many intermediate states are chemically implausible or unstable. Existing discrete diffusion and flow-based methods rely on reversing fixed corruption processes or following prescribed probability paths, which can force generation through low-likelihood regions and require countless sampling steps. We introduce Minimal-action discrete Schr\\\"odinger Bridge Matching (MadSBM), a rate-based generative framework for peptide design that formulates generation as a controlled continuous-time Markov process on the amino-acid edit graph. To yield probability trajectories that remain near high-likelihood sequence neighborhoods throughout generation, MadSBM 1) defines generation relative to a biologically informed reference process derived from pre-trained protein language model logits and 2) learns a time-dependent control field that biases transition rates to produce low-action transport paths from a masked prior to the data distribution. We finally introduce guidance to the MadSBM sampling procedure towards a specific functional objective, expanding the design space of therapeutic peptides; to our knowledge, this represents the first-ever application of discrete classifier guidance to Schr\\\"odinger bridge-based generative models.", "date": "2026-01-29", "categories": []}
{"id": "2601.23212", "title": "Disentangling multispecific antibody function with graph neural networks", "abstract": "Multispecific antibodies offer transformative therapeutic potential by engaging multiple epitopes simultaneously, yet their efficacy is an emergent property governed by complex molecular architectures. Rational design is often bottlenecked by the inability to predict how subtle changes in domain topology influence functional outcomes, a challenge exacerbated by the scarcity of comprehensive experimental data. Here, we introduce a computational framework to address part of this gap. First, we present a generative method for creating large-scale, realistic synthetic functional landscapes that capture non-linear interactions where biological activity depends on domain connectivity. Second, we propose a graph neural network architecture that explicitly encodes these topological constraints, distinguishing between format configurations that appear identical to sequence-only models. We demonstrate that this model, trained on synthetic landscapes, recapitulates complex functional properties and, via transfer learning, has the potential to achieve high predictive accuracy on limited biological datasets. We showcase the model's utility by optimizing trade-offs between efficacy and toxicity in trispecific T-cell engagers and retrieving optimal common light chains. This work provides a robust benchmarking environment for disentangling the combinatorial complexity of multispecifics, accelerating the design of next-generation therapeutics.", "date": "2026-01-30", "categories": []}
{"id": "2503.00143", "title": "RecCrysFormer: Refined Protein Structural Prediction from 3D Patterson Maps via Recycling Training Runs", "abstract": "Determining protein structures at an atomic level remains a significant challenge in structural biology. We introduce $\\texttt{RecCrysFormer}$, a hybrid model that exploits the strengths of transformers with the aim of integrating experimental and ML approaches to protein structure determination from crystallographic data. $\\texttt{RecCrysFormer}$ leverages Patterson maps and incorporates known standardized partial structures of amino acid residues to directly predict electron density maps, which are essential for constructing detailed atomic models through crystallographic refinement processes. $\\texttt{RecCrysFormer}$ benefits from a ``recycling'' training regimen that iteratively incorporates results from crystallographic refinements and previous training runs as additional inputs in the form of template maps. Using a preliminary dataset of synthetic peptide fragments based on Protein Data Bank, $\\texttt{RecCrysFormer}$ achieves good accuracy in structural predictions and shows robustness against variations in crystal parameters, such as unit cell dimensions and angles.", "date": "2025-02-28", "categories": []}
{"id": "2601.18604", "title": "LaCoGSEA: Unsupervised deep learning for pathway analysis via latent correlation", "abstract": "Motivation: Pathway enrichment analysis is widely used to interpret gene expression data. Standard approaches, such as GSEA, rely on predefined phenotypic labels and pairwise comparisons, which limits their applicability in unsupervised settings. Existing unsupervised extensions, including single-sample methods, provide pathway-level summaries but primarily capture linear relationships and do not explicitly model gene-pathway associations. More recently, deep learning models have been explored to capture non-linear transcriptomic structure. However, their interpretation has typically relied on generic explainable AI (XAI) techniques designed for feature-level attribution. As these methods are not designed for pathway-level interpretation in unsupervised transcriptomic analyses, their effectiveness in this setting remains limited.\n  Results: To bridge this gap, we introduce LaCoGSEA (Latent Correlation GSEA), an unsupervised framework that integrates deep representation learning with robust pathway statistics. LaCoGSEA employs an autoencoder to capture non-linear manifolds and proposes a global gene-latent correlation metric as a proxy for differential expression, generating dense gene rankings without prior labels. We demonstrate that LaCoGSEA offers three key advantages: (i) it achieves improved clustering performance in distinguishing cancer subtypes compared to existing unsupervised baselines; (ii) it recovers a broader range of biologically meaningful pathways at higher ranks compared with linear dimensionality reduction and gradient-based XAI methods; and (iii) it maintains high robustness and consistency across varying experimental protocols and dataset sizes. Overall, LaCoGSEA provides state-of-the-art performance in unsupervised pathway enrichment analysis.\n  Availability and implementation: https://github.com/willyzzz/LaCoGSEA", "date": "2026-01-26", "categories": []}
{"id": "2601.18640", "title": "TwinPurify: Purifying gene expression data to reveal tumor-intrinsic transcriptional programs via self-supervised learning", "abstract": "Advances in single-cell and spatial transcriptomic technologies have transformed tumor ecosystem profiling at cellular resolution. However, large scale studies on patient cohorts continue to rely on bulk transcriptomic data, where variation in tumor purity obscures tumor-intrinsic transcriptional signals and constrains downstream discovery. Many deconvolution methods report strong performance on synthetic bulk mixtures but fail to generalize to real patient cohorts because of unmodeled biological and technical variation.\n  Here, we introduce TwinPurify, a representation learning framework that adapts the Barlow Twins self-supervised objective, representing a fundamental departure from the deconvolution paradigm. Rather than resolving the bulk mixture into discrete cell-type fractions, TwinPurify instead learns continuous, high-dimensional tumor embeddings by leveraging adjacent-normal profiles within the same cohort as \"background\" guidance, enabling the disentanglement of tumor-specific signals without relying on any external reference.\n  Benchmarked against multiple large cancer cohorts across RNA-seq and microarray platforms, TwinPurify outperforms conventional representation learning baselines like auto-encoders in recovering tumor-intrinsic and immune signals. The purified embeddings improve molecular subtype and grade classification, enhance survival model concordance, and uncover biologically meaningful pathway activities compared to raw bulk profiles. By providing a transferable framework for decontaminating bulk transcriptomics, TwinPurify extends the utility of existing clinical datasets for molecular discovery.", "date": "2026-01-26", "categories": []}
{"id": "2601.19037", "title": "XIMP: Cross Graph Inter-Message Passing for Molecular Property Prediction", "abstract": "Accurate molecular property prediction is central to drug discovery, yet graph neural networks often underperform in data-scarce regimes and fail to surpass traditional fingerprints. We introduce cross-graph inter-message passing (XIMP), which performs message passing both within and across multiple related graph representations. For small molecules, we combine the molecular graph with scaffold-aware junction trees and pharmacophore-encoding extended reduced graphs, integrating complementary abstractions. While prior work is either limited to a single abstraction or non-iterative communication across graphs, XIMP supports an arbitrary number of abstractions and both direct and indirect communication between them in each layer. Across ten diverse molecular property prediction tasks, XIMP outperforms state-of-the-art baselines in most cases, leveraging interpretable abstractions as an inductive bias that guides learning toward established chemical concepts, enhancing generalization in low-data settings.", "date": "2026-01-26", "categories": []}
{"id": "2601.19205", "title": "EnzyPGM: Pocket-conditioned Generative Model for Substrate-specific Enzyme Design", "abstract": "Designing enzymes with substrate-binding pockets is a critical challenge in protein engineering, as catalytic activity depends on the precise interaction between pockets and substrates. Currently, generative models dominate functional protein design but cannot model pocket-substrate interactions, which limits the generation of enzymes with precise catalytic environments. To address this issue, we propose EnzyPGM, a unified framework that jointly generates enzymes and substrate-binding pockets conditioned on functional priors and substrates, with a particular focus on learning accurate pocket-substrate interactions. At its core, EnzyPGM includes two main modules: a Residue-atom Bi-scale Attention (RBA) that jointly models intra-residue dependencies and fine-grained interactions between pocket residues and substrate atoms, and a Residue Function Fusion (RFF) that incorporates enzyme function priors into residue representations. Also, we curate EnzyPock, an enzyme-pocket dataset comprising 83,062 enzyme-substrate pairs across 1,036 four-level enzyme families. Extensive experiments demonstrate that EnzyPGM achieves state-of-the-art performance on EnzyPock. Notably, EnzyPGM reduces the average binding energy of 0.47 kcal/mol over EnzyGen, showing its superior performance on substrate-specific enzyme design. The code and dataset will be released later.", "date": "2026-01-27", "categories": []}
{"id": "2601.19232", "title": "Structure-based RNA Design by Step-wise Optimization of Latent Diffusion Model", "abstract": "RNA inverse folding, designing sequences to form specific 3D structures, is critical for therapeutics, gene regulation, and synthetic biology. Current methods, focused on sequence recovery, struggle to address structural objectives like secondary structure consistency (SS), minimum free energy (MFE), and local distance difference test (LDDT), leading to suboptimal structural accuracy. To tackle this, we propose a reinforcement learning (RL) framework integrated with a latent diffusion model (LDM). Drawing inspiration from the success of diffusion models in RNA inverse folding, which adeptly model complex sequence-structure interactions, we develop an LDM incorporating pre-trained RNA-FM embeddings from a large-scale RNA model. These embeddings capture co-evolutionary patterns, markedly improving sequence recovery accuracy. However, existing approaches, including diffusion-based methods, cannot effectively handle non-differentiable structural objectives. By contrast, RL excels in this task by using policy-driven reward optimization to navigate complex, non-gradient-based objectives, offering a significant advantage over traditional methods. In summary, we propose the Step-wise Optimization of Latent Diffusion Model (SOLD), a novel RL framework that optimizes single-step noise without sampling the full diffusion trajectory, achieving efficient refinement of multiple structural objectives. Experimental results demonstrate SOLD surpasses its LDM baseline and state-of-the-art methods across all metrics, establishing a robust framework for RNA inverse folding with profound implications for biotechnological and therapeutic applications.", "date": "2026-01-27", "categories": []}
{"id": "2409.02588", "title": "Multiview Random Vector Functional Link Network for Predicting DNA-Binding Proteins", "abstract": "The identification of DNA-binding proteins (DBPs) is essential due to their significant impact on various biological activities. Understanding the mechanisms underlying protein-DNA interactions is essential for elucidating various life activities. In recent years, machine learning-based models have been prominently utilized for DBP prediction. In this paper, to predict DBPs, we propose a novel framework termed a multiview random vector functional link (MvRVFL) network, which fuses neural network architecture with multiview learning. The MvRVFL model integrates both late and early fusion advantages, enabling separate regularization parameters for each view, while utilizing a closed-form solution for efficiently determining unknown parameters. The primal objective function incorporates a coupling term aimed at minimizing a composite of errors stemming from all views. From each of the three protein views of the DBP datasets, we extract five features. These features are then fused together by incorporating a hidden feature during the model training process. The performance of the proposed MvRVFL model on the DBP dataset surpasses that of baseline models, demonstrating its superior effectiveness. We further validate the practicality of the proposed model across diverse benchmark datasets, and both theoretical analysis and empirical results consistently demonstrate its superior generalization performance over baseline models.", "date": "2024-09-04", "categories": []}
{"id": "2601.18713", "title": "Point transformer for protein structural heterogeneity analysis using CryoEM", "abstract": "Structural dynamics of macromolecules is critical to their structural-function relationship. Cryogenic electron microscopy (CryoEM) provides snapshots of vitrified protein at different compositional and conformational states, and the structural heterogeneity of proteins can be characterized through computational analysis of the images. For protein systems with multiple degrees of freedom, it is still challenging to disentangle and interpret the different modes of dynamics. Here, by implementing Point Transformer, a self-attention network designed for point cloud analysis, we are able to improve the performance of heterogeneity analysis on CryoEM data, and characterize the dynamics of highly complex protein systems in a more human-interpretable way.", "date": "2026-01-26", "categories": []}
{"id": "2601.18716", "title": "Conditioned Generative Modeling of Molecular Glues: A Realistic AI Approach for Synthesizable Drug-like Molecules", "abstract": "Alzheimer's disease (AD) is marked by the pathological accumulation of amyloid beta-42 (Abeta-42), contributing to synaptic dysfunction and neurodegeneration. While extracellular amyloid plaques are well-studied, increasing evidence highlights intracellular Abeta-42 as an early and toxic driver of disease progression. In this study, we present a novel, AI-assisted drug design approach to promote targeted degradation of Abeta-42 via the ubiquitin-proteasome system (UPS), using E3 ligase-directed molecular glues. We systematically evaluated the ternary complex formation potential of Abeta-42 with three E3 ligases: CRBN, VHL, and MDM2, through structure-based modeling, ADMET screening, and docking. We then developed a Ligase-Conditioned Junction Tree Variational Autoencoder (LC-JT-VAE) to generate ligase-specific small molecules, incorporating protein sequence embeddings and torsional angle-aware molecular graphs. Our results demonstrate that this generative model can produce chemically valid, novel, and target-specific molecular glues capable of facilitating Abeta-42 degradation. This integrated approach offers a promising framework for designing UPS-targeted therapies for neurodegenerative diseases.", "date": "2026-01-26", "categories": []}
{"id": "2510.13018", "title": "Escaping Local Optima in the Waddington Landscape: A Two-Stage TRPO-PPO Approach for Single-Cell Perturbation Analysis", "abstract": "Modeling cellular responses to genetic and chemical perturbations remains a central challenge in single-cell biology. Existing data-driven frameworks have advanced perturbation prediction through variational autoencoders, chemically conditioned autoencoders, and large-scale transformer pretraining. However, most existing models rely exclusively on either in silico perturbation data or experimental perturbation data but rarely integrate both, limiting their ability to generalize and validate predictions across simulated and real biological contexts in a digital twin system. Moreover, the models are prone to local optima in the nonconvex Waddington landscape of cell fate decisions, where poor initialization can trap trajectories in spurious lineages. In this work, we introduce a two-stage reinforcement learning algorithm for modeling single-cell perturbation. We first compute an explicit natural gradient update using Fisher-vector products and a conjugate gradient solver, scaled by a KL trust-region constraint to provide a safe, curvature-aware first step for the policy. Starting with these preconditioned parameters, we then apply a second phase of proximal policy optimization (PPO) with a KL penalty, exploiting minibatch efficiency to refine the policy. We demonstrate that this initialization strategy substantially improves generalization on Single-cell RNA sequencing (scRNA-seq) perturbation analysis in a digital twin system.", "date": "2025-10-14", "categories": []}
{"id": "2601.02530", "title": "Multi-scale Graph Autoregressive Modeling: Molecular Property Prediction via Next Token Prediction", "abstract": "We present Connection-Aware Motif Sequencing (CamS), a graph-to-sequence representation that enables decoder-only Transformers to learn molecular graphs via standard next-token prediction (NTP). For molecular property prediction, SMILES-based NTP scales well but lacks explicit topology, whereas graph-native masked modeling captures connectivity but risks disrupting the pivotal chemical details (e.g., activity cliffs). CamS bridges this gap by serializing molecular graphs into structure-rich causal sequences. CamS first mines data-driven connection-aware motifs. It then serializes motifs via scaffold-rooted breadth-first search (BFS) to establish a stable core-to-periphery order. Crucially, CamS enables hierarchical modeling by concatenating sequences from fine to coarse motif scales, allowing the model to condition global scaffolds on dense, uncorrupted local structural evidence. We instantiate CamS-LLaMA by pre-training a vanilla LLaMA backbone on CamS sequences. It achieves state-of-the-art performance on MoleculeNet and the activity-cliff benchmark MoleculeACE, outperforming both SMILES-based language models and strong graph baselines. Interpretability analysis confirms that our multi-scale causal serialization effectively drives attention toward cliff-determining differences.", "date": "2026-01-05", "categories": []}
{"id": "2505.05577", "title": "PyTDC: A multimodal machine learning training, evaluation, and inference platform for biomedical foundation models", "abstract": "Existing biomedical benchmarks do not provide end-to-end infrastructure for training, evaluation, and inference of models that integrate multimodal biological data and a broad range of machine learning tasks in therapeutics. We present PyTDC, an open-source machine-learning platform providing streamlined training, evaluation, and inference software for multimodal biological AI models. PyTDC unifies distributed, heterogeneous, continuously updated data sources and model weights and standardizes benchmarking and inference endpoints. This paper discusses the components of PyTDC's architecture and, to our knowledge, the first-of-its-kind case study on the introduced single-cell drug-target nomination ML task. We find state-of-the-art methods in graph representation learning and domain-specific methods from graph theory perform poorly on this task. Though we find a context-aware geometric deep learning method that outperforms the evaluated SoTA and domain-specific baseline methods, the model is unable to generalize to unseen cell types or incorporate additional modalities, highlighting PyTDC's capacity to facilitate an exciting avenue of research developing multimodal, context-aware, foundation models for open problems in biomedical AI.", "date": "2025-05-08", "categories": []}
{"id": "2601.15504", "title": "SAGE-FM: A lightweight and interpretable spatial transcriptomics foundation model", "abstract": "Spatial transcriptomics enables spatial gene expression profiling, motivating computational models that capture spatially conditioned regulatory relationships. We introduce SAGE-FM, a lightweight spatial transcriptomics foundation model based on graph convolutional networks (GCNs) trained with a masked central spot prediction objective. Trained on 416 human Visium samples spanning 15 organs, SAGE-FM learns spatially coherent embeddings that robustly recover masked genes, with 91% of masked genes showing significant correlations (p < 0.05). The embeddings generated by SAGE-FM outperform MOFA and existing spatial transcriptomics methods in unsupervised clustering and preservation of biological heterogeneity. SAGE-FM generalizes to downstream tasks, enabling 81% accuracy in pathologist-defined spot annotation in oropharyngeal squamous cell carcinoma and improving glioblastoma subtype prediction relative to MOFA. In silico perturbation experiments further demonstrate that the model captures directional ligand-receptor and upstream-downstream regulatory effects consistent with ground truth. These results demonstrate that simple, parameter-efficient GCNs can serve as biologically interpretable and spatially aware foundation models for large-scale spatial transcriptomics.", "date": "2026-01-21", "categories": []}
{"id": "2601.15771", "title": "Rethinking Drug-Drug Interaction Modeling as Generalizable Relation Learning", "abstract": "Drug-drug interaction (DDI) prediction is central to drug discovery and clinical development, particularly in the context of increasingly prevalent polypharmacy. Although existing computational methods achieve strong performance on standard benchmarks, they often fail to generalize to realistic deployment scenarios, where most candidate drug pairs involve previously unseen drugs and validated interactions are scarce. We demonstrate that proximity in the embedding spaces of prevailing molecule-centric DDI models does not reliably correspond to interaction labels, and that simply scaling up model capacity therefore fails to improve generalization. To address these limitations, we propose GenRel-DDI, a generalizable relation learning framework that reformulates DDI prediction as a relation-centric learning problem, in which interaction representations are learned independently of drug identities. This relation-level abstraction enables the capture of transferable interaction patterns that generalize to unseen drugs and novel drug pairs. Extensive experiments across multiple benchmark demonstrate that GenRel-DDI consistently and significantly outperforms state-of-the-art methods, with particularly large gains on strict entity-disjoint evaluations, highlighting the effectiveness and practical utility of relation learning for robust DDI prediction. The code is available at https://github.com/SZU-ADDG/GenRel-DDI.", "date": "2026-01-22", "categories": []}
{"id": "2601.14285", "title": "A Comparison of Polynomial-Based Tree Clustering Methods", "abstract": "Tree structures appear in many fields of the life sciences, including phylogenetics, developmental biology and nucleic acid structures. Trees can be used to represent RNA secondary structures, which directly relate to the function of non-coding RNAs. Recent developments in sequencing technology and artificial intelligence have yielded numerous biological data that can be represented with tree structures. This requires novel methods for tree structure data analytics. Tree polynomials provide a computationally efficient, interpretable and comprehensive way to encode tree structures as matrices, which are compatible with most data analytics tools. Machine learning methods based on the Canberra distance between tree polynomials have been introduced to analyze phylogenies and nucleic acid structures. In this paper, we compare the performance of different distances in tree clustering methods based on a tree distinguishing polynomial. We also implement two basic autoencoder models for clustering trees using the polynomial. We find that the distance based methods with entry-level normalized distances have the highest clustering accuracy among the compared methods.", "date": "2026-01-13", "categories": []}
{"id": "2601.14349", "title": "MARBLE: Multi-Agent Reasoning for Bioinformatics Learning and Evolution", "abstract": "Motivation: Developing high-performing bioinformatics models typically requires repeated cycles of hypothesis formulation, architectural redesign, and empirical validation, making progress slow, labor-intensive, and difficult to reproduce. Although recent LLM-based assistants can automate isolated steps, they lack performance-grounded reasoning and stability-aware mechanisms required for reliable, iterative model improvement in bioinformatics workflows. Results: We introduce MARBLE, an execution-stable autonomous model refinement framework for bioinformatics models. MARBLE couples literature-aware reference selection with structured, debate-driven architectural reasoning among role-specialized agents, followed by autonomous execution, evaluation, and memory updates explicitly grounded in empirical performance. Across spatial transcriptomics domain segmentation, drug-target interaction prediction, and drug response prediction, MARBLE consistently achieves sustained performance improvements over strong baselines across multiple refinement cycles, while maintaining high execution robustness and low regression rates. Framework-level analyses demonstrate that structured debate, balanced evidence selection, and performance-grounded memory are critical for stable, repeatable model evolution, rather than single-run or brittle gains. Availability: Source code, data and Supplementary Information are available at https://github.com/PRISM-DGU/MARBLE.", "date": "2026-01-20", "categories": []}
{"id": "2601.14841", "title": "MTFlow: Time-Conditioned Flow Matching for Microtubule Segmentation in Noisy Microscopy Images", "abstract": "Microtubules are cytoskeletal filaments that play essential roles in many cellular processes and are key therapeutic targets in several diseases. Accurate segmentation of microtubule networks is critical for studying their organization and dynamics but remains challenging due to filament curvature, dense crossings, and image noise. We present MTFlow, a novel time-conditioned flow-matching model for microtubule segmentation. Unlike conventional U-Net variants that predict masks in a single pass, MTFlow learns vector fields that iteratively transport noisy masks toward the ground truth, enabling interpretable, trajectory-based refinement. Our architecture combines a U-Net backbone with temporal embeddings, allowing the model to capture the dynamics of uncertainty resolution along filament boundaries. We trained and evaluated MTFlow on synthetic and real microtubule datasets and assessed its generalization capability on public biomedical datasets of curvilinear structures such as retinal blood vessels and nerves. MTFlow achieves competitive segmentation accuracy comparable to state-of-the-art models, offering a powerful and time-efficient tool for filamentous structure analysis with more precise annotations than manual or semi-automatic approaches.", "date": "2026-01-21", "categories": []}
{"id": "2307.10634", "title": "Generative Language Models on Nucleotide Sequences of Human Genes", "abstract": "Language models, especially transformer-based ones, have achieved colossal success in NLP. To be precise, studies like BERT for NLU and works like GPT-3 for NLG are very important. If we consider DNA sequences as a text written with an alphabet of four letters representing the nucleotides, they are similar in structure to natural languages. This similarity has led to the development of discriminative language models such as DNABert in the field of DNA-related bioinformatics. To our knowledge, however, the generative side of the coin is still largely unexplored. Therefore, we have focused on the development of an autoregressive generative language model such as GPT-3 for DNA sequences. Since working with whole DNA sequences is challenging without extensive computational resources, we decided to conduct our study on a smaller scale and focus on nucleotide sequences of human genes rather than the whole DNA. This decision has not changed the structure of the problem, as both DNA and genes can be considered as 1D sequences consisting of four different nucleotides without losing much information and without oversimplification. Firstly, we systematically studied an almost entirely unexplored problem and observed that RNNs perform best, while simple techniques such as N-grams are also promising. Another beneficial point was learning how to work with generative models on languages we do not understand, unlike natural languages. The importance of using real-world tasks beyond classical metrics such as perplexity was noted. In addition, we examined whether the data-hungry nature of these models can be altered by selecting a language with minimal vocabulary size, four due to four different types of nucleotides. The reason for reviewing this was that choosing such a language might make the problem easier. However, in this study, we found that this did not change the amount of data required very much.", "date": "2023-07-20", "categories": []}
{"id": "2408.16245", "title": "Large-Scale Multi-omic Biosequence Transformers for Modeling Protein-Nucleic Acid Interactions", "abstract": "The transformer architecture has revolutionized bioinformatics and driven progress in the understanding and prediction of the properties of biomolecules. To date, most biosequence transformers have been trained on single-omic data - either proteins or nucleic acids - and have seen incredible success in downstream tasks in each domain, with particularly noteworthy breakthroughs in protein structural modeling. However, single-omic pretraining limits the ability of these models to capture cross-modal interactions. Here we present OmniBioTE, the largest open-source multi-omic model trained on over 250 billion tokens of mixed protein and nucleic acid data. We show that despite only being trained on unlabeled sequence data, OmniBioTE learns joint representations mapping genes to their corresponding protein sequences. We further demonstrate that OmniBioTE achieves state-of-the-art results predicting the change in Gibbs free energy ({\\Delta G}) of the binding interaction between a given nucleic acid and protein. Remarkably, we show that multi-omic biosequence transformers emergently learn useful structural information without any a priori structural training, allowing us to predict which protein residues are most involved in the protein-nucleic acid binding interaction. Compared to single-omic controls trained with identical compute, OmniBioTE also demonstrates superior performance-per-FLOP across both multi-omic and single-omic benchmarks. Together, these results highlight the power of a unified modeling approach for biological sequences and establish OmniBioTE as a foundation model for multi-omic discovery.", "date": "2024-08-29", "categories": []}
{"id": "2411.12853", "title": "Integrating Secondary Structures Information into Triangular Spatial Relationships (TSR) for Advanced Protein Classification", "abstract": "Protein structures represent the key to deciphering biological functions. The more detailed form of similarity among these proteins is sometimes overlooked by the conventional structural comparison methods. In contrast, further advanced methods, such as Triangular Spatial Relationship (TSR), have been demonstrated to make finer differentiations. Still, the classical implementation of TSR does not provide for the integration of secondary structure information, which is important for a more detailed understanding of the folding pattern of a protein. To overcome these limitations, we developed the SSE-TSR approach. The proposed method integrates secondary structure elements (SSEs) into TSR-based protein representations. This allows an enriched representation of protein structures by considering 18 different combinations of helix, strand, and coil arrangements. Our results show that using SSEs improves the accuracy and reliability of protein classification to varying degrees. We worked with two large protein datasets of 9.2K and 7.8K samples, respectively. We applied the SSE-TSR approach and used a neural network model for classification. Interestingly, introducing SSEs improved performance statistics for Dataset 1, with accuracy moving from 96.0% to 98.3%. For Dataset 2, where the performance statistics were already good, further small improvements were found with the introduction of SSE, giving an accuracy of 99.5% compared to 99.4%. These results show that SSE integration can dramatically improve TSR key discrimination, with significant benefits in datasets with low initial accuracies and only incremental gains in those with high baseline performance. Thus, SSE-TSR is a powerful bioinformatics tool that improves protein classification and understanding of protein function and interaction.", "date": "2024-11-19", "categories": []}
{"id": "2411.16332", "title": "Cluster-based human-in-the-loop strategy for improving machine learning-based circulating tumor cell detection in liquid biopsy", "abstract": "Detection and differentiation of circulating tumor cells (CTCs) and non-CTCs in blood draws of cancer patients pose multiple challenges. While the gold standard relies on tedious manual evaluation of an automatically generated selection of images, machine learning (ML) techniques offer the potential to automate these processes. However, human assessment remains indispensable when the ML system arrives at uncertain or wrong decisions due to an insufficient set of labeled training data. This study introduces a human-in-the-loop (HiL) strategy for improving ML-based CTC detection. We combine self-supervised deep learning and a conventional ML-based classifier and propose iterative targeted sampling and labeling of new unlabeled training samples by human experts. The sampling strategy is based on the classification performance of local latent space clusters. The advantages of the proposed approach compared to naive random sampling are demonstrated for liquid biopsy data from patients with metastatic breast cancer.", "date": "2024-11-25", "categories": []}
{"id": "2601.12219", "title": "Persistent Sheaf Laplacian Analysis of Protein Stability and Solubility Changes upon Mutation", "abstract": "Genetic mutations frequently disrupt protein structure, stability, and solubility, acting as primary drivers for a wide spectrum of diseases. Despite the critical importance of these molecular alterations, existing computational models often lack interpretability, and fail to integrate essential physicochemical interaction. To overcome these limitations, we propose SheafLapNet, a unified predictive framework grounded in the mathematical theory of Topological Deep Learning (TDL) and Persistent Sheaf Laplacian (PSL). Unlike standard Topological Data Analysis (TDA) tools such as persistent homology, which are often insensitive to heterogeneous information, PSL explicitly encodes specific physical and chemical information such as partial charges directly into the topological analysis. SheafLapNet synergizes these sheaf-theoretic invariants with advanced protein transformer features and auxiliary physical descriptors to capture intrinsic molecular interactions in a multiscale and mechanistic manner. To validate our framework, we employ rigorous benchmarks for both regression and classification tasks. For stability prediction, we utilize the comprehensive S2648 and S350 datasets. For solubility prediction, we employ the PON-Sol2 dataset, which provides annotations for increased, decreased, or neutral solubility changes. By integrating these multi-perspective features, SheafLapNet achieves state-of-the-art performance across these diverse benchmarks, demonstrating that sheaf-theoretic modeling significantly enhances both interpretability and generalizability in predicting mutation-induced structural and functional changes.", "date": "2026-01-18", "categories": []}
{"id": "2601.13693", "title": "End-to-End Reverse Screening Identifies Protein Targets of Small Molecules Using HelixFold3", "abstract": "Identifying protein targets for small molecules, or reverse screening, is essential for understanding drug action, guiding compound repurposing, predicting off-target effects, and elucidating the molecular mechanisms of bioactive compounds. Despite its critical role, reverse screening remains challenging because accurately capturing interactions between a small molecule and structurally diverse proteins is inherently complex, and conventional step-wise workflows often propagate errors across decoupled steps such as target structure modeling, pocket identification, docking, and scoring. Here, we present an end-to-end reverse screening strategy leveraging HelixFold3, a high-accuracy biomolecular structure prediction model akin to AlphaFold3, which simultaneously models the folding of proteins from a protein library and the docking of small-molecule ligands within a unified framework. We validate this approach on a diverse and representative set of approximately one hundred small molecules. Compared with conventional reverse docking, our method improves screening accuracy and demonstrates enhanced structural fidelity, binding-site precision, and target prioritization. By systematically linking small molecules to their protein targets, this framework establishes a scalable and straightforward platform for dissecting molecular mechanisms, exploring off-target interactions, and supporting rational drug discovery.", "date": "2026-01-20", "categories": []}
{"id": "2407.06703", "title": "HERMES: Holographic Equivariant neuRal network model for Mutational Effect and Stability prediction", "abstract": "Predicting the stability and fitness effects of amino acid mutations in proteins is a cornerstone of biological discovery and engineering. Various experimental techniques have been developed to measure mutational effects, providing us with extensive datasets across a diverse range of proteins. By training on these data, traditional computational modeling and more recent machine learning approaches have advanced significantly in predicting mutational effects. Here, we introduce HERMES, a 3D rotationally equivariant structure-based neural network model for mutational effect and stability prediction. Pre-trained to predict amino acid propensity from its surrounding 3D structure, HERMES can be fine-tuned for mutational effects using our open-source code. We present a suite of HERMES models, pre-trained with different strategies, and fine-tuned to predict the stability effect of mutations. Benchmarking against other models shows that HERMES often outperforms or matches their performance in predicting mutational effect on stability, binding, and fitness. HERMES offers versatile tools for evaluating mutational effects and can be fine-tuned for specific predictive objectives.", "date": "2024-07-09", "categories": []}
{"id": "2505.04823", "title": "ProteinGuide: On-the-fly property guidance for protein sequence generative models", "abstract": "Sequence generative models are transforming protein engineering. However, no principled framework exists for conditioning these models on auxiliary information, such as experimental data, without additional training of a generative model. Herein, we present ProteinGuide, a method for such \"on-the-fly\" conditioning, amenable to a broad class of protein generative models including Masked Language Models (e.g. ESM3), any-order auto-regressive models (e.g. ProteinMPNN) as well as diffusion and flow matching models (e.g. MultiFlow). ProteinGuide stems from our unifying view of these model classes under a single statistical framework. As proof of principle, we perform several in silico experiments. We first guide pre-trained generative models to design proteins with user-specified properties, such as higher stability or activity. Next, we design for optimizing two desired properties that are in tension with each other. Finally, we apply our method in the wet lab, using ProteinGuide to increase the editing activity of an adenine base editor in vivo with data from only a single pooled library of 2,000 variants. We find that a single round of ProteinGuide achieves a higher editing efficiency than was previously achieved using seven rounds of directed evolution.", "date": "2025-05-07", "categories": []}
{"id": "2601.10131", "title": "M^4olGen: Multi-Agent, Multi-Stage Molecular Generation under Precise Multi-Property Constraints", "abstract": "Generating molecules that satisfy precise numeric constraints over multiple physicochemical properties is critical and challenging. Although large language models (LLMs) are expressive, they struggle with precise multi-objective control and numeric reasoning without external structure and feedback. We introduce \\textbf{M olGen}, a fragment-level, retrieval-augmented, two-stage framework for molecule generation under multi-property constraints. Stage I : Prototype generation: a multi-agent reasoner performs retrieval-anchored, fragment-level edits to produce a candidate near the feasible region. Stage II : RL-based fine-grained optimization: a fragment-level optimizer trained with Group Relative Policy Optimization (GRPO) applies one- or multi-hop refinements to explicitly minimize the property errors toward our target while regulating edit complexity and deviation from the prototype. A large, automatically curated dataset with reasoning chains of fragment edits and measured property deltas underpins both stages, enabling deterministic, reproducible supervision and controllable multi-hop reasoning. Unlike prior work, our framework better reasons about molecules by leveraging fragments and supports controllable refinement toward numeric targets. Experiments on generation under two sets of property constraints (QED, LogP, Molecular Weight and HOMO, LUMO) show consistent gains in validity and precise satisfaction of multi-property targets, outperforming strong LLMs and graph-based algorithms.", "date": "2026-01-15", "categories": []}
{"id": "2601.11012", "title": "Efficient Protein Optimization via Structure-aware Hamiltonian Dynamics", "abstract": "The ability to engineer optimized protein variants has transformative potential for biotechnology and medicine. Prior sequence-based optimization methods struggle with the high-dimensional complexities due to the epistasis effect and the disregard for structural constraints. To address this, we propose HADES, a Bayesian optimization method utilizing Hamiltonian dynamics to efficiently sample from a structure-aware approximated posterior. Leveraging momentum and uncertainty in the simulated physical movements, HADES enables rapid transition of proposals toward promising areas. A position discretization procedure is introduced to propose discrete protein sequences from such a continuous state system. The posterior surrogate is powered by a two-stage encoder-decoder framework to determine the structure and function relationships between mutant neighbors, consequently learning a smoothed landscape to sample from. Extensive experiments demonstrate that our method outperforms state-of-the-art baselines in in-silico evaluations across most metrics. Remarkably, our approach offers a unique advantage by leveraging the mutual constraints between protein structure and sequence, facilitating the design of protein sequences with similar structures and optimized properties. The code and data are publicly available at https://github.com/GENTEL-lab/HADES.", "date": "2026-01-16", "categories": []}
{"id": "2601.11028", "title": "AVP-Pro: An Adaptive Multi-Modal Fusion and Contrastive Learning Approach for Comprehensive Two-Stage Antiviral Peptide Identification", "abstract": "The accurate identification of antiviral peptides (AVPs) is crucial for novel drug development. However, existing methods still have limitations in capturing complex sequence dependencies and distinguishing confusing samples with high similarity. To address these challenges, we propose AVP-Pro, a novel two-stage predictive framework that integrates adaptive feature fusion and contrastive learning. To comprehensively capture the physicochemical properties and deep-seated patterns of peptide sequences, we constructed a panoramic feature space encompassing 10 distinct descriptors and designed a hierarchical fusion architecture. This architecture integrates self-attention and adaptive gating mechanisms to dynamically modulate the weights of local motifs extracted by CNNs and global dependencies captured by BiLSTMs based on sequence context. Targeting the blurred decision boundary caused by the high similarity between positive and negative sample sequences, we adopted an Online Hard Example Mining (OHEM)-driven contrastive learning strategy enhanced by BLOSUM62. This approach significantly sharpened the model's discriminative power. Model evaluation results show that in the first stage of general AVP identification, the model achieved an accuracy of 0.9531 and an MCC of 0.9064, outperforming existing state-of-the-art (SOTA) methods. In the second stage of functional subtype prediction, combined with a transfer learning strategy, the model realized accurate classification of 6 viral families and 8 specific viruses under small-sample conditions. AVP-Pro provides a powerful and interpretable new tool for the high-throughput screening of antiviral drugs. To further enhance accessibility for users, we have developed a user-friendly web interface, which is available at https://wwwy1031-avp-pro.hf.space.", "date": "2026-01-16", "categories": []}
{"id": "2401.02501", "title": "A Kolmogorov metric embedding for live cell microscopy signaling patterns", "abstract": "We present a metric embedding that captures spatiotemporal patterns of cell signaling dynamics in 5-D $(x,y,z,channel,time)$ live cell microscopy movies. The embedding uses a metric distance called the normalized information distance (NID) based on Kolmogorov complexity theory, an absolute measure of information content between digital objects. The NID uses statistics of lossless compression to compute a theoretically optimal metric distance between pairs of 5-D movies, requiring no a priori knowledge of expected pattern dynamics, and no training data. The cell signaling structure function (SSF) is defined using a class of metric 3-D image filters that compute at each spatiotemporal cell centroid the voxel intensity configuration of the nucleus w.r.t. the surrounding cytoplasm, or a functional output e.g. velocity. The only parameter is the expected cell radii ($\\mu m$). The SSF can be optionally combined with segmentation and tracking algorithms. The resulting lossless compression pipeline represents each 5-D input movie as a single point in a metric embedding space. The utility of a metric embedding follows from Euclidean distance between any points in the embedding space approximating optimally the pattern difference, as measured by the NID, between corresponding pairs of 5-D movies. This is true throughout the embedding space, not only at points corresponding to input images. Examples are shown for synthetic data, for 2-D+time movies of ERK and AKT signaling under different oncogenic mutations in human epithelial (MCF10A) cells, for 3-D MCF10A spheroids under optogenetic manipulation of ERK, and for ERK dynamics during colony differentiation in human stem cells.", "date": "2024-01-04", "categories": []}
{"id": "2412.16276", "title": "SGAC: A Graph Neural Network Framework for Imbalanced and Structure-Aware AMP Classification", "abstract": "Classifying Antimicrobial Peptides (AMPs) from the vast collection of peptides derived from metagenomic sequencing offers a promising avenue for combating antibiotic resistance. However, most existing AMP classification methods rely primarily on sequence-based representations and fail to capture the spatial structural information critical for accurate identification. Although recent graph-based approaches attempt to incorporate structural information, they typically construct residue- or atom-level graphs that introduce redundant atomic details and increase structural complexity. Furthermore, the class imbalance between the small number of known AMPs and the abundant non-AMPs significantly hinders predictive performance. To address these challenges, we employ lightweight OmegaFold to predict the three-dimensional structures of peptides and construct peptide graphs using C {\\alpha} atoms to capture their backbone geometry and spatial topology. Building on this representation, we propose the Spatial GNN-based AMP Classifier (SGAC), a novel framework that leverages Graph Neural Networks (GNNs) to extract structural features and generate discriminative graph representations. To handle class imbalance, SGAC incorporates Weight-enhanced Contrastive Learning to cluster structurally similar peptides and separate dissimilar ones through adaptive weighting, and applies Weight-enhanced Pseudo-label Distillation to generate high-confidence pseudo labels for unlabeled samples, achieving balanced and consistent representation learning. Experiments on publicly available AMP and non-AMP datasets demonstrate that SGAC significantly achieves state-of-the-art performance compared to baselines.", "date": "2024-12-20", "categories": []}
{"id": "2601.08336", "title": "Tissue Classification and Whole-Slide Images Analysis via Modeling of the Tumor Microenvironment and Biological Pathways", "abstract": "Automatic integration of whole slide images (WSIs) and gene expression profiles has demonstrated substantial potential in precision clinical diagnosis and cancer progression studies. However, most existing studies focus on individual gene sequences and slide level classification tasks, with limited attention to spatial transcriptomics and patch level applications. To address this limitation, we propose a multimodal network, BioMorphNet, which automatically integrates tissue morphological features and spatial gene expression to support tissue classification and differential gene analysis. For considering morphological features, BioMorphNet constructs a graph to model the relationships between target patches and their neighbors, and adjusts the response strength based on morphological and molecular level similarity, to better characterize the tumor microenvironment. In terms of multimodal interactions, BioMorphNet derives clinical pathway features from spatial transcriptomic data based on a predefined pathway database, serving as a bridge between tissue morphology and gene expression. In addition, a novel learnable pathway module is designed to automatically simulate the biological pathway formation process, providing a complementary representation to existing clinical pathways. Compared with the latest morphology gene multimodal methods, BioMorphNet's average classification metrics improve by 2.67%, 5.48%, and 6.29% for prostate cancer, colorectal cancer, and breast cancer datasets, respectively. BioMorphNet not only classifies tissue categories within WSIs accurately to support tumor localization, but also analyzes differential gene expression between tissue categories based on prediction confidence, contributing to the discovery of potential tumor biomarkers.", "date": "2026-01-13", "categories": []}
{"id": "2512.00521", "title": "Rep3Net: An Approach Exploiting Multimodal Representation for Molecular Bioactivity Prediction", "abstract": "Accurate prediction of compound potency accelerates early-stage drug discovery by prioritizing candidates for experimental testing. However, many Quantitative Structure-Activity Relationship (QSAR) approaches for this prediction are constrained by their choice of molecular representation: handcrafted descriptors capture global properties but miss local topology, graph neural networks encode structure but often lack broader chemical context, and SMILES-based language models provide contextual patterns learned from large corpora but are seldom combined with structural features. To exploit these complementary signals, we introduce Rep3Net, a unified multimodal architecture that fuses RDKit molecular descriptors, graph-derived features from a residual graph-convolutional backbone, and ChemBERTa SMILES embeddings. We evaluate Rep3Net on a curated ChEMBL subset for Human PARP1 using fivefold cross validation. Rep3Net attains an MSE of $0.83\\pm0.06$, RMSE of $0.91\\pm0.03$, $R^{2}=0.43\\pm0.01$, and yields Pearson and Spearman correlations of $0.66\\pm0.01$ and $0.67\\pm0.01$, respectively, substantially improving over several strong GNN baselines. In addition, Rep3Net achieves a favorable latency-to-parameter trade-off thanks to a single-layer GCN backbone and parallel frozen encoders. Ablations show that graph topology, ChemBERTa semantics, and handcrafted descriptors each contribute complementary information, with full fusion providing the largest error reduction. These results demonstrate that multimodal representation fusion can improve potency prediction for PARP1 and provide a scalable framework for virtual screening in early-stage drug discovery.", "date": "2025-11-29", "categories": []}
{"id": "2601.01089", "title": "Central Dogma Transformer: Towards Mechanism-Oriented AI for Cellular Understanding", "abstract": "Understanding cellular mechanisms requires integrating information across DNA, RNA, and protein - the three molecular systems linked by the Central Dogma of molecular biology. While domain-specific foundation models have achieved success for each modality individually, they remain isolated, limiting our ability to model integrated cellular processes. Here we present the Central Dogma Transformer (CDT), an architecture that integrates pre-trained language models for DNA, RNA, and protein following the directional logic of the Central Dogma. CDT employs directional cross-attention mechanisms - DNA-to-RNA attention models transcriptional regulation, while RNA-to-Protein attention models translational relationships - producing a unified Virtual Cell Embedding that integrates all three modalities. We validate CDT v1 - a proof-of-concept implementation using fixed (non-cell-specific) RNA and protein embeddings - on CRISPRi enhancer perturbation data from K562 cells, achieving a Pearson correlation of 0.503, representing 63% of the theoretical ceiling set by cross-experiment variability (r = 0.797). Attention and gradient analyses provide complementary interpretive windows: in detailed case studies, these approaches highlight largely distinct genomic regions, with gradient analysis identifying a CTCF binding site that Hi-C data showed as physically contacting both enhancer and target gene. These results suggest that AI architectures aligned with biological information flow can achieve both predictive accuracy and mechanistic interpretability.", "date": "2026-01-03", "categories": []}
{"id": "2601.06214", "title": "Dynamics-inspired Structure Hallucination for Protein-protein Interaction Modeling", "abstract": "Protein-protein interaction (PPI) represents a central challenge within the biology field, and accurately predicting the consequences of mutations in this context is crucial for drug design and protein engineering. Deep learning (DL) has shown promise in forecasting the effects of such mutations, but is hindered by two primary constraints. First, the structures of mutant proteins are often elusive to acquire. Secondly, PPI takes place dynamically, which is rarely integrated into the DL architecture design. To address these obstacles, we present a novel framework named Refine-PPI with two key enhancements. First, we introduce a structure refinement module trained by a mask mutation modeling (MMM) task on available wild-type structures, which is then transferred to produce the inaccessible mutant structures. Second, we employ a new kind of geometric network, called the probability density cloud network (PDC-Net), to capture 3D dynamic variations and encode the atomic uncertainty associated with PPI. Comprehensive experiments on SKEMPI.v2 substantiate the superiority of Refine-PPI over all existing tools for predicting free energy change. These findings underscore the effectiveness of our hallucination strategy and the PDC module in addressing the absence of mutant protein structure and modeling geometric uncertainty.", "date": "2026-01-08", "categories": []}
{"id": "2601.06381", "title": "Hierarchical Pooling and Explainability in Graph Neural Networks for Tumor and Tissue-of-Origin Classification Using RNA-seq Data", "abstract": "This study explores the use of graph neural networks (GNNs) with hierarchical pooling and multiple convolution layers for cancer classification based on RNA-seq data. We combine gene expression data from The Cancer Genome Atlas (TCGA) with a precomputed STRING protein-protein interaction network to classify tissue origin and distinguish between normal and tumor samples. The model employs Chebyshev graph convolutions (K=2) and weighted pooling layers, aggregating gene clusters into 'supernodes' across multiple coarsening levels. This approach enables dimensionality reduction while preserving meaningful interactions. Saliency methods were applied to interpret the model by identifying key genes and biological processes relevant to cancer. Our findings reveal that increasing the number of convolution and pooling layers did not enhance classification performance. The highest F1-macro score (0.978) was achieved with a single pooling layer. However, adding more layers resulted in over-smoothing and performance degradation. However, the model proved highly interpretable through gradient methods, identifying known cancer-related genes and highlighting enriched biological processes, and its hierarchical structure can be used to develop new explainable architectures. Overall, while deeper GNN architectures did not improve performance, the hierarchical pooling structure provided valuable insights into tumor biology, making GNNs a promising tool for cancer biomarker discovery and interpretation", "date": "2026-01-10", "categories": []}
{"id": "2601.07261", "title": "Pseudodata-guided Invariant Representation Learning Boosts the Out-of-Distribution Generalization in Enzymatic Kinetic Parameter Prediction", "abstract": "Accurate prediction of enzyme kinetic parameters is essential for understanding catalytic mechanisms and guiding enzyme engineering.However, existing deep learning-based enzyme-substrate interaction (ESI) predictors often exhibit performance degradation on sequence-divergent, out-of-distribution (OOD) cases, limiting robustness under biologically relevant perturbations.We propose O$^2$DENet, a lightweight, plug-and-play module that enhances OOD generalization via biologically and chemically informed perturbation augmentation and invariant representation learning.O$^2$DENet introduces enzyme-substrate perturbations and enforces consistency between original and augmented enzyme-substrate-pair representations to encourage invariance to distributional shifts.When integrated with representative ESI models, O$^2$DENet consistently improves predictive performance for both $k_{cat}$ and $K_m$ across stringent sequence-identity-based OOD benchmarks, achieving state-of-the-art results among the evaluated methods in terms of accuracy and robustness metrics.Overall, O$^2$DENet provides a general and effective strategy to enhance the stability and deployability of data-driven enzyme kinetics predictors for real-world enzyme engineering applications.", "date": "2026-01-12", "categories": []}
{"id": "2601.07415", "title": "PLANET v2.0: A comprehensive Protein-Ligand Affinity Prediction Model Based on Mixture Density Network", "abstract": "Drug discovery represents a time-consuming and financially intensive process, and virtual screening can accelerate it. Scoring functions, as one of the tools guiding virtual screening, have their precision closely tied to screening efficiency. In our previous study, we developed a graph neural network model called PLANET (Protein-Ligand Affinity prediction NETwork), but it suffers from the defect in representing protein-ligand contact maps. Incorrect binding modes inevitably lead to poor affinity predictions, so accurate prediction of the protein-ligand contact map is desired to improve PLANET. In this study, we have proposed PLANET v2.0 as an upgraded version. The model is trained via multi-objective training strategy and incorporates the Mixture Density Network to predict binding modes. Except for the probability density distributions of non-covalent interactions, we innovatively employ another Gaussian mixture model to describe the relationship between distance and energy of each interaction pair and predict protein-ligand affinity like calculating the mathematical expectation. As on the CASF-2016 benchmark, PLANET v2.0 demonstrates excellent scoring power, ranking power, and docking power. The screening power of PLANET v2.0 gets notably improved compared to PLANET and Glide SP and it demonstrates robust validation on a commercial ultra-large-scale dataset. Given its efficiency and accuracy, PLANET v2.0 can hopefully become one of the practical tools for virtual screening workflows. PLANET v2.0 is freely available at https://www.pdbbind-plus.org.cn/planetv2.", "date": "2026-01-12", "categories": []}
{"id": "2601.05330", "title": "Improving Enzyme Prediction with Chemical Reaction Equations by Hypergraph-Enhanced Knowledge Graph Embeddings", "abstract": "Predicting enzyme-substrate interactions has long been a fundamental problem in biochemistry and metabolic engineering. While existing methods could leverage databases of expert-curated enzyme-substrate pairs for models to learn from known pair interactions, the databases are often sparse, i.e., there are only limited and incomplete examples of such pairs, and also labor-intensive to maintain. This lack of sufficient training data significantly hinders the ability of traditional enzyme prediction models to generalize to unseen interactions. In this work, we try to exploit chemical reaction equations from domain-specific databases, given their easier accessibility and denser, more abundant data. However, interactions of multiple compounds, e.g., educts and products, with the same enzymes create complex relational data patterns that traditional models cannot easily capture. To tackle that, we represent chemical reaction equations as triples of (educt, enzyme, product) within a knowledge graph, such that we can take advantage of knowledge graph embedding (KGE) to infer missing enzyme-substrate pairs for graph completion. Particularly, in order to capture intricate relationships among compounds, we propose our knowledge-enhanced hypergraph model for enzyme prediction, i.e., Hyper-Enz, which integrates a hypergraph transformer with a KGE model to learn representations of the hyper-edges that involve multiple educts and products. Also, a multi-expert paradigm is introduced to guide the learning of enzyme-substrate interactions with both the proposed model and chemical reaction equations. Experimental results show a significant improvement, with up to a 88% relative improvement in average enzyme retrieval accuracy and 30% improvement in pair-level prediction compared to traditional models, demonstrating the effectiveness of our approach.", "date": "2026-01-08", "categories": []}
{"id": "2601.05531", "title": "DNATokenizer: A GPU-First Byte-to-Identifier Tokenizer for High-Throughput DNA Language Models", "abstract": "Tokenization sits at the boundary between high-throughput genomic input and GPU compute, posing challenges in both algorithm design and system throughput. Overlapping k-mer tokenization can introduce information leakage under masked language modeling (MLM) and may degrade downstream accuracy. Single-nucleotide tokenization avoids leakage and preserves per-base fidelity, but it greatly increases sequence length for attention-based architectures. Non-overlapping k-mers and byte-pair encoding (BPE) provide compression and avoid leakage, at the cost of boundary sensitivity or reduced interpretability. Empirically, the choice of tokenization interacts strongly with model architecture and task requirements. At the system level, however, standard string tokenizers and host-bound vocabulary lookups dominate wall-clock time once inputs reach billions of bases, regardless of the tokenization algorithm. We present DNATok, a high-performance, GPU-first tokenization system that replaces general-purpose string processing with byte lookup table (LUT)-based identifier streaming and an overlapped host-to-device (H2D)/compute pipeline using pinned memory and architectural parallelism. DNATok is vocabulary-agnostic: it accelerates single-nucleotide, non-overlapping k-mer, and BPE tokenization, and integrates as a drop-in systems layer beneath genomic foundation models. DNATok achieves 84-95x higher encoding throughput than optimized Hugging Face baselines and up to 1.9x higher H2D throughput. End-to-end streaming reaches 1.27-1.84e8 tokens/s depending on configuration, effectively removing tokenization as a bottleneck for production-scale training and inference.", "date": "2026-01-09", "categories": []}
{"id": "2601.05792", "title": "Tensor-DTI: Enhancing Biomolecular Interaction Prediction with Contrastive Embedding Learning", "abstract": "Accurate drug-target interaction (DTI) prediction is essential for computational drug discovery, yet existing models often rely on single-modality predefined molecular descriptors or sequence-based embeddings with limited representativeness. We propose Tensor-DTI, a contrastive learning framework that integrates multimodal embeddings from molecular graphs, protein language models, and binding-site predictions to improve interaction modeling. Tensor-DTI employs a siamese dual-encoder architecture, enabling it to capture both chemical and structural interaction features while distinguishing interacting from non-interacting pairs. Evaluations on multiple DTI benchmarks demonstrate that Tensor-DTI outperforms existing sequence-based and graph-based models. We also conduct large-scale inference experiments on CDK2 across billion-scale chemical libraries, where Tensor-DTI produces chemically plausible hit distributions even when CDK2 is withheld from training. In enrichment studies against Glide docking and Boltz-2 co-folder, Tensor-DTI remains competitive on CDK2 and improves the screening budget required to recover moderate fractions of high-affinity ligands on out-of-family targets under strict family-holdout splits. Additionally, we explore its applicability to protein-RNA and peptide-protein interactions. Our findings highlight the benefits of integrating multimodal information with contrastive objectives to enhance interaction-prediction accuracy and to provide more interpretable and reliability-aware models for virtual screening.", "date": "2026-01-09", "categories": []}
{"id": "2601.04506", "title": "Surface-based Molecular Design with Multi-modal Flow Matching", "abstract": "Therapeutic peptides show promise in targeting previously undruggable binding sites, with recent advancements in deep generative models enabling full-atom peptide co-design for specific protein receptors. However, the critical role of molecular surfaces in protein-protein interactions (PPIs) has been underexplored. To bridge this gap, we propose an omni-design peptides generation paradigm, called SurfFlow, a novel surface-based generative algorithm that enables comprehensive co-design of sequence, structure, and surface for peptides. SurfFlow employs a multi-modality conditional flow matching (CFM) architecture to learn distributions of surface geometries and biochemical properties, enhancing peptide binding accuracy. Evaluated on the comprehensive PepMerge benchmark, SurfFlow consistently outperforms full-atom baselines across all metrics. These results highlight the advantages of considering molecular surfaces in de novo peptide discovery and demonstrate the potential of integrating multiple protein modalities for more effective therapeutic peptide discovery.", "date": "2026-01-08", "categories": []}
{"id": "2511.09900", "title": "Boosting In-Silicon Directed Evolution with Fine-Tuned Protein Language Model and Tree Search", "abstract": "Protein evolution through amino acid mutations is a cornerstone of life sciences. Recent advances in protein language models have shown rich evolutionary patterns, offering unprecedented potential for in-silicon directed evolution. However, existing directed evolution methods largely rely on heuristic evolution strategies and have yet to efficiently integrate the transformative protein language models with advanced optimization techniques, such as reinforcement learning, to adaptively learn superior evolution policies. To bridge this gap, we propose AlphaDE, a novel framework that evolves protein sequences by harnessing the innovative paradigms of large language models, such as fine-tuning and test-time inference. First, AlphaDE fine-tunes pretrained protein language models using masked language modeling on homologous protein sequences to activate the evolutionary plausibility of the interested protein family. Second, AlphaDE introduces test-time inference based on Monte Carlo tree search, which effectively evolves proteins with evolutionary guidance from the fine-tuned protein language model. Extensive benchmark experiments show that AlphaDE remarkably outperforms previous state-of-the-art methods even with few-shot fine-tuning. A case study further demonstrates that AlphaDE supports condensing the protein sequence space of avGFP through computational evolution.", "date": "2025-11-13", "categories": []}
{"id": "2601.03295", "title": "MetagenBERT: a Transformer-based Architecture using Foundational genomic Large Language Models for novel Metagenome Representation", "abstract": "Metagenomic disease prediction commonly relies on species abundance tables derived from large, incomplete reference catalogs, constraining resolution and discarding valuable information contained in DNA reads. To overcome these limitations, we introduce MetagenBERT, a Transformer based framework that produces end to end metagenome embeddings directly from raw DNA sequences, without taxonomic or functional annotations. Reads are embedded using foundational genomic language models (DNABERT2 and the microbiome specialized DNABERTMS), then aggregated through a scalable clustering strategy based on FAISS accelerated KMeans. Each metagenome is represented as a cluster abundance vector summarizing the distribution of its embedded reads. We evaluate this approach on five benchmark gut microbiome datasets (Cirrhosis, T2D, Obesity, IBD, CRC). MetagenBERT achieves competitive or superior AUC performance relative to species abundance baselines across most tasks. Concatenating both representations further improves prediction, demonstrating complementarity between taxonomic and embedding derived signals. Clustering remains robust when applied to as little as 10% of reads, highlighting substantial redundancy in metagenomes and enabling major computational gains. We additionally introduce MetagenBERT Glob Mcardis, a cross cohort variant trained on the large, phenotypically diverse MetaCardis cohort and transferred to other datasets, retaining predictive signal including for unseen phenotypes, indicating the feasibility of a foundation model for metagenome representation. Robustness analyses (PERMANOVA, PERMDISP, entropy) show consistent separation of different states across subsamples. Overall, MetagenBERT provides a scalable, annotation free representation of metagenomes pointing toward future phenotype aware generalization across heterogeneous cohorts and sequencing technologies.", "date": "2026-01-05", "categories": []}
{"id": "2310.08061", "title": "ETDock: A Novel Equivariant Transformer for Protein-Ligand Docking", "abstract": "Predicting the docking between proteins and ligands is a crucial and challenging task for drug discovery. However, traditional docking methods mainly rely on scoring functions, and deep learning-based docking approaches usually neglect the 3D spatial information of proteins and ligands, as well as the graph-level features of ligands, which limits their performance. To address these limitations, we propose an equivariant transformer neural network for protein-ligand docking pose prediction. Our approach involves the fusion of ligand graph-level features by feature processing, followed by the learning of ligand and protein representations using our proposed TAMformer module. Additionally, we employ an iterative optimization approach based on the predicted distance matrix to generate refined ligand poses. The experimental results on real datasets show that our model can achieve state-of-the-art performance.", "date": "2023-10-12", "categories": []}
{"id": "2601.01347", "title": "From Classification to Generation: An Open-Ended Paradigm for Adverse Drug Reaction Prediction Based on Graph-Motif Feature Fusion", "abstract": "Computational biology offers immense potential for reducing the high costs and protracted cycles of new drug development through adverse drug reaction (ADR) prediction. However, current methods remain impeded by drug data scarcity-induced cold-start challenge, closed label sets, and inadequate modeling of label dependencies. Here we propose an open-ended ADR prediction paradigm based on Graph-Motif feature fusion and Multi-Label Generation (GM-MLG). Leveraging molecular structure as an intrinsic and inherent feature, GM-MLG constructs a dual-graph representation architecture spanning the atomic level, the local molecular level (utilizing fine-grained motifs dynamically extracted via the BRICS algorithm combined with additional fragmentation rules), and the global molecular level. Uniquely, GM-MLG pioneers transforming ADR prediction from multi-label classification into Transformer Decoder-based multi-label generation. By treating ADR labels as discrete token sequences, it employs positional embeddings to explicitly capture dependencies and co-occurrence relationships within large-scale label spaces, generating predictions via autoregressive decoding to dynamically expand the prediction space. Experiments demonstrate GM-MLG achieves up to 38% improvement and an average gain of 20%, expanding the prediction space from 200 to over 10,000 types. Furthermore, it elucidates non-linear structure-activity relationships between ADRs and motifs via retrosynthetic motif analysis, providing interpretable and innovative support for systematic risk reduction in drug safety.", "date": "2026-01-04", "categories": []}
{"id": "2601.01358", "title": "A New Framework for Explainable Rare Cell Identification in Single-Cell Transcriptomics Data", "abstract": "The detection of rare cell types in single-cell transcriptomics data is crucial for elucidating disease pathogenesis and tissue development dynamics. However, a critical gap that persists in current methods is their inability to provide an explanation based on genes for each cell they have detected as rare. We identify three primary sources of this deficiency. First, the anomaly detectors often function as \"black boxes\", designed to detect anomalies but unable to explain why a cell is anomalous. Second, the standard analytical framework hinders interpretability by relying on dimensionality reduction techniques, such as Principal Component Analysis (PCA), which transform meaningful gene expression data into abstract, uninterpretable features. Finally, existing explanation algorithms cannot be readily applied to this domain, as single-cell data is characterized by high dimensionality, noise, and substantial sparsity. To overcome these limitations, we introduce a framework for explainable anomaly detection in single-cell transcriptomics data which not only identifies individual anomalies, but also provides a visual explanation based on genes that makes an instance anomalous. This framework has two key ingredients that are not existed in current methods applied in this domain. First, it eliminates the PCA step which is deemed to be an essential component in previous studies. Second, it employs the state-of-art anomaly detector and explainer as the efficient and effective means to find each rare cell and the relevant gene subspace in order to provide explanations for each rare cell as well as the typical normal cell associated with the rare cell's closest normal cells.", "date": "2026-01-04", "categories": []}
{"id": "2601.00143", "title": "MethConvTransformer: A Deep Learning Framework for Cross-Tissue Alzheimer's Disease Detection", "abstract": "Alzheimer's disease (AD) is a multifactorial neurodegenerative disorder characterized by progressive cognitive decline and widespread epigenetic dysregulation in the brain. DNA methylation, as a stable yet dynamic epigenetic modification, holds promise as a noninvasive biomarker for early AD detection. However, methylation signatures vary substantially across tissues and studies, limiting reproducibility and translational utility. To address these challenges, we develop MethConvTransformer, a transformer-based deep learning framework that integrates DNA methylation profiles from both brain and peripheral tissues to enable biomarker discovery. The model couples a CpG-wise linear projection with convolutional and self-attention layers to capture local and long-range dependencies among CpG sites, while incorporating subject-level covariates and tissue embeddings to disentangle shared and region-specific methylation effects. In experiments across six GEO datasets and an independent ADNI validation cohort, our model consistently outperforms conventional machine-learning baselines, achieving superior discrimination and generalization. Moreover, interpretability analyses using linear projection, SHAP, and Grad-CAM++ reveal biologically meaningful methylation patterns aligned with AD-associated pathways, including immune receptor signaling, glycosylation, lipid metabolism, and endomembrane (ER/Golgi) organization. Together, these results indicate that MethConvTransformer delivers robust, cross-tissue epigenetic biomarkers for AD while providing multi-resolution interpretability, thereby advancing reproducible methylation-based diagnostics and offering testable hypotheses on disease mechanisms.", "date": "2026-01-01", "categories": []}
{"id": "2601.00277", "title": "Benchmarking Preprocessing and Integration Methods in Single-Cell Genomics", "abstract": "Single-cell data analysis has the potential to revolutionize personalized medicine by characterizing disease-associated molecular changes at the single-cell level. Advanced single-cell multimodal assays can now simultaneously measure various molecules (e.g., DNA, RNA, Protein) across hundreds of thousands of individual cells, providing a comprehensive molecular readout. A significant analytical challenge is integrating single-cell measurements across different modalities. Various methods have been developed to address this challenge, but there has been no systematic evaluation of these techniques with different preprocessing strategies. This study examines a general pipeline for single-cell data analysis, which includes normalization, data integration, and dimensionality reduction. The performance of different algorithm combinations often depends on the dataset sizes and characteristics. We evaluate six datasets across diverse modalities, tissues, and organisms using three metrics: Silhouette Coefficient Score, Adjusted Rand Index, and Calinski-Harabasz Index. Our experiments involve combinations of seven normalization methods, four dimensional reduction methods, and five integration methods. The results show that Seurat and Harmony excel in data integration, with Harmony being more time-efficient, especially for large datasets. UMAP is the most compatible dimensionality reduction method with the integration techniques, and the choice of normalization method varies depending on the integration method used.", "date": "2026-01-01", "categories": []}
{"id": "2601.00647", "title": "Physio-DPO: Aligning Large Language Models with the Protein Energy Landscape to Eliminate Structural Hallucinations", "abstract": "Large Protein Language Models have shown strong potential for generative protein design, yet they frequently produce structural hallucinations, generating sequences with high linguistic likelihood that fold into thermodynamically unstable conformations. Existing alignment approaches such as Direct Preference Optimization are limited in this setting, as they model preferences as binary labels and ignore the continuous structure of the physical energy landscape. We propose Physio-DPO, a physics informed alignment framework that grounds protein language models in thermodynamic stability. Physio-DPO introduces a magnitude aware objective that scales optimization updates according to the energy gap between native structures and physics perturbed hard negatives. Experiments show that Physio-DPO consistently outperforms strong baselines including SFT, PPO, and standard DPO, reducing self consistency RMSD to 1.28 \\AA\\ and increasing foldability to 92.8%. Qualitative analysis further demonstrates that Physio-DPO effectively mitigates structural hallucinations by recovering biophysical interactions such as hydrophobic core packing and hydrogen bond networks.", "date": "2026-01-02", "categories": []}
{"id": "2408.00200", "title": "UnPaSt: unsupervised patient stratification by biclustering of omics data", "abstract": "Unsupervised patient stratification is essential for disease subtype discovery, yet, despite growing evidence of molecular heterogeneity of non-oncological diseases, popular methods are benchmarked primarily using cancers with mutually exclusive molecular subtypes well-differentiated by numerous biomarkers. Evaluating 22 unsupervised methods, including clustering and biclustering, using simulated and real transcriptomics data revealed their inefficiency in scenarios with non-mutually exclusive subtypes or subtypes discriminated only by few biomarkers. To address these limitations and advance precision medicine, we developed UnPaSt, a novel biclustering algorithm for unsupervised patient stratification based on differentially expressed biclusters. UnPaSt outperformed widely used patient stratification approaches in the de novo identification of known subtypes of breast cancer and asthma. In addition, it detected many biologically insightful patterns across bulk transcriptomics, proteomics, single-cell, spatial transcriptomics, and multi-omics datasets, enabling a more nuanced and interpretable view of high-throughput data heterogeneity than traditionally used methods.", "date": "2024-07-31", "categories": []}
{"id": "2512.24810", "title": "DTI-GP: Bayesian operations for drug-target interactions using deep kernel Gaussian processes", "abstract": "Precise probabilistic information about drug-target interaction (DTI) predictions is vital for understanding limitations and boosting predictive performance. Gaussian processes (GP) offer a scalable framework to integrate state-of-the-art DTI representations and Bayesian inference, enabling novel operations, such as Bayesian classification with rejection, top-$K$ selection, and ranking. We propose a deep kernel learning-based GP architecture (DTI-GP), which incorporates a combined neural embedding module for chemical compounds and protein targets, and a GP module. The workflow continues with sampling from the predictive distribution to estimate a Bayesian precedence matrix, which is used in fast and accurate selection and ranking operations. DTI-GP outperforms state-of-the-art solutions, and it allows (1) the construction of a Bayesian accuracy-confidence enrichment score, (2) rejection schemes for improved enrichment, and (3) estimation and search for top-$K$ selections and ranking with high expected utility.", "date": "2025-12-31", "categories": []}
{"id": "2410.13264", "title": "Constraint Decoupled Latent Diffusion for Protein Backmapping", "abstract": "Coarse-grained (CG) molecular dynamics simulations enable efficient exploration of protein conformational ensembles. However, reconstructing atomic details from CG structures (backmapping) remains a challenging problem. Current approaches face an inherent trade-off between maintaining atomistic accuracy and exploring diverse conformations, often necessitating complex constraint handling or extensive refinement steps. To address these challenges, we introduce a novel two-stage framework, named CODLAD (COnstraint Decoupled LAtent Diffusion). This framework first compresses atomic structures into discrete latent representations, explicitly embedding structural constraints, thereby decoupling constraint handling from generation. Subsequently, it performs efficient denoising diffusion in this latent space to produce structurally valid and diverse all-atom conformations. Comprehensive evaluations on diverse protein datasets demonstrate that CODLAD achieves state-of-the-art performance in atomistic accuracy, conformational diversity, and computational efficiency while exhibiting strong generalization across different protein systems. Code is available at https://github.com/xiaoxiaokuye/CODLAD.", "date": "2024-10-17", "categories": []}
{"id": "2512.23175", "title": "HELM-BERT: A Transformer for Medium-sized Peptide Property Prediction", "abstract": "Therapeutic peptides have emerged as a pivotal modality in modern drug discovery, occupying a chemically and topologically rich space. While accurate prediction of their physicochemical properties is essential for accelerating peptide development, existing molecular language models rely on representations that fail to capture this complexity. Atom-level SMILES notation generates long token sequences and obscures cyclic topology, whereas amino-acid-level representations cannot encode the diverse chemical modifications central to modern peptide design. To bridge this representational gap, the Hierarchical Editing Language for Macromolecules (HELM) offers a unified framework enabling precise description of both monomer composition and connectivity, making it a promising foundation for peptide language modeling. Here, we propose HELM-BERT, the first encoder-based peptide language model trained on HELM notation. Based on DeBERTa, HELM-BERT is specifically designed to capture hierarchical dependencies within HELM sequences. The model is pre-trained on a curated corpus of 39,079 chemically diverse peptides spanning linear and cyclic structures. HELM-BERT significantly outperforms state-of-the-art SMILES-based language models in downstream tasks, including cyclic peptide membrane permeability prediction and peptide-protein interaction prediction. These results demonstrate that HELM's explicit monomer- and topology-aware representations offer substantial data-efficiency advantages for modeling therapeutic peptides, bridging a long-standing gap between small-molecule and protein language models.", "date": "2025-12-29", "categories": []}
{"id": "2505.01433", "title": "Enhancing TCR-Peptide Interaction Prediction with Pretrained Language Models and Molecular Representations", "abstract": "Understanding the binding specificity between T-cell receptors (TCRs) and peptide-major histocompatibility complexes (pMHCs) is central to immunotherapy and vaccine development. However, current predictive models struggle with generalization, especially in data-scarce settings and when faced with novel epitopes. We present LANTERN (Large lAnguage model-powered TCR-Enhanced Recognition Network), a deep learning framework that combines large-scale protein language models with chemical representations of peptides. By encoding TCR \\b{eta}-chain sequences using ESM-1b and transforming peptide sequences into SMILES strings processed by MolFormer, LANTERN captures rich biological and chemical features critical for TCR-peptide recognition. Through extensive benchmarking against existing models such as ChemBERTa, TITAN, and NetTCR, LANTERN demonstrates superior performance, particularly in zero-shot and few-shot learning scenarios. Our model also benefits from a robust negative sampling strategy and shows significant clustering improvements via embedding analysis. These results highlight the potential of LANTERN to advance TCR-pMHC binding prediction and support the development of personalized immunotherapies.", "date": "2025-04-22", "categories": []}
{"id": "2505.04891", "title": "Clustering with Communication: A Variational Framework for Single Cell Representation Learning", "abstract": "Single-cell RNA sequencing (scRNA-seq) has revealed complex cellular heterogeneity, but recent studies emphasize that understanding biological function also requires modeling cell-cell communication (CCC), the signaling interactions mediated by ligand-receptor pairs that coordinate cellular behavior. Tools like CellChat have demonstrated that CCC plays a critical role in processes such as cell differentiation, tissue regeneration, and immune response, and that transcriptomic data inherently encodes rich information about intercellular signaling. We propose CCCVAE, a novel variational autoencoder framework that incorporates CCC signals into single-cell representation learning. By leveraging a communication-aware kernel derived from ligand-receptor interactions and a sparse Gaussian process, CCCVAE encodes biologically informed priors into the latent space. Unlike conventional VAEs that treat each cell independently, CCCVAE encourages latent embeddings to reflect both transcriptional similarity and intercellular signaling context. Empirical results across four scRNA-seq datasets show that CCCVAE improves clustering performance, achieving higher evaluation scores than standard VAE baselines. This work demonstrates the value of embedding biological priors into deep generative models for unsupervised single-cell analysis.", "date": "2025-05-08", "categories": []}
{"id": "2510.23620", "title": "Genotype-Phenotype Integration through Machine Learning and Personalized Gene Regulatory Networks for Cancer Metastasis Prediction", "abstract": "Metastasis is the leading cause of cancer-related mortality, yet most predictive models rely on shallow architectures and neglect patient-specific regulatory mechanisms. Here, we integrate classical machine learning and deep learning to predict metastatic potential across multiple cancer types. Gene expression profiles from the Cancer Cell Line Encyclopedia were combined with a transcription factor-target prior from DoRothEA, focusing on nine metastasis-associated regulators. After selecting differential genes using the Kruskal-Wallis test, ElasticNet, Random Forest, and XGBoost models were trained for benchmarking. Personalized gene regulatory networks were then constructed using PANDA and LIONESS and analyzed through a graph attention neural network (GATv2) to learn topological and expression-based representations. While XGBoost achieved the highest AUROC (0.7051), the GNN captured non-linear regulatory dependencies at the patient level. These results demonstrate that combining traditional machine learning with graph-based deep learning enables a scalable and interpretable framework for metastasis risk prediction in precision oncology.", "date": "2025-10-22", "categories": []}
{"id": "2511.00977", "title": "Modeling Microenvironment Trajectories on Spatial Transcriptomics with NicheFlow", "abstract": "Understanding the evolution of cellular microenvironments in spatiotemporal data is essential for deciphering tissue development and disease progression. While experimental techniques like spatial transcriptomics now enable high-resolution mapping of tissue organization across space and time, current methods that model cellular evolution operate at the single-cell level, overlooking the coordinated development of cellular states in a tissue. We introduce NicheFlow, a flow-based generative model that infers the temporal trajectory of cellular microenvironments across sequential spatial slides. By representing local cell neighborhoods as point clouds, NicheFlow jointly models the evolution of cell states and spatial coordinates using optimal transport and Variational Flow Matching. Our approach successfully recovers both global spatial architecture and local microenvironment composition across diverse spatiotemporal datasets, from embryonic to brain development.", "date": "2025-11-02", "categories": []}
{"id": "2512.21544", "title": "AVP-Fusion: Adaptive Multi-Modal Fusion and Contrastive Learning for Two-Stage Antiviral Peptide Identification", "abstract": "Accurate identification of antiviral peptides (AVPs) is critical for accelerating novel drug development. However, current computational methods struggle to capture intricate sequence dependencies and effectively handle ambiguous, hard-to-classify samples. To address these challenges, we propose AVP-Fusion, a novel two-stage deep learning framework integrating adaptive feature fusion and contrastive learning. Unlike traditional static feature concatenation, we construct a panoramic feature space using 10 distinct descriptors and introduce an Adaptive Gating Mechanism.This mechanism dynamically regulates the weights of local motifs extracted by CNNs and global dependencies captured by BiLSTMs based on sequence context. Furthermore, to address data distribution challenges, we employ a contrastive learning strategy driven by Online Hard Example Mining (OHEM) and BLOSUM62-based data augmentation, which significantly sharpens the model's decision boundaries. Experimental results on the benchmark Set 1 dataset demonstrate that AVP-Fusion achieves an accuracy of 0.9531 and an MCC of 0.9064, significantly outperforming state-of-the-art methods. In the second stage, leveraging transfer learning, the model enables precise subclass prediction for six viral families and eight specific viruses, even under limited sample sizes. In summary, AVP-Fusion serves as a robust and interpretable tool for high-throughput antiviral drug screening.", "date": "2025-12-25", "categories": []}
{"id": "2512.22007", "title": "DuaDeep-SeqAffinity: Dual-Stream Deep Learning Framework for Sequence-Only Antigen-Antibody Affinity Prediction", "abstract": "Predicting the binding affinity between antigens and antibodies is fundamental to drug discovery and vaccine development. Traditional computational approaches often rely on experimentally determined 3D structures, which are scarce and computationally expensive to obtain. This paper introduces DuaDeep-SeqAffinity, a novel sequence-only deep learning framework that predicts affinity scores solely from their amino acid sequences using a dual-stream hybrid architecture. Our approach leverages pre-trained ESM-2 protein language model embeddings, combining 1D Convolutional Neural Networks (CNNs) for local motif detection with Transformer encoders for global contextual representation. A subsequent fusion module integrates these multi-faceted features, which are then passed to a fully connected network for final score regression. Experimental results demonstrate that DuaDeep-SeqAffinity significantly outperforms individual architectural components and existing state-of-the-art (SOTA) methods. DuaDeep achieved a superior Pearson correlation of 0.688, an R^2 of 0.460, and a Root Mean Square Error (RMSE) of 0.737, surpassing single-branch variants ESM-CNN and ESM-Transformer. Notably, the model achieved an Area Under the Curve (AUC) of 0.890, outperforming sequence-only benchmarks and even surpassing structure-sequence hybrid models. These findings prove that high-fidelity sequence embeddings can capture essential binding patterns typically reserved for structural modeling. By eliminating the reliance on 3D structures, DuaDeep-SeqAffinity provides a highly scalable and efficient solution for high-throughput screening of vast sequence libraries, significantly accelerating the therapeutic discovery pipeline.", "date": "2025-12-26", "categories": []}
{"id": "2412.07201", "title": "A Review on the Applications of Transformer-based language models for Nucleotide Sequence Analysis", "abstract": "In recent times, Transformer-based language models are making quite an impact in the field of natural language processing. As relevant parallels can be drawn between biological sequences and natural languages, the models used in NLP can be easily extended and adapted for various applications in bioinformatics. In this regard, this paper introduces the major developments of Transformer-based models in the recent past in the context of nucleotide sequences. We have reviewed and analysed a large number of application-based papers on this subject, giving evidence of the main characterizing features and to different approaches that may be adopted to customize such powerful computational machines. We have also provided a structured description of the functioning of Transformers, that may enable even first time users to grab the essence of such complex architectures. We believe this review will help the scientific community in understanding the various applications of Transformer-based language models to nucleotide sequences. This work will motivate the readers to build on these methodologies to tackle also various other problems in the field of bioinformatics.", "date": "2024-12-10", "categories": []}
{"id": "2512.20135", "title": "MolAct: An Agentic RL Framework for Molecular Editing and Property Optimization", "abstract": "Molecular editing and optimization are multi-step problems that require iteratively improving properties while keeping molecules chemically valid and structurally similar. We frame both tasks as sequential, tool-guided decisions and introduce MolAct, an agentic reinforcement learning framework that employs a two-stage training paradigm: first building editing capability, then optimizing properties while reusing the learned editing behaviors. To the best of our knowledge, this is the first work to formalize molecular design as an Agentic Reinforcement Learning problem, where an LLM agent learns to interleave reasoning, tool-use, and molecular optimization. The framework enables agents to interact in multiple turns, invoking chemical tools for validity checking, property assessment, and similarity control, and leverages their feedback to refine subsequent edits. We instantiate the MolAct framework to train two model families: MolEditAgent for molecular editing tasks and MolOptAgent for molecular optimization tasks. In molecular editing, MolEditAgent-7B delivers 100, 95, and 98 valid add, delete, and substitute edits, outperforming strong closed \"thinking\" baselines such as DeepSeek-R1; MolEditAgent-3B approaches the performance of much larger open \"thinking\" models like Qwen3-32B-think. In molecular optimization, MolOptAgent-7B (trained on MolEditAgent-7B) surpasses the best closed \"thinking\" baseline (e.g., Claude 3.7) on LogP and remains competitive on solubility, while maintaining balanced performance across other objectives. These results highlight that treating molecular design as a multi-step, tool-augmented process is key to reliable and interpretable improvements.", "date": "2025-12-23", "categories": []}
{"id": "2512.20954", "title": "Reflection Pretraining Enables Token-Level Self-Correction in Biological Sequence Models", "abstract": "Chain-of-Thought (CoT) prompting has significantly advanced task-solving capabilities in natural language processing with large language models. Unlike standard prompting, CoT encourages the model to generate intermediate reasoning steps, non-answer tokens, that help guide the model toward more accurate final outputs. These intermediate steps enable more complex reasoning processes such as error correction, memory management, future planning, and self-reflection. However, applying CoT to non-natural language domains, such as protein and RNA language models, is not yet possible, primarily due to the limited expressiveness of their token spaces (e.g., amino acid tokens). In this work, we propose and define the concept of language expressiveness: the ability of a given language, using its tokens and grammar, to encode information. We show that the limited expressiveness of protein language severely restricts the applicability of CoT-style reasoning. To overcome this, we introduce reflection pretraining, for the first time in a biological sequence model, which enables the model to engage in intermediate reasoning through the generation of auxiliary \"thinking tokens\" beyond simple answer tokens. Theoretically, we demonstrate that our augmented token set significantly enhances biological language expressiveness, thereby improving the overall reasoning capacity of the model. Experimentally, our pretraining approach teaches protein models to self-correct and leads to substantial performance gains compared to standard pretraining.", "date": "2025-12-24", "categories": []}
{"id": "2508.12029", "title": "BConformeR: A Conformer Based on Mutual Sampling for Unified Prediction of Continuous and Discontinuous Antibody Binding Sites", "abstract": "Accurate prediction of antibody-binding sites (epitopes) on antigens is crucial for vaccine design, immunodiagnostics, therapeutic antibody development, antibody engineering, research into autoimmune and allergic diseases, and advancing our understanding of immune responses. Despite in silico methods that have been proposed to predict both linear (continuous) and conformational (discontinuous) epitopes, they consistently underperform in predicting conformational epitopes. In this work, we propose Conformer-based models trained separately on AlphaFold-predicted structures and experimentally determined structures, leveraging convolutional neural networks (CNNs) to extract local features and Transformers to capture long-range dependencies within antigen sequences. Ablation studies demonstrate that CNN enhances the prediction of linear epitopes, and the Transformer module improves the prediction of conformational epitopes. Experimental results show that our model outperforms existing baselines in terms of MCC, ROC-AUC, PR-AUC, and F1 scores on both linear and conformational epitopes.", "date": "2025-08-16", "categories": []}
{"id": "2512.17146", "title": "Biosecurity-Aware AI: Agentic Risk Auditing of Soft Prompt Attacks on ESM-Based Variant Predictors", "abstract": "Genomic Foundation Models (GFMs), such as Evolutionary Scale Modeling (ESM), have demonstrated remarkable success in variant effect prediction. However, their security and robustness under adversarial manipulation remain largely unexplored. To address this gap, we introduce the Secure Agentic Genomic Evaluator (SAGE), an agentic framework for auditing the adversarial vulnerabilities of GFMs. SAGE functions through an interpretable and automated risk auditing loop. It injects soft prompt perturbations, monitors model behavior across training checkpoints, computes risk metrics such as AUROC and AUPR, and generates structured reports with large language model-based narrative explanations. This agentic process enables continuous evaluation of embedding-space robustness without modifying the underlying model. Using SAGE, we find that even state-of-the-art GFMs like ESM2 are sensitive to targeted soft prompt attacks, resulting in measurable performance degradation. These findings reveal critical and previously hidden vulnerabilities in genomic foundation models, showing the importance of agentic risk auditing in securing biomedical applications such as clinical variant interpretation.", "date": "2025-12-19", "categories": []}
{"id": "2512.17527", "title": "SafeBench-Seq: A Homology-Clustered, CPU-Only Baseline for Protein Hazard Screening with Physicochemical/Composition Features and Cluster-Aware Confidence Intervals", "abstract": "Foundation models for protein design raise concrete biosecurity risks, yet the community lacks a simple, reproducible baseline for sequence-level hazard screening that is explicitly evaluated under homology control and runs on commodity CPUs. We introduce SafeBench-Seq, a metadata-only, reproducible benchmark and baseline classifier built entirely from public data (SafeProtein hazards and UniProt benigns) and interpretable features (global physicochemical descriptors and amino-acid composition). To approximate \"never-before-seen\" threats, we homology-cluster the combined dataset at <=40% identity and perform cluster-level holdouts (no cluster overlap between train/test). We report discrimination (AUROC/AUPRC) and screening-operating points (TPR@1% FPR; FPR@95% TPR) with 95% bootstrap confidence intervals (n=200), and we provide calibrated probabilities via CalibratedClassifierCV (isotonic for Logistic Regression / Random Forest; Platt sigmoid for Linear SVM). We quantify probability quality using Brier score, Expected Calibration Error (ECE; 15 bins), and reliability diagrams. Shortcut susceptibility is probed via composition-preserving residue shuffles and length-/composition-only ablations. Empirically, random splits substantially overestimate robustness relative to homology-clustered evaluation; calibrated linear models exhibit comparatively good calibration, while tree ensembles retain slightly higher Brier/ECE. SafeBench-Seq is CPU-only, reproducible, and releases metadata only (accessions, cluster IDs, split labels), enabling rigorous evaluation without distributing hazardous sequences.", "date": "2025-12-19", "categories": []}
{"id": "2411.01098", "title": "Artificial Intelligence for Microbiology and Microbiome Research", "abstract": "Advancements in artificial intelligence (AI) have transformed many scientific fields, with microbiology and microbiome research now experiencing significant breakthroughs through machine learning applications. This review provides a comprehensive overview of AI-driven approaches tailored for microbiology and microbiome studies, emphasizing both technical advancements and biological insights. We begin with an introduction to foundational AI techniques, including primary machine learning paradigms and various deep learning architectures, and offer guidance on choosing between traditional machine learning and sophisticated deep learning methods based on specific research goals. The primary section on application scenarios spans diverse research areas, from taxonomic profiling, functional annotation \\& prediction, microbe-X interactions, microbial ecology, metabolic modeling, precision nutrition, clinical microbiology, to prevention \\& therapeutics. Finally, we discuss challenges in this field and highlight some recent breakthroughs. Together, this review underscores AI's transformative role in microbiology and microbiome research, paving the way for innovative methodologies and applications that enhance our understanding of microbial life and its impact on our planet and our health.", "date": "2024-11-02", "categories": []}
{"id": "2504.16941", "title": "Mathematical Insights into Protein Architecture: Persistent Homology and Machine Learning Applied to the Flagellar Motor", "abstract": "We present a machine learning approach that leverages persistent homology to classify bacterial flagellar motors into two functional states: rotated and stalled. By embedding protein structural data into a topological framework, we extract multiscale features from filtered simplicial complexes constructed over atomic coordinates. These topological invariants, specifically persistence diagrams and barcodes, capture critical geometric and connectivity patterns that correlate with motor function. The extracted features are vectorized and integrated into a machine learning pipeline that includes dimensionality reduction and supervised classification. Applied to a curated dataset of experimentally characterized flagellar motors from diverse bacterial species, our model demonstrates high classification accuracy and robustness to structural variation. This approach highlights the power of topological data analysis in revealing functionally relevant patterns beyond the reach of traditional geometric descriptors, offering a novel computational tool for protein function prediction.", "date": "2025-04-08", "categories": []}
{"id": "2512.15900", "title": "Boosting t-SNE Efficiency for Sequencing Data: Insights from Kernel Selection", "abstract": "Dimensionality reduction techniques are essential for visualizing and analyzing high-dimensional biological sequencing data. t-distributed Stochastic Neighbor Embedding (t-SNE) is widely used for this purpose, traditionally employing the Gaussian kernel to compute pairwise similarities. However, the Gaussian kernel's lack of data-dependence and computational overhead limit its scalability and effectiveness for categorical biological sequences. Recent work proposed the isolation kernel as an alternative, yet it may not optimally capture sequence similarities. In this study, we comprehensively evaluate nine different kernel functions for t-SNE applied to molecular sequences, using three embedding methods: One-Hot Encoding, Spike2Vec, and minimizers. Through both subjective visualization and objective metrics (including neighborhood preservation scores), we demonstrate that the cosine similarity kernel in general outperforms other kernels, including Gaussian and isolation kernels, achieving superior runtime efficiency and better preservation of pairwise distances in low-dimensional space. We further validate our findings through extensive classification and clustering experiments across six diverse biological datasets (Spike7k, Host, ShortRead, Rabies, Genome, and Breast Cancer), employing multiple machine learning algorithms and evaluation metrics. Our results show that kernel selection significantly impacts not only visualization quality but also downstream analytical tasks, with the cosine similarity kernel providing the most robust performance across different data types and embedding strategies, making it particularly suitable for large-scale biological sequence analysis.", "date": "2025-12-17", "categories": []}
{"id": "2512.15931", "title": "BarcodeMamba+: Advancing State-Space Models for Fungal Biodiversity Research", "abstract": "Accurate taxonomic classification from DNA barcodes is a cornerstone of global biodiversity monitoring, yet fungi present extreme challenges due to sparse labelling and long-tailed taxa distributions. Conventional supervised learning methods often falter in this domain, struggling to generalize to unseen species and to capture the hierarchical nature of the data. To address these limitations, we introduce BarcodeMamba+, a foundation model for fungal barcode classification built on a powerful and efficient state-space model architecture. We employ a pretrain and fine-tune paradigm, which utilizes partially labelled data and we demonstrate this is substantially more effective than traditional fully-supervised methods in this data-sparse environment. During fine-tuning, we systematically integrate and evaluate a suite of enhancements--including hierarchical label smoothing, a weighted loss function, and a multi-head output layer from MycoAI--to specifically tackle the challenges of fungal taxonomy. Our experiments show that each of these components yields significant performance gains. On a challenging fungal classification benchmark with distinct taxonomic distribution shifts from the broad training set, our final model outperforms a range of existing methods across all taxonomic levels. Our work provides a powerful new tool for genomics-based biodiversity research and establishes an effective and scalable training paradigm for this challenging domain. Our code is publicly available at https://github.com/bioscan-ml/BarcodeMamba.", "date": "2025-12-17", "categories": []}
{"id": "2503.01459", "title": "Primer C-VAE: An interpretable deep learning primer design method to detect emerging virus variants", "abstract": "Motivation: PCR is more economical and quicker than Next Generation Sequencing for detecting target organisms, with primer design being a critical step. In epidemiology with rapidly mutating viruses, designing effective primers is challenging. Traditional methods require substantial manual intervention and struggle to ensure effective primer design across different strains. For organisms with large, similar genomes like Escherichia coli and Shigella flexneri, differentiating between species is also difficult but crucial.\n  Results: We developed Primer C-VAE, a model based on a Variational Auto-Encoder framework with Convolutional Neural Networks to identify variants and generate specific primers. Using SARS-CoV-2, our model classified variants (alpha, beta, gamma, delta, omicron) with 98% accuracy and generated variant-specific primers. These primers appeared with >95% frequency in target variants and <5% in others, showing good performance in in-silico PCR tests. For Alpha, Delta, and Omicron, our primer pairs produced fragments <200 bp, suitable for qPCR detection. The model also generated effective primers for organisms with longer gene sequences like E. coli and S. flexneri.\n  Conclusion: Primer C-VAE is an interpretable deep learning approach for developing specific primer pairs for target organisms. This flexible, semi-automated and reliable tool works regardless of sequence completeness and length, allowing for qPCR applications and can be applied to organisms with large and highly similar genomes.", "date": "2025-03-03", "categories": []}
{"id": "2410.13597", "title": "Text-guided multi-property molecular optimization with a diffusion language model", "abstract": "Molecular optimization (MO) is a crucial stage in drug discovery in which task-oriented generated molecules are optimized to meet practical industrial requirements. Existing mainstream MO approaches primarily utilize external property predictors to guide iterative property optimization. However, learning all molecular samples in the vast chemical space is unrealistic for predictors. As a result, errors and noise are inevitably introduced during property prediction due to the nature of approximation. This leads to discrepancy accumulation, generalization reduction and suboptimal molecular candidates. In this paper, we propose a text-guided multi-property molecular optimization method utilizing transformer-based diffusion language model (TransDLM). TransDLM leverages standardized chemical nomenclature as semantic representations of molecules and implicitly embeds property requirements into textual descriptions, thereby mitigating error propagation during diffusion process. By fusing physically and chemically detailed textual semantics with specialized molecular representations, TransDLM effectively integrates diverse information sources to guide precise optimization, which enhances the model's ability to balance structural retention and property enhancement. Additionally, the success of a case study further demonstrates TransDLM's ability to solve practical problems. Experimentally, our approach surpasses state-of-the-art methods in maintaining molecular structural similarity and enhancing chemical properties on the benchmark dataset.", "date": "2024-10-17", "categories": []}
{"id": "2506.01983", "title": "Improvement of AMPs Identification with Generative Adversarial Network and Ensemble Classification", "abstract": "Identification of antimicrobial peptides is an important and necessary issue in today's era. Antimicrobial peptides are essential as an alternative to antibiotics for biomedical applications and many other practical applications. These oligopeptides are useful in drug design and cause innate immunity against microorganisms. Artificial intelligence algorithms have played a significant role in the ease of identifying these peptides.This research is improved by improving proposed method in the field of antimicrobial peptides prediction. Suggested method is improved by combining the best coding method from different perspectives, In the following a deep neural network to balance the imbalanced combined datasets. The results of this research show that the proposed method have a significant improvement in the accuracy and efficiency of the prediction of antimicrobial peptides and are able to provide the best results compared to the existing methods. These development in the field of prediction and classification of antimicrobial peptides, basically in the fields of medicine and pharmaceutical industries, have high effectiveness and application.", "date": "2025-05-16", "categories": []}
{"id": "2506.14920", "title": "Q2SAR: A Quantum Multiple Kernel Learning Approach for Drug Discovery", "abstract": "Quantitative Structure-Activity Relationship (QSAR) modeling is a cornerstone of computational drug discovery. This research demonstrates the successful application of a Quantum Multiple Kernel Learning (QMKL) framework to enhance QSAR classification, showing a notable performance improvement over classical methods. We apply this methodology to a dataset for identifying DYRK1A kinase inhibitors. The workflow involves converting SMILES representations into numerical molecular descriptors, reducing dimensionality via Principal Component Analysis (PCA), and employing a Support Vector Machine (SVM) trained on an optimized combination of multiple quantum and classical kernels. By benchmarking the QMKL-SVM against a classical Gradient Boosting model, we show that the quantum-enhanced approach achieves a superior AUC score, highlighting its potential to provide a quantum advantage in challenging cheminformatics classification tasks.", "date": "2025-06-17", "categories": []}
{"id": "2508.00969", "title": "Masked Omics Modeling for Multimodal Representation Learning across Histopathology and Molecular Profiles", "abstract": "Self-supervised learning (SSL) has driven major advances in computational pathology by enabling the learning of rich representations from histopathology data. Yet, tissue analysis alone may fall short in capturing broader molecular complexity, as key complementary information resides in high-dimensional omics profiles such as transcriptomics, methylomics, and genomics. To address this gap, we introduce MORPHEUS, the first multimodal pre-training strategy that integrates histopathology images and multi-omics data within a shared transformer-based architecture. At its core, MORPHEUS relies on a novel masked omics modeling objective that encourages the model to learn meaningful cross-modal relationships. This yields a general-purpose pre-trained encoder that can be applied to histopathology alone or in combination with any subset of omics modalities. Beyond inference, MORPHEUS also supports flexible any-to-any omics reconstruction, enabling one or more omics profiles to be reconstructed from any modality subset that includes histopathology. Pre-trained on a large pan-cancer cohort, MORPHEUS shows substantial improvements over supervised and SSL baselines across diverse tasks and modality combinations. Together, these capabilities position it as a promising direction for the development of multimodal foundation models in oncology. Code is publicly available at https://github.com/Lucas-rbnt/MORPHEUS", "date": "2025-08-01", "categories": []}
{"id": "2512.01885", "title": "TransientTrack: Advanced Multi-Object Tracking and Classification of Cancer Cells with Transient Fluorescent Signals", "abstract": "Tracking cells in time-lapse videos is an essential technique for monitoring cell population dynamics at a single-cell level. Current methods for cell tracking are developed on videos with mostly single, constant signals and do not detect pivotal events such as cell death. Here, we present TransientTrack, a deep learning-based framework for cell tracking in multi-channel microscopy video data with transient fluorescent signals that fluctuate over time following processes such as the circadian rhythm of cells. By identifying key cellular events - mitosis (cell division) and apoptosis (cell death) our method allows us to build complete trajectories, including cell lineage information. TransientTrack is lightweight and performs matching on cell detection embeddings directly, without the need for quantification of tracking-specific cell features. Furthermore, our approach integrates Transformer Networks, multi-stage matching using all detection boxes, and the interpolation of missing tracklets with the Kalman Filter. This unified framework achieves strong performance across diverse conditions, effectively tracking cells and capturing cell division and death. We demonstrate the use of TransientTrack in an analysis of the efficacy of a chemotherapeutic drug at a single-cell level. The proposed framework could further advance quantitative studies of cancer cell dynamics, enabling detailed characterization of treatment response and resistance mechanisms. The code is available at https://github.com/bozeklab/TransientTrack.", "date": "2025-12-01", "categories": []}
{"id": "2512.14011", "title": "Accelerating MHC-II Epitope Discovery via Multi-Scale Prediction in Antigen Presentation", "abstract": "Antigenic epitope presented by major histocompatibility complex II (MHC-II) proteins plays an essential role in immunotherapy. However, compared to the more widely studied MHC-I in computational immunotherapy, the study of MHC-II antigenic epitope poses significantly more challenges due to its complex binding specificity and ambiguous motif patterns. Consequently, existing datasets for MHC-II interactions are smaller and less standardized than those available for MHC-I. To address these challenges, we present a well-curated dataset derived from the Immune Epitope Database (IEDB) and other public sources. It not only extends and standardizes existing peptide-MHC-II datasets, but also introduces a novel antigen-MHC-II dataset with richer biological context. Leveraging this dataset, we formulate three major machine learning (ML) tasks of peptide binding, peptide presentation, and antigen presentation, which progressively capture the broader biological processes within the MHC-II antigen presentation pathway. We further employ a multi-scale evaluation framework to benchmark existing models, along with a comprehensive analysis over various modeling designs to this problem with a modular framework. Overall, this work serves as a valuable resource for advancing computational immunotherapy, providing a foundation for future research in ML guided epitope discovery and predictive modeling of immune responses.", "date": "2025-12-16", "categories": []}
{"id": "2512.14019", "title": "EXAONE Path 2.5: Pathology Foundation Model with Multi-Omics Alignment", "abstract": "Cancer progression arises from interactions across multiple biological layers, especially beyond morphological and across molecular layers that remain invisible to image-only models. To capture this broader biological landscape, we present EXAONE Path 2.5, a pathology foundation model that jointly models histologic, genomic, epigenetic and transcriptomic modalities, producing an integrated patient representation that reflects tumor biology more comprehensively. Our approach incorporates three key components: (1) multimodal SigLIP loss enabling all-pairwise contrastive learning across heterogeneous modalities, (2) a fragment-aware rotary positional encoding (F-RoPE) module that preserves spatial structure and tissue-fragment topology in WSI, and (3) domain-specialized internal foundation models for both WSI and RNA-seq to provide biologically grounded embeddings for robust multimodal alignment. We evaluate EXAONE Path 2.5 against six leading pathology foundation models across two complementary benchmarks: an internal real-world clinical dataset and the Patho-Bench benchmark covering 80 tasks. Our framework demonstrates high data and parameter efficiency, achieving on-par performance with state-of-the-art foundation models on Patho-Bench while exhibiting the highest adaptability in the internal clinical setting. These results highlight the value of biologically informed multimodal design and underscore the potential of integrated genotype-to-phenotype modeling for next-generation precision oncology.", "date": "2025-12-16", "categories": []}
{"id": "2407.12051", "title": "Dy-mer: An Explainable DNA Sequence Representation Scheme using Dictionary Learning", "abstract": "DNA sequences encode critical genetic information, yet their variable length and discrete nature impede direct utilization in deep learning models. Existing DNA representation schemes convert sequences into numerical vectors but fail to capture structural features of local subsequences and often suffer from limited interpretability and poor generalization on small datasets. To address these limitations, we propose Dy-mer, an interpretable and robust DNA representation scheme based on dictionary learning. Dy-mer formulates an optimization problem in tensor format, which ensures computational efficiency in batch processing. Our scheme reconstructs DNA sequences as concatenations of dynamic-length subsequences (dymers) through a convolution operation and simultaneously optimize a learnable dymer dictionary and sparse representations. Our method achieves state-of-the-art performance in downstream tasks such as DNA promoter classification and motif detection. Experiments further show that the learned dymers match known DNA motifs and clustering using Dy-mer yields semantically meaningful phylogenetic trees. These results demonstrate that the proposed approach achieves both strong predictive performance and high interpretability, making it well suited for biological research applications.", "date": "2024-07-06", "categories": []}
{"id": "2512.00696", "title": "Hierarchical Molecular Language Models (HMLMs)", "abstract": "Artificial intelligence (AI) is reshaping computational and network biology by enabling new approaches to decode cellular communication networks. We introduce Hierarchical Molecular Language Models (HMLMs), a novel framework that models cellular signaling as a specialized molecular language, where signaling molecules function as tokens, protein interactions define syntax, and functional consequences constitute semantics. HMLMs employ a transformer-based architecture adapted to accommodate graph-structured signaling networks through information transducers, mathematical entities that capture how molecules receive, process, and transmit signals. The architecture integrates multi-modal data sources across molecular, pathway, and cellular scales through hierarchical attention mechanisms and scale-bridging operators that enable information flow across biological hierarchies. Applied to a complex network of cardiac fibroblast signaling, HMLMs outperformed traditional approaches in temporal dynamics prediction, particularly under sparse sampling conditions. Attention-based analysis revealed biologically meaningful crosstalk patterns, including previously uncharacterized interactions between signaling pathways. By bridging molecular mechanisms with cellular phenotypes through AI-driven molecular language representation, HMLMs establish a foundation for biology-oriented large language models (LLMs) that could be pre-trained on comprehensive pathway datasets and applied across diverse signaling systems and tissues, advancing precision medicine and therapeutic discovery.", "date": "2025-11-30", "categories": []}
{"id": "2512.08772", "title": "De novo generation of functional terpene synthases using TpsGPT", "abstract": "Terpene synthases (TPS) are a key family of enzymes responsible for generating the diverse terpene scaffolds that underpin many natural products, including front-line anticancer drugs such as Taxol. However, de novo TPS design through directed evolution is costly and slow. We introduce TpsGPT, a generative model for scalable TPS protein design, built by fine-tuning the protein language model ProtGPT2 on 79k TPS sequences mined from UniProt. TpsGPT generated de novo enzyme candidates in silico and we evaluated them using multiple validation metrics, including EnzymeExplorer classification, ESMFold structural confidence (pLDDT), sequence diversity, CLEAN classification, InterPro domain detection, and Foldseek structure alignment. From an initial pool of 28k generated sequences, we identified seven putative TPS enzymes that satisfied all validation criteria. Experimental validation confirmed TPS enzymatic activity in at least two of these sequences. Our results show that fine-tuning of a protein language model on a carefully curated, enzyme-class-specific dataset, combined with rigorous filtering, can enable the de novo generation of functional, evolutionarily distant enzymes.", "date": "2025-12-09", "categories": []}
{"id": "2512.11927", "title": "Gene regulatory network inference algorithm based on spectral signed directed graph convolution", "abstract": "Accurately reconstructing Gene Regulatory Networks (GRNs) is crucial for understanding gene functions and disease mechanisms. Single-cell RNA sequencing (scRNA-seq) technology provides vast data for computational GRN reconstruction. Since GRNs are ideally modeled as signed directed graphs to capture activation/inhibition relationships, the most intuitive and reasonable approach is to design feature extractors based on the topological structure of GRNs to extract structural features, then combine them with biological characteristics for research. However, traditional spectral graph convolution struggles with this representation. Thus, we propose MSGRNLink, a novel framework that explicitly models GRNs as signed directed graphs and employs magnetic signed Laplacian convolution. Experiments across simulated and real datasets demonstrate that MSGRNLink outperforms all baseline models in AUROC. Parameter sensitivity analysis and ablation studies confirmed its robustness and the importance of each module. In a bladder cancer case study, MSGRNLink predicted more known edges and edge signs than benchmark models, further validating its biological relevance.", "date": "2025-12-12", "categories": []}
{"id": "2512.11928", "title": "MONET -- Virtual Cell Painting of Brightfield Images and Time Lapses Using Reference Consistent Diffusion", "abstract": "Cell painting is a popular technique for creating human-interpretable, high-contrast images of cell morphology. There are two major issues with cell paint: (1) it is labor-intensive and (2) it requires chemical fixation, making the study of cell dynamics impossible. We train a diffusion model (Morphological Observation Neural Enhancement Tool, or MONET) on a large dataset to predict cell paint channels from brightfield images. We show that model quality improves with scale. The model uses a consistency architecture to generate time-lapse videos, despite the impossibility of obtaining cell paint video training data. In addition, we show that this architecture enables a form of in-context learning, allowing the model to partially transfer to out-of-distribution cell lines and imaging protocols. Virtual cell painting is not intended to replace physical cell painting completely, but to act as a complementary tool enabling novel workflows in biological research.", "date": "2025-12-12", "categories": []}
{"id": "2512.12134", "title": "Modeling Dabrafenib Response Using Multi-Omics Modality Fusion and Protein Network Embeddings Based on Graph Convolutional Networks", "abstract": "Cancer cell response to targeted therapy arises from complex molecular interactions, making single omics insufficient for accurate prediction. This study develops a model to predict Dabrafenib sensitivity by integrating multiple omics layers (genomics, transcriptomics, proteomics, epigenomics, and metabolomics) with protein network embeddings generated using Graph Convolutional Networks (GCN). Each modality is encoded into low dimensional representations through neural network preprocessing. Protein interaction information from STRING is incorporated using GCN to capture biological topology. An attention based fusion mechanism assigns adaptive weights to each modality according to its relevance. Using GDSC cancer cell line data, the model shows that selective integration of two modalities, especially proteomics and transcriptomics, achieves the best test performance (R2 around 0.96), outperforming all single omics and full multimodal settings. Genomic and epigenomic data were less informative, while proteomic and transcriptomic layers provided stronger phenotypic signals related to MAPK inhibitor activity. These results show that attention guided multi omics fusion combined with GCN improves drug response prediction and reveals complementary molecular determinants of Dabrafenib sensitivity. The approach offers a promising computational framework for precision oncology and predictive modeling of targeted therapies.", "date": "2025-12-13", "categories": []}
{"id": "2512.12198", "title": "MolGuidance: Advanced Guidance Strategies for Conditional Molecular Generation with Flow Matching", "abstract": "Key objectives in conditional molecular generation include ensuring chemical validity, aligning generated molecules with target properties, promoting structural diversity, and enabling efficient sampling for discovery. Recent advances in computer vision introduced a range of new guidance strategies for generative models, many of which can be adapted to support these goals. In this work, we integrate state-of-the-art guidance methods -- including classifier-free guidance, autoguidance, and model guidance -- in a leading molecule generation framework built on an SE(3)-equivariant flow matching process. We propose a hybrid guidance strategy that separately guides continuous and discrete molecular modalities -- operating on velocity fields and predicted logits, respectively -- while jointly optimizing their guidance scales via Bayesian optimization. Our implementation, benchmarked on the QM9 and QMe14S datasets, achieves new state-of-the-art performance in property alignment for de novo molecular generation. The generated molecules also exhibit high structural validity. Furthermore, we systematically compare the strengths and limitations of various guidance methods, offering insights into their broader applicability.", "date": "2025-12-13", "categories": []}
{"id": "2512.12272", "title": "Accurate de novo sequencing of the modified proteome with OmniNovo", "abstract": "Post-translational modifications (PTMs) serve as a dynamic chemical language regulating protein function, yet current proteomic methods remain blind to a vast portion of the modified proteome. Standard database search algorithms suffer from a combinatorial explosion of search spaces, limiting the identification of uncharacterized or complex modifications. Here we introduce OmniNovo, a unified deep learning framework for reference-free sequencing of unmodified and modified peptides directly from tandem mass spectra. Unlike existing tools restricted to specific modification types, OmniNovo learns universal fragmentation rules to decipher diverse PTMs within a single coherent model. By integrating a mass-constrained decoding algorithm with rigorous false discovery rate estimation, OmniNovo achieves state-of-the-art accuracy, identifying 51\\% more peptides than standard approaches at a 1\\% false discovery rate. Crucially, the model generalizes to biological sites unseen during training, illuminating the dark matter of the proteome and enabling unbiased comprehensive analysis of cellular regulation.", "date": "2025-12-13", "categories": []}
{"id": "2503.08838", "title": "PUMA: Discovery of Protein Units via Mutation-Aware Merging", "abstract": "Proteins are the essential drivers of biological processes. At the molecular level, they are chains of amino acids that can be viewed through a linguistic lens where the twenty standard residues serve as an alphabet combining to form a complex language, referred to as the language of life. To understand this language, we must first identify its fundamental units. Analogous to words, these units are hypothesized to represent an intermediate layer between single residues and larger domains. Crucially, just as protein diversity arises from evolution, these units should inherently reflect evolutionary relationships. We introduce PUMA (Protein Units via Mutation-Aware Merging) to discover these evolutionarily meaningful units. PUMA employs an iterative merging algorithm guided by substitution matrices to identify protein units and organize them into families linked by plausible mutations. This process creates a hierarchical genealogy where parent units and their mutational variants coexist, simultaneously producing a unit vocabulary and the genealogical structure connecting them. We validate that PUMA families are biologically meaningful; mutations within a PUMA family correlate with clinically benign variants and with high-scoring mutations in high-throughput assays. Furthermore, these units align with the contextual preferences of protein language models and map to known functional annotations. PUMA's genealogical framework provides evolutionarily grounded units, offering a structured approach for understanding the language of life.", "date": "2025-03-11", "categories": []}
{"id": "2510.27097", "title": "Hierarchical Bayesian Model for Gene Deconvolution and Functional Analysis in Human Endometrium Across the Menstrual Cycle", "abstract": "Bulk tissue RNA sequencing of heterogeneous samples provides averaged gene expression profiles, obscuring cell type-specific dynamics. To address this, we present a probabilistic hierarchical Bayesian model that deconvolves bulk RNA-seq data into constituent cell-type expression profiles and proportions, leveraging a high-resolution single-cell reference. We apply our model to human endometrial tissue across the menstrual cycle, a context characterized by dramatic hormone-driven cellular composition changes. Our extended framework provides a principled inference of cell type proportions and cell-specific gene expression changes across cycle phases. We demonstrate the model's structure, priors, and inference strategy in detail, and we validate its performance with simulations and comparisons to existing methods. The results reveal dynamic shifts in epithelial, stromal, and immune cell fractions between menstrual phases, and identify cell-type-specific differential gene expression associated with endometrial function (e.g., decidualization markers in stromal cells during the secretory phase). We further conduct robustness tests and show that our Bayesian approach is resilient to reference mismatches and noise. Finally, we discuss the biological significance of our findings, potential clinical implications for fertility and endometrial disorders, and future directions, including integration of spatial transcriptomics.", "date": "2025-10-31", "categories": []}
{"id": "2510.08169", "title": "Bidirectional Representations Augmented Autoregressive Biological Sequence Generation", "abstract": "Autoregressive (AR) models, common in sequence generation, are limited in many biological tasks such as de novo peptide sequencing and protein modeling by their unidirectional nature, failing to capture crucial global bidirectional token dependencies. Non-Autoregressive (NAR) models offer holistic, bidirectional representations but face challenges with generative coherence and scalability. To transcend this, we propose a hybrid framework enhancing AR generation by dynamically integrating rich contextual information from non-autoregressive mechanisms. Our approach couples a shared input encoder with two decoders: a non-autoregressive one learning latent bidirectional biological features, and an AR decoder synthesizing the biological sequence by leveraging these bidirectional features. A novel cross-decoder attention module enables the AR decoder to iteratively query and integrate these bidirectional features, enriching its predictions. This synergy is cultivated via a tailored training strategy with importance annealing for balanced objectives and cross-decoder gradient blocking for stable, focused learning. Evaluations on a demanding nine-species benchmark of de novo peptide sequencing show that our model substantially surpasses AR and NAR baselines. It uniquely harmonizes AR stability with NAR contextual awareness, delivering robust, superior performance on diverse downstream data. This research advances biological sequence modeling techniques and contributes a novel architectural paradigm for augmenting AR models with enhanced bidirectional understanding for complex sequence generation. Code is available at https://github.com/BEAM-Labs/denovo.", "date": "2025-10-09", "categories": []}
{"id": "2512.04629", "title": "BioMedGPT-Mol: Multi-task Learning for Molecular Understanding and Generation", "abstract": "Molecules play a crucial role in biomedical research and discovery, particularly in the field of small molecule drug development. Given the rapid advancements in large language models, especially the recent emergence of reasoning models, it is natural to explore how a general-purpose language model can be efficiently adapted for molecular science applications. In this work, we introduce BioMedGPT-Mol, a molecular language model designed to support molecular understanding and generation tasks. By curating and unifying existing public instruction datasets, we have assembled a large-scale, comprehensive, and high-quality training dataset. The model is then fine-tuned through a meticulously designed multi-task learning framework. On a consolidated benchmark derived from LlaSMol, TOMG-Bench, and MuMOInstruct, BioMedGPT-Mol achieves remarkable performance. Our experimental results demonstrate that a general-purpose reasoning model can be effectively and efficiently post-trained into a professional molecular language model through a well-structured multi-task curriculum. Leveraging these capabilities, we further apply the model to multi-step retrosynthetic planning, achieving state-of-the-art performance on RetroBench and demonstrating its superior efficacy as an end-to-end retrosynthetic planner. We anticipate that our approach can be extended to other biomedical scientific domains.", "date": "2025-12-04", "categories": []}
{"id": "2512.10034", "title": "DynaMate: An Autonomous Agent for Protein-Ligand Molecular Dynamics Simulations", "abstract": "Force field-based molecular dynamics (MD) simulations are indispensable for probing the structure, dynamics, and functions of biomolecular systems, including proteins and protein-ligand complexes. Despite their broad utility in drug discovery and protein engineering, the technical complexity of MD setup, encompassing parameterization, input preparation, and software configuration, remains a major barrier for widespread and efficient usage. Agentic LLMs have demonstrated their capacity to autonomously execute multi-step scientific processes, and to date, they have not successfully been used to automate protein-ligand MD workflows. Here, we present DynaMate, a modular multi-agent framework that autonomously designs and executes complete MD workflows for both protein and protein-ligand systems, and offers free energy binding affinity calculations with the MM/PB(GB)SA method. The framework integrates dynamic tool use, web search, PaperQA, and a self-correcting behavior. DynaMate comprises three specialized modules, interacting to plan the experiment, perform the simulation, and analyze the results. We evaluated its performance across twelve benchmark systems of varying complexity, assessing success rate, efficiency, and adaptability. DynaMate reliably performed full MD simulations, corrected runtime errors through iterative reasoning, and produced meaningful analyses of protein-ligand interactions. This automated framework paves the way toward standardized, scalable, and time-efficient molecular modeling pipelines for future biomolecular and drug design applications.", "date": "2025-12-10", "categories": []}
{"id": "2512.10066", "title": "Classifying Metamorphic versus Single-Fold Proteins with Statistical Learning and AlphaFold2", "abstract": "The remarkable success of AlphaFold2 in providing accurate atomic-level prediction of protein structures from their amino acid sequence has transformed approaches to the protein folding problem. However, its core paradigm of mapping one sequence to one structure may only be appropriate for single-fold proteins with one stable conformation. Metamorphic proteins, which can adopt multiple distinct conformations, have conformational diversity that cannot be adequately modeled by AlphaFold2. Hence, classifying whether a given protein is metamorphic or single-fold remains a critical challenge for both laboratory experiments and computational methods. To address this challenge, we developed a novel classification framework by re-purposing AlphaFold2 to generate conformational ensembles via a multiple sequence alignment sampling method. From these ensembles, we extract a comprehensive set of features characterizing the conformational ensemble's modality and structural dispersion. A random forest classifier trained on a carefully curated benchmark dataset of known metamorphic and single-fold proteins achieves a mean AUC of 0.869 with cross-validation, demonstrating the effectiveness of our integrated approach. Furthermore, by applying our classifier to 600 randomly sampled proteins from the Protein Data Bank, we identified several potential metamorphic protein candidates -- including the 40S ribosomal protein S30, whose conformational change is crucial for its secondary function in antimicrobial defense. By combining AI-driven protein structure prediction with statistical learning, our work provides a powerful new approach for discovering metamorphic proteins and deepens our understanding of their role in their molecular function.", "date": "2025-12-10", "categories": []}
{"id": "2512.10141", "title": "Sequence-to-Image Transformation for Sequence Classification Using Rips Complex Construction and Chaos Game Representation", "abstract": "Traditional feature engineering approaches for molecular sequence classification suffer from sparsity issues and computational complexity, while deep learning models often underperform on tabular biological data. This paper introduces a novel topological approach that transforms molecular sequences into images by combining Chaos Game Representation (CGR) with Rips complex construction from algebraic topology. Our method maps sequence elements to 2D coordinates via CGR, computes pairwise distances, and constructs Rips complexes to capture both local structural and global topological features. We provide formal guarantees on representation uniqueness, topological stability, and information preservation. Extensive experiments on anticancer peptide datasets demonstrate superior performance over vector-based, sequence language models, and existing image-based methods, achieving 86.8\\% and 94.5\\% accuracy on breast and lung cancer datasets, respectively. The topological representation preserves critical sequence information while enabling effective utilization of vision-based deep learning architectures for molecular sequence analysis.", "date": "2025-12-10", "categories": []}
{"id": "2512.10147", "title": "Murmur2Vec: A Hashing Based Solution For Embedding Generation Of COVID-19 Spike Sequences", "abstract": "Early detection and characterization of coronavirus disease (COVID-19), caused by SARS-CoV-2, remain critical for effective clinical response and public-health planning. The global availability of large-scale viral sequence data presents significant opportunities for computational analysis; however, existing approaches face notable limitations. Phylogenetic tree-based methods are computationally intensive and do not scale efficiently to today's multi-million-sequence datasets. Similarly, current embedding-based techniques often rely on aligned sequences or exhibit suboptimal predictive performance and high runtime costs, creating barriers to practical large-scale analysis. In this study, we focus on the most prevalent SARS-CoV-2 lineages associated with the spike protein region and introduce a scalable embedding method that leverages hashing to generate compact, low-dimensional representations of spike sequences. These embeddings are subsequently used to train a variety of machine learning models for supervised lineage classification. We conduct an extensive evaluation comparing our approach with multiple baseline and state-of-the-art biological sequence embedding methods across diverse metrics. Our results demonstrate that the proposed embeddings offer substantial improvements in efficiency, achieving up to 86.4\\% classification accuracy while reducing embedding generation time by as much as 99.81\\%. This highlights the method's potential as a fast, effective, and scalable solution for large-scale viral sequence analysis.", "date": "2025-12-10", "categories": []}
{"id": "2502.02182", "title": "Sequence models for continuous cell cycle stage prediction from brightfield images", "abstract": "Understanding cell cycle dynamics is crucial for studying biological processes such as growth, development and disease progression. While fluorescent protein reporters like the Fucci system allow live monitoring of cell cycle phases, they require genetic engineering and occupy additional fluorescence channels, limiting broader applicability in complex experiments. In this study, we conduct a comprehensive evaluation of deep learning methods for predicting continuous Fucci signals using non-fluorescence brightfield imaging, a widely available label-free modality. To that end, we generated a large dataset of 1.3 M images of dividing RPE1 cells with full cell cycle trajectories to quantitatively compare the predictive performance of distinct model categories including single time-frame models, causal state space models and bidirectional transformer models. We show that both causal and transformer-based models significantly outperform single- and fixed frame approaches, enabling the prediction of visually imperceptible transitions like G1/S within 1h resolution. Our findings underscore the importance of sequence models for accurate predictions of cell cycle dynamics and highlight their potential for label-free imaging.", "date": "2025-02-04", "categories": []}
{"id": "2509.18480", "title": "SimpleFold: Folding Proteins is Simpler than You Think", "abstract": "Protein folding models have achieved groundbreaking results typically via a combination of integrating domain knowledge into the architectural blocks and training pipelines. Nonetheless, given the success of generative models across different but related problems, it is natural to question whether these architectural designs are a necessary condition to build performant models. In this paper, we introduce SimpleFold, the first flow-matching based protein folding model that solely uses general purpose transformer blocks. Protein folding models typically employ computationally expensive modules involving triangular updates, explicit pair representations or multiple training objectives curated for this specific domain. Instead, SimpleFold employs standard transformer blocks with adaptive layers and is trained via a generative flow-matching objective with an additional structural term. We scale SimpleFold to 3B parameters and train it on approximately 9M distilled protein structures together with experimental PDB data. On standard folding benchmarks, SimpleFold-3B achieves competitive performance compared to state-of-the-art baselines, in addition SimpleFold demonstrates strong performance in ensemble prediction which is typically difficult for models trained via deterministic reconstruction objectives. Due to its general-purpose architecture, SimpleFold shows efficiency in deployment and inference on consumer-level hardware. SimpleFold challenges the reliance on complex domain-specific architectures designs in protein folding, opening up an alternative design space for future progress.", "date": "2025-09-23", "categories": []}
{"id": "2512.09365", "title": "KGOT: Unified Knowledge Graph and Optimal Transport Pseudo-Labeling for Molecule-Protein Interaction Prediction", "abstract": "Predicting molecule-protein interactions (MPIs) is a fundamental task in computational biology, with crucial applications in drug discovery and molecular function annotation. However, existing MPI models face two major challenges. First, the scarcity of labeled molecule-protein pairs significantly limits model performance, as available datasets capture only a small fraction of biological relevant interactions. Second, most methods rely solely on molecular and protein features, ignoring broader biological context such as genes, metabolic pathways, and functional annotations that could provide essential complementary information. To address these limitations, our framework first aggregates diverse biological datasets, including molecular, protein, genes and pathway-level interactions, and then develop an optimal transport-based approach to generate high-quality pseudo-labels for unlabeled molecule-protein pairs, leveraging the underlying distribution of known interactions to guide label assignment. By treating pseudo-labeling as a mechanism for bridging disparate biological modalities, our approach enables the effective use of heterogeneous data to enhance MPI prediction. We evaluate our framework on multiple MPI datasets including virtual screening tasks and protein retrieval tasks, demonstrating substantial improvements over state-of-the-art methods in prediction accuracies and zero shot ability across unseen interactions. Beyond MPI prediction, our approach provides a new paradigm for leveraging diverse biological data sources to tackle problems traditionally constrained by single- or bi-modal learning, paving the way for future advances in computational biology and drug discovery.", "date": "2025-12-10", "categories": []}
{"id": "2512.09894", "title": "Exploring Protein Language Model Architecture-Induced Biases for Antibody Comprehension", "abstract": "Recent advances in protein language models (PLMs) have demonstrated remarkable capabilities in understanding protein sequences. However, the extent to which different model architectures capture antibody-specific biological properties remains unexplored. In this work, we systematically investigate how architectural choices in PLMs influence their ability to comprehend antibody sequence characteristics and functions. We evaluate three state-of-the-art PLMs-AntiBERTa, BioBERT, and ESM2--against a general-purpose language model (GPT-2) baseline on antibody target specificity prediction tasks. Our results demonstrate that while all PLMs achieve high classification accuracy, they exhibit distinct biases in capturing biological features such as V gene usage, somatic hypermutation patterns, and isotype information. Through attention attribution analysis, we show that antibody-specific models like AntiBERTa naturally learn to focus on complementarity-determining regions (CDRs), while general protein models benefit significantly from explicit CDR-focused training strategies. These findings provide insights into the relationship between model architecture and biological feature extraction, offering valuable guidance for future PLM development in computational antibody design.", "date": "2025-12-10", "categories": []}
{"id": "2309.02071", "title": "BeeTLe: An Imbalance-Aware Deep Sequence Model for Linear B-Cell Epitope Prediction and Classification with Logit-Adjusted Losses", "abstract": "The process of identifying and characterizing B-cell epitopes, which are the portions of antigens recognized by antibodies, is important for our understanding of the immune system, and for many applications including vaccine development, therapeutics, and diagnostics. Computational epitope prediction is challenging yet rewarding as it significantly reduces the time and cost of laboratory work. Most of the existing tools do not have satisfactory performance and only discriminate epitopes from non-epitopes. This paper presents a new deep learning-based multi-task framework for linear B-cell epitope prediction as well as antibody type-specific epitope classification. Specifically, a sequenced-based neural network model using recurrent layers and Transformer blocks is developed. We propose an amino acid encoding method based on eigen decomposition to help the model learn the representations of epitopes. We introduce modifications to standard cross-entropy loss functions by extending a logit adjustment technique to cope with the class imbalance. Experimental results on data curated from the largest public epitope database demonstrate the validity of the proposed methods and the superior performance compared to competing ones.", "date": "2023-09-05", "categories": []}
{"id": "2501.06039", "title": "AI-powered virtual tissues from spatial proteomics for clinical diagnostics and biomedical discovery", "abstract": "Spatial proteomics technologies have transformed our understanding of complex tissue architecture in cancer but present unique challenges for computational analysis. Each study uses a different marker panel and protocol, and most methods are tailored to single cohorts, which limits knowledge transfer and robust biomarker discovery. Here we present Virtual Tissues (VirTues), a general-purpose foundation model for spatial proteomics that learns marker-aware, multi-scale representations of proteins, cells, niches and tissues directly from multiplex imaging data. From a single pretrained backbone, VirTues supports marker reconstruction, cell typing and niche annotation, spatial biomarker discovery, and patient stratification, including zero-shot annotation across heterogeneous panels and datasets. In triple-negative breast cancer, VirTues-derived biomarkers predict anti-PD-L1 chemo-immunotherapy response and stratify disease-free survival in an independent cohort, outperforming state-of-the-art biomarkers derived from the same datasets and current clinical stratification schemes.", "date": "2025-01-10", "categories": []}
{"id": "2507.19523", "title": "Language Models for Controllable DNA Sequence Design", "abstract": "We consider controllable DNA sequence design, where sequences are generated by conditioning on specific biological properties. While language models (LMs) such as GPT and BERT have achieved remarkable success in natural language generation, their application to DNA sequence generation remains largely underexplored. In this work, we introduce ATGC-Gen, an Automated Transformer Generator for Controllable Generation, which leverages cross-modal encoding to integrate diverse biological signals. ATGC-Gen is instantiated with both decoder-only and encoder-only transformer architectures, allowing flexible training and generation under either autoregressive or masked recovery objectives. We evaluate ATGC-Gen on representative tasks including promoter and enhancer sequence design, and further introduce a new dataset based on ChIP-Seq experiments for modeling protein binding specificity. Our experiments demonstrate that ATGC-Gen can generate fluent, diverse, and biologically relevant sequences aligned with the desired properties. Compared to prior methods, our model achieves notable improvements in controllability and functional relevance, highlighting the potential of language models in advancing programmable genomic design. The source code is released at (https://github.com/divelab/AIRS/blob/main/OpenBio/ATGC_Gen).", "date": "2025-07-19", "categories": []}
{"id": "2512.08732", "title": "Neural Ordinary Differential Equations for Simulating Metabolic Pathway Dynamics from Time-Series Multiomics Data", "abstract": "The advancement of human healthspan and bioengineering relies heavily on predicting the behavior of complex biological systems. While high-throughput multiomics data is becoming increasingly abundant, converting this data into actionable predictive models remains a bottleneck. High-capacity, datadriven simulation systems are critical in this landscape; unlike classical mechanistic models restricted by prior knowledge, these architectures can infer latent interactions directly from observational data, allowing for the simulation of temporal trajectories and the anticipation of downstream intervention effects in personalized medicine and synthetic biology. To address this challenge, we introduce Neural Ordinary Differential Equations (NODEs) as a dynamic framework for learning the complex interplay between the proteome and metabolome. We applied this framework to time-series data derived from engineered Escherichia coli strains, modeling the continuous dynamics of metabolic pathways. The proposed NODE architecture demonstrates superior performance in capturing system dynamics compared to traditional machine learning pipelines. Our results show a greater than 90% improvement in root mean squared error over baselines across both Limonene (up to 94.38% improvement) and Isopentenol (up to 97.65% improvement) pathway datasets. Furthermore, the NODE models demonstrated a 1000x acceleration in inference time, establishing them as a scalable, high-fidelity tool for the next generation of metabolic engineering and biological discovery.", "date": "2025-12-09", "categories": []}
{"id": "2507.14670", "title": "Gene-DML: Dual-Pathway Multi-Level Discrimination for Gene Expression Prediction from Histopathology Images", "abstract": "Accurately predicting gene expression from histopathology images offers a scalable and non-invasive approach to molecular profiling, with significant implications for precision medicine and computational pathology. However, existing methods often underutilize the cross-modal representation alignment between histopathology images and gene expression profiles across multiple representational levels, thereby limiting their prediction performance. To address this, we propose Gene-DML, a unified framework that structures latent space through Dual-pathway Multi-Level discrimination to enhance correspondence between morphological and transcriptional modalities. The multi-scale instance-level discrimination pathway aligns hierarchical histopathology representations extracted at local, neighbor, and global levels with gene expression profiles, capturing scale-aware morphological-transcriptional relationships. In parallel, the cross-level instance-group discrimination pathway enforces structural consistency between individual (image/gene) instances and modality-crossed (gene/image, respectively) groups, strengthening the alignment across modalities. By jointly modeling fine-grained and structural-level discrimination, Gene-DML is able to learn robust cross-modal representations, enhancing both predictive accuracy and generalization across diverse biological contexts. Extensive experiments on public spatial transcriptomics datasets demonstrate that Gene-DML achieves state-of-the-art performance in gene expression prediction. The code and processed datasets are available at https://github.com/YXSong000/Gene-DML.", "date": "2025-07-19", "categories": []}
{"id": "2507.17876", "title": "Look the Other Way: Designing 'Positive' Molecules with Negative Data via Task Arithmetic", "abstract": "The scarcity of molecules with desirable properties (i.e., `positive' molecules) is an inherent bottleneck for generative molecule design. To sidestep such obstacle, here we propose molecular task arithmetic: training a model on diverse and abundant negative examples to learn 'property directions' - without accessing any positively labeled data - and moving models in the opposite property directions to generate positive molecules. When analyzed on 33 design experiments with distinct molecular entities (small molecules, proteins), model architectures, and scales, molecular task arithmetic generated more diverse and successful designs than models trained on positive molecules in general. Moreover, we employed molecular task arithmetic in dual-objective and few-shot design tasks. We find that molecular task arithmetic can consistently increase the diversity of designs while maintaining desirable complex design properties, such as good docking scores to a protein. With its simplicity, data efficiency, and performance, molecular task arithmetic bears the potential to become the de facto transfer learning strategy for de novo molecule design.", "date": "2025-07-23", "categories": []}
{"id": "2512.04333", "title": "RGE-GCN: Recursive Gene Elimination with Graph Convolutional Networks for RNA-seq based Early Cancer Detection", "abstract": "Early detection of cancer plays a key role in improving survival rates, but identifying reliable biomarkers from RNA-seq data is still a major challenge. The data are high-dimensional, and conventional statistical methods often fail to capture the complex relationships between genes. In this study, we introduce RGE-GCN (Recursive Gene Elimination with Graph Convolutional Networks), a framework that combines feature selection and classification in a single pipeline. Our approach builds a graph from gene expression profiles, uses a Graph Convolutional Network to classify cancer versus normal samples, and applies Integrated Gradients to highlight the most informative genes. By recursively removing less relevant genes, the model converges to a compact set of biomarkers that are both interpretable and predictive. We evaluated RGE-GCN on synthetic data as well as real-world RNA-seq cohorts of lung, kidney, and cervical cancers. Across all datasets, the method consistently achieved higher accuracy and F1-scores than standard tools such as DESeq2, edgeR, and limma-voom. Importantly, the selected genes aligned with well-known cancer pathways including PI3K-AKT, MAPK, SUMOylation, and immune regulation. These results suggest that RGE-GCN shows promise as a generalizable approach for RNA-seq based early cancer detection and biomarker discovery (https://rce-gcn.streamlit.app/ ).", "date": "2025-12-03", "categories": []}
{"id": "2512.06496", "title": "PRIMRose: Insights into the Per-Residue Energy Metrics of Proteins with Double InDel Mutations using Deep Learning", "abstract": "Understanding how protein mutations affect protein structure is essential for advancements in computational biology and bioinformatics. We introduce PRIMRose, a novel approach that predicts energy values for each residue given a mutated protein sequence. Unlike previous models that assess global energy shifts, our method analyzes the localized energetic impact of double amino acid insertions or deletions (InDels) at the individual residue level, enabling residue-specific insights into structural and functional disruption. We implement a Convolutional Neural Network architecture to predict the energy changes of each residue in a protein mutation. We train our model on datasets constructed from nine proteins, grouped into three categories: one set with exhaustive double InDel mutations, another with approximately 145k randomly sampled double InDel mutations, and a third with approximately 80k randomly sampled double InDel mutations. Our model achieves high predictive accuracy across a range of energy metrics as calculated by the Rosetta molecular modeling suite and reveals localized patterns that influence model performance, such as solvent accessibility and secondary structure context. This per-residue analysis offers new insights into the mutational tolerance of specific regions within proteins and provides higher interpretable and biologically meaningful predictions of InDels' effects.", "date": "2025-12-06", "categories": []}
{"id": "2512.06592", "title": "On fine-tuning Boltz-2 for protein-protein affinity prediction", "abstract": "Accurate prediction of protein-protein binding affinity is vital for understanding molecular interactions and designing therapeutics. We adapt Boltz-2, a state-of-the-art structure-based protein-ligand affinity predictor, for protein-protein affinity regression and evaluate it on two datasets, TCR3d and PPB-affinity. Despite high structural accuracy, Boltz-2-PPI underperforms relative to sequence-based alternatives in both small- and larger-scale data regimes. Combining embeddings from Boltz-2-PPI with sequence-based embeddings yields complementary improvements, particularly for weaker sequence models, suggesting different signals are learned by sequence- and structure-based models. Our results echo known biases associated with training with structural data and suggest that current structure-based representations are not primed for performant affinity prediction.", "date": "2025-12-06", "categories": []}
{"id": "2512.06612", "title": "Learning Relative Gene Expression Trends from Pathology Images in Spatial Transcriptomics", "abstract": "Gene expression estimation from pathology images has the potential to reduce the RNA sequencing cost. Point-wise loss functions have been widely used to minimize the discrepancy between predicted and absolute gene expression values. However, due to the complexity of the sequencing techniques and intrinsic variability across cells, the observed gene expression contains stochastic noise and batch effects, and estimating the absolute expression values accurately remains a significant challenge. To mitigate this, we propose a novel objective of learning relative expression patterns rather than absolute levels. We assume that the relative expression levels of genes exhibit consistent patterns across independent experiments, even when absolute expression values are affected by batch effects and stochastic noise in tissue samples. Based on the assumption, we model the relation and propose a novel loss function called STRank that is robust to noise and batch effects. Experiments using synthetic datasets and real datasets demonstrate the effectiveness of the proposed method. The code is available at https://github.com/naivete5656/STRank.", "date": "2025-12-07", "categories": []}
{"id": "2512.07040", "title": "Transformation of Biological Networks into Images via Semantic Cartography for Visual Interpretation and Scalable Deep Analysis", "abstract": "Complex biological networks are fundamental to biomedical science, capturing interactions among molecules, cells, genes, and tissues. Deciphering these networks is critical for understanding health and disease, yet their scale and complexity represent a daunting challenge for current computational methods. Traditional biological network analysis methods, including deep learning approaches, while powerful, face inherent challenges such as limited scalability, oversmoothing long-range dependencies, difficulty in multimodal integration, expressivity bounds, and poor interpretability. We present Graph2Image, a framework that transforms large biological networks into sets of two-dimensional images by spatially arranging representative network nodes on a 2D grid. This transformation decouples the nodes as images, enabling the use of convolutional neural networks (CNNs) with global receptive fields and multi-scale pyramids, thus overcoming limitations of existing biological network analysis methods in scalability, memory efficiency, and long-range context capture. Graph2Image also facilitates seamless integration with other imaging and omics modalities and enhances interpretability through direct visualization of node-associated images. When applied to several large-scale biological network datasets, Graph2Image improved classification accuracy by up to 67.2% over existing methods and provided interpretable visualizations that revealed biologically coherent patterns. It also allows analysis of very large biological networks (nodes > 1 billion) on a personal computer. Graph2Image thus provides a scalable, interpretable, and multimodal-ready approach for biological network analysis, offering new opportunities for disease diagnosis and the study of complex biological systems.", "date": "2025-12-07", "categories": []}
{"id": "2512.07113", "title": "PlantBiMoE: A Bidirectional Foundation Model with SparseMoE for Plant Genomes", "abstract": "Understanding the underlying linguistic rules of plant genomes remains a fundamental challenge in computational biology. Recent advances including AgroNT and PDLLMs have made notable progress although, they suffer from excessive parameter size and limited ability to model the bidirectional nature of DNA strands respectively. To address these limitations, we propose PlantBiMoE, a lightweight and expressive plant genome language model that integrates bidirectional Mamba and a Sparse Mixture-of-Experts (SparseMoE) framework. The bidirectional Mamba enables the model to effectively capture structural dependencies across both the forward and reverse DNA strands, while SparseMoE significantly reduces the number of active parameters, improving computational efficiency without sacrificing modeling capacity. We evaluated and tested our model on the Modified Plants Genome Benchmark (MPGB), an enhanced genomic benchmark, which consolidates 31 datasets across 11 representative tasks, with input sequence lengths ranging from 50 to 6,000 bp. Experimental results demonstrate that PlantBiMoE achieves the best performance on 20 out of 31 datasets and the average best when comparing with existing models. In summary, all above results demonstrate that our model can effectively represent plant genomic sequences, serving as a robust computational tool for diverse genomic tasks, while making substantive contributions to plant genomics, gene editing, and synthetic biology. The code is available at: https://github.com/HUST-Keep-Lin/PlantBiMoE", "date": "2025-12-08", "categories": []}
{"id": "2507.08355", "title": "scE2TM improves single-cell embedding interpretability and reveals cellular perturbation signatures", "abstract": "Single-cell RNA sequencing technologies have revolutionized our understanding of cellular heterogeneity, yet computational methods often struggle to balance performance with biological interpretability. Embedded topic models have been widely used for interpretable single-cell embedding learning. However, these models suffer from the potential problem of interpretation collapse, where topics semantically collapse towards each other, resulting in redundant topics and incomplete capture of biological variation. Furthermore, the rise of single-cell foundation models creates opportunities to harness external biological knowledge for guiding model embeddings. Here, we present scE2TM, an external knowledge-guided embedded topic model that provides a high-quality cell embedding and interpretation for scRNA-seq analysis. Through embedding clustering regularization method, each topic is constrained to be the center of a separately aggregated gene cluster, enabling it to capture unique biological information. Across 20 scRNA-seq datasets, scE2TM achieves superior clustering performance compared with seven state-of-the-art methods. A comprehensive interpretability benchmark further shows that scE2TM-learned topics exhibit higher diversity and stronger consistency with underlying biological pathways. Modeling interferon-stimulated PBMCs, scE2TM simulates topic perturbations that drive control cells toward stimulated-like transcriptional states, faithfully mirroring experimental interferon responses. In melanoma, scE2TM identifies malignant-specific topics and extrapolates them to unseen patient data, revealing gene programs associated with patient survival.", "date": "2025-07-11", "categories": []}
{"id": "2512.05287", "title": "DMAGT: Unveiling miRNA-Drug Associations by Integrating SMILES and RNA Sequence Structures through Graph Transformer Models", "abstract": "MiRNAs, due to their role in gene regulation, have paved a new pathway for pharmacology, focusing on drug development that targets miRNAs. However, traditional wet lab experiments are limited by efficiency and cost constraints, making it difficult to extensively explore potential associations between developed drugs and target miRNAs. Therefore, we have designed a novel machine learning model based on a multi-layer transformer-based graph neural network, DMAGT, specifically for predicting associations between drugs and miRNAs. This model transforms drug-miRNA associations into graphs, employs Word2Vec for embedding features of drug molecular structures and miRNA base structures, and leverages a graph transformer model to learn from embedded features and relational structures, ultimately predicting associations between drugs and miRNAs. To evaluate DMAGT, we tested its performance on three datasets composed of drug-miRNA associations: ncDR, RNAInter, and SM2miR, achieving up to AUC of $95.24\\pm0.05$. DMAGT demonstrated superior performance in comparative experiments tackling similar challenges. To validate its practical efficacy, we specifically focused on two drugs, namely 5-Fluorouracil and Oxaliplatin. Of the 20 potential drug-miRNA associations identified as the most likely, 14 were successfully validated. The above experiments demonstrate that DMAGT has an excellent performance and stability in predicting drug-miRNA associations, providing a new shortcut for miRNA drug development.", "date": "2025-12-04", "categories": []}
{"id": "2512.05502", "title": "GRASP: Graph Reasoning Agents for Systems Pharmacology with Human-in-the-Loop", "abstract": "Quantitative Systems Pharmacology (QSP) modeling is essential for drug development but it requires significant time investment that limits the throughput of domain experts. We present \\textbf{GRASP} -- a multi-agent, graph-reasoning framework with a human-in-the-loop conversational interface -- that encodes QSP models as typed biological knowledge graphs and compiles them to executable MATLAB/SimBiology code while preserving units, mass balance, and physiological constraints. A two-phase workflow -- \\textsc{Understanding} (graph reconstruction of legacy code) and \\textsc{Action} (constraint-checked, language-driven modification) -- is orchestrated by a state machine with iterative validation. GRASP performs breadth-first parameter-alignment around new entities to surface dependent quantities and propose biologically plausible defaults, and it runs automatic execution/diagnostics until convergence. In head-to-head evaluations using LLM-as-judge, GRASP outperforms SME-guided CoT and ToT baselines across biological plausibility, mathematical correctness, structural fidelity, and code quality (\\(\\approx\\)9--10/10 vs.\\ 5--7/10). BFS alignment achieves F1 = 0.95 for dependency discovery, units, and range. These results demonstrate that graph-structured, agentic workflows can make QSP model development both accessible and rigorous, enabling domain experts to specify mechanisms in natural language without sacrificing biomedical fidelity.", "date": "2025-12-05", "categories": []}
{"id": "2502.15610", "title": "A general language model for peptide function identification", "abstract": "Accurate identification of bioactive peptides (BPs) and protein post-translational modifications (PTMs) is essential for understanding protein function and advancing therapeutic discovery. However, most computational methods remain limited in their generalizability across diverse peptide functions. Here, we present PDeepPP, a unified deep learning framework that integrates pretrained protein language models with a hybrid transformer-CNN architecture, enabling robust identification across diverse peptide classes and PTM sites. We curated comprehensive benchmark datasets and implemented strategies to address data imbalance, allowing PDeepPP to systematically extract both global and local sequence features. Through extensive analyses including dimensionality reduction and comparison studies, PDeepPP demonstrates strong, interpretable peptide representations and achieves state-of-the-art performance in 25 of the 33 biological identification tasks. Notably, PDeepPP attains high accuracy in antimicrobial (0.9726) and phosphorylation site (0.9984) identification, with 99.5% specificity in glycosylation site prediction and substantial reduction in false negatives in antimalarial tasks. By enabling large-scale, accurate peptide analysis, PDeepPP supports biomedical research and the discovery of novel therapeutic targets for disease treatment. All code, datasets, and pretrained models are publicly available via GitHub (https://github.com/fondress/PDeepPP) and Hugging Face (https://huggingface.co/fondress/PDeppPP)", "date": "2025-02-21", "categories": []}
{"id": "2510.18870", "title": "Triangle Multiplication Is All You Need For Biomolecular Structure Representations", "abstract": "AlphaFold has transformed protein structure prediction, but emerging applications such as virtual ligand screening, proteome-wide folding, and de novo binder design demand predictions at a massive scale, where runtime and memory costs become prohibitive. A major bottleneck lies in the Pairformer backbone of AlphaFold3-style models, which relies on computationally expensive triangular primitives-especially triangle attention-for pairwise reasoning. We introduce Pairmixer, a streamlined alternative that eliminates triangle attention while preserving higher-order geometric reasoning capabilities that are critical for structure prediction. Pairmixer substantially improves computational efficiency, matching state-of-the-art structure predictors across folding and docking benchmarks, delivering up to 4x faster inference on long sequences while reducing training cost by 34%. Its efficiency alleviates the computational burden of downstream applications such as modeling large protein complexes, high-throughput ligand and binder screening, and hallucination-based design. Within BoltzDesign, for example, Pairmixer delivers over 2x faster sampling and scales to sequences ~30% longer than the memory limits of Pairformer. Code is available at https://github.com/genesistherapeutics/pairmixer.", "date": "2025-10-21", "categories": []}
{"id": "2512.05080", "title": "OMTRA: A Multi-Task Generative Model for Structure-Based Drug Design", "abstract": "Structure-based drug design (SBDD) focuses on designing small-molecule ligands that bind to specific protein pockets. Computational methods are integral in modern SBDD workflows and often make use of virtual screening methods via docking or pharmacophore search. Modern generative modeling approaches have focused on improving novel ligand discovery by enabling de novo design. In this work, we recognize that these tasks share a common structure and can therefore be represented as different instantiations of a consistent generative modeling framework. We propose a unified approach in OMTRA, a multi-modal flow matching model that flexibly performs many tasks relevant to SBDD, including some with no analogue in conventional workflows. Additionally, we curate a dataset of 500M 3D molecular conformers, complementing protein-ligand data and expanding the chemical diversity available for training. OMTRA obtains state of the art performance on pocket-conditioned de novo design and docking; however, the effects of large-scale pretraining and multi-task training are modest. All code, trained models, and dataset for reproducing this work are available at https://github.com/gnina/OMTRA", "date": "2025-12-04", "categories": []}
{"id": "2505.17478", "title": "ConfRover: Simultaneous Modeling of Protein Conformation and Dynamics via Autoregression", "abstract": "Understanding protein dynamics is critical for elucidating their biological functions. The increasing availability of molecular dynamics (MD) data enables the training of deep generative models to efficiently explore the conformational space of proteins. However, existing approaches either fail to explicitly capture the temporal dependencies between conformations or do not support direct generation of time-independent samples. To address these limitations, we introduce ConfRover, an autoregressive model that simultaneously learns protein conformation and dynamics from MD trajectories, supporting both time-dependent and time-independent sampling. At the core of our model is a modular architecture comprising: (i) an encoding layer, adapted from protein folding models, that embeds protein-specific information and conformation at each time frame into a latent space; (ii) a temporal module, a sequence model that captures conformational dynamics across frames; and (iii) an SE(3) diffusion model as the structure decoder, generating conformations in continuous space. Experiments on ATLAS, a large-scale protein MD dataset of diverse structures, demonstrate the effectiveness of our model in learning conformational dynamics and supporting a wide range of downstream tasks. ConfRover is the first model to sample both protein conformations and trajectories within a single framework, offering a novel and flexible approach for learning from protein MD data. Project website: https://bytedance-seed.github.io/ConfRover.", "date": "2025-05-23", "categories": []}
{"id": "2506.00821", "title": "SafeGenes: Evaluating the Adversarial Robustness of Genomic Foundation Models", "abstract": "Genomic Foundation Models (GFMs), such as Evolutionary Scale Modeling (ESM), have demonstrated significant success in variant effect prediction. However, their adversarial robustness remains largely unexplored. To address this gap, we propose SafeGenes: a framework for Secure analysis of genomic foundation models, leveraging adversarial attacks to evaluate robustness against both engineered near-identical adversarial Genes and embedding-space manipulations. In this study, we assess the adversarial vulnerabilities of GFMs using two approaches: the Fast Gradient Sign Method (FGSM) and a soft prompt attack. FGSM introduces minimal perturbations to input sequences, while the soft prompt attack optimizes continuous embeddings to manipulate model predictions without modifying the input tokens. By combining these techniques, SafeGenes provides a comprehensive assessment of GFM susceptibility to adversarial manipulation. Targeted soft prompt attacks induced severe degradation in MLM-based shallow architectures such as ProteinBERT, while still producing substantial failure modes even in high-capacity foundation models such as ESM1b and ESM1v. These findings expose critical vulnerabilities in current foundation models, opening new research directions toward improving their security and robustness in high-stakes genomic applications such as variant effect prediction.", "date": "2025-06-01", "categories": []}
{"id": "2512.03111", "title": "PanFoMa: A Lightweight Foundation Model and Benchmark for Pan-Cancer", "abstract": "Single-cell RNA sequencing (scRNA-seq) is essential for decoding tumor heterogeneity. However, pan-cancer research still faces two key challenges: learning discriminative and efficient single-cell representations, and establishing a comprehensive evaluation benchmark. In this paper, we introduce PanFoMa, a lightweight hybrid neural network that combines the strengths of Transformers and state-space models to achieve a balance between performance and efficiency. PanFoMa consists of a front-end local-context encoder with shared self-attention layers to capture complex, order-independent gene interactions; and a back-end global sequential feature decoder that efficiently integrates global context using a linear-time state-space model. This modular design preserves the expressive power of Transformers while leveraging the scalability of Mamba to enable transcriptome modeling, effectively capturing both local and global regulatory signals. To enable robust evaluation, we also construct a large-scale pan-cancer single-cell benchmark, PanFoMaBench, containing over 3.5 million high-quality cells across 33 cancer subtypes, curated through a rigorous preprocessing pipeline. Experimental results show that PanFoMa outperforms state-of-the-art models on our pan-cancer benchmark (+4.0\\%) and across multiple public tasks, including cell type annotation (+7.4\\%), batch integration (+4.0\\%) and multi-omics integration (+3.1\\%). The code is available at https://github.com/Xiaoshui-Huang/PanFoMa.", "date": "2025-12-02", "categories": []}
{"id": "2512.03158", "title": "Contrastive Deep Learning for Variant Detection in Wastewater Genomic Sequencing", "abstract": "Wastewater-based genomic surveillance has emerged as a powerful tool for population-level viral monitoring, offering comprehensive insights into circulating viral variants across entire communities. However, this approach faces significant computational challenges stemming from high sequencing noise, low viral coverage, fragmented reads, and the complete absence of labeled variant annotations. Traditional reference-based variant calling pipelines struggle with novel mutations and require extensive computational resources. We present a comprehensive framework for unsupervised viral variant detection using Vector-Quantized Variational Autoencoders (VQ-VAE) that learns discrete codebooks of genomic patterns from k-mer tokenized sequences without requiring reference genomes or variant labels. Our approach extends the base VQ-VAE architecture with masked reconstruction pretraining for robustness to missing data and contrastive learning for highly discriminative embeddings. Evaluated on SARS-CoV-2 wastewater sequencing data comprising approximately 100,000 reads, our VQ-VAE achieves 99.52% mean token-level accuracy and 56.33% exact sequence match rate while maintaining 19.73% codebook utilization (101 of 512 codes active), demonstrating efficient discrete representation learning. Contrastive fine-tuning with different projection dimensions yields substantial clustering improvements: 64-dimensional embeddings achieve +35% Silhouette score improvement (0.31 to 0.42), while 128-dimensional embeddings achieve +42% improvement (0.31 to 0.44), clearly demonstrating the impact of embedding dimensionality on variant discrimination capability. Our reference-free framework provides a scalable, interpretable approach to genomic surveillance with direct applications to public health monitoring.", "date": "2025-12-02", "categories": []}
{"id": "2512.03592", "title": "Harnessing Hypergraphs in Geometric Deep Learning for 3D RNA Inverse Folding", "abstract": "The RNA inverse folding problem, a key challenge in RNA design, involves identifying nucleotide sequences that can fold into desired secondary structures, which are critical for ensuring molecular stability and function. The inherent complexity of this task stems from the intricate relationship between sequence and structure, making it particularly challenging. In this paper, we propose a framework, named HyperRNA, a generative model with an encoder-decoder architecture that leverages hypergraphs to design RNA sequences. Specifically, our HyperRNA model consists of three main components: preprocessing, encoding and decoding.\n  In the preprocessing stage, graph structures are constructed by extracting the atom coordinates of RNA backbone based on 3-bead coarse-grained representation. The encoding stage processes these graphs, capturing higher order dependencies and complex biomolecular interactions using an attention embedding module and a hypergraph-based encoder. Finally, the decoding stage generates the RNA sequence in an autoregressive manner. We conducted quantitative and qualitative experiments on the PDBBind and RNAsolo datasets to evaluate the inverse folding task for RNA sequence generation and RNA-protein complex sequence generation. The experimental results demonstrate that HyperRNA not only outperforms existing RNA design methods but also highlights the potential of leveraging hypergraphs in RNA engineering.", "date": "2025-12-03", "categories": []}
{"id": "2510.12498", "title": "Artificial Intelligence Virtual Cells: From Measurements to Decisions across Modality, Scale, Dynamics, and Evaluation", "abstract": "Artificial Intelligence Virtual Cells (AIVCs) aim to learn executable, decision-relevant models of cell state from multimodal, multiscale measurements. Recent studies have introduced single-cell and spatial foundation models, improved cross-modality alignment, scaled perturbation atlases, and explored pathway-level readouts. Nevertheless, although held-out validation is standard practice, evaluations remain predominantly within single datasets and settings; evidence indicates that transport across laboratories and platforms is often limited, that some data splits are vulnerable to leakage and coverage bias, and that dose, time and combination effects are not yet systematically handled. Cross-scale coupling also remains constrained, as anchors linking molecular, cellular and tissue levels are sparse, and alignment to scientific or clinical readouts varies across studies. We propose a model-agnostic Cell-State Latent (CSL) perspective that organizes learning via an operator grammar: measurement, lift/project for cross-scale coupling, and intervention for dosing and scheduling. This view motivates a decision-aligned evaluation blueprint across modality, scale, context and intervention, and emphasizes function-space readouts such as pathway activity, spatial neighborhoods and clinically relevant endpoints. We recommend operator-aware data design, leakage-resistant partitions, and transparent calibration and reporting to enable reproducible, like-for-like comparisons.", "date": "2025-10-14", "categories": []}
{"id": "2510.24974", "title": "Conformational Rank Conditioned Committees for Machine Learning-Assisted Directed Evolution", "abstract": "Machine Learning-assisted directed evolution (MLDE) is a powerful tool for efficiently navigating antibody fitness landscapes. Many structure-aware MLDE pipelines rely on a single conformation or a single committee across all conformations, limiting their ability to separate conformational uncertainty from epistemic uncertainty. Here, we introduce a rank -conditioned committee (RCC) framework that leverages ranked conformations to assign a deep neural network committee per rank. This design enables a principled separation between epistemic uncertainty and conformational uncertainty. We validate our RCC-MLDE approach on SARS-CoV-2 antibody docking, demonstrating significant improvements over baseline strategies. Our results offer a scalable route for therapeutic antibody discovery while directly addressing the challenge of modeling conformational uncertainty.", "date": "2025-10-28", "categories": []}
{"id": "2512.02030", "title": "Generative design and validation of therapeutic peptides for glioblastoma based on a potential target ATP5A", "abstract": "Glioblastoma (GBM) remains the most aggressive tumor, urgently requiring novel therapeutic strategies. Here, we present a dry-to-wet framework combining generative modeling and experimental validation to optimize peptides targeting ATP5A, a potential peptide-binding protein for GBM. Our framework introduces the first lead-conditioned generative model, which focuses exploration on geometrically relevant regions around lead peptides and mitigates the combinatorial complexity of de novo methods. Specifically, we propose POTFlow, a \\underline{P}rior and \\underline{O}ptimal \\underline{T}ransport-based \\underline{Flow}-matching model for peptide optimization. POTFlow employs secondary structure information (e.g., helix, sheet, loop) as geometric constraints, which are further refined by optimal transport to produce shorter flow paths. With this design, our method achieves state-of-the-art performance compared with five popular approaches. When applied to GBM, our method generates peptides that selectively inhibit cell viability and significantly prolong survival in a patient-derived xenograft (PDX) model. As the first lead peptide-conditioned flow matching model, POTFlow holds strong potential as a generalizable framework for therapeutic peptide design.", "date": "2025-11-19", "categories": []}
{"id": "2512.02033", "title": "CONFIDE: Hallucination Assessment for Reliable Biomolecular Structure Prediction and Design", "abstract": "Reliable evaluation of protein structure predictions remains challenging, as metrics like pLDDT capture energetic stability but often miss subtle errors such as atomic clashes or conformational traps reflecting topological frustration within the protein folding energy landscape. We present CODE (Chain of Diffusion Embeddings), a self evaluating metric empirically found to quantify topological frustration directly from the latent diffusion embeddings of the AlphaFold3 series of structure predictors in a fully unsupervised manner. Integrating this with pLDDT, we propose CONFIDE, a unified evaluation framework that combines energetic and topological perspectives to improve the reliability of AlphaFold3 and related models. CODE strongly correlates with protein folding rates driven by topological frustration, achieving a correlation of 0.82 compared to pLDDT's 0.33 (a relative improvement of 148\\%). CONFIDE significantly enhances the reliability of quality evaluation in molecular glue structure prediction benchmarks, achieving a Spearman correlation of 0.73 with RMSD, compared to pLDDT's correlation of 0.42, a relative improvement of 73.8\\%. Beyond quality assessment, our approach applies to diverse drug design tasks, including all-atom binder design, enzymatic active site mapping, mutation induced binding affinity prediction, nucleic acid aptamer screening, and flexible protein modeling. By combining data driven embeddings with theoretical insight, CODE and CONFIDE outperform existing metrics across a wide range of biomolecular systems, offering robust and versatile tools to refine structure predictions, advance structural biology, and accelerate drug discovery.", "date": "2025-11-20", "categories": []}
{"id": "2512.02223", "title": "On the Approximation of Phylogenetic Distance Functions by Artificial Neural Networks", "abstract": "Inferring the phylogenetic relationships among a sample of organisms is a fundamental problem in modern biology. While distance-based hierarchical clustering algorithms achieved early success on this task, these have been supplanted by Bayesian and maximum likelihood search procedures based on complex models of molecular evolution. In this work we describe minimal neural network architectures that can approximate classic phylogenetic distance functions and the properties required to learn distances under a variety of molecular evolutionary models. In contrast to model-based inference (and recently proposed model-free convolutional and transformer networks), these architectures have a small computational footprint and are scalable to large numbers of taxa and molecular characters. The learned distance functions generalize well and, given an appropriate training dataset, achieve results comparable to state-of-the art inference methods.", "date": "2025-12-01", "categories": []}
{"id": "2512.02471", "title": "scCluBench: Comprehensive Benchmarking of Clustering Algorithms for Single-Cell RNA Sequencing", "abstract": "Cell clustering is crucial for uncovering cellular heterogeneity in single-cell RNA sequencing (scRNA-seq) data by identifying cell types and marker genes. Despite its importance, benchmarks for scRNA-seq clustering methods remain fragmented, often lacking standardized protocols and failing to incorporate recent advances in artificial intelligence. To fill these gaps, we present scCluBench, a comprehensive benchmark of clustering algorithms for scRNA-seq data. First, scCluBench provides 36 scRNA-seq datasets collected from diverse public sources, covering multiple tissues, which are uniformly processed and standardized to ensure consistency for systematic evaluation and downstream analyses. To evaluate performance, we collect and reproduce a range of scRNA-seq clustering methods, including traditional, deep learning-based, graph-based, and biological foundation models. We comprehensively evaluate each method both quantitatively and qualitatively, using core performance metrics as well as visualization analyses. Furthermore, we construct representative downstream biological tasks, such as marker gene identification and cell type annotation, to further assess the practical utility. scCluBench then investigates the performance differences and applicability boundaries of various clustering models across diverse analytical tasks, systematically assessing their robustness and scalability in real-world scenarios. Overall, scCluBench offers a standardized and user-friendly benchmark for scRNA-seq clustering, with curated datasets, unified evaluation protocols, and transparent analyses, facilitating informed method selection and providing valuable insights into model generalizability and application scope.", "date": "2025-12-02", "categories": []}
{"id": "2502.07299", "title": "Life-Code: Central Dogma Modeling with Multi-Omics Sequence Unification", "abstract": "The interactions between DNA, RNA, and proteins are fundamental to biological processes, as illustrated by the central dogma of molecular biology. Although modern biological pre-trained models have achieved great success in analyzing these macromolecules individually, their interconnected nature remains underexplored. This paper follows the guidance of the central dogma to redesign both the data and model pipeline and offers a comprehensive framework, Life-Code, that spans different biological functions. As for data flow, we propose a unified pipeline to integrate multi-omics data by reverse-transcribing RNA and reverse-translating amino acids into nucleotide-based sequences. As for the model, we design a codon tokenizer and a hybrid long-sequence architecture to encode the interactions between coding and non-coding regions through masked modeling pre-training. To model the translation and folding process with coding sequences, Life-Code learns protein structures of the corresponding amino acids by knowledge distillation from off-the-shelf protein language models. Such designs enable Life-Code to capture complex interactions within genetic sequences, providing a more comprehensive understanding of multi-omics with the central dogma. Extensive experiments show that Life-Code achieves state-of-the-art results on various tasks across three omics, highlighting its potential for advancing multi-omics analysis and interpretation.", "date": "2025-02-11", "categories": []}
{"id": "2506.04490", "title": "Multiscale guidance of protein structure prediction with heterogeneous cryo-EM data", "abstract": "Protein structure prediction models are now capable of generating accurate 3D structural hypotheses from sequence alone. However, they routinely fail to capture the conformational diversity of dynamic biomolecular complexes, often requiring heuristic MSA subsampling approaches for generating alternative states. In parallel, cryo-electron microscopy (cryo-EM) has emerged as a powerful tool for imaging near-native structural heterogeneity, but is challenged by arduous pipelines to transform raw experimental data into atomic models. Here, we bridge the gap between these modalities, combining cryo-EM density maps with the rich sequence and biophysical priors learned by protein structure prediction models. Our method, CryoBoltz, guides the sampling trajectory of a pretrained biomolecular structure prediction model using both global and local structural constraints derived from density maps, driving predictions towards conformational states consistent with the experimental data. We demonstrate that this flexible yet powerful inference-time approach allows us to build atomic models into heterogeneous cryo-EM maps across a variety of dynamic biomolecular systems including transporters and antibodies. Code is available at https://github.com/ml-struct-bio/cryoboltz .", "date": "2025-06-04", "categories": []}
{"id": "2512.00376", "title": "Layer Probing Improves Kinase Functional Prediction with Protein Language Models", "abstract": "Protein language models (PLMs) have transformed sequence-based protein analysis, yet most applications rely only on final-layer embeddings, which may overlook biologically meaningful information encoded in earlier layers. We systematically evaluate all 33 layers of ESM-2 for kinase functional prediction using both unsupervised clustering and supervised classification. We show that mid-to-late transformer layers (layers 20-33) outperform the final layer by 32 percent in unsupervised Adjusted Rand Index and improve homology-aware supervised accuracy to 75.7 percent. Domain-level extraction, calibrated probability estimates, and a reproducible benchmarking pipeline further strengthen reliability. Our results demonstrate that transformer depth contains functionally distinct biological signals and that principled layer selection significantly improves kinase function prediction.", "date": "2025-11-29", "categories": []}
{"id": "2512.00379", "title": "EnzyCLIP: A Cross-Attention Dual Encoder Framework with Contrastive Learning for Predicting Enzyme Kinetic Constants", "abstract": "Accurate prediction of enzyme kinetic parameters is crucial for drug discovery, metabolic engineering, and synthetic biology applications. Current computational approaches face limitations in capturing complex enzyme-substrate interactions and often focus on single parameters while neglecting the joint prediction of catalytic turnover numbers (Kcat) and Michaelis-Menten constants (Km). We present EnzyCLIP, a novel dual-encoder framework that leverages contrastive learning and cross-attention mechanisms to predict enzyme kinetic parameters from protein sequences and substrate molecular structures. Our approach integrates ESM-2 protein language model embeddings with ChemBERTa chemical representations through a CLIP-inspired architecture enhanced with bidirectional cross-attention for dynamic enzyme-substrate interaction modeling. EnzyCLIP combines InfoNCE contrastive loss with Huber regression loss to learn aligned multimodal representations while predicting log10-transformed kinetic parameters. The model is trained on the CatPred-DB database containing 23,151 Kcat and 41,174 Km experimentally validated measurements, and achieved competitive performance with R2 scores of 0.593 for Kcat and 0.607 for Km prediction. XGBoost ensemble methods applied to the learned embeddings further improved Km prediction (R2 = 0.61) while maintaining robust Kcat performance.", "date": "2025-11-29", "categories": []}
{"id": "2512.00708", "title": "Towards Precision Protein-Ligand Affinity Prediction Benchmark: A Complete and Modification-Aware DAVIS Dataset", "abstract": "Advancements in AI for science unlocks capabilities for critical drug discovery tasks such as protein-ligand binding affinity prediction. However, current models overfit to existing oversimplified datasets that does not represent naturally occurring and biologically relevant proteins with modifications. In this work, we curate a complete and modification-aware version of the widely used DAVIS dataset by incorporating 4,032 kinase-ligand pairs involving substitutions, insertions, deletions, and phosphorylation events. This enriched dataset enables benchmarking of predictive models under biologically realistic conditions. Based on this new dataset, we propose three benchmark settings-Augmented Dataset Prediction, Wild-Type to Modification Generalization, and Few-Shot Modification Generalization-designed to assess model robustness in the presence of protein modifications. Through extensive evaluation of both docking-free and docking-based methods, we find that docking-based model generalize better in zero-shot settings. In contrast, docking-free models tend to overfit to wild-type proteins and struggle with unseen modifications but show notable improvement when fine-tuned on a small set of modified examples. We anticipate that the curated dataset and benchmarks offer a valuable foundation for developing models that better generalize to protein modifications, ultimately advancing precision medicine in drug discovery. The benchmark is available at: https://github.com/ZhiGroup/DAVIS-complete", "date": "2025-11-30", "categories": []}
{"id": "2512.01199", "title": "Know Thyself by Knowing Others: Learning Neuron Identity from Population Context", "abstract": "Neurons process information in ways that depend on their cell type, connectivity, and the brain region in which they are embedded. However, inferring these factors from neural activity remains a significant challenge. To build general-purpose representations that allow for resolving information about a neuron's identity, we introduce NuCLR, a self-supervised framework that aims to learn representations of neural activity that allow for differentiating one neuron from the rest. NuCLR brings together views of the same neuron observed at different times and across different stimuli and uses a contrastive objective to pull these representations together. To capture population context without assuming any fixed neuron ordering, we build a spatiotemporal transformer that integrates activity in a permutation-equivariant manner. Across multiple electrophysiology and calcium imaging datasets, a linear decoding evaluation on top of NuCLR representations achieves a new state-of-the-art for both cell type and brain region decoding tasks, and demonstrates strong zero-shot generalization to unseen animals. We present the first systematic scaling analysis for neuron-level representation learning, showing that increasing the number of animals used during pretraining consistently improves downstream performance. The learned representations are also label-efficient, requiring only a small fraction of labeled samples to achieve competitive performance. These results highlight how large, diverse neural datasets enable models to recover information about neuron identity that generalize across animals. Code is available at https://github.com/nerdslab/nuclr.", "date": "2025-12-01", "categories": []}
{"id": "2403.01528", "title": "Leveraging Biomolecule and Natural Language through Multi-Modal Learning: A Survey", "abstract": "The integration of biomolecular modeling with natural language (BL) has emerged as a promising interdisciplinary area at the intersection of artificial intelligence, chemistry and biology. This approach leverages the rich, multifaceted descriptions of biomolecules contained within textual data sources to enhance our fundamental understanding and enable downstream computational tasks such as biomolecule property prediction. The fusion of the nuanced narratives expressed through natural language with the structural and functional specifics of biomolecules described via various molecular modeling techniques opens new avenues for comprehensively representing and analyzing biomolecules. By incorporating the contextual language data that surrounds biomolecules into their modeling, BL aims to capture a holistic view encompassing both the symbolic qualities conveyed through language as well as quantitative structural characteristics. In this review, we provide an extensive analysis of recent advancements achieved through cross modeling of biomolecules and natural language. (1) We begin by outlining the technical representations of biomolecules employed, including sequences, 2D graphs, and 3D structures. (2) We then examine in depth the rationale and key objectives underlying effective multi-modal integration of language and molecular data sources. (3) We subsequently survey the practical applications enabled to date in this developing research area. (4) We also compile and summarize the available resources and datasets to facilitate future work. (5) Looking ahead, we identify several promising research directions worthy of further exploration and investment to continue advancing the field. The related resources and contents are updating in https://github.com/QizhiPei/Awesome-Biomolecule-Language-Cross-Modeling.", "date": "2024-03-03", "categories": []}
{"id": "2503.02058", "title": "RiboGen: RNA Sequence and Structure Co-Generation with Equivariant MultiFlow", "abstract": "Ribonucleic acid (RNA) plays fundamental roles in biological systems, from carrying genetic information to performing enzymatic function. Understanding and designing RNA can enable novel therapeutic application and biotechnological innovation. To enhance RNA design, in this paper we introduce RiboGen, the first deep learning model to simultaneously generate RNA sequence and all-atom 3D structure. RiboGen leverages the standard Flow Matching with Discrete Flow Matching in a multimodal data representation. RiboGen is based on Euclidean Equivariant neural networks for efficiently processing and learning three-dimensional geometry. Our experiments show that RiboGen can efficiently generate chemically plausible and self-consistent RNA samples, suggesting that co-generation of sequence and structure is a competitive approach for modeling RNA.", "date": "2025-03-03", "categories": []}
{"id": "2509.02196", "title": "Beyond Ensembles: Simulating All-Atom Protein Dynamics in a Learned Latent Space", "abstract": "Simulating the long-timescale dynamics of biomolecules is a central challenge in computational science. While enhanced sampling methods can accelerate these simulations, they rely on pre-defined collective variables that are often difficult to identify, restricting their ability to model complex switching mechanisms between metastable states. A recent generative model, LD-FPG, demonstrated that this problem could be bypassed by learning to sample the static equilibrium ensemble as all-atom deformations from a reference structure, establishing a powerful method for all-atom ensemble generation. However, while this approach successfully captures a system's probable conformations, it does not model the temporal evolution between them. We introduce the Graph Latent Dynamics Propagator (GLDP), a modular component for simulating dynamics within the learned latent space of LD-FPG. We then compare three classes of propagators: (i) score-guided Langevin dynamics, (ii) Koopman-based linear operators, and (iii) autoregressive neural networks. Within a unified encoder-propagator-decoder framework, we evaluate long-horizon stability, backbone and side-chain ensemble fidelity, and temporal kinetics via TICA. Benchmarks on systems ranging from small peptides to mixed-topology proteins and large GPCRs reveal that autoregressive neural networks deliver the most robust long rollouts and coherent physical timescales; score-guided Langevin best recovers side-chain thermodynamics when the score is well learned; and Koopman provides an interpretable, lightweight baseline that tends to damp fluctuations. These results clarify the trade-offs among propagators and offer practical guidance for latent-space simulators of all-atom protein dynamics.", "date": "2025-09-02", "categories": []}
{"id": "2511.21781", "title": "BeeRNA: tertiary structure-based RNA inverse folding using Artificial Bee Colony", "abstract": "The Ribonucleic Acid (RNA) inverse folding problem, designing nucleotide sequences that fold into specific tertiary structures, is a fundamental computational biology problem with important applications in synthetic biology and bioengineering. The design of complex three-dimensional RNA architectures remains computationally demanding and mostly unresolved, as most existing approaches focus on secondary structures. In order to address tertiary RNA inverse folding, we present BeeRNA, a bio-inspired method that employs the Artificial Bee Colony (ABC) optimization algorithm. Our approach combines base-pair distance filtering with RMSD-based structural assessment using RhoFold for structure prediction, resulting in a two-stage fitness evaluation strategy. To guarantee biologically plausible sequences with balanced GC content, the algorithm takes thermodynamic constraints and adaptive mutation rates into consideration. In this work, we focus primarily on short and medium-length RNAs ($<$ 100 nucleotides), a biologically significant regime that includes microRNAs (miRNAs), aptamers, and ribozymes, where BeeRNA achieves high structural fidelity with practical CPU runtimes. The lightweight, training-free implementation will be publicly released for reproducibility, offering a promising bio-inspired approach for RNA design in therapeutics and biotechnology.", "date": "2025-11-26", "categories": []}
{"id": "2511.22292", "title": "Adaptive tumor growth forecasting via neural & universal ODEs", "abstract": "Forecasting tumor growth is critical for optimizing treatment. Classical growth models such as the Gompertz and Bertalanffy equations capture general tumor dynamics but may fail to adapt to patient-specific variability, particularly with limited data available. In this study, we leverage Neural Ordinary Differential Equations (Neural ODEs) and Universal Differential Equations (UDEs), two pillars of Scientific Machine Learning (SciML), to construct adaptive tumor growth models capable of learning from experimental data. Using the Gompertz model as a baseline, we replace rigid terms with adaptive neural networks to capture hidden dynamics through robust modeling in the Julia programming language. We use our models to perform forecasting under data constraints and symbolic recovery to transform the learned dynamics into explicit mathematical expressions. Our approach has the potential to improve predictive accuracy, guiding dynamic and effective treatment strategies for improved clinical outcomes.", "date": "2025-11-27", "categories": []}
{"id": "2511.22735", "title": "Integrated Transcriptomic-proteomic Biomarker Identification for Radiation Response Prediction in Non-small Cell Lung Cancer Cell Lines", "abstract": "To develop an integrated transcriptome-proteome framework for identifying concurrent biomarkers predictive of radiation response, as measured by survival fraction at 2 Gy (SF2), in non-small cell lung cancer (NSCLC) cell lines. RNA sequencing (RNA-seq) and data-independent acquisition mass spectrometry (DIA-MS) proteomic data were collected from 73 and 46 NSCLC cell lines, respectively. Following preprocessing, 1,605 shared genes were retained for analysis. Feature selection was performed using least absolute shrinkage and selection operator (Lasso) regression with a frequency-based ranking criterion under five-fold cross-validation repeated ten times. Support vector regression (SVR) models were constructed using transcriptome-only, proteome-only, and combined transcriptome-proteome feature sets. Model performance was assessed by the coefficient of determination (R2) and root mean square error (RMSE). Correlation analyses evaluated concordance between RNA and protein expression and the relationships of selected biomarkers with SF2. RNA-protein expression exhibited significant positive correlations (median Pearson's r = 0.363). Independent pipelines identified 20 prioritized gene signatures from transcriptomic, proteomic, and combined datasets. Models trained on single-omic features achieved limited cross-omic generalizability, while the combined model demonstrated balanced predictive accuracy in both datasets (R2=0.461, RMSE=0.120 for transcriptome; R2=0.604, RMSE=0.111 for proteome). This study presents the first proteotranscriptomic framework for SF2 prediction in NSCLC, highlighting the complementary value of integrating transcriptomic and proteomic data. The identified concurrent biomarkers capture both transcriptional regulation and functional protein activity, offering mechanistic insights and translational potential.", "date": "2025-11-27", "categories": []}
{"id": "2511.00209", "title": "Diffusion Models at the Drug Discovery Frontier: A Review on Generating Small Molecules versus Therapeutic Peptides", "abstract": "Diffusion models have emerged as a leading framework in generative modeling, poised to transform the traditionally slow and costly process of drug discovery. This review provides a systematic comparison of their application in designing two principal therapeutic modalities: small molecules and therapeutic peptides. We dissect how the unified framework of iterative denoising is adapted to the distinct molecular representations, chemical spaces, and design objectives of each modality. For small molecules, these models excel at structure-based design, generating novel, pocket-fitting ligands with desired physicochemical properties, yet face the critical hurdle of ensuring chemical synthesizability. Conversely, for therapeutic peptides, the focus shifts to generating functional sequences and designing de novo structures, where the primary challenges are achieving biological stability against proteolysis, ensuring proper folding, and minimizing immunogenicity. Despite these distinct challenges, both domains face shared hurdles: the scarcity of high-quality experimental data, the reliance on inaccurate scoring functions for validation, and the crucial need for experimental validation. We conclude that the full potential of diffusion models will be unlocked by bridging these modality-specific gaps and integrating them into automated, closed-loop Design-Build-Test-Learn (DBTL) platforms, thereby shifting the paradigm from mere chemical exploration to the on-demand engineering of novel~therapeutics.", "date": "2025-10-31", "categories": []}
{"id": "2511.12489", "title": "SculptDrug : A Spatial Condition-Aware Bayesian Flow Model for Structure-based Drug Design", "abstract": "Structure-Based drug design (SBDD) has emerged as a popular approach in drug discovery, leveraging three-dimensional protein structures to generate drug ligands. However, existing generative models encounter several key challenges: (1) incorporating boundary condition constraints, (2) integrating hierarchical structural conditions, and (3) ensuring spatial modeling fidelity. To address these limitations, we propose SculptDrug, a spatial condition-aware generative model based on Bayesian flow networks (BFNs). First, SculptDrug follows a BFN-based framework and employs a progressive denoising strategy to ensure spatial modeling fidelity, iteratively refining atom positions while enhancing local interactions for precise spatial alignment. Second, we introduce a Boundary Awareness Block that incorporates protein surface constraints into the generative process to ensure that generated ligands are geometrically compatible with the target protein. Third, we design a Hierarchical Encoder that captures global structural context while preserving fine-grained molecular interactions, ensuring overall consistency and accurate ligand-protein conformations. We evaluate SculptDrug on the CrossDocked dataset, and experimental results demonstrate that SculptDrug outperforms state-of-the-art baselines, highlighting the effectiveness of spatial condition-aware modeling.", "date": "2025-11-16", "categories": []}
{"id": "2511.20510", "title": "FRAGMENTA: End-to-end Fragmentation-based Generative Model with Agentic Tuning for Drug Lead Optimization", "abstract": "Molecule generation using generative AI is vital for drug discovery, yet class-specific datasets often contain fewer than 100 training examples. While fragment-based models handle limited data better than atom-based approaches, existing heuristic fragmentation limits diversity and misses key fragments. Additionally, model tuning typically requires slow, indirect collaboration between medicinal chemists and AI engineers. We introduce FRAGMENTA, an end-to-end framework for drug lead optimization comprising: 1) a novel generative model that reframes fragmentation as a \"vocabulary selection\" problem, using dynamic Q-learning to jointly optimize fragmentation and generation; and 2) an agentic AI system that refines objectives via conversational feedback from domain experts. This system removes the AI engineer from the loop and progressively learns domain knowledge to eventually automate tuning. In real-world cancer drug discovery experiments, FRAGMENTA's Human-Agent configuration identified nearly twice as many high-scoring molecules as baselines. Furthermore, the fully autonomous Agent-Agent system outperformed traditional Human-Human tuning, demonstrating the efficacy of agentic tuning in capturing expert intent.", "date": "2025-11-25", "categories": []}
{"id": "2511.21566", "title": "A decoupled alignment kernel for peptide membrane permeability predictions", "abstract": "Cyclic peptides are promising modalities for targeting intracellular sites; however, cell-membrane permeability remains a key bottleneck, exacerbated by limited public data and the need for well-calibrated uncertainty. Instead of relying on data-eager complex deep learning architecture, we propose a monomer-aware decoupled global alignment kernel (MD-GAK), which couples chemically meaningful residue-residue similarity with sequence alignment while decoupling local matches from gap penalties. MD-GAK is a relatively simple kernel. To further demonstrate the robustness of our framework, we also introduce a variant, PMD-GAK, which incorporates a triangular positional prior. As we will show in the experimental section, PMD-GAK can offer additional advantages over MD-GAK, particularly in reducing calibration errors. Since our focus is on uncertainty estimation, we use Gaussian Processes as the predictive model, as both MD-GAK and PMD-GAK can be directly applied within this framework. We demonstrate the effectiveness of our methods through an extensive set of experiments, comparing our fully reproducible approach against state-of-the-art models, and show that it outperforms them across all metrics.", "date": "2025-11-26", "categories": []}
{"id": "2503.01347", "title": "From Spots to Pixels: Dense Spatial Gene Expression Prediction from Histology Images", "abstract": "Spatial transcriptomics (ST) measures gene expression at fine-grained spatial resolution, offering insights into tissue molecular landscapes. Previous methods for spatial gene expression prediction typically crop spots of interest from histopathology slide images, and train models to map each spot to a corresponding gene expression profile. However, these methods inherently lose the spatial resolution in gene expression: 1) each spot often contains multiple cells with distinct gene expression profiles; 2) spots are typically defined at fixed spatial resolutions, limiting the ability to predict gene expression at varying scales. To address these limitations, this paper presents PixNet, a dense prediction network capable of predicting spatially resolved gene expression across spots of varying sizes and scales directly from histopathology slide images. Different from previous methods that map individual spots to gene expression values, we generate a spatially dense continuous gene expression map from the histopathology slide image, and aggregate values within spots of interest to predict the gene expression. Our PixNet outperforms state-of-the-art methods on four common ST datasets in multiple spatial scales. The source code will be publicly available.", "date": "2025-03-03", "categories": []}
{"id": "2511.14613", "title": "3D-Guided Scalable Flow Matching for Generating Volumetric Tissue Spatial Transcriptomics from Serial Histology", "abstract": "A scalable and robust 3D tissue transcriptomics profile can enable a holistic understanding of tissue organization and provide deeper insights into human biology and disease. Most predictive algorithms that infer ST directly from histology treat each section independently and ignore 3D structure, while existing 3D-aware approaches are not generative and do not scale well. We present Holographic Tissue Expression Inpainting and Analysis (HoloTea), a 3D-aware flow-matching framework that imputes spot-level gene expression from H&E while explicitly using information from adjacent sections. Our key idea is to retrieve morphologically corresponding spots on neighboring slides in a shared feature space and fuse this cross section context into a lightweight ControlNet, allowing conditioning to follow anatomical continuity. To better capture the count nature of the data, we introduce a 3D-consistent prior for flow matching that combines a learned zero-inflated negative binomial (ZINB) prior with a spatial-empirical prior constructed from neighboring sections. A global attention block introduces 3D H&E scaling linearly with the number of spots in the slide, enabling training and inference on large 3D ST datasets. Across three spatial transcriptomics datasets spanning different tissue types and resolutions, HoloTea consistently improves 3D expression accuracy and generalization compared to 2D and 3D baselines. We envision HoloTea advancing the creation of accurate 3D virtual tissues, ultimately accelerating biomarker discovery and deepening our understanding of disease.", "date": "2025-11-18", "categories": []}
{"id": "2506.10879", "title": "A Goemans-Williamson type algorithm for identifying subcohorts in clinical trials", "abstract": "We design an efficient algorithm that outputs tests for identifying predominantly homogeneous subcohorts of patients from large in-homogeneous datasets. Our theoretical contribution is a rounding technique, similar to that of Goemans and Wiliamson (1995), that approximates the optimal solution within a factor of $0.82$. As an application, we use our algorithm to trade-off sensitivity for specificity to systematically identify clinically interesting homogeneous subcohorts of patients in the RNA microarray dataset for breast cancer from Curtis et al. (2012). One such clinically interesting subcohort suggests a link between LXR over-expression and BRCA2 and MSH6 methylation levels for patients in that subcohort.", "date": "2025-06-12", "categories": []}
{"id": "2507.04981", "title": "Classification of autoimmune diseases from Peripheral blood TCR repertoires by multimodal multi-instance learning", "abstract": "T cell receptor (TCR) repertoires encode critical immunological signatures for autoimmune diseases, yet their clinical application remains limited by sequence sparsity and low witness rates. We developed EAMil, a multi-instance deep learning framework that leverages TCR sequencing data to diagnose systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) with exceptional accuracy. By integrating PrimeSeq feature extraction with ESMonehot encoding and enhanced gate attention mechanisms, our model achieved state-of-the-art performance with AUCs of 98.95% for SLE and 97.76% for RA. EAMil successfully identified disease-associated genes with over 90% concordance with established differential analyses and effectively distinguished disease-specific TCR genes. The model demonstrated robustness in classifying multiple disease categories, utilizing the SLEDAI score to stratify SLE patients by disease severity as well as to diagnose the site of damage in SLE patients, and effectively controlling for confounding factors such as age and gender. This interpretable framework for immune receptor analysis provides new insights for autoimmune disease detection and classification with broad potential clinical applications across immune-mediated conditions.", "date": "2025-07-07", "categories": []}
{"id": "2508.00578", "title": "Learning Potential Energy Surfaces of Hydrogen Atom Transfer Reactions in Peptides", "abstract": "Hydrogen atom transfer (HAT) reactions are essential in many biological processes, such as radical migration in damaged proteins, but their mechanistic pathways remain incompletely understood. Simulating HAT is challenging due to the need for quantum chemical accuracy at biologically relevant scales; thus, neither classical force fields nor DFT-based molecular dynamics are applicable. Machine-learned potentials offer an alternative, able to learn potential energy surfaces (PESs) with near-quantum accuracy. However, training these models to generalize across diverse HAT configurations, especially at radical positions in proteins, requires tailored data generation and careful model selection. Here, we systematically generate HAT configurations in peptides to build large datasets using semiempirical methods and DFT. We benchmark three graph neural network architectures (SchNet, Allegro, and MACE) on their ability to learn HAT PESs and indirectly predict reaction barriers from energy predictions. MACE consistently outperforms the others in energy, force, and barrier prediction, achieving a mean absolute error of 1.13 kcal/mol on out-of-distribution DFT barrier predictions. Using molecular dynamics, we show our MACE potential is stable, reactive, and generalizes beyond training data to model HAT barriers in collagen I. This accuracy enables integration of ML potentials into large-scale collagen simulations to compute reaction rates from predicted barriers, advancing mechanistic understanding of HAT and radical migration in peptides. We analyze scaling laws, model transferability, and cost-performance trade-offs, and outline strategies for improvement by combining ML potentials with transition state search algorithms and active learning. Our approach is generalizable to other biomolecular systems, enabling quantum-accurate simulations of chemical reactivity in complex environments.", "date": "2025-08-01", "categories": []}
{"id": "2508.08312", "title": "CFM-GP: Unified Conditional Flow Matching to Learn Gene Perturbation Across Cell Types", "abstract": "Understanding gene perturbation effects across diverse cellular contexts is a central challenge in functional genomics, with important implications for therapeutic discovery and precision medicine. Single-cell technologies enable high-resolution measurement of transcriptional responses, but collecting such data is costly and time-consuming, especially when repeated for each cell type. Existing computational methods often require separate models per cell type, limiting scalability and generalization. We present CFM-GP, a method for cell type-agnostic gene perturbation prediction. CFM-GP learns a continuous, time-dependent transformation between unperturbed and perturbed gene expression distributions, conditioned on cell type, allowing a single model to predict across all cell types. Unlike prior approaches that use discrete modeling, CFM-GP employs a flow matching objective to capture perturbation dynamics in a scalable manner. We evaluate on five datasets: SARS-CoV-2 infection, IFN-beta stimulated PBMCs, glioblastoma treated with Panobinostat, lupus under IFN-beta stimulation, and Statefate progenitor fate mapping. CFM-GP consistently outperforms state-of-the-art baselines in R-squared and Spearman correlation, and pathway enrichment analysis confirms recovery of key biological pathways. These results demonstrate the robustness and biological fidelity of CFM-GP as a scalable solution for cross-cell type gene perturbation prediction.", "date": "2025-08-09", "categories": []}
{"id": "2508.15480", "title": "Learning Protein-Ligand Binding in Hyperbolic Space", "abstract": "Protein-ligand binding prediction is central to virtual screening and affinity ranking, two fundamental tasks in drug discovery. While recent retrieval-based methods embed ligands and protein pockets into Euclidean space for similarity-based search, the geometry of Euclidean embeddings often fails to capture the hierarchical structure and fine-grained affinity variations intrinsic to molecular interactions. In this work, we propose HypSeek, a hyperbolic representation learning framework that embeds ligands, protein pockets, and sequences into Lorentz-model hyperbolic space. By leveraging the exponential geometry and negative curvature of hyperbolic space, HypSeek enables expressive, affinity-sensitive embeddings that can effectively model both global activity and subtle functional differences-particularly in challenging cases such as activity cliffs, where structurally similar ligands exhibit large affinity gaps. Our mode unifies virtual screening and affinity ranking in a single framework, introducing a protein-guided three-tower architecture to enhance representational structure. HypSeek improves early enrichment in virtual screening on DUD-E from 42.63 to 51.44 (+20.7%) and affinity ranking correlation on JACS from 0.5774 to 0.7239 (+25.4%), demonstrating the benefits of hyperbolic geometry across both tasks and highlighting its potential as a powerful inductive bias for protein-ligand modeling.", "date": "2025-08-21", "categories": []}
{"id": "2511.07390", "title": "A Diffusion Model to Shrink Proteins While Maintaining Their Function", "abstract": "Many proteins useful in modern medicine or bioengineering are challenging to make in the lab, fuse with other proteins in cells, or deliver to tissues in the body, because their sequences are too long. Shortening these sequences typically involves costly, time-consuming experimental campaigns. Ideally, we could instead use modern models of massive databases of sequences from nature to learn how to propose shrunken proteins that resemble sequences found in nature. Unfortunately, these models struggle to efficiently search the combinatorial space of all deletions, and are not trained with inductive biases to learn how to delete. To address this gap, we propose SCISOR, a novel discrete diffusion model that deletes letters from sequences to generate protein samples that resemble those found in nature. To do so, SCISOR trains a de-noiser to reverse a forward noising process that adds random insertions to natural sequences. As a generative model, SCISOR fits evolutionary sequence data competitively with previous large models. In evaluation, SCISOR achieves state-of-the-art predictions of the functional effects of deletions on ProteinGym. Finally, we use the SCISOR de-noiser to shrink long protein sequences, and show that its suggested deletions result in significantly more realistic proteins and more often preserve functional motifs than previous models of evolutionary sequences.", "date": "2025-11-10", "categories": []}
{"id": "2511.18287", "title": "TRIDENT: A Trimodal Cascade Generative Framework for Drug and RNA-Conditioned Cellular Morphology Synthesis", "abstract": "Accurately modeling the relationship between perturbations, transcriptional responses, and phenotypic changes is essential for building an AI Virtual Cell (AIVC). However, existing methods typically constrained to modeling direct associations, such as Perturbation $\\rightarrow$ RNA or Perturbation $\\rightarrow$ Morphology, overlook the crucial causal link from RNA to morphology. To bridge this gap, we propose TRIDENT, a cascade generative framework that synthesizes realistic cellular morphology by conditioning on both the perturbation and the corresponding gene expression profile. To train and evaluate this task, we construct MorphoGene, a new dataset pairing L1000 gene expression with Cell Painting images for 98 compounds. TRIDENT significantly outperforms state-of-the-art approaches, achieving up to 7-fold improvement with strong generalization to unseen compounds. In a case study on docetaxel, we validate that RNA-guided synthesis accurately produces the corresponding phenotype. An ablation study further confirms that this RNA conditioning is essential for the model's high fidelity. By explicitly modeling transcriptome-phenome mapping, TRIDENT provides a powerful in silico tool and moves us closer to a predictive virtual cell.", "date": "2025-11-23", "categories": []}
{"id": "2511.18336", "title": "Auxiliary Gene Learning: Spatial Gene Expression Estimation by Auxiliary Gene Selection", "abstract": "Spatial transcriptomics (ST) is a novel technology that enables the observation of gene expression at the resolution of individual spots within pathological tissues. ST quantifies the expression of tens of thousands of genes in a tissue section; however, heavy observational noise is often introduced during measurement. In prior studies, to ensure meaningful assessment, both training and evaluation have been restricted to only a small subset of highly variable genes, and genes outside this subset have also been excluded from the training process. However, since there are likely co-expression relationships between genes, low-expression genes may still contribute to the estimation of the evaluation target. In this paper, we propose $Auxiliary \\ Gene \\ Learning$ (AGL) that utilizes the benefit of the ignored genes by reformulating their expression estimation as auxiliary tasks and training them jointly with the primary tasks. To effectively leverage auxiliary genes, we must select a subset of auxiliary genes that positively influence the prediction of the target genes. However, this is a challenging optimization problem due to the vast number of possible combinations. To overcome this challenge, we propose Prior-Knowledge-Based Differentiable Top-$k$ Gene Selection via Bi-level Optimization (DkGSB), a method that ranks genes by leveraging prior knowledge and relaxes the combinatorial selection problem into a differentiable top-$k$ selection problem. The experiments confirm the effectiveness of incorporating auxiliary genes and show that the proposed method outperforms conventional auxiliary task learning approaches.", "date": "2025-11-23", "categories": []}
{"id": "2511.16912", "title": "PepEVOLVE: Position-Aware Dynamic Peptide Optimization via Group-Relative Advantage", "abstract": "Macrocyclic peptides are an emerging modality that combines biologics-like affinity with small-molecule-like developability, but their vast combinatorial space and multi-parameter objectives make lead optimization slow and challenging. Prior generative approaches such as PepINVENT require chemists to pre-specify mutable positions for optimization, choices that are not always known a priori, and rely on static pretraining and optimization algorithms that limit the model's ability to generalize and effectively optimize peptide sequences. We introduce PepEVOLVE, a position-aware, dynamic framework that learns both where to edit and how to dynamically optimize peptides for multi-objective improvement. PepEVOLVE (i) augments pretraining with dynamic masking and CHUCKLES shifting to improve generalization, (ii) uses a context-free multi-armed bandit router that discovers high-reward residues, and (iii) couples a novel evolving optimization algorithm with group-relative advantage to stabilize reinforcement updates. During in silico evaluations, the router policy reliably learns and concentrates probability on chemically meaningful sites that influence the peptide's properties. On a therapeutically motivated Rev-binding macrocycle benchmark, PepEVOLVE outperformed PepINVENT by reaching higher mean scores (approximately 0.8 vs. 0.6), achieving best candidates with a score of 0.95 (vs. 0.87), and converging in fewer steps under the task of optimizing permeability and lipophilicity with structural constraints. Overall, PepEVOLVE offers a practical, reproducible path to peptide lead optimization when optimal edit sites are unknown, enabling more efficient exploration and improving design quality across multiple objectives.", "date": "2025-11-21", "categories": []}
{"id": "2511.16923", "title": "A Hybrid Computational Intelligence Framework for scRNA-seq Imputation: Integrating scRecover and Random Forests", "abstract": "Single-cell RNA sequencing (scRNA-seq) enables transcriptomic profiling at cellular resolution but suffers from pervasive dropout events that obscure biological signals. We present SCR-MF, a modular two-stage workflow that combines principled dropout detection using scRecover with robust non-parametric imputation via missForest. Across public and simulated datasets, SCR-MF achieves robust and interpretable performance comparable to or exceeding existing imputation methods in most cases, while preserving biological fidelity and transparency. Runtime analysis demonstrates that SCR-MF provides a competitive balance between accuracy and computational efficiency, making it suitable for mid-scale single-cell datasets.", "date": "2025-11-21", "categories": []}
{"id": "2511.17078", "title": "Hash Collisions in Molecular Fingerprints: Effects on Property Prediction and Bayesian Optimization", "abstract": "Molecular fingerprinting methods use hash functions to create fixed-length vector representations of molecules. However, hash collisions cause distinct substructures to be represented with the same feature, leading to overestimates in molecular similarity calculations. We investigate whether using exact fingerprints improves accuracy compared to standard compressed fingerprints in molecular property prediction and Bayesian optimization where the underlying predictive model is a Gaussian process. We find that using exact fingerprints yields a small yet consistent improvement in predictive accuracy on five molecular property prediction benchmarks from the DOCKSTRING dataset. However, these gains did not translate to significant improvements in Bayesian optimization performance.", "date": "2025-11-21", "categories": []}
{"id": "2511.15906", "title": "Unified all-atom molecule generation with neural fields", "abstract": "Generative models for structure-based drug design are often limited to a specific modality, restricting their broader applicability. To address this challenge, we introduce FuncBind, a framework based on computer vision to generate target-conditioned, all-atom molecules across atomic systems. FuncBind uses neural fields to represent molecules as continuous atomic densities and employs score-based generative models with modern architectures adapted from the computer vision literature. This modality-agnostic representation allows a single unified model to be trained on diverse atomic systems, from small to large molecules, and handle variable atom/residue counts, including non-canonical amino acids. FuncBind achieves competitive in silico performance in generating small molecules, macrocyclic peptides, and antibody complementarity-determining region loops, conditioned on target structures. FuncBind also generated in vitro novel antibody binders via de novo redesign of the complementarity-determining region H3 loop of two chosen co-crystal structures. As a final contribution, we introduce a new dataset and benchmark for structure-conditioned macrocyclic peptide generation. The code is available at https://github.com/prescient-design/funcbind.", "date": "2025-11-19", "categories": []}
{"id": "2511.12969", "title": "HiFusion: Hierarchical Intra-Spot Alignment and Regional Context Fusion for Spatial Gene Expression Prediction from Histopathology", "abstract": "Spatial transcriptomics (ST) bridges gene expression and tissue morphology but faces clinical adoption barriers due to technical complexity and prohibitive costs. While computational methods predict gene expression from H&E-stained whole-slide images (WSIs), existing approaches often fail to capture the intricate biological heterogeneity within spots and are susceptible to morphological noise when integrating contextual information from surrounding tissue. To overcome these limitations, we propose HiFusion, a novel deep learning framework that integrates two complementary components. First, we introduce the Hierarchical Intra-Spot Modeling module that extracts fine-grained morphological representations through multi-resolution sub-patch decomposition, guided by a feature alignment loss to ensure semantic consistency across scales. Concurrently, we present the Context-aware Cross-scale Fusion module, which employs cross-attention to selectively incorporate biologically relevant regional context, thereby enhancing representational capacity. This architecture enables comprehensive modeling of both cellular-level features and tissue microenvironmental cues, which are essential for accurate gene expression prediction. Extensive experiments on two benchmark ST datasets demonstrate that HiFusion achieves state-of-the-art performance across both 2D slide-wise cross-validation and more challenging 3D sample-specific scenarios. These results underscore HiFusion's potential as a robust, accurate, and scalable solution for ST inference from routine histopathology.", "date": "2025-11-17", "categories": []}
{"id": "2511.14516", "title": "Full-Atom Peptide Design via Riemannian-Euclidean Bayesian Flow Networks", "abstract": "Diffusion and flow matching models have recently emerged as promising approaches for peptide binder design. Despite their progress, these models still face two major challenges. First, categorical sampling of discrete residue types collapses their continuous parameters into onehot assignments, while continuous variables (e.g., atom positions) evolve smoothly throughout the generation process. This mismatch disrupts the update dynamics and results in suboptimal performance. Second, current models assume unimodal distributions for side-chain torsion angles, which conflicts with the inherently multimodal nature of side chain rotameric states and limits prediction accuracy. To address these limitations, we introduce PepBFN, the first Bayesian flow network for full atom peptide design that directly models parameter distributions in fully continuous space. Specifically, PepBFN models discrete residue types by learning their continuous parameter distributions, enabling joint and smooth Bayesian updates with other continuous structural parameters. It further employs a novel Gaussian mixture based Bayesian flow to capture the multimodal side chain rotameric states and a Matrix Fisher based Riemannian flow to directly model residue orientations on the $\\mathrm{SO}(3)$ manifold. Together, these parameter distributions are progressively refined via Bayesian updates, yielding smooth and coherent peptide generation. Experiments on side chain packing, reverse folding, and binder design tasks demonstrate the strong potential of PepBFN in computational peptide design.", "date": "2025-11-18", "categories": []}
{"id": "2511.14781", "title": "Quantifying the Role of OpenFold Components in Protein Structure Prediction", "abstract": "Models such as AlphaFold2 and OpenFold have transformed protein structure prediction, yet their inner workings remain poorly understood. We present a methodology to systematically evaluate the contribution of individual OpenFold components to structure prediction accuracy. We identify several components that are critical for most proteins, while others vary in importance across proteins. We further show that the contribution of several components is correlated with protein length. These findings provide insight into how OpenFold achieves accurate predictions and highlight directions for interpreting protein prediction networks more broadly.", "date": "2025-11-06", "categories": []}
{"id": "2511.15139", "title": "CASPER: Cross-modal Alignment of Spatial and single-cell Profiles for Expression Recovery", "abstract": "Spatial Transcriptomics enables mapping of gene expression within its native tissue context, but current platforms measure only a limited set of genes due to experimental constraints and excessive costs. To overcome this, computational models integrate Single-Cell RNA Sequencing data with Spatial Transcriptomics to predict unmeasured genes. We propose CASPER, a cross-attention based framework that predicts unmeasured gene expression in Spatial Transcriptomics by leveraging centroid-level representations from Single-Cell RNA Sequencing. We performed rigorous testing over four state-of-the-art Spatial Transcriptomics/Single-Cell RNA Sequencing dataset pairs across four existing baseline models. CASPER shows significant improvement in nine out of the twelve metrics for our experiments. This work paves the way for further work in Spatial Transcriptomics to Single-Cell RNA Sequencing modality translation. The code for CASPER is available at https://github.com/AI4Med-Lab/CASPER.", "date": "2025-11-19", "categories": []}
{"id": "2503.09649", "title": "Technical and Legal Aspects of Federated Learning in Bioinformatics: Applications, Challenges and Opportunities", "abstract": "Federated learning leverages data across institutions to improve clinical discovery while complying with data-sharing restrictions and protecting patient privacy. This paper provides a gentle introduction to this approach in bioinformatics, and is the first to review key applications in proteomics, genome-wide association studies (GWAS), single-cell and multi-omics studies in their legal as well as methodological and infrastructural challenges. As the evolution of biobanks in genetics and systems biology has proved, accessing more extensive and varied data pools leads to a faster and more robust exploration and translation of results. More widespread use of federated learning may have a similar impact in bioinformatics, allowing academic and clinical institutions to access many combinations of genotypic, phenotypic and environmental information that are undercovered or not included in existing biobanks.", "date": "2025-03-12", "categories": []}
{"id": "2511.13762", "title": "Gene Incremental Learning for Single-Cell Transcriptomics", "abstract": "Classes, as fundamental elements of Computer Vision, have been extensively studied within incremental learning frameworks. In contrast, tokens, which play essential roles in many research fields, exhibit similar characteristics of growth, yet investigations into their incremental learning remain significantly scarce. This research gap primarily stems from the holistic nature of tokens in language, which imposes significant challenges on the design of incremental learning frameworks for them. To overcome this obstacle, in this work, we turn to a type of token, gene, for a large-scale biological dataset--single-cell transcriptomics--to formulate a pipeline for gene incremental learning and establish corresponding evaluations. We found that the forgetting problem also exists in gene incremental learning, thus we adapted existing class incremental learning methods to mitigate the forgetting of genes. Through extensive experiments, we demonstrated the soundness of our framework design and evaluations, as well as the effectiveness of our method adaptations. Finally, we provide a complete benchmark for gene incremental learning in single-cell transcriptomics.", "date": "2025-11-14", "categories": []}
{"id": "2511.13786", "title": "CellStream: Dynamical Optimal Transport Informed Embeddings for Reconstructing Cellular Trajectories from Snapshots Data", "abstract": "Single-cell RNA sequencing (scRNA-seq), especially temporally resolved datasets, enables genome-wide profiling of gene expression dynamics at single-cell resolution across discrete time points. However, current technologies provide only sparse, static snapshots of cell states and are inherently influenced by technical noise, complicating the inference and representation of continuous transcriptional dynamics. Although embedding methods can reduce dimensionality and mitigate technical noise, the majority of existing approaches typically treat trajectory inference separately from embedding construction, often neglecting temporal structure. To address this challenge, here we introduce CellStream, a novel deep learning framework that jointly learns embedding and cellular dynamics from single-cell snapshot data by integrating an autoencoder with unbalanced dynamical optimal transport. Compared to existing methods, CellStream generates dynamics-informed embeddings that robustly capture temporal developmental processes while maintaining high consistency with the underlying data manifold. We demonstrate CellStream's effectiveness on both simulated datasets and real scRNA-seq data, including spatial transcriptomics. Our experiments indicate significant quantitative improvements over state-of-the-art methods in representing cellular trajectories with enhanced temporal coherence and reduced noise sensitivity. Overall, CellStream provides a new tool for learning and representing continuous streams from the noisy, static snapshots of single-cell gene expression.", "date": "2025-11-16", "categories": []}
{"id": "2511.14559", "title": "Apo2Mol: 3D Molecule Generation via Dynamic Pocket-Aware Diffusion Models", "abstract": "Deep generative models are rapidly advancing structure-based drug design, offering substantial promise for generating small molecule ligands that bind to specific protein targets. However, most current approaches assume a rigid protein binding pocket, neglecting the intrinsic flexibility of proteins and the conformational rearrangements induced by ligand binding, limiting their applicability in practical drug discovery. Here, we propose Apo2Mol, a diffusion-based generative framework for 3D molecule design that explicitly accounts for conformational flexibility in protein binding pockets. To support this, we curate a dataset of over 24,000 experimentally resolved apo-holo structure pairs from the Protein Data Bank, enabling the characterization of protein structure changes associated with ligand binding. Apo2Mol employs a full-atom hierarchical graph-based diffusion model that simultaneously generates 3D ligand molecules and their corresponding holo pocket conformations from input apo states. Empirical studies demonstrate that Apo2Mol can achieve state-of-the-art performance in generating high-affinity ligands and accurately capture realistic protein pocket conformational changes.", "date": "2025-11-18", "categories": []}
{"id": "2511.14744", "title": "Measuring AI Progress in Drug Discovery: A Reproducible Leaderboard for the Tox21 Challenge", "abstract": "Deep learning's rise since the early 2010s has transformed fields like computer vision and natural language processing and strongly influenced biomedical research. For drug discovery specifically, a key inflection - akin to vision's \"ImageNet moment\" - arrived in 2015, when deep neural networks surpassed traditional approaches on the Tox21 Data Challenge. This milestone accelerated the adoption of deep learning across the pharmaceutical industry, and today most major companies have integrated these methods into their research pipelines. After the Tox21 Challenge concluded, its dataset was included in several established benchmarks, such as MoleculeNet and the Open Graph Benchmark. However, during these integrations, the dataset was altered and labels were imputed or manufactured, resulting in a loss of comparability across studies. Consequently, the extent to which bioactivity and toxicity prediction methods have improved over the past decade remains unclear. To this end, we introduce a reproducible leaderboard, hosted on Hugging Face with the original Tox21 Challenge dataset, together with a set of baseline and representative methods. The current version of the leaderboard indicates that the original Tox21 winner - the ensemble-based DeepTox method - and the descriptor-based self-normalizing neural networks introduced in 2017, continue to perform competitively and rank among the top methods for toxicity prediction, leaving it unclear whether substantial progress in toxicity prediction has been achieved over the past decade. As part of this work, we make all baselines and evaluated models publicly accessible for inference via standardized API calls to Hugging Face Spaces.", "date": "2025-11-18", "categories": []}
{"id": "2507.12574", "title": "Assay2Mol: large language model-based drug design using BioAssay context", "abstract": "Scientific databases aggregate vast amounts of quantitative data alongside descriptive text. In biochemistry, molecule screening assays evaluate candidate molecules' functional responses against disease targets. Unstructured text that describes the biological mechanisms through which these targets operate, experimental screening protocols, and other attributes of assays offer rich information for drug discovery campaigns but has been untapped because of that unstructured format. We present Assay2Mol, a large language model-based workflow that can capitalize on the vast existing biochemical screening assays for early-stage drug discovery. Assay2Mol retrieves existing assay records involving targets similar to the new target and generates candidate molecules using in-context learning with the retrieved assay screening data. Assay2Mol outperforms recent machine learning approaches that generate candidate ligand molecules for target protein structures, while also promoting more synthesizable molecule generation.", "date": "2025-07-16", "categories": []}
{"id": "2509.25225", "title": "MSCoD: An Enhanced Bayesian Updating Framework with Multi-Scale Information Bottleneck and Cooperative Attention for Structure-Based Drug Design", "abstract": "Structure-Based Drug Design (SBDD) is a powerful strategy in computational drug discovery, utilizing three-dimensional protein structures to guide the design of molecules with improved binding affinity. However, capturing complex protein-ligand interactions across multiple scales remains challenging, as current methods often overlook the hierarchical organization and intrinsic asymmetry of these interactions. To address these limitations, we propose MSCoD, a novel Bayesian updating-based generative framework for structure-based drug design. In our MSCoD, Multi-Scale Information Bottleneck (MSIB) was developed, which enables semantic compression at multiple abstraction levels for efficient hierarchical feature extraction. Furthermore, a multi-head cooperative attention (MHCA) mechanism was developed, which employs asymmetric protein-to-ligand attention to capture diverse interaction types while addressing the dimensionality disparity between proteins and ligands. Empirical studies showed that MSCoD outperforms state-of-the-art methods on the benchmark dataset. Its real-world applicability is confirmed by case studies on difficult targets like KRAS G12D (7XKJ). Additionally, the MSIB and MHCA modules prove transferable, boosting the performance of GraphDTA on standard drug target affinity prediction benchmarks (Davis and Kiba). The code and data underlying this article are freely available at https://github.com/xulong0826/MSCoD.", "date": "2025-09-24", "categories": []}
{"id": "2511.09512", "title": "GenePheno: Interpretable Gene Knockout-Induced Phenotype Abnormality Prediction from Gene Sequences", "abstract": "Exploring how genetic sequences shape phenotypes is a fundamental challenge in biology and a key step toward scalable, hypothesis-driven experimentation. The task is complicated by the large modality gap between sequences and phenotypes, as well as the pleiotropic nature of gene-phenotype relationships. Existing sequence-based efforts focus on the degree to which variants of specific genes alter a limited set of phenotypes, while general gene knockout induced phenotype abnormality prediction methods heavily rely on curated genetic information as inputs, which limits scalability and generalizability. As a result, the task of broadly predicting the presence of multiple phenotype abnormalities under gene knockout directly from gene sequences remains underexplored. We introduce GenePheno, the first interpretable multi-label prediction framework that predicts knockout induced phenotypic abnormalities from gene sequences. GenePheno employs a contrastive multi-label learning objective that captures inter-phenotype correlations, complemented by an exclusive regularization that enforces biological consistency. It further incorporates a gene function bottleneck layer, offering human interpretable concepts that reflect functional mechanisms behind phenotype formation. To support progress in this area, we curate four datasets with canonical gene sequences as input and multi-label phenotypic abnormalities induced by gene knockouts as targets. Across these datasets, GenePheno achieves state-of-the-art gene-centric $F_{\\text{max}}$ and phenotype-centric AUC, and case studies demonstrate its ability to reveal gene functional mechanisms.", "date": "2025-11-12", "categories": []}
{"id": "2511.11717", "title": "Multiscale Grassmann Manifolds for Single-Cell Data Analysis", "abstract": "Single-cell data analysis seeks to characterize cellular heterogeneity based on high-dimensional gene expression profiles. Conventional approaches represent each cell as a vector in Euclidean space, which limits their ability to capture intrinsic correlations and multiscale geometric structures. We propose a multiscale framework based on Grassmann manifolds that integrates machine learning with subspace geometry for single-cell data analysis. By generating embeddings under multiple representation scales, the framework combines their features from different geometric views into a unified Grassmann manifold. A power-based scale sampling function is introduced to control the selection of scales and balance in- formation across resolutions. Experiments on nine benchmark single-cell RNA-seq datasets demonstrate that the proposed approach effectively preserves meaningful structures and provides stable clustering performance, particularly for small to medium-sized datasets. These results suggest that Grassmann manifolds offer a coherent and informative foundation for analyzing single cell data.", "date": "2025-11-12", "categories": []}
{"id": "2511.12805", "title": "Practical Causal Evaluation Metrics for Biological Networks", "abstract": "Estimating causal networks from biological data is a critical step in systems biology. When evaluating the inferred network, assessing the networks based on their intervention effects is particularly important for downstream probabilistic reasoning and the identification of potential drug targets. In the context of gene regulatory network inference, biological databases are often used as reference sources. These databases typically describe relationships in a qualitative rather than quantitative manner. However, few evaluation metrics have been developed that take this qualitative nature into account. To address this, we developed a metric, the sign-augmented Structural Intervention Distance (sSID), and a weighted sSID that incorporates the net effects of the intervention. Through simulations and analyses of real transcriptomic datasets, we found that our proposed metrics could identify a different algorithm as optimal compared to conventional metrics, and the network selected by sSID had a superior performance in the classification task of clinical covariates using transcriptomic data. This suggests that sSID can distinguish networks that are structurally correct but functionally incorrect, highlighting its potential as a more biologically meaningful and practical evaluation metric.", "date": "2025-11-16", "categories": []}
{"id": "2511.13124", "title": "Departures: Distributional Transport for Single-Cell Perturbation Prediction with Neural Schr\\\"odinger Bridges", "abstract": "Predicting single-cell perturbation outcomes directly advances gene function analysis and facilitates drug candidate selection, making it a key driver of both basic and translational biomedical research. However, a major bottleneck in this task is the unpaired nature of single-cell data, as the same cell cannot be observed both before and after perturbation due to the destructive nature of sequencing. Although some neural generative transport models attempt to tackle unpaired single-cell perturbation data, they either lack explicit conditioning or depend on prior spaces for indirect distribution alignment, limiting precise perturbation modeling. In this work, we approximate Schr\\\"odinger Bridge (SB), which defines stochastic dynamic mappings recovering the entropy-regularized optimal transport (OT), to directly align the distributions of control and perturbed single-cell populations across different perturbation conditions. Unlike prior SB approximations that rely on bidirectional modeling to infer optimal source-target sample coupling, we leverage Minibatch-OT based pairing to avoid such bidirectional inference and the associated ill-posedness of defining the reverse process. This pairing directly guides bridge learning, yielding a scalable approximation to the SB. We approximate two SB models, one modeling discrete gene activation states and the other continuous expression distributions. Joint training enables accurate perturbation modeling and captures single-cell heterogeneity. Experiments on public genetic and drug perturbation datasets show that our model effectively captures heterogeneous single-cell responses and achieves state-of-the-art performance.", "date": "2025-11-17", "categories": []}
{"id": "2511.13705", "title": "Rare Genomic Subtype Discovery from RNA-seq via Autoencoder Embeddings and Stability-Aware Clustering", "abstract": "Unsupervised learning on high-dimensional RNA-seq data can reveal molecular subtypes beyond standard labels. We combine an autoencoder-based representation with clustering and stability analysis to search for rare but reproducible genomic subtypes. On the UCI \"Gene Expression Cancer RNA-Seq\" dataset (801 samples, 20,531 genes; BRCA, COAD, KIRC, LUAD, PRAD), a pan-cancer analysis shows clusters aligning almost perfectly with tissue of origin (Cramer's V = 0.887), serving as a negative control. We therefore reframe the problem within KIRC (n = 146): we select the top 2,000 highly variable genes, standardize them, train a feed-forward autoencoder (128-dimensional latent space), and run k-means for k = 2-10. While global indices favor small k, scanning k with a pre-specified discovery rule (rare < 10 percent and stable with Jaccard >= 0.60 across 20 seeds after Hungarian alignment) yields a simple solution at k = 5 (silhouette = 0.129, DBI = 2.045) with a rare cluster C0 (6.85 percent of patients) that is highly stable (Jaccard = 0.787). Cluster-vs-rest differential expression (Welch's t-test, Benjamini-Hochberg FDR) identifies coherent markers. Overall, pan-cancer clustering is dominated by tissue of origin, whereas a stability-aware within-cancer approach reveals a rare, reproducible KIRC subtype.", "date": "2025-11-17", "categories": []}
{"id": "2511.10590", "title": "Pretrained Joint Predictions for Scalable Batch Bayesian Optimization of Molecular Designs", "abstract": "Batched synthesis and testing of molecular designs is the key bottleneck of drug development. There has been great interest in leveraging biomolecular foundation models as surrogates to accelerate this process. In this work, we show how to obtain scalable probabilistic surrogates of binding affinity for use in Batch Bayesian Optimization (Batch BO). This demands parallel acquisition functions that hedge between designs and the ability to rapidly sample from a joint predictive density to approximate them. Through the framework of Epistemic Neural Networks (ENNs), we obtain scalable joint predictive distributions of binding affinity on top of representations taken from large structure-informed models. Key to this work is an investigation into the importance of prior networks in ENNs and how to pretrain them on synthetic data to improve downstream performance in Batch BO. Their utility is demonstrated by rediscovering known potent EGFR inhibitors on a semi-synthetic benchmark in up to 5x fewer iterations, as well as potent inhibitors from a real-world small-molecule library in up to 10x fewer iterations, offering a promising solution for large-scale drug discovery applications.", "date": "2025-11-13", "categories": []}
{"id": "2511.11380", "title": "When Genes Speak: A Semantic-Guided Framework for Spatially Resolved Transcriptomics Data Clustering", "abstract": "Spatial transcriptomics enables gene expression profiling with spatial context, offering unprecedented insights into the tissue microenvironment. However, most computational models treat genes as isolated numerical features, ignoring the rich biological semantics encoded in their symbols. This prevents a truly deep understanding of critical biological characteristics. To overcome this limitation, we present SemST, a semantic-guided deep learning framework for spatial transcriptomics data clustering. SemST leverages Large Language Models (LLMs) to enable genes to \"speak\" through their symbolic meanings, transforming gene sets within each tissue spot into biologically informed embeddings. These embeddings are then fused with the spatial neighborhood relationships captured by Graph Neural Networks (GNNs), achieving a coherent integration of biological function and spatial structure. We further introduce the Fine-grained Semantic Modulation (FSM) module to optimally exploit these biological priors. The FSM module learns spot-specific affine transformations that empower the semantic embeddings to perform an element-wise calibration of the spatial features, thus dynamically injecting high-order biological knowledge into the spatial context. Extensive experiments on public spatial transcriptomics datasets show that SemST achieves state-of-the-art clustering performance. Crucially, the FSM module exhibits plug-and-play versatility, consistently improving the performance when integrated into other baseline methods.", "date": "2025-11-14", "categories": []}
{"id": "2511.09564", "title": "Mamba-driven multi-perspective structural understanding for molecular ground-state conformation prediction", "abstract": "A comprehensive understanding of molecular structures is important for the prediction of molecular ground-state conformation involving property information. Meanwhile, state space model (e.g., Mamba) has recently emerged as a promising mechanism for long sequence modeling and has achieved remarkable results in various language and vision tasks. However, towards molecular ground-state conformation prediction, exploiting Mamba to understand molecular structure is underexplored. To this end, we strive to design a generic and efficient framework with Mamba to capture critical components. In general, molecular structure could be considered to consist of three elements, i.e., atom types, atom positions, and connections between atoms. Thus, considering the three elements, an approach of Mamba-driven multi-perspective structural understanding (MPSU-Mamba) is proposed to localize molecular ground-state conformation. Particularly, for complex and diverse molecules, three different kinds of dedicated scanning strategies are explored to construct a comprehensive perception of corresponding molecular structures. And a bright-channel guided mechanism is defined to discriminate the critical conformation-related atom information. Experimental results on QM9 and Molecule3D datasets indicate that MPSU-Mamba significantly outperforms existing methods. Furthermore, we observe that for the case of few training samples, MPSU-Mamba still achieves superior performance, demonstrating that our method is indeed beneficial for understanding molecular structures.", "date": "2025-11-10", "categories": []}
{"id": "2511.09576", "title": "Prostate-VarBench: A Benchmark with Interpretable TabNet Framework for Prostate Cancer Variant Classification", "abstract": "Variants of Uncertain Significance (VUS) limit the clinical utility of prostate cancer genomics by delaying diagnosis and therapy when evidence for pathogenicity or benignity is incomplete. Progress is further limited by inconsistent annotations across sources and the absence of a prostate-specific benchmark for fair comparison. We introduce Prostate-VarBench, a curated pipeline for creating prostate-specific benchmarks that integrates COSMIC (somatic cancer mutations), ClinVar (expert-curated clinical variants), and TCGA-PRAD (prostate tumor genomics from The Cancer Genome Atlas) into a harmonized dataset of 193,278 variants supporting patient- or gene-aware splits to prevent data leakage. To ensure data integrity, we corrected a Variant Effect Predictor (VEP) issue that merged multiple transcript records, introducing ambiguity in clinical significance fields. We then standardized 56 interpretable features across eight clinically relevant tiers, including population frequency, variant type, and clinical context. AlphaMissense pathogenicity scores were incorporated to enhance missense variant classification and reduce VUS uncertainty. Building on this resource, we trained an interpretable TabNet model to classify variant pathogenicity, whose step-wise sparse masks provide per-case rationales consistent with molecular tumor board review practices. On the held-out test set, the model achieved 89.9% accuracy with balanced class metrics, and the VEP correction yields an 6.5% absolute reduction in VUS.", "date": "2025-11-12", "categories": []}
{"id": "2511.10056", "title": "From Static Structures to Ensembles: Studying and Harnessing Protein Structure Tokenization", "abstract": "Protein structure tokenization converts 3D structures into discrete or vectorized representations, enabling the integration of structural and sequence data. Despite many recent works on structure tokenization, the properties of the underlying discrete representations are not well understood. In this work, we first demonstrate that the successful utilization of structural tokens in a language model for structure prediction depends on using rich, pre-trained sequence embeddings to bridge the semantic gap between the sequence and structural \"language\". The analysis of the structural vocabulary itself then reveals significant semantic redundancy, where multiple distinct tokens correspond to nearly identical local geometries, acting as \"structural synonyms\". This redundancy, rather than being a flaw, can be exploited with a simple \"synonym swap\" strategy to generate diverse conformational ensembles by perturbing a predicted structure with its structural synonyms. This computationally lightweight method accurately recapitulates protein flexibility, performing competitively with state-of-the-art models. Our study provides fundamental insights into the nature of discrete protein structure representations and introduces a powerful, near-instantaneous method for modeling protein dynamics. Source code is available in https://github.com/IDEA-XL/TokenMD.", "date": "2025-11-13", "categories": []}
{"id": "2511.10165", "title": "EPO: Diverse and Realistic Protein Ensemble Generation via Energy Preference Optimization", "abstract": "Accurate exploration of protein conformational ensembles is essential for uncovering function but remains hard because molecular-dynamics (MD) simulations suffer from high computational costs and energy-barrier trapping. This paper presents Energy Preference Optimization (EPO), an online refinement algorithm that turns a pretrained protein ensemble generator into an energy-aware sampler without extra MD trajectories. Specifically, EPO leverages stochastic differential equation sampling to explore the conformational landscape and incorporates a novel energy-ranking mechanism based on list-wise preference optimization. Crucially, EPO introduces a practical upper bound to efficiently approximate the intractable probability of long sampling trajectories in continuous-time generative models, making it easily adaptable to existing pretrained generators. On Tetrapeptides, ATLAS, and Fast-Folding benchmarks, EPO successfully generates diverse and physically realistic ensembles, establishing a new state-of-the-art in nine evaluation metrics. These results demonstrate that energy-only preference signals can efficiently steer generative models toward thermodynamically consistent conformational ensembles, providing an alternative to long MD simulations and widening the applicability of learned potentials in structural biology and drug discovery.", "date": "2025-11-13", "categories": []}
{"id": "2511.10244", "title": "PepTriX: A Framework for Explainable Peptide Analysis through Protein Language Models", "abstract": "Peptide classification tasks, such as predicting toxicity and HIV inhibition, are fundamental to bioinformatics and drug discovery. Traditional approaches rely heavily on handcrafted encodings of one-dimensional (1D) peptide sequences, which can limit generalizability across tasks and datasets. Recently, protein language models (PLMs), such as ESM-2 and ESMFold, have demonstrated strong predictive performance. However, they face two critical challenges. First, fine-tuning is computationally costly. Second, their complex latent representations hinder interpretability for domain experts. Additionally, many frameworks have been developed for specific types of peptide classification, lacking generalization. These limitations restrict the ability to connect model predictions to biologically relevant motifs and structural properties. To address these limitations, we present PepTriX, a novel framework that integrates one dimensional (1D) sequence embeddings and three-dimensional (3D) structural features via a graph attention network enhanced with contrastive training and cross-modal co-attention. PepTriX automatically adapts to diverse datasets, producing task-specific peptide vectors while retaining biological plausibility. After evaluation by domain experts, we found that PepTriX performs remarkably well across multiple peptide classification tasks and provides interpretable insights into the structural and biophysical motifs that drive predictions. Thus, PepTriX offers both predictive robustness and interpretable validation, bridging the gap between performance-driven peptide-level models (PLMs) and domain-level understanding in peptide research.", "date": "2025-11-13", "categories": []}
{"id": "2511.10440", "title": "Completion of partial structures using Patterson maps with the CrysFormer machine learning model", "abstract": "Protein structure determination has long been one of the primary challenges of structural biology, to which deep machine learning (ML)-based approaches have increasingly been applied. However, these ML models generally do not incorporate the experimental measurements directly, such as X-ray crystallographic diffraction data. To this end, we explore an approach that more tightly couples these traditional crystallographic and recent ML-based methods, by training a hybrid 3-d vision transformer and convolutional network on inputs from both domains. We make use of two distinct input constructs / Patterson maps, which are directly obtainable from crystallographic data, and ``partial structure'' template maps derived from predicted structures deposited in the AlphaFold Protein Structure Database with subsequently omitted residues. With these, we predict electron density maps that are then post-processed into atomic models through standard crystallographic refinement processes. Introducing an initial dataset of small protein fragments taken from Protein Data Bank entries and placing them in hypothetical crystal settings, we demonstrate that our method is effective at both improving the phases of the crystallographic structure factors and completing the regions missing from partial structure templates, as well as improving the agreement of the electron density maps with the ground truth atomic structures.", "date": "2025-11-13", "categories": []}
{"id": "2503.05738", "title": "Learning conformational ensembles of proteins based on backbone geometry", "abstract": "Deep generative models have recently been proposed for sampling protein conformations from the Boltzmann distribution, as an alternative to often prohibitively expensive Molecular Dynamics simulations. However, current state-of-the-art approaches rely on fine-tuning pre-trained folding models and evolutionary sequence information, limiting their applicability and efficiency, and introducing potential biases. In this work, we propose a flow matching model for sampling protein conformations based solely on backbone geometry - BBFlow. We introduce a geometric encoding of the backbone equilibrium structure as input and propose to condition not only the flow but also the prior distribution on the respective equilibrium structure, eliminating the need for evolutionary information. The resulting model is orders of magnitudes faster than current state-of-the-art approaches at comparable accuracy, is transferable to multi-chain proteins, and can be trained from scratch in a few GPU days. In our experiments, we demonstrate that the proposed model achieves competitive performance with reduced inference time, across not only an established benchmark of naturally occurring proteins but also de novo proteins, for which evolutionary information is scarce or absent. BBFlow is available at https://github.com/graeter-group/bbflow.", "date": "2025-02-19", "categories": []}
{"id": "2509.00200", "title": "Simulation-based inference of yeast centromeres", "abstract": "The chromatin folding and the spatial arrangement of chromosomes in the cell play a crucial role in DNA replication and genes expression. An improper chromatin folding could lead to malfunctions and, over time, diseases. For eukaryotes, centromeres are essential for proper chromosome segregation and folding. Despite extensive research using de novo sequencing of genomes and annotation analysis, centromere locations in yeasts remain difficult to infer and are still unknown in most species. Recently, genome-wide chromosome conformation capture coupled with next-generation sequencing (Hi-C) has become one of the leading methods to investigate chromosome structures. Some recent studies have used Hi-C data to give a point estimate of each centromere, but those approaches highly rely on a good pre-localization. Here, we present a novel approach that infers in a stochastic manner the locations of all centromeres in budding yeast based on both the experimental Hi-C map and simulated contact maps.", "date": "2025-08-29", "categories": []}
{"id": "2511.09026", "title": "DeepVRegulome: DNABERT-based deep-learning framework for predicting the functional impact of short genomic variants on the human regulome", "abstract": "Whole-genome sequencing (WGS) has revealed numerous non-coding short variants whose functional impacts remain poorly understood. Despite recent advances in deep-learning genomic approaches, accurately predicting and prioritizing clinically relevant mutations in gene regulatory regions remains a major challenge. Here we introduce Deep VRegulome, a deep-learning method for prediction and interpretation of functionally disruptive variants in the human regulome, which combines 700 DNABERT fine-tuned models, trained on vast amounts of ENCODE gene regulatory regions, with variant scoring, motif analysis, attention-based visualization, and survival analysis. We showcase its application on TCGA glioblastoma WGS dataset in prioritizing survival-associated mutations and regulatory regions. The analysis identified 572 splice-disrupting and 9,837 transcription-factor binding site altering mutations occurring in greater than 10% of glioblastoma samples. Survival analysis linked 1352 mutations and 563 disrupted regulatory regions to patient outcomes, enabling stratification via non-coding mutation signatures. All the code, fine-tuned models, and an interactive data portal are publicly available.", "date": "2025-11-12", "categories": []}
{"id": "2511.08080", "title": "Hierarchical Structure-Property Alignment for Data-Efficient Molecular Generation and Editing", "abstract": "Property-constrained molecular generation and editing are crucial in AI-driven drug discovery but remain hindered by two factors: (i) capturing the complex relationships between molecular structures and multiple properties remains challenging, and (ii) the narrow coverage and incomplete annotations of molecular properties weaken the effectiveness of property-based models. To tackle these limitations, we propose HSPAG, a data-efficient framework featuring hierarchical structure-property alignment. By treating SMILES and molecular properties as complementary modalities, the model learns their relationships at atom, substructure, and whole-molecule levels. Moreover, we select representative samples through scaffold clustering and hard samples via an auxiliary variational auto-encoder (VAE), substantially reducing the required pre-training data. In addition, we incorporate a property relevance-aware masking mechanism and diversified perturbation strategies to enhance generation quality under sparse annotations. Experiments demonstrate that HSPAG captures fine-grained structure-property relationships and supports controllable generation under multiple property constraints. Two real-world case studies further validate the editing capabilities of HSPAG.", "date": "2025-11-11", "categories": []}
{"id": "2406.16995", "title": "tcrLM: a lightweight protein language model for predicting T cell receptor and epitope binding specificity", "abstract": "The anti-cancer immune response relies on the bindings between T-cell receptors (TCRs) and antigens, which elicits adaptive immunity to eliminate tumor cells. This ability of the immune system to respond to novel various neoantigens arises from the immense diversity of TCR repository. However, TCR diversity poses a significant challenge on accurately predicting antigen-TCR bindings. In this study, we introduce a lightweight masked language model, termed tcrLM, to address this challenge. Our approach involves randomly masking segments of TCR sequences and training tcrLM to infer the masked segments, thereby enabling the extraction of expressive features from TCR sequences. To further enhance robustness, we incorporate virtual adversarial training into tcrLM. We construct the largest TCR CDR3 sequence set with more than 100 million distinct sequences, and pretrain tcrLM on these sequences. The pre-trained encoder is subsequently applied to predict TCR-antigen binding specificity. We evaluate model performance on three test datasets: independent, external, and COVID-19 test set. The results demonstrate that tcrLM not only surpasses existing TCR-antigen binding prediction methods, but also outperforms other mainstream protein language models. More interestingly, tcrLM effectively captures the biochemical properties and positional preference of amino acids within TCR sequences. Additionally, the predicted TCR-neoantigen binding scores indicates the immunotherapy responses and clinical outcomes in a melanoma cohort. These findings demonstrate the potential of tcrLM in predicting TCR-antigen binding specificity, with significant implications for advancing immunotherapy and personalized medicine.", "date": "2024-06-24", "categories": []}
{"id": "2505.13413", "title": "Joint Velocity-Growth Flow Matching for Single-Cell Dynamics Modeling", "abstract": "Learning the underlying dynamics of single cells from snapshot data has gained increasing attention in scientific and machine learning research. The destructive measurement technique and cell proliferation/death result in unpaired and unbalanced data between snapshots, making the learning of the underlying dynamics challenging. In this paper, we propose joint Velocity-Growth Flow Matching (VGFM), a novel paradigm that jointly learns state transition and mass growth of single-cell populations via flow matching. VGFM builds an ideal single-cell dynamics containing velocity of state and growth of mass, driven by a presented two-period dynamic understanding of the static semi-relaxed optimal transport, a mathematical tool that seeks the coupling between unpaired and unbalanced data. To enable practical usage, we approximate the ideal dynamics using neural networks, forming our joint velocity and growth matching framework. A distribution fitting loss is also employed in VGFM to further improve the fitting performance for snapshot data. Extensive experimental results on both synthetic and real datasets demonstrate that VGFM can capture the underlying biological dynamics accounting for mass and state variations over time, outperforming existing approaches for single-cell dynamics modeling.", "date": "2025-05-19", "categories": []}
{"id": "2509.25884", "title": "scUnified: An AI-Ready Standardized Resource for Single-Cell RNA Sequencing Analysis", "abstract": "Single-cell RNA sequencing (scRNA-seq) technology enables systematic delineation of cellular states and interactions, providing crucial insights into cellular heterogeneity. Building on this potential, numerous computational methods have been developed for tasks such as cell clustering, cell type annotation, and marker gene identification. To fully assess and compare these methods, standardized, analysis-ready datasets are essential. However, such datasets remain scarce, and variations in data formats, preprocessing workflows, and annotation strategies hinder reproducibility and complicate systematic evaluation of existing methods. To address these challenges, we present scUnified, an AI-ready standardized resource for single-cell RNA sequencing data that consolidates 13 high-quality datasets spanning two species (human and mouse) and nine tissue types. All datasets undergo standardized quality control and preprocessing and are stored in a uniform format to enable direct application in diverse computational analyses without additional data cleaning. We further demonstrate the utility of scUnified through experimental analyses of representative biological tasks, providing a reproducible foundation for the standardized evaluation of computational methods on a unified dataset.", "date": "2025-09-30", "categories": []}
{"id": "2510.08662", "title": "DPCformer: An Interpretable Deep Learning Model for Genomic Prediction in Crops", "abstract": "Genomic Selection (GS) uses whole-genome information to predict crop phenotypes and accelerate breeding. Traditional GS methods, however, struggle with prediction accuracy for complex traits and large datasets. We propose DPCformer, a deep learning model integrating convolutional neural networks with a self-attention mechanism to model complex genotype-phenotype relationships. We applied DPCformer to 13 traits across five crops (maize, cotton, tomato, rice, chickpea). Our approach uses an 8-dimensional one-hot encoding for SNP data, ordered by chromosome, and employs the PMF algorithm for feature selection. Evaluations show DPCformer outperforms existing methods. In maize datasets, accuracy for traits like days to tasseling and plant height improved by up to 2.92%. For cotton, accuracy gains for fiber traits reached 8.37%. On small-sample tomato data, the Pearson Correlation Coefficient for a key trait increased by up to 57.35%. In chickpea, the yield correlation was boosted by 16.62%. DPCformer demonstrates superior accuracy, robustness in small-sample scenarios, and enhanced interpretability, providing a powerful tool for precision breeding and addressing global food security challenges.", "date": "2025-10-09", "categories": []}
{"id": "2511.05510", "title": "TEMPO: Temporal Multi-scale Autoregressive Generation of Protein Conformational Ensembles", "abstract": "Understanding the dynamic behavior of proteins is critical to elucidating their functional mechanisms, yet generating realistic, temporally coherent trajectories of protein ensembles remains a significant challenge. In this work, we introduce a novel hierarchical autoregressive framework for modeling protein dynamics that leverages the intrinsic multi-scale organization of molecular motions. Unlike existing methods that focus on generating static conformational ensembles or treat dynamic sampling as an independent process, our approach characterizes protein dynamics as a Markovian process. The framework employs a two-scale architecture: a low-resolution model captures slow, collective motions driving major conformational transitions, while a high-resolution model generates detailed local fluctuations conditioned on these large-scale movements. This hierarchical design ensures that the causal dependencies inherent in protein dynamics are preserved, enabling the generation of temporally coherent and physically realistic trajectories. By bridging high-level biophysical principles with state-of-the-art generative modeling, our approach provides an efficient framework for simulating protein dynamics that balances computational efficiency with physical accuracy.", "date": "2025-10-24", "categories": []}
{"id": "2511.05726", "title": "GastroDL-Fusion: A Dual-Modal Deep Learning Framework Integrating Protein-Ligand Complexes and Gene Sequences for Gastrointestinal Disease Drug Discovery", "abstract": "Accurate prediction of protein-ligand binding affinity plays a pivotal role in accelerating the discovery of novel drugs and vaccines, particularly for gastrointestinal (GI) diseases such as gastric ulcers, Crohn's disease, and ulcerative colitis. Traditional computational models often rely on structural information alone and thus fail to capture the genetic determinants that influence disease mechanisms and therapeutic responses. To address this gap, we propose GastroDL-Fusion, a dual-modal deep learning framework that integrates protein-ligand complex data with disease-associated gene sequence information for drug and vaccine development. In our approach, protein-ligand complexes are represented as molecular graphs and modeled using a Graph Isomorphism Network (GIN), while gene sequences are encoded into biologically meaningful embeddings via a pre-trained Transformer (ProtBERT/ESM). These complementary modalities are fused through a multi-layer perceptron to enable robust cross-modal interaction learning. We evaluate the model on benchmark datasets of GI disease-related targets, demonstrating that GastroDL-Fusion significantly improves predictive performance over conventional methods. Specifically, the model achieves a mean absolute error (MAE) of 1.12 and a root mean square error (RMSE) of 1.75, outperforming CNN, BiLSTM, GIN, and Transformer-only baselines. These results confirm that incorporating both structural and genetic features yields more accurate predictions of binding affinities, providing a reliable computational tool for accelerating the design of targeted therapies and vaccines in the context of gastrointestinal diseases.", "date": "2025-11-07", "categories": []}
{"id": "2511.06269", "title": "LLM$^3$-DTI: A Large Language Model and Multi-modal data co-powered framework for Drug-Target Interaction prediction", "abstract": "Drug-target interaction (DTI) prediction is of great significance for drug discovery and drug repurposing. With the accumulation of a large volume of valuable data, data-driven methods have been increasingly harnessed to predict DTIs, reducing costs across various dimensions. Therefore, this paper proposes a $\\textbf{L}$arge $\\textbf{L}$anguage $\\textbf{M}$odel and $\\textbf{M}$ulti-$\\textbf{M}$odel data co-powered $\\textbf{D}$rug $\\textbf{T}$arget $\\textbf{I}$nteraction prediction framework, named LLM$^3$-DTI. LLM$^3$-DTI constructs multi-modal data embedding to enhance DTI prediction performance. In this framework, the text semantic embeddings of drugs and targets are encoded by a domain-specific LLM. To effectively align and fuse multi-modal embedding. We propose the dual cross-attention mechanism and the TSFusion module. Finally, these multi-modal data are utilized for the DTI task through an output network. The experimental results indicate that LLM$^3$-DTI can proficiently identify validated DTIs, surpassing the performance of the models employed for comparison across diverse scenarios. Consequently, LLM$^3$-DTI is adept at fulfilling the task of DTI prediction with excellence. The data and code are available at https://github.com/chaser-gua/LLM3DTI.", "date": "2025-11-09", "categories": []}
{"id": "2511.06585", "title": "Learning Biomolecular Motion: The Physics-Informed Machine Learning Paradigm", "abstract": "The convergence of statistical learning and molecular physics is transforming our approach to modeling biomolecular systems. Physics-informed machine learning (PIML) offers a systematic framework that integrates data-driven inference with physical constraints, resulting in models that are accurate, mechanistic, generalizable, and able to extrapolate beyond observed domains. This review surveys recent advances in physics-informed neural networks and operator learning, differentiable molecular simulation, and hybrid physics-ML potentials, with emphasis on long-timescale kinetics, rare events, and free-energy estimation. We frame these approaches as solutions to the \"biomolecular closure problem\", recovering unresolved interactions beyond classical force fields while preserving thermodynamic consistency and mechanistic interpretability. We examine theoretical foundations, tools and frameworks, computational trade-offs, and unresolved issues, including model expressiveness and stability. We outline prospective research avenues at the intersection of machine learning, statistical physics, and computational chemistry, contending that future advancements will depend on mechanistic inductive biases, and integrated differentiable physical learning frameworks for biomolecular simulation and discovery.", "date": "2025-11-10", "categories": []}
{"id": "2511.07006", "title": "S$^2$Drug: Bridging Protein Sequence and 3D Structure in Contrastive Representation Learning for Virtual Screening", "abstract": "Virtual screening (VS) is an essential task in drug discovery, focusing on the identification of small-molecule ligands that bind to specific protein pockets. Existing deep learning methods, from early regression models to recent contrastive learning approaches, primarily rely on structural data while overlooking protein sequences, which are more accessible and can enhance generalizability. However, directly integrating protein sequences poses challenges due to the redundancy and noise in large-scale protein-ligand datasets. To address these limitations, we propose \\textbf{S$^2$Drug}, a two-stage framework that explicitly incorporates protein \\textbf{S}equence information and 3D \\textbf{S}tructure context in protein-ligand contrastive representation learning. In the first stage, we perform protein sequence pretraining on ChemBL using an ESM2-based backbone, combined with a tailored data sampling strategy to reduce redundancy and noise on both protein and ligand sides. In the second stage, we fine-tune on PDBBind by fusing sequence and structure information through a residue-level gating module, while introducing an auxiliary binding site prediction task. This auxiliary task guides the model to accurately localize binding residues within the protein sequence and capture their 3D spatial arrangement, thereby refining protein-ligand matching. Across multiple benchmarks, S$^2$Drug consistently improves virtual screening performance and achieves strong results on binding site prediction, demonstrating the value of bridging sequence and structure in contrastive learning.", "date": "2025-11-10", "categories": []}
{"id": "2410.15040", "title": "Retrieval Augmented Diffusion Model for Structure-informed Antibody Design and Optimization", "abstract": "Antibodies are essential proteins responsible for immune responses in organisms, capable of specifically recognizing antigen molecules of pathogens. Recent advances in generative models have significantly enhanced rational antibody design. However, existing methods mainly create antibodies from scratch without template constraints, leading to model optimization challenges and unnatural sequences. To address these issues, we propose a retrieval-augmented diffusion framework, termed RADAb, for efficient antibody design. Our method leverages a set of structural homologous motifs that align with query structural constraints to guide the generative model in inversely optimizing antibodies according to desired design criteria. Specifically, we introduce a structure-informed retrieval mechanism that integrates these exemplar motifs with the input backbone through a novel dual-branch denoising module, utilizing both structural and evolutionary information. Additionally, we develop a conditional diffusion model that iteratively refines the optimization process by incorporating both global context and local evolutionary conditions. Our approach is agnostic to the choice of generative models. Empirical experiments demonstrate that our method achieves state-of-the-art performance in multiple antibody inverse folding and optimization tasks, offering a new perspective on biomolecular generative models.", "date": "2024-10-19", "categories": []}
{"id": "2511.04814", "title": "A Standardized Benchmark for Multilabel Antimicrobial Peptide Classification", "abstract": "Antimicrobial peptides have emerged as promising molecules to combat antimicrobial resistance. However, fragmented datasets, inconsistent annotations, and the lack of standardized benchmarks hinder computational approaches and slow down the discovery of new candidates. To address these challenges, we present the Expanded Standardized Collection for Antimicrobial Peptide Evaluation (ESCAPE), an experimental framework integrating over 80.000 peptides from 27 validated repositories. Our dataset separates antimicrobial peptides from negative sequences and incorporates their functional annotations into a biologically coherent multilabel hierarchy, capturing activities across antibacterial, antifungal, antiviral, and antiparasitic classes. Building on ESCAPE, we propose a transformer-based model that leverages sequence and structural information to predict multiple functional activities of peptides. Our method achieves up to a 2.56% relative average improvement in mean Average Precision over the second-best method adapted for this task, establishing a new state-of-the-art multilabel peptide classification. ESCAPE provides a comprehensive and reproducible evaluation framework to advance AI-driven antimicrobial peptide research.", "date": "2025-11-06", "categories": []}
{"id": "2411.06635", "title": "scMEDAL for the interpretable analysis of single-cell transcriptomics data with batch effect visualization using a deep mixed effects autoencoder", "abstract": "Single-cell RNA sequencing enables high-resolution analysis of cellular heterogeneity, yet disentangling biological signal from batch effects remains a major challenge. Existing batch-correction algorithms suppress or discard batch-related variation rather than modeling it. We propose scMEDAL, single-cell Mixed Effects Deep Autoencoder Learning, a framework that separately models batch-invariant and batch-specific effects using two complementary subnetworks. The principal innovation, scMEDAL-RE, is a random-effects Bayesian autoencoder that learns batch-specific representations while preserving biologically meaningful information confounded with batch effects signal often lost under standard correction. Complementing it, the fixed-effects subnetwork, scMEDAL-FE, trained via adversarial learning provides a default batch-correction component. Evaluations across diverse conditions (autism, leukemia, cardiovascular), cell types, and technical and biological effects show that scMEDAL-RE produces interpretable, batch-specific embeddings that complement both scMEDAL-FE and established correction methods (scVI, Scanorama, Harmony, SAUCIE), yielding more accurate prediction of disease status, donor group, and tissue. scMEDAL also provides generative visualizations, including counterfactual reconstructions of a cell's expression as if acquired in another batch. The framework allows substitution of the fixed-effects component with other correction methods, while retaining scMEDAL-RE's enhanced predictive power and visualization. Overall, scMEDAL is a versatile, interpretable framework that complements existing correction, providing enhanced insight into cellular heterogeneity and data acquisition.", "date": "2024-11-11", "categories": []}
{"id": "2511.03976", "title": "PETRA: Pretrained Evolutionary Transformer for SARS-CoV-2 Mutation Prediction", "abstract": "Since its emergence, SARS-CoV-2 has demonstrated a rapid and unpredictable evolutionary trajectory, characterized by the continual emergence of immune-evasive variants. This poses persistent challenges to public health and vaccine development.\n  While large-scale generative pre-trained transformers (GPTs) have revolutionized the modeling of sequential data, their direct applications to noisy viral genomic sequences are limited. In this paper, we introduce PETRA(Pretrained Evolutionary TRAnsformer), a novel transformer approach based on evolutionary trajectories derived from phylogenetic trees rather than raw RNA sequences. This method effectively mitigates sequencing noise and captures the hierarchical structure of viral evolution.\n  With a weighted training framework to address substantial geographical and temporal imbalances in global sequence data, PETRA excels in predicting future SARS-CoV-2 mutations, achieving a weighted recall@1 of 9.45% for nucleotide mutations and 17.10\\% for spike amino-acid mutations, compared to 0.49% and 6.64% respectively for the best baseline. PETRA also demonstrates its ability to aid in the real-time mutation prediction of major clades like 24F(XEC) and 25A(LP.8.1). The code is open sourced on https://github.com/xz-keg/PETra", "date": "2025-11-06", "categories": []}
{"id": "2511.04040", "title": "Enhancing Multimodal Protein Function Prediction Through Dual-Branch Dynamic Selection with Reconstructive Pre-Training", "abstract": "Multimodal protein features play a crucial role in protein function prediction. However, these features encompass a wide range of information, ranging from structural data and sequence features to protein attributes and interaction networks, making it challenging to decipher their complex interconnections. In this work, we propose a multimodal protein function prediction method (DSRPGO) by utilizing dynamic selection and reconstructive pre-training mechanisms. To acquire complex protein information, we introduce reconstructive pre-training to mine more fine-grained information with low semantic levels. Moreover, we put forward the Bidirectional Interaction Module (BInM) to facilitate interactive learning among multimodal features. Additionally, to address the difficulty of hierarchical multi-label classification in this task, a Dynamic Selection Module (DSM) is designed to select the feature representation that is most conducive to current protein function prediction. Our proposed DSRPGO model improves significantly in BPO, MFO, and CCO on human datasets, thereby outperforming other benchmark models.", "date": "2025-11-06", "categories": []}
{"id": "2511.04239", "title": "seqme: a Python library for evaluating biological sequence design", "abstract": "Recent advances in computational methods for designing biological sequences have sparked the development of metrics to evaluate these methods performance in terms of the fidelity of the designed sequences to a target distribution and their attainment of desired properties. However, a single software library implementing these metrics was lacking. In this work we introduce seqme, a modular and highly extendable open-source Python library, containing model-agnostic metrics for evaluating computational methods for biological sequence design. seqme considers three groups of metrics: sequence-based, embedding-based, and property-based, and is applicable to a wide range of biological sequences: small molecules, DNA, ncRNA, mRNA, peptides and proteins. The library offers a number of embedding and property models for biological sequences, as well as diagnostics and visualization functions to inspect the results. seqme can be used to evaluate both one-shot and iterative computational design methods.", "date": "2025-11-06", "categories": []}
{"id": "2510.19090", "title": "Learning noisy tissue dynamics across time scales", "abstract": "Tissue dynamics play a crucial role in biological processes ranging from inflammation to morphogenesis. However, these noisy multicellular dynamics are notoriously hard to predict. Here, we introduce a biomimetic machine learning framework capable of inferring noisy multicellular dynamics directly from experimental movies. This generative model combines graph neural networks, normalizing flows and WaveNet algorithms to represent tissues as neural stochastic differential equations where cells are edges of an evolving graph. Cell interactions are encoded in a dual signaling graph capable of handling signaling cascades. The dual graph architecture of our neural networks reflects the architecture of the underlying biological tissues, substantially reducing the amount of data needed for training, compared to convolutional or fully-connected neural networks. Taking epithelial tissue experiments as a case study, we show that our model not only captures stochastic cell motion but also predicts the evolution of cell states in their division cycle. Finally, we demonstrate that our method can accurately generate the experimental dynamics of developmental systems, such as the fly wing, and cell signaling processes mediated by stochastic ERK waves, paving the way for its use as a digital twin in bioengineering and clinical contexts.", "date": "2025-10-21", "categories": []}
{"id": "2511.02888", "title": "NABench: Large-Scale Benchmarks of Nucleotide Foundation Models for Fitness Prediction", "abstract": "Nucleotide sequence variation can induce significant shifts in functional fitness. Recent nucleotide foundation models promise to predict such fitness effects directly from sequence, yet heterogeneous datasets and inconsistent preprocessing make it difficult to compare methods fairly across DNA and RNA families. Here we introduce NABench, a large-scale, systematic benchmark for nucleic acid fitness prediction. NABench aggregates 162 high-throughput assays and curates 2.6 million mutated sequences spanning diverse DNA and RNA families, with standardized splits and rich metadata. We show that NABench surpasses prior nucleotide fitness benchmarks in scale, diversity, and data quality. Under a unified evaluation suite, we rigorously assess 29 representative foundation models across zero-shot, few-shot prediction, transfer learning, and supervised settings. The results quantify performance heterogeneity across tasks and nucleic-acid types, demonstrating clear strengths and failure modes for different modeling choices and establishing strong, reproducible baselines. We release NABench to advance nucleic acid modeling, supporting downstream applications in RNA/DNA design, synthetic biology, and biochemistry. Our code is available at https://github.com/mrzzmrzz/NABench.", "date": "2025-11-04", "categories": []}
{"id": "2511.03113", "title": "FP-AbDiff: Improving Score-based Antibody Design by Capturing Nonequilibrium Dynamics through the Underlying Fokker-Planck Equation", "abstract": "Computational antibody design holds immense promise for therapeutic discovery, yet existing generative models are fundamentally limited by two core challenges: (i) a lack of dynamical consistency, which yields physically implausible structures, and (ii) poor generalization due to data scarcity and structural bias. We introduce FP-AbDiff, the first antibody generator to enforce Fokker-Planck Equation (FPE) physics along the entire generative trajectory. Our method minimizes a novel FPE residual loss over the mixed manifold of CDR geometries (R^3 x SO(3)), compelling locally-learned denoising scores to assemble into a globally coherent probability flow. This physics-informed regularizer is synergistically integrated with deep biological priors within a state-of-the-art SE(3)-equivariant diffusion framework. Rigorous evaluation on the RAbD benchmark confirms that FP-AbDiff establishes a new state-of-the-art. In de novo CDR-H3 design, it achieves a mean Root Mean Square Deviation of 0.99 {\\AA} when superposing on the variable region, a 25% improvement over the previous state-of-the-art model, AbX, and the highest reported Contact Amino Acid Recovery of 39.91%. This superiority is underscored in the more challenging six-CDR co-design task, where our model delivers consistently superior geometric precision, cutting the average full-chain Root Mean Square Deviation by ~15%, and crucially, achieves the highest full-chain Amino Acid Recovery on the functionally dominant CDR-H3 loop (45.67%). By aligning generative dynamics with physical laws, FP-AbDiff enhances robustness and generalizability, establishing a principled approach for physically faithful and functionally viable antibody design.", "date": "2025-11-05", "categories": []}
{"id": "2511.03464", "title": "POEMS: Product of Experts for Interpretable Multi-omic Integration using Sparse Decoding", "abstract": "Integrating different molecular layers, i.e., multiomics data, is crucial for unraveling the complexity of diseases; yet, most deep generative models either prioritize predictive performance at the expense of interpretability or enforce interpretability by linearizing the decoder, thereby weakening the network's nonlinear expressiveness. To overcome this tradeoff, we introduce POEMS: Product Of Experts for Interpretable Multiomics Integration using Sparse Decoding, an unsupervised probabilistic framework that preserves predictive performance while providing interpretability. POEMS provides interpretability without linearizing any part of the network by 1) mapping features to latent factors using sparse connections, which directly translates to biomarker discovery, 2) allowing for cross-omic associations through a shared latent space using product of experts model, and 3) reporting contributions of each omic by a gating network that adaptively computes their influence in the representation learning. Additionally, we present an efficient sparse decoder. In a cancer subtyping case study, POEMS achieves competitive clustering and classification performance while offering our novel set of interpretations, demonstrating that biomarker based insight and predictive accuracy can coexist in multiomics representation learning.", "date": "2025-11-05", "categories": []}
{"id": "2410.09964", "title": "Lower-dimensional projections of cellular expression improves cell type classification from single-cell RNA sequencing", "abstract": "Single-cell RNA sequencing (scRNA-seq) enables the study of cellular diversity at single cell level. It provides a global view of cell-type specification during the onset of biological mechanisms such as developmental processes and human organogenesis. Various statistical, machine and deep learning-based methods have been proposed for cell-type classification. Most of the methods utilizes unsupervised lower dimensional projections obtained from for a large reference data. In this work, we proposed a reference-based method for cell type classification, called EnProCell. The EnProCell, first, computes lower dimensional projections that capture both the high variance and class separability through an ensemble of principle component analysis and multiple discriminant analysis. In the second phase, EnProCell trains a deep neural network on the lower dimensional representation of data to classify cell types. The proposed method outperformed the existing state-of-the-art methods when tested on four different data sets produced from different single-cell sequencing technologies. The EnProCell showed higher accuracy (98.91) and F1 score (98.64) than other methods for predicting reference from reference datasets. Similarly, EnProCell also showed better performance than existing methods in predicting cell types for data with unknown cell types (query) from reference datasets (accuracy:99.52; F1 score: 99.07). In addition to improved performance, the proposed methodology is simple and does not require more computational resources and time. the EnProCell is available at https://github.com/umar1196/EnProCell.", "date": "2024-10-13", "categories": []}
{"id": "2501.10124", "title": "Gene Regulatory Network Inference in the Presence of Selection Bias and Latent Confounders", "abstract": "Gene regulatory network inference (GRNI) aims to discover how genes causally regulate each other from gene expression data. It is well-known that statistical dependencies in observed data do not necessarily imply causation, as spurious dependencies may arise from latent confounders, such as non-coding RNAs. Numerous GRNI methods have thus been proposed to address this confounding issue. However, dependencies may also result from selection--only cells satisfying certain survival or inclusion criteria are observed--while these selection-induced spurious dependencies are frequently overlooked in gene expression data analyses. In this work, we show that such selection is ubiquitous and, when ignored or conflated with true regulations, can lead to flawed causal interpretation and misguided intervention recommendations. To address this challenge, a fundamental question arises: can we distinguish dependencies due to regulation, confounding, and crucially, selection? We show that gene perturbations offer a simple yet effective answer: selection-induced dependencies are symmetric under perturbation, while those from regulation or confounding are not. Building on this motivation, we propose GISL (Gene regulatory network Inference in the presence of Selection bias and Latent confounders), a principled algorithm that leverages perturbation data to uncover both true gene regulatory relations and non-regulatory mechanisms of selection and confounding up to the equivalence class. Experiments on synthetic and real-world gene expression data demonstrate the effectiveness of our method.", "date": "2025-01-17", "categories": []}
{"id": "2501.18456", "title": "adabmDCA 2.0 -- a flexible but easy-to-use package for Direct Coupling Analysis", "abstract": "In this methods article, we provide a flexible but easy-to-use implementation of Direct Coupling Analysis (DCA) based on Boltzmann machine learning, together with a tutorial on how to use it. The package \\texttt{adabmDCA 2.0} is available in different programming languages (C++, Julia, Python) usable on different architectures (single-core and multi-core CPU, GPU) using a common front-end interface. In addition to several learning protocols for dense and sparse generative DCA models, it allows to directly address common downstream tasks like residue-residue contact prediction, mutational-effect prediction, scoring of sequence libraries and generation of artificial sequences for sequence design. It is readily applicable to protein and RNA sequence data.", "date": "2025-01-30", "categories": []}
{"id": "2505.00316", "title": "Surrogate modeling of Cellular-Potts Agent-Based Models as a segmentation task using the U-Net neural network architecture", "abstract": "The Cellular-Potts model is a powerful and ubiquitous framework for developing computational models for simulating complex multicellular biological systems. Cellular-Potts models (CPMs) are often computationally expensive due to the explicit modeling of interactions among large numbers of individual model agents and diffusive fields described by partial differential equations (PDEs). In this work, we develop a convolutional neural network (CNN) surrogate model using a U-Net architecture that accounts for periodic boundary conditions. We use this model to accelerate the evaluation of a mechanistic CPM previously used to investigate in vitro vasculogenesis. The surrogate model was trained to predict 100 computational steps ahead (Monte-Carlo steps, MCS), accelerating simulation evaluations by a factor of 590 times compared to CPM code execution. Over multiple recursive evaluations, our model effectively captures the emergent behaviors demonstrated by the original Cellular-Potts model of such as vessel sprouting, extension and anastomosis, and contraction of vascular lacunae. This approach demonstrates the potential for deep learning to serve as efficient surrogate models for CPM simulations, enabling faster evaluation of computationally expensive CPM of biological processes at greater spatial and temporal scales.", "date": "2025-05-01", "categories": []}
{"id": "2505.24759", "title": "Unsupervised Evolutionary Cell Type Matching via Entropy-Minimized Optimal Transport", "abstract": "Identifying evolutionary correspondences between cell types across species is a fundamental challenge in comparative genomics and evolutionary biology. Existing approaches often rely on either reference-based matching, which imposes asymmetry by designating one species as the reference, or projection-based matching, which may increase computational complexity and obscure biological interpretability at the cell-type level. Here, we present OT-MESH, an unsupervised computational framework leveraging entropy-regularized optimal transport (OT) to systematically determine cross-species cell type homologies. Our method uniquely integrates the Minimize Entropy of Sinkhorn (MESH) technique to refine the OT plan, transforming diffuse transport matrices into sparse, interpretable correspondences. Through systematic evaluation on synthetic datasets, we demonstrate that OT-MESH achieves near-optimal matching accuracy with computational efficiency, while maintaining remarkable robustness to noise. Compared to other OT-based methods like RefCM, OT-MESH provides speedup while achieving comparable accuracy. Applied to retinal bipolar cells (BCs) and retinal ganglion cells (RGCs) from mouse and macaque, OT-MESH accurately recovers known evolutionary relationships and uncovers novel correspondences, one of which was independently validated experimentally. Thus, our framework offers a principled, scalable, and interpretable solution for evolutionary cell type mapping, facilitating deeper insights into cellular specialization and conservation across species.", "date": "2025-05-30", "categories": []}
{"id": "2509.01486", "title": "Prior-Guided Flow Matching for Target-Aware Molecule Design with Learnable Atom Number", "abstract": "Structure-based drug design (SBDD), aiming to generate 3D molecules with high binding affinity toward target proteins, is a vital approach in novel drug discovery. Although recent generative models have shown great potential, they suffer from unstable probability dynamics and mismatch between generated molecule size and the protein pockets geometry, resulting in inconsistent quality and off-target effects. We propose PAFlow, a novel target-aware molecular generation model featuring prior interaction guidance and a learnable atom number predictor. PAFlow adopts the efficient flow matching framework to model the generation process and constructs a new form of conditional flow matching for discrete atom types. A protein-ligand interaction predictor is incorporated to guide the vector field toward higher-affinity regions during generation, while an atom number predictor based on protein pocket information is designed to better align generated molecule size with target geometry. Extensive experiments on the CrossDocked2020 benchmark show that PAFlow achieves a new state-of-the-art in binding affinity (up to -8.31 Avg. Vina Score), simultaneously maintains favorable molecular properties.", "date": "2025-09-01", "categories": []}
{"id": "2510.16013", "title": "AGNES: Adaptive Graph Neural Network and Dynamic Programming Hybrid Framework for Real-Time Nanopore Seed Chaining", "abstract": "Nanopore sequencing enables real-time long-read DNA sequencing with reads exceeding 10 kilobases, but inherent error rates of 12-15 percent present significant computational challenges for read alignment. The critical seed chaining step must connect exact k-mer matches between reads and reference genomes while filtering spurious matches, yet state-of-the-art methods rely on fixed gap penalty functions unable to adapt to varying genomic contexts including tandem repeats and structural variants. This paper presents RawHash3, a hybrid framework combining graph neural networks with classical dynamic programming for adaptive seed chaining that maintains real-time performance while providing statistical guarantees. We formalize seed chaining as graph learning where seeds constitute nodes with 12-dimensional feature vectors and edges encode 8-dimensional spatial relationships including gap consistency. Our architecture employs three-layer EdgeConv GNN with confidence-based method selection that dynamically switches between learned guidance and algorithmic fallback. Comprehensive evaluation on 1,000 synthetic nanopore reads with 5,200 test seeds demonstrates RawHash3 achieves 99.94 percent precision and 40.07 percent recall, representing statistically significant 25.0 percent relative improvement over baseline with p less than 0.001. The system maintains median inference latency of 1.59ms meeting real-time constraints, while demonstrating superior robustness with 100 percent success rate under 20 percent label corruption versus baseline degradation to 30.3 percent. Cross-validation confirms stability establishing graph neural networks as viable approach for production genomics pipelines.", "date": "2025-10-15", "categories": []}
{"id": "2511.02332", "title": "Biological Regulatory Network Inference through Circular Causal Structure Learning", "abstract": "Biological networks are pivotal in deciphering the complexity and functionality of biological systems. Causal inference, which focuses on determining the directionality and strength of interactions between variables rather than merely relying on correlations, is considered a logical approach for inferring biological networks. Existing methods for causal structure inference typically assume that causal relationships between variables can be represented by directed acyclic graphs (DAGs). However, this assumption is at odds with the reality of widespread feedback loops in biological systems, making these methods unsuitable for direct use in biological network inference. In this study, we propose a new framework named SCALD (Structural CAusal model for Loop Diagram), which employs a nonlinear structure equation model and a stable feedback loop conditional constraint through continuous optimization to infer causal regulatory relationships under feedback loops. We observe that SCALD outperforms state-of-the-art methods in inferring both transcriptional regulatory networks and signaling transduction networks. SCALD has irreplaceable advantages in identifying feedback regulation. Through transcription factor (TF) perturbation data analysis, we further validate the accuracy and sensitivity of SCALD. Additionally, SCALD facilitates the discovery of previously unknown regulatory relationships, which we have subsequently confirmed through ChIP-seq data analysis. Furthermore, by utilizing SCALD, we infer the key driver genes that facilitate the transformation from colon inflammation to cancer by examining the dynamic changes within regulatory networks during the process.", "date": "2025-11-04", "categories": []}
{"id": "2508.04180", "title": "One Small Step with Fingerprints, One Giant Leap for De Novo Molecule Generation from Mass Spectra", "abstract": "A common approach to the de novo molecular generation problem from mass spectra involves a two-stage pipeline: (1) encoding mass spectra into molecular fingerprints, followed by (2) decoding these fingerprints into molecular structures. In our work, we adopt MIST (Goldman et. al., 2023) as the encoder and MolForge (Ucak et. al., 2023) as the decoder, leveraging additional training data to enhance performance. We also threshold the probabilities of each fingerprint bit to focus on the presence of substructures. This results in a tenfold improvement over previous state-of-the-art methods, generating top-1 31% / top-10 40% of molecular structures correctly from mass spectra in MassSpecGym (Bushuiev et. al., 2024). We position this as a strong baseline for future research in de novo molecule elucidation from mass spectra.", "date": "2025-08-06", "categories": []}
{"id": "2511.01277", "title": "Identification of Capture Phases in Nanopore Protein Sequencing Data Using a Deep Learning Model", "abstract": "Nanopore protein sequencing produces long, noisy ionic current traces in which key molecular phases, such as protein capture and translocation, are embedded. Capture phases mark the successful entry of a protein into the pore and serve as both a checkpoint and a signal that a channel merits further analysis. However, manual identification of capture phases is time-intensive, often requiring several days for expert reviewers to annotate the data due to the need for domain-specific interpretation of complex signal patterns. To address this, a lightweight one-dimensional convolutional neural network (1D CNN) was developed and trained to detect capture phases in down-sampled signal windows. Evaluated against CNN-LSTM (Long Short-Term Memory) hybrids, histogram-based classifiers, and other CNN variants using run-level data splits, our best model, CaptureNet-Deep, achieved an F1 score of 0.94 and precision of 93.39% on held-out test data. The model supports low-latency inference and is integrated into a dashboard for Oxford Nanopore experiments, reducing the total analysis time from several days to under thirty minutes. These results show that efficient, real-time capture detection is possible using simple, interpretable architectures and suggest a broader role for lightweight ML models in sequencing workflows.", "date": "2025-11-03", "categories": []}
{"id": "2511.01555", "title": "Fast, memory-efficient genomic interval tokenizers for modern machine learning", "abstract": "Introduction: Epigenomic datasets from high-throughput sequencing experiments are commonly summarized as genomic intervals. As the volume of this data grows, so does interest in analyzing it through deep learning. However, the heterogeneity of genomic interval data, where each dataset defines its own regions, creates barriers for machine learning methods that require consistent, discrete vocabularies. Methods: We introduce gtars-tokenizers, a high-performance library that maps genomic intervals to a predefined universe or vocabulary of regions, analogous to text tokenization in natural language processing. Built in Rust with bindings for Python, R, CLI, and WebAssembly, gtars-tokenizers implements two overlap methods (BITS and AIList) and integrates seamlessly with modern ML frameworks through Hugging Face-compatible APIs. Results: The gtars-tokenizers package achieves top efficiency for large-scale datasets, while enabling genomic intervals to be processed using standard ML workflows in PyTorch and TensorFlow without ad hoc preprocessing. This token-based approach bridges genomics and machine learning, supporting scalable and standardized analysis of interval data across diverse computational environments. Availability: PyPI and GitHub: https://github.com/databio/gtars.", "date": "2025-11-03", "categories": []}
{"id": "2503.20278", "title": "The cell as a token: high-dimensional geometry in language models and cell embeddings", "abstract": "Single-cell sequencing technology maps cells to a high-dimensional space encoding their internal activity. Recently-proposed virtual cell models extend this concept, enriching cells' representations based on patterns learned from pretraining on vast cell atlases. This review explores how advances in understanding the structure of natural language embeddings informs ongoing efforts to analyze single-cell datasets. Both fields process unstructured data by partitioning datasets into tokens embedded within a high-dimensional vector space. We discuss how the context of tokens influences the geometry of embedding space, and how low-dimensional manifolds shape this space's robustness and interpretation. We highlight how new developments in foundation models for language, such as interpretability probes and in-context reasoning, can inform efforts to construct cell atlases and train virtual cell models.", "date": "2025-03-26", "categories": []}
{"id": "2506.11420", "title": "PPDiff: Diffusing in Hybrid Sequence-Structure Space for Protein-Protein Complex Design", "abstract": "Designing protein-binding proteins with high affinity is critical in biomedical research and biotechnology. Despite recent advancements targeting specific proteins, the ability to create high-affinity binders for arbitrary protein targets on demand, without extensive rounds of wet-lab testing, remains a significant challenge. Here, we introduce PPDiff, a diffusion model to jointly design the sequence and structure of binders for arbitrary protein targets in a non-autoregressive manner. PPDiffbuilds upon our developed Sequence Structure Interleaving Network with Causal attention layers (SSINC), which integrates interleaved self-attention layers to capture global amino acid correlations, k-nearest neighbor (kNN) equivariant graph layers to model local interactions in three-dimensional (3D) space, and causal attention layers to simplify the intricate interdependencies within the protein sequence. To assess PPDiff, we curate PPBench, a general protein-protein complex dataset comprising 706,360 complexes from the Protein Data Bank (PDB). The model is pretrained on PPBenchand finetuned on two real-world applications: target-protein mini-binder complex design and antigen-antibody complex design. PPDiffconsistently surpasses baseline methods, achieving success rates of 50.00%, 23.16%, and 16.89% for the pretraining task and the two downstream applications, respectively. The code, data and models are available at https://github.com/JocelynSong/PPDiff.", "date": "2025-06-13", "categories": []}
{"id": "2510.27040", "title": "GeoPep: A geometry-aware masked language model for protein-peptide binding site prediction", "abstract": "Multimodal approaches that integrate protein structure and sequence have achieved remarkable success in protein-protein interface prediction. However, extending these methods to protein-peptide interactions remains challenging due to the inherent conformational flexibility of peptides and the limited availability of structural data that hinder direct training of structure-aware models. To address these limitations, we introduce GeoPep, a novel framework for peptide binding site prediction that leverages transfer learning from ESM3, a multimodal protein foundation model. GeoPep fine-tunes ESM3's rich pre-learned representations from protein-protein binding to address the limited availability of protein-peptide binding data. The fine-tuned model is further integrated with a parameter-efficient neural network architecture capable of learning complex patterns from sparse data. Furthermore, the model is trained using distance-based loss functions that exploit 3D structural information to enhance binding site prediction. Comprehensive evaluations demonstrate that GeoPep significantly outperforms existing methods in protein-peptide binding site prediction by effectively capturing sparse and heterogeneous binding patterns.", "date": "2025-10-30", "categories": []}
{"id": "2510.27281", "title": "HiF-DTA: Hierarchical Feature Learning Network for Drug-Target Affinity Prediction", "abstract": "Accurate prediction of Drug-Target Affinity (DTA) is crucial for reducing experimental costs and accelerating early screening in computational drug discovery. While sequence-based deep learning methods avoid reliance on costly 3D structures, they still overlook simultaneous modeling of global sequence semantic features and local topological structural features within drugs and proteins, and represent drugs as flat sequences without atomic-level, substructural-level, and molecular-level multi-scale features. We propose HiF-DTA, a hierarchical network that adopts a dual-pathway strategy to extract both global sequence semantic and local topological features from drug and protein sequences, and models drugs multi-scale to learn atomic, substructural, and molecular representations fused via a multi-scale bilinear attention module. Experiments on Davis, KIBA, and Metz datasets show HiF-DTA outperforms state-of-the-art baselines, with ablations confirming the importance of global-local extraction and multi-scale fusion.", "date": "2025-10-31", "categories": []}
{"id": "2510.27671", "title": "MolChord: Structure-Sequence Alignment for Protein-Guided Drug Design", "abstract": "Structure-based drug design (SBDD), which maps target proteins to candidate molecular ligands, is a fundamental task in drug discovery. Effectively aligning protein structural representations with molecular representations, and ensuring alignment between generated drugs and their pharmacological properties, remains a critical challenge. To address these challenges, we propose MolChord, which integrates two key techniques: (1) to align protein and molecule structures with their textual descriptions and sequential representations (e.g., FASTA for proteins and SMILES for molecules), we leverage NatureLM, an autoregressive model unifying text, small molecules, and proteins, as the molecule generator, alongside a diffusion-based structure encoder; and (2) to guide molecules toward desired properties, we curate a property-aware dataset by integrating preference data and refine the alignment process using Direct Preference Optimization (DPO). Experimental results on CrossDocked2020 demonstrate that our approach achieves state-of-the-art performance on key evaluation metrics, highlighting its potential as a practical tool for SBDD.", "date": "2025-10-31", "categories": []}
{"id": "2410.01755", "title": "Integrating Protein Sequence and Expression Level to Analysis Molecular Characterization of Breast Cancer Subtypes", "abstract": "Breast cancer's complexity and variability pose significant challenges in understanding its progression and guiding effective treatment. This study aims to integrate protein sequence data with expression levels to improve the molecular characterization of breast cancer subtypes and predict clinical outcomes. Using ProtGPT2, a language model specifically designed for protein sequences, we generated embeddings that capture the functional and structural properties of proteins. These embeddings were integrated with protein expression levels to form enriched biological representations, which were analyzed using machine learning methods, such as ensemble K-means for clustering and XGBoost for classification. Our approach enabled the successful clustering of patients into biologically distinct groups and accurately predicted clinical outcomes such as survival and biomarker status, achieving high performance metrics, notably an F1 score of 0.88 for survival and 0.87 for biomarker status prediction. Feature importance analysis identified KMT2C, CLASP2, and MYO1B as key proteins involved in hormone signaling, cytoskeletal remodeling, and therapy resistance in hormone receptor-positive and triple-negative breast cancer, with potential influence on breast cancer subtype behavior and progression. Furthermore, protein-protein interaction networks and correlation analyses revealed functional interdependencies among proteins that may influence the behavior and progression of breast cancer subtypes. These findings suggest that integrating protein sequence and expression data provides valuable insights into tumor biology and has significant potential to enhance personalized treatment strategies in breast cancer care.", "date": "2024-10-02", "categories": []}
{"id": "2510.22033", "title": "Linearized Optimal Transport for Analysis of High-Dimensional Point-Cloud and Single-Cell Data", "abstract": "Single-cell technologies generate high-dimensional point clouds of cells, enabling detailed characterization of complex patient states and treatment responses. Yet each patient is represented by an irregular point cloud rather than a simple vector, making it difficult to directly quantify and compare biological differences between individuals. Nonlinear methods such as kernels and neural networks achieve predictive accuracy but act as black boxes, offering little biological interpretability.\n  To address these limitations, we adapt the Linear Optimal Transport (LOT) framework to this setting, embedding irregular point clouds into a fixed-dimensional Euclidean space while preserving distributional structure. This embedding provides a principled linear representation that preserves optimal transport geometry while enabling downstream analysis. It also forms a registration between any two patients, enabling direct comparison of their cellular distributions. Within this space, LOT enables: (i) \\textbf{accurate and interpretable classification} of COVID-19 patient states, where classifier weights map back to specific markers and spatial regions driving predictions; and (ii) \\textbf{synthetic data generation} for patient-derived organoids, exploiting the linearity of the LOT embedding. LOT barycenters yield averaged cellular profiles representing combined conditions or samples, supporting drug interaction testing.\n  Together, these results establish LOT as a unified framework that bridges predictive performance, interpretability, and generative modeling. By transforming heterogeneous point clouds into structured embeddings directly traceable to the original data, LOT opens new opportunities for understanding immune variation and treatment effects in high-dimensional biological systems.", "date": "2025-10-24", "categories": []}
{"id": "2510.25807", "title": "Discovering Interpretable Biological Concepts in Single-cell RNA-seq Foundation Models", "abstract": "Single-cell RNA-seq foundation models achieve strong performance on downstream tasks but remain black boxes, limiting their utility for biological discovery. Recent work has shown that sparse dictionary learning can extract concepts from deep learning models, with promising applications in biomedical imaging and protein models. However, interpreting biological concepts remains challenging, as biological sequences are not inherently human-interpretable. We introduce a novel concept-based interpretability framework for single-cell RNA-seq models with a focus on concept interpretation and evaluation. We propose an attribution method with counterfactual perturbations that identifies genes that influence concept activation, moving beyond correlational approaches like differential expression analysis. We then provide two complementary interpretation approaches: an expert-driven analysis facilitated by an interactive interface and an ontology-driven method with attribution-based biological pathway enrichment. Applying our framework to two well-known single-cell RNA-seq models from the literature, we interpret concepts extracted by Top-K Sparse Auto-Encoders trained on two immune cell datasets. With a domain expert in immunology, we show that concepts improve interpretability compared to individual neurons while preserving the richness and informativeness of the latent representations. This work provides a principled framework for interpreting what biological knowledge foundation models have encoded, paving the way for their use for hypothesis generation and discovery.", "date": "2025-10-29", "categories": []}
{"id": "2510.26715", "title": "LSM-MS2: A Foundation Model Bridging Spectral Identification and Biological Interpretation", "abstract": "A vast majority of mass spectrometry data remains uncharacterized, leaving much of its biological and chemical information untapped. Recent advances in machine learning have begun to address this gap, particularly for tasks such as spectral identification in tandem mass spectrometry data. Here, we present the latest generation of LSM-MS2, a large-scale deep learning foundation model trained on millions of spectra to learn a semantic chemical space. LSM-MS2 achieves state-of-the-art performance in spectral identification, improving on existing methods by 30% in accuracy of identifying challenging isomeric compounds, yielding 42% more correct identifications in complex biological samples, and maintaining robustness under low-concentration conditions. Furthermore, LSM-MS2 produces rich spectral embeddings that enable direct biological interpretation from minimal downstream data, successfully differentiating disease states and predicting clinical outcomes across diverse translational applications.", "date": "2025-10-30", "categories": []}
{"id": "2408.09604", "title": "Circuit design in biology and machine learning. I. Random networks and dimensional reduction", "abstract": "A biological circuit is a neural or biochemical cascade, taking inputs and producing outputs. How have biological circuits learned to solve environmental challenges over the history of life? The answer certainly follows Dobzhansky's famous quote that ``nothing in biology makes sense except in the light of evolution.'' But that quote leaves out the mechanistic basis by which natural selection's trial-and-error learning happens, which is exactly what we have to understand. How does the learning process that designs biological circuits actually work? How much insight can we gain about the form and function of biological circuits by studying the processes that have made those circuits? Because life's circuits must often solve the same problems as those faced by machine learning, such as environmental tracking, homeostatic control, dimensional reduction, or classification, we can begin by considering how machine learning designs computational circuits to solve problems. We can then ask: How much insight do those computational circuits provide about the design of biological circuits? How much does biology differ from computers in the particular circuit designs that it uses to solve problems? This article steps through two classic machine learning models to set the foundation for analyzing broad questions about the design of biological circuits. One insight is the surprising power of randomly connected networks. Another is the central role of internal models of the environment embedded within biological circuits, illustrated by a model of dimensional reduction and trend prediction. Overall, many challenges in biology have machine learning analogs, suggesting hypotheses about how biology's circuits are designed.", "date": "2024-08-18", "categories": []}
{"id": "2504.17247", "title": "OmegAMP: Targeted AMP Discovery through Biologically Informed Generation", "abstract": "Deep learning-based antimicrobial peptide (AMP) discovery faces critical challenges such as limited controllability, lack of representations that efficiently model antimicrobial properties, and low experimental hit rates. To address these challenges, we introduce OmegAMP, a framework designed for reliable AMP generation with increased controllability. Its diffusion-based generative model leverages a novel conditioning mechanism to achieve fine-grained control over desired physicochemical properties and to direct generation towards specific activity profiles, including species-specific effectiveness. This is further enhanced by a biologically informed encoding space that significantly improves overall generative performance. Complementing these generative capabilities, OmegAMP leverages a novel synthetic data augmentation strategy to train classifiers for AMP filtering, drastically reducing false positive rates and thereby increasing the likelihood of experimental success. Our in silico experiments demonstrate that OmegAMP delivers state-of-the-art performance across key stages of the AMP discovery pipeline, enabling us to achieve an unprecedented success rate in wet lab experiments. We tested 25 candidate peptides, 24 of them (96%) demonstrated antimicrobial activity, proving effective even against multi-drug resistant strains. Our findings underscore OmegAMP's potential to significantly advance computational frameworks in the fight against antimicrobial resistance.", "date": "2025-04-24", "categories": []}
{"id": "2510.24670", "title": "Pearl: A Foundation Model for Placing Every Atom in the Right Location", "abstract": "Accurately predicting the three-dimensional structures of protein-ligand complexes remains a fundamental challenge in computational drug discovery that limits the pace and success of therapeutic design. Deep learning methods have recently shown strong potential as structural prediction tools, achieving promising accuracy across diverse biomolecular systems. However, their performance and utility are constrained by scarce experimental data, inefficient architectures, physically invalid poses, and the limited ability to exploit auxiliary information available at inference. To address these issues, we introduce Pearl (Placing Every Atom in the Right Location), a foundation model for protein-ligand cofolding at scale. Pearl addresses these challenges with three key innovations: (1) training recipes that include large-scale synthetic data to overcome data scarcity; (2) architectures that incorporate an SO(3)-equivariant diffusion module to inherently respect 3D rotational symmetries, improving generalization and sample efficiency, and (3) controllable inference, including a generalized multi-chain templating system supporting both protein and non-polymeric components as well as dual unconditional/conditional modes. Pearl establishes a new state-of-the-art performance in protein-ligand cofolding. On the key metric of generating accurate (RMSD < 2 \\r{A}) and physically valid poses, Pearl surpasses AlphaFold 3 and other open source baselines on the public Runs N' Poses and PoseBusters benchmarks, delivering 14.5% and 14.2% improvements, respectively, over the next best model. In the pocket-conditional cofolding regime, Pearl delivers $3.6\\times$ improvement on a proprietary set of challenging, real-world drug targets at the more rigorous RMSD < 1 \\r{A} threshold. Finally, we demonstrate that model performance correlates directly with synthetic dataset size used in training.", "date": "2025-10-28", "categories": []}
{"id": "2510.24732", "title": "Flows, straight but not so fast: Exploring the design space of Rectified Flows in Protein Design", "abstract": "Generative modeling techniques such as Diffusion and Flow Matching have achieved significant successes in generating designable and diverse protein backbones. However, many current models are computationally expensive, requiring hundreds or even thousands of function evaluations (NFEs) to yield samples of acceptable quality, which can become a bottleneck in practical design campaigns that often generate $10^4\\ -\\ 10^6$ designs per target. In image generation, Rectified Flows (ReFlow) can significantly reduce the required NFEs for a given target quality, but their application in protein backbone generation has been less studied. We apply ReFlow to improve the low NFE performance of pretrained SE(3) flow matching models for protein backbone generation and systematically study ReFlow design choices in the context of protein generation in data curation, training and inference time settings. In particular, we (1) show that ReFlow in the protein domain is particularly sensitive to the choice of coupling generation and annealing, (2) demonstrate how useful design choices for ReFlow in the image domain do not directly translate to better performance on proteins, and (3) make improvements to ReFlow methodology for proteins.", "date": "2025-10-13", "categories": []}
{"id": "2510.24826", "title": "Augmenting Biological Fitness Prediction Benchmarks with Landscapes Features from GraphFLA", "abstract": "Machine learning models increasingly map biological sequence-fitness landscapes to predict mutational effects. Effective evaluation of these models requires benchmarks curated from empirical data. Despite their impressive scales, existing benchmarks lack topographical information regarding the underlying fitness landscapes, which hampers interpretation and comparison of model performance beyond averaged scores. Here, we introduce GraphFLA, a Python framework that constructs and analyzes fitness landscapes from mutagensis data in diverse modalities (e.g., DNA, RNA, protein, and beyond) with up to millions of mutants. GraphFLA calculates 20 biologically relevant features that characterize 4 fundamental aspects of landscape topography. By applying GraphFLA to over 5,300 landscapes from ProteinGym, RNAGym, and CIS-BP, we demonstrate its utility in interpreting and comparing the performance of dozens of fitness prediction models, highlighting factors influencing model accuracy and respective advantages of different models. In addition, we release 155 combinatorially complete empirical fitness landscapes, encompassing over 2.2 million sequences across various modalities. All the codes and datasets are available at https://github.com/COLA-Laboratory/GraphFLA.", "date": "2025-10-28", "categories": []}
{"id": "2504.09060", "title": "Multimodal 3D Genome Pre-training", "abstract": "Deep learning techniques have driven significant progress in various analytical tasks within 3D genomics in computational biology. However, a holistic understanding of 3D genomics knowledge remains underexplored. Here, we propose MIX-HIC, the first multimodal foundation model of 3D genome that integrates both 3D genome structure and epigenomic tracks, which obtains unified and comprehensive semantics. For accurate heterogeneous semantic fusion, we design the cross-modal interaction and mapping blocks for robust unified representation, yielding the accurate aggregation of 3D genome knowledge. Besides, we introduce the first large-scale dataset comprising over 1 million pairwise samples of Hi-C contact maps and epigenomic tracks for high-quality pre-training, enabling the exploration of functional implications in 3D genomics. Extensive experiments show that MIX-HIC can significantly surpass existing state-of-the-art methods in diverse downstream tasks. This work provides a valuable resource for advancing 3D genomics research.", "date": "2025-04-12", "categories": []}
{"id": "2505.17257", "title": "JanusDNA: A Powerful Bi-directional Hybrid DNA Foundation Model", "abstract": "Large language models (LLMs) have revolutionized natural language processing and are increasingly applied to other sequential data types, including genetic sequences. However, adapting LLMs to genomics presents significant challenges. Capturing complex genomic interactions requires modeling long-range dependencies within DNA sequences, where interactions often span over 10,000 base pairs, even within a single gene, posing substantial computational burdens under conventional model architectures and training paradigms. Moreover, standard LLM training approaches are suboptimal for DNA: autoregressive training, while efficient, supports only unidirectional understanding. However, DNA is inherently bidirectional, e.g., bidirectional promoters regulate transcription in both directions and account for nearly 11% of human gene expression. Masked language models (MLMs) allow bidirectional understanding but are inefficient, as only masked tokens contribute to the loss per step. To address these limitations, we introduce JanusDNA, the first bidirectional DNA foundation model built upon a novel pretraining paradigm that combines the optimization efficiency of autoregressive modeling with the bidirectional comprehension of masked modeling. JanusDNA adopts a hybrid Mamba, Attention and Mixture of Experts (MoE) architecture, combining long-range modeling of Attention with efficient sequential learning of Mamba. MoE layers further scale model capacity via sparse activation while keeping computational cost low. Notably, JanusDNA processes up to 1 million base pairs at single nucleotide resolution on a single 80GB GPU. Extensive experiments and ablations show JanusDNA achieves new SOTA results on three genomic representation benchmarks, outperforming models with 250x more activated parameters. Code: https://github.com/Qihao-Duan/JanusDNA", "date": "2025-05-22", "categories": []}
{"id": "2510.23786", "title": "Relaxed Sequence Sampling for Diverse Protein Design", "abstract": "Protein design using structure prediction models such as AlphaFold2 has shown remarkable success, but existing approaches like relaxed sequence optimization (RSO) rely on single-path gradient descent and ignore sequence-space constraints, limiting diversity and designability. We introduce Relaxed Sequence Sampling (RSS), a Markov chain Monte Carlo (MCMC) framework that integrates structural and evolutionary information for protein design. RSS operates in continuous logit space, combining gradient-guided exploration with protein language model-informed jumps. Its energy function couples AlphaFold2-derived structural objectives with ESM2-derived sequence priors, balancing accuracy and biological plausibility. In an in silico protein binder design task, RSS produces 5$\\times$ more designable structures and 2-3$\\times$ greater structural diversity than RSO baselines, at equal computational cost. These results highlight RSS as a principled approach for efficiently exploring the protein design landscape.", "date": "2025-10-27", "categories": []}
{"id": "2510.24053", "title": "Low-N Protein Activity Optimization with FolDE", "abstract": "Proteins are traditionally optimized through the costly construction and measurement of many mutants. Active Learning-assisted Directed Evolution (ALDE) alleviates that cost by predicting the best improvements and iteratively testing mutants to inform predictions. However, existing ALDE methods face a critical limitation: selecting the highest-predicted mutants in each round yields homogeneous training data insufficient for accurate prediction models in subsequent rounds. Here we present FolDE, an ALDE method designed to maximize end-of-campaign success. In simulations across 20 protein targets, FolDE discovers 23% more top 10% mutants than the best baseline ALDE method (p=0.005) and is 55% more likely to find top 1% mutants. FolDE achieves this primarily through naturalness-based warm-starting, which augments limited activity measurements with protein language model outputs to improve activity prediction. We also introduce a constant-liar batch selector, which improves batch diversity; this is important in multi-mutation campaigns but had limited effect in our benchmarks. The complete workflow is freely available as open-source software, making efficient protein optimization accessible to any laboratory.", "date": "2025-10-28", "categories": []}
{"id": "2403.06890", "title": "Application of Quantum Tensor Networks for Protein Classification", "abstract": "We show that protein sequences can be thought of as sentences in natural language processing and can be parsed using the existing Quantum Natural Language framework into parameterized quantum circuits of reasonable qubits, which can be trained to solve various protein-related machine-learning problems. We classify proteins based on their subcellular locations, a pivotal task in bioinformatics that is key to understanding biological processes and disease mechanisms. Leveraging the quantum-enhanced processing capabilities, we demonstrate that Quantum Tensor Networks (QTN) can effectively handle the complexity and diversity of protein sequences. We present a detailed methodology that adapts QTN architectures to the nuanced requirements of protein data, supported by comprehensive experimental results. We demonstrate two distinct QTNs, inspired by classical recurrent neural networks (RNN) and convolutional neural networks (CNN), to solve the binary classification task mentioned above. Our top-performing quantum model has achieved a 94% accuracy rate, which is comparable to the performance of a classical model that uses the ESM2 protein language model embeddings. It's noteworthy that the ESM2 model is extremely large, containing 8 million parameters in its smallest configuration, whereas our best quantum model requires only around 800 parameters. We demonstrate that these hybrid models exhibit promising performance, showcasing their potential to compete with classical models of similar complexity.", "date": "2024-03-11", "categories": []}
{"id": "2408.10609", "title": "PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis", "abstract": "We introduce a comprehensive framework for modeling single cell transcriptomic responses to perturbations, aimed at standardizing benchmarking in this rapidly evolving field. Our approach includes a modular and user-friendly model development and evaluation platform, a collection of diverse perturbational datasets, and a set of metrics designed to fairly compare models and dissect their performance. Through extensive evaluation of both published and baseline models across diverse datasets, we highlight the limitations of widely used models, such as mode collapse. We also demonstrate the importance of rank metrics which complement traditional model fit measures, such as RMSE, for validating model effectiveness. Notably, our results show that while no single model architecture clearly outperforms others, simpler architectures are generally competitive and scale well with larger datasets. Overall, this benchmarking exercise sets new standards for model evaluation, supports robust model development, and furthers the use of these models to simulate genetic and chemical screens for therapeutic discovery.", "date": "2024-08-20", "categories": []}
{"id": "2505.11197", "title": "Modeling Cell Dynamics and Interactions with Unbalanced Mean Field Schr\\\"odinger Bridge", "abstract": "Modeling the dynamics from sparsely time-resolved snapshot data is crucial for understanding complex cellular processes and behavior. Existing methods leverage optimal transport, Schr\\\"odinger bridge theory, or their variants to simultaneously infer stochastic, unbalanced dynamics from snapshot data. However, these approaches remain limited in their ability to account for cell-cell interactions. This integration is essential in real-world scenarios since intercellular communications are fundamental life processes and can influence cell state-transition dynamics. To address this challenge, we formulate the Unbalanced Mean-Field Schr\\\"odinger Bridge (UMFSB) framework to model unbalanced stochastic interaction dynamics from snapshot data. Inspired by this framework, we further propose CytoBridge, a deep learning algorithm designed to approximate the UMFSB problem. By explicitly modeling cellular transitions, proliferation, and interactions through neural networks, CytoBridge offers the flexibility to learn these processes directly from data. The effectiveness of our method has been extensively validated using both synthetic gene regulatory data and real scRNA-seq datasets. Compared to existing methods, CytoBridge identifies growth, transition, and interaction patterns, eliminates false transitions, and reconstructs the developmental landscape with greater accuracy. Code is available at: https://github.com/zhenyiizhang/CytoBridge-NeurIPS.", "date": "2025-05-16", "categories": []}
{"id": "2505.12638", "title": "ChromFound: Towards A Universal Foundation Model for Single-Cell Chromatin Accessibility Data", "abstract": "The advent of single-cell Assay for Transposase-Accessible Chromatin using sequencing (scATAC-seq) offers an innovative perspective for deciphering regulatory mechanisms by assembling a vast repository of single-cell chromatin accessibility data. While foundation models have achieved significant success in single-cell transcriptomics, there is currently no foundation model for scATAC-seq that supports zero-shot high-quality cell identification and comprehensive multi-omics analysis simultaneously. Key challenges lie in the high dimensionality and sparsity of scATAC-seq data, as well as the lack of a standardized schema for representing open chromatin regions (OCRs). Here, we present ChromFound, a foundation model tailored for scATAC-seq. ChromFound utilizes a hybrid architecture and genome-aware tokenization to effectively capture genome-wide long contexts and regulatory signals from dynamic chromatin landscapes. Pretrained on 1.97 million cells from 30 tissues and 6 disease conditions, ChromFound demonstrates broad applicability across 6 diverse tasks. Notably, it achieves robust zero-shot performance in generating universal cell representations and exhibits excellent transferability in cell type annotation and cross-omics prediction. By uncovering enhancer-gene links undetected by existing computational methods, ChromFound offers a promising framework for understanding disease risk variants in the noncoding genome.", "date": "2025-05-19", "categories": []}
{"id": "2505.22494", "title": "ProSpero: Active Learning for Robust Protein Design Beyond Wild-Type Neighborhoods", "abstract": "Designing protein sequences of both high fitness and novelty is a challenging task in data-efficient protein engineering. Exploration beyond wild-type neighborhoods often leads to biologically implausible sequences or relies on surrogate models that lose fidelity in novel regions. Here, we propose ProSpero, an active learning framework in which a frozen pre-trained generative model is guided by a surrogate updated from oracle feedback. By integrating fitness-relevant residue selection with biologically-constrained Sequential Monte Carlo sampling, our approach enables exploration beyond wild-type neighborhoods while preserving biological plausibility. We show that our framework remains effective even when the surrogate is misspecified. ProSpero consistently outperforms or matches existing methods across diverse protein engineering tasks, retrieving sequences of both high fitness and novelty.", "date": "2025-05-28", "categories": []}
{"id": "2510.22008", "title": "A Multimodal Human Protein Embeddings Database: DeepDrug Protein Embeddings Bank (DPEB)", "abstract": "Computationally predicting protein-protein interactions (PPIs) is challenging due to the lack of integrated, multimodal protein representations. DPEB is a curated collection of 22,043 human proteins that integrates four embedding types: structural (AlphaFold2), transformer-based sequence (BioEmbeddings), contextual amino acid patterns (ESM-2: Evolutionary Scale Modeling), and sequence-based n-gram statistics (ProtVec]). AlphaFold2 protein structures are available through public databases (e.g., AlphaFold2 Protein Structure Database), but the internal neural network embeddings are not. DPEB addresses this gap by providing AlphaFold2-derived embeddings for computational modeling. Our benchmark evaluations show GraphSAGE with BioEmbedding achieved the highest PPI prediction performance (87.37% AUROC, 79.16% accuracy). The framework also achieved 77.42% accuracy for enzyme classification and 86.04% accuracy for protein family classification. DPEB supports multiple graph neural network methods for PPI prediction, enabling applications in systems biology, drug target identification, pathway analysis, and disease mechanism studies.", "date": "2025-10-24", "categories": []}
{"id": "2402.01744", "title": "Unveiling Molecular Moieties through Hierarchical Grad-CAM Graph Explainability", "abstract": "Background: Virtual Screening (VS) has become an essential tool in drug discovery, enabling the rapid and cost-effective identification of potential bioactive molecules. Among recent advancements, Graph Neural Networks (GNNs) have gained prominence for their ability to model complex molecular structures using graph-based representations. However, the integration of explainable methods to elucidate the specific contributions of molecular substructures to biological activity remains a significant challenge. This limitation hampers both the interpretability of predictive models and the rational design of novel therapeutics. Results: We trained 20 GNN models on a dataset of small molecules with the goal of predicting their activity on 20 distinct protein targets from the Kinase family. These classifiers achieved state-of-the-art performance in virtual screening tasks, demonstrating high accuracy and robustness on different targets. Building upon these models, we implemented the Hierarchical Grad-CAM graph Explainer (HGE) framework, enabling an in-depth analysis of the molecular moieties driving protein-ligand binding stabilization. HGE exploits Grad-CAM explanations at the atom, ring, and whole-molecule levels, leveraging the message-passing mechanism to highlight the most relevant chemical moieties. Validation against experimental data from the literature confirmed the ability of the explainer to recognize a molecular pattern of drugs and correctly annotate them to the known target. Conclusion: Our approach may represent a valid support to shorten both the screening and the hit discovery process. Detailed knowledge of the molecular substructures that play a role in the binding process can help the computational chemist to gain insights into the structure optimization, as well as in drug repurposing tasks.", "date": "2024-01-29", "categories": []}
{"id": "2510.21337", "title": "Morphologically Intelligent Perturbation Prediction with FORM", "abstract": "Understanding how cells respond to external stimuli is a central challenge in biomedical research and drug development. Current computational frameworks for modelling cellular responses remain restricted to two-dimensional representations, limiting their capacity to capture the complexity of cell morphology under perturbation. This dimensional constraint poses a critical bottleneck for the development of accurate virtual cell models. Here, we present FORM, a machine learning framework for predicting perturbation-induced changes in three-dimensional cellular structure. FORM consists of two components: a morphology encoder, trained end-to-end via a novel multi-channel VQGAN to learn compact 3D representations of cell shape, and a diffusion-based perturbation trajectory module that captures how morphology evolves across perturbation conditions. Trained on a large-scale dataset of over 65,000 multi-fluorescence 3D cell volumes spanning diverse chemical and genetic perturbations, FORM supports both unconditional morphology synthesis and conditional simulation of perturbed cell states. Beyond generation, FORM can predict downstream signalling activity, simulate combinatorial perturbation effects, and model morphodynamic transitions between states of unseen perturbations. To evaluate performance, we introduce MorphoEval, a benchmarking suite that quantifies perturbation-induced morphological changes in structural, statistical, and biological dimensions. Together, FORM and MorphoEval work toward the realisation of the 3D virtual cell by linking morphology, perturbation, and function through high-resolution predictive simulation.", "date": "2025-10-24", "categories": []}
{"id": "2510.21663", "title": "Self-Supervised Learning of Synapse Types from EM Images", "abstract": "Separating synapses into different classes based on their appearance in EM images has many applications in biology. Examples may include assigning a neurotransmitter to a particular class, or separating synapses whose strength can be modulated from those whose strength is fixed. Traditionally, this has been done in a supervised manner, giving the classification algorithm examples of the different classes. Here we instead separate synapses into classes based only on the observation that nearby synapses in the same neuron are likely more similar than synapses chosen randomly from different cells. We apply our methodology to data from {\\it Drosophila}. Our approach has the advantage that the number of synapse types does not need to be known in advance. It may also provide a principled way to select ground-truth that spans the range of synapse structure.", "date": "2025-10-24", "categories": []}
{"id": "2410.14621", "title": "JAMUN: Bridging Smoothed Molecular Dynamics and Score-Based Learning for Conformational Ensembles", "abstract": "Conformational ensembles of protein structures are immensely important both for understanding protein function and drug discovery in novel modalities such as cryptic pockets. Current techniques for sampling ensembles such as molecular dynamics (MD) are computationally inefficient, while many recent machine learning methods do not transfer to systems outside their training data. We propose JAMUN which performs MD in a smoothed, noised space of all-atom 3D conformations of molecules by utilizing the framework of walk-jump sampling. JAMUN enables ensemble generation for small peptides at rates of an order of magnitude faster than traditional molecular dynamics. The physical priors in JAMUN enables transferability to systems outside of its training data, even to peptides that are longer than those originally trained on. Our model, code and weights are available at https://github.com/prescient-design/jamun.", "date": "2024-10-18", "categories": []}
{"id": "2412.20984", "title": "Pareto-Optimal Energy Alignment for Designing Nature-Like Antibodies", "abstract": "We present a three-stage framework for training deep learning models specializing in antibody sequence-structure co-design. We first pre-train a language model using millions of antibody sequence data. Then, we employ the learned representations to guide the training of a diffusion model for joint optimization over both sequence and structure of antibodies. During the final alignment stage, we optimize the model to favor antibodies with low repulsion and high attraction to the antigen binding site, enhancing the rationality and functionality of the designs. To mitigate conflicting energy preferences, we extend AbDPO (Antibody Direct Preference Optimization) to guide the model toward Pareto optimality under multiple energy-based alignment objectives. Furthermore, we adopt an iterative learning paradigm with temperature scaling, enabling the model to benefit from diverse online datasets without requiring additional data. In practice, our proposed methods achieve high stability and efficiency in producing a better Pareto front of antibody designs compared to top samples generated by baselines and previous alignment techniques. Through extensive experiments, we showcase the superior performance of our methods in generating nature-like antibodies with high binding affinity.", "date": "2024-12-30", "categories": []}
{"id": "2505.09630", "title": "Generative diffusion model surrogates for mechanistic agent-based biological models", "abstract": "Mechanistic, multicellular, agent-based models are commonly used to investigate tissue, organ, and organism-scale biology at single-cell resolution. The Cellular-Potts Model (CPM) is a powerful and popular framework for developing and interrogating these models. CPMs become computationally expensive at large space- and time- scales making application and investigation of developed models difficult. Surrogate models may allow for the accelerated evaluation of CPMs of complex biological systems. However, the stochastic nature of these models means each set of parameters may give rise to different model configurations, complicating surrogate model development. In this work, we leverage denoising diffusion probabilistic models to train a generative AI surrogate of a CPM used to investigate in vitro vasculogenesis. We describe the use of an image classifier to learn the characteristics that define unique areas of a 2-dimensional parameter space. We then apply this classifier to aid in surrogate model selection and verification. Our CPM model surrogate generates model configurations 20,000 timesteps ahead of a reference configuration and demonstrates approximately a 22x reduction in computational time as compared to native code execution. Our work represents a step towards the implementation of DDPMs to develop digital twins of stochastic biological systems.", "date": "2025-05-01", "categories": []}
{"id": "2510.12384", "title": "Phenome-Wide Multi-Omics Integration Uncovers Distinct Archetypes of Human Aging", "abstract": "Aging is a highly complex and heterogeneous process that progresses at different rates across individuals, making biological age (BA) a more accurate indicator of physiological decline than chronological age. While previous studies have built aging clocks using single-omics data, they often fail to capture the full molecular complexity of human aging. In this work, we leveraged the Human Phenotype Project, a large-scale cohort of 10,000 adults aged 40-70 years, with extensive longitudinal profiling that includes clinical, behavioral, environmental, and multi-omics datasets spanning transcriptomics, lipidomics, metabolomics, and the microbiome. By employing advanced machine learning frameworks capable of modeling nonlinear biological dynamics, we developed and rigorously validated a multi-omics aging clock that robustly predicts diverse health outcomes and future disease risk. Unsupervised clustering of the integrated molecular profiles from multi-omics uncovered distinct biological subtypes of aging, revealing striking heterogeneity in aging trajectories and pinpointing pathway-specific alterations associated with different aging patterns. These findings demonstrate the power of multi-omics integration to decode the molecular landscape of aging and lay the groundwork for personalized healthspan monitoring and precision strategies to prevent age-related diseases.", "date": "2025-10-14", "categories": []}
{"id": "2510.19870", "title": "Transforming Multi-Omics Integration with GANs: Applications in Alzheimer's and Cancer", "abstract": "Multi-omics data integration is crucial for understanding complex diseases, yet limited sample sizes, noise, and heterogeneity often reduce predictive power. To address these challenges, we introduce Omics-GAN, a Generative Adversarial Network (GAN)-based framework designed to generate high-quality synthetic multi-omics profiles while preserving biological relationships. We evaluated Omics-GAN on three omics types (mRNA, miRNA, and DNA methylation) using the ROSMAP cohort for Alzheimer's disease (AD) and TCGA datasets for colon and liver cancer. A support vector machine (SVM) classifier with repeated 5-fold cross-validation demonstrated that synthetic datasets consistently improved prediction accuracy compared to original omics profiles. The AUC of SVM for mRNA improved from 0.72 to 0.74 in AD, and from 0.68 to 0.72 in liver cancer. Synthetic miRNA enhanced classification in colon cancer from 0.59 to 0.69, while synthetic methylation data improved performance in liver cancer from 0.64 to 0.71. Boxplot analyses confirmed that synthetic data preserved statistical distributions while reducing noise and outliers. Feature selection identified significant genes overlapping with original datasets and revealed additional candidates validated by GO and KEGG enrichment analyses. Finally, molecular docking highlighted potential drug repurposing candidates, including Nilotinib for AD, Atovaquone for liver cancer, and Tecovirimat for colon cancer. Omics-GAN enhances disease prediction, preserves biological fidelity, and accelerates biomarker and drug discovery, offering a scalable strategy for precision medicine applications.", "date": "2025-10-22", "categories": []}
{"id": "2510.19887", "title": "Compressing Biology: Evaluating the Stable Diffusion VAE for Phenotypic Drug Discovery", "abstract": "High-throughput phenotypic screens generate vast microscopy image datasets that push the limits of generative models due to their large dimensionality. Despite the growing popularity of general-purpose models trained on natural images for microscopy data analysis, their suitability in this domain has not been quantitatively demonstrated. We present the first systematic evaluation of Stable Diffusion's variational autoencoder (SD-VAE) for reconstructing Cell Painting images, assessing performance across a large dataset with diverse molecular perturbations and cell types. We find that SD-VAE reconstructions preserve phenotypic signals with minimal loss, supporting its use in microscopy workflows. To benchmark reconstruction quality, we compare pixel-level, embedding-based, latent-space, and retrieval-based metrics for a biologically informed evaluation. We show that general-purpose feature extractors like InceptionV3 match or surpass publicly available bespoke models in retrieval tasks, simplifying future pipelines. Our findings offer practical guidelines for evaluating generative models on microscopy data and support the use of off-the-shelf models in phenotypic drug discovery.", "date": "2025-10-22", "categories": []}
{"id": "2309.16519", "title": "AtomSurf : Surface Representation for Learning on Protein Structures", "abstract": "While there has been significant progress in evaluating and comparing different representations for learning on protein data, the role of surface-based learning approaches remains not well-understood. In particular, there is a lack of direct and fair benchmark comparison between the best available surface-based learning methods against alternative representations such as graphs. Moreover, the few existing surface-based approaches either use surface information in isolation or, at best, perform global pooling between surface and graph-based architectures.\n  In this work, we fill this gap by first adapting a state-of-the-art surface encoder for protein learning tasks. We then perform a direct and fair comparison of the resulting method against alternative approaches within the Atom3D benchmark, highlighting the limitations of pure surface-based learning. Finally, we propose an integrated approach, which allows learned feature sharing between graphs and surface representations on the level of nodes and vertices across all layers.\n  We demonstrate that the resulting architecture achieves state-of-the-art results on all tasks in the Atom3D benchmark, while adhering to the strict benchmark protocol, as well as more broadly on binding site identification and binding pocket classification. Furthermore, we use coarsened surfaces and optimize our approach for efficiency, making our tool competitive in training and inference time with existing techniques. Code can be found online: https://github.com/Vincentx15/atomsurf", "date": "2023-09-28", "categories": []}
{"id": "2503.21681", "title": "A Comprehensive Benchmark for RNA 3D Structure-Function Modeling", "abstract": "The relationship between RNA structure and function has recently attracted interest within the deep learning community, a trend expected to intensify as nucleic acid structure models advance. Despite this momentum, the lack of standardized, accessible benchmarks for applying deep learning to RNA 3D structures hinders progress. To this end, we introduce a collection of seven benchmarking datasets specifically designed to support RNA structure-function prediction. Built on top of the established Python package rnaglib, our library streamlines data distribution and encoding, provides tools for dataset splitting and evaluation, and offers a comprehensive, user-friendly environment for model comparison. The modular and reproducible design of our datasets encourages community contributions and enables rapid customization. To demonstrate the utility of our benchmarks, we report baseline results for all tasks using a relational graph neural network.", "date": "2025-03-27", "categories": []}
{"id": "2504.08437", "title": "Customizing Spider Silk: Generative Models with Mechanical Property Conditioning for Protein Engineering", "abstract": "The remarkable mechanical properties of spider silk, including its tensile strength and extensibility, are primarily governed by the repetitive regions of the proteins that constitute the fiber, the major ampullate spidroins (MaSps). However, establishing correlations between mechanical characteristics and repeat sequences is challenging due to the intricate sequence-structure-function relationships of MaSps and the limited availability of annotated datasets. In this study, we present a novel computational framework for designing MaSp repeat sequences with customizable mechanical properties. To achieve this, we developed a lightweight GPT-based generative model by distilling the pre-trained ProtGPT2 protein language model. The distilled model was subjected to multilevel fine-tuning using curated subsets of the Spider Silkome dataset. Specifically, we adapt the model for MaSp repeat generation using 6,000 MaSp repeat sequences and further refine it with 572 repeats associated with experimentally determined fiber-level mechanical properties. Our model generates biologically plausible MaSp repeat regions tailored to specific mechanical properties while also predicting those properties for given sequences. Validation includes sequence-level analysis, assessing physicochemical attributes and expected distribution of key motifs as well as secondary structure compositions. A correlation study using BLAST on the Spider Silkome dataset and a test set of MaSp repeats with known mechanical properties further confirmed the predictive accuracy of the model. This framework advances the rational design of spider silk-inspired biomaterials, offering a versatile tool for engineering protein sequences with tailored mechanical attributes.", "date": "2025-04-11", "categories": []}
{"id": "2509.24262", "title": "LAMP-PRo: Label-aware Attention for Multi-label Prediction of DNA- and RNA-binding Proteins using Protein Language Models", "abstract": "Identifying DNA- (DBPs) and RNA-binding proteins (RBPs) is crucial for the understanding of cell function, molecular interactions as well as regulatory functions. Owing to their high similarity, most of the existing approaches face challenges in differentiating between DBPs and RBPs leading to high cross-prediction errors. Moreover, identifying proteins which bind to both DNA and RNA (DRBPs) is also quite a challenging task. In this regard, we propose a novel framework viz. LAMP-PRo which is based on pre-trained protein language model (PLM), attention mechanisms and multi-label learning to mitigate these issues. First, pre-trained PLM such ESM-2 is used for embedding the protein sequences followed by convolutional neural network (CNN). Subsequently multi-head self-attention mechanism is applied for the contextual information while label-aware attention is used to compute class-specific representations by attending to the sequence in a way that is tailored to each label (DBP, RBP and non-NABP) in a multi-label setup. We have also included a novel cross-label attention mechanism to explicitly capture dependencies between DNA- and RNA-binding proteins, enabling more accurate prediction of DRBP. Finally, a linear layer followed by a sigmoid function are used for the final prediction. Extensive experiments are carried out to compare LAMP-PRo with the existing methods wherein the proposed model shows consistent competent performance. Furthermore, we also provide visualization to showcase model interpretability, highlighting which parts of the sequence are most relevant for a predicted label. The original datasets are available at http://bliulab.net/iDRBP\\_MMC and the codes are available at https://github.com/NimishaGhosh/LAMP-PRo.", "date": "2025-09-29", "categories": []}
{"id": "2510.17826", "title": "Speak to a Protein: An Interactive Multimodal Co-Scientist for Protein Analysis", "abstract": "Building a working mental model of a protein typically requires weeks of reading, cross-referencing crystal and predicted structures, and inspecting ligand complexes, an effort that is slow, unevenly accessible, and often requires specialized computational skills. We introduce \\emph{Speak to a Protein}, a new capability that turns protein analysis into an interactive, multimodal dialogue with an expert co-scientist. The AI system retrieves and synthesizes relevant literature, structures, and ligand data; grounds answers in a live 3D scene; and can highlight, annotate, manipulate and see the visualization. It also generates and runs code when needed, explaining results in both text and graphics. We demonstrate these capabilities on relevant proteins, posing questions about binding pockets, conformational changes, or structure-activity relationships to test ideas in real-time. \\emph{Speak to a Protein} reduces the time from question to evidence, lowers the barrier to advanced structural analysis, and enables hypothesis generation by tightly coupling language, code, and 3D structures. \\emph{Speak to a Protein} is freely accessible at https://open.playmolecule.org.", "date": "2025-10-01", "categories": []}
{"id": "2510.18571", "title": "A Multi-Evidence Framework Rescues Low-Power Prognostic Signals and Rejects Statistical Artifacts in Cancer Genomics", "abstract": "Motivation: Standard genome-wide association studies in cancer genomics rely on statistical significance with multiple testing correction, but systematically fail in underpowered cohorts. In TCGA breast cancer (n=967, 133 deaths), low event rates (13.8%) create severe power limitations, producing false negatives for known drivers and false positives for large passenger genes. Results: We developed a five-criteria computational framework integrating causal inference (inverse probability weighting, doubly robust estimation) with orthogonal biological validation (expression, mutation patterns, literature evidence). Applied to TCGA-BRCA mortality analysis, standard Cox+FDR detected zero genes at FDR<0.05, confirming complete failure in underpowered settings. Our framework correctly identified RYR2 -- a cardiac gene with no cancer function -- as a false positive despite nominal significance (p=0.024), while identifying KMT2C as a complex candidate requiring validation despite marginal significance (p=0.047, q=0.954). Power analysis revealed median power of 15.1% across genes, with KMT2C achieving only 29.8% power (HR=1.55), explaining borderline statistical significance despite strong biological evidence. The framework distinguished true signals from artifacts through mutation pattern analysis: RYR2 showed 29.8% silent mutations (passenger signature) with no hotspots, while KMT2C showed 6.7% silent mutations with 31.4% truncating variants (driver signature). This multi-evidence approach provides a template for analyzing underpowered cohorts, prioritizing biological interpretability over purely statistical significance.\n  Availability: All code and analysis pipelines available at github.com/akarlaraytu/causal-inference-for-cancer-genomics", "date": "2025-10-21", "categories": []}
{"id": "2411.04863", "title": "OneProt: Towards Multi-Modal Protein Foundation Models", "abstract": "Recent advances in Artificial Intelligence have enabled multi-modal systems to model and translate diverse information spaces. Extending beyond text and vision, we introduce OneProt, a multi-modal AI for proteins that integrates structural, sequence, text, and binding site data. Using the ImageBind framework, OneProt aligns the latent spaces of protein modality encoders in a lightweight fine-tuning scheme that focuses on pairwise alignment with sequence data rather than requiring full matches. This novel approach comprises a mix of Graph Neural Networks and transformer architectures. It demonstrates strong performance in retrieval tasks and showcases the efficacy of multi-modal systems in Protein Machine Learning through a broad spectrum of downstream baselines, including enzyme function prediction and binding site analysis. Furthermore, OneProt enables the transfer of representational information from specialized encoders to the sequence encoder, enhancing capabilities for distinguishing evolutionarily related and unrelated sequences and exhibiting representational properties where evolutionarily related proteins align in similar directions within the latent space. In addition, we extensively investigate modality ablations to identify the encoders that contribute most to predictive performance, highlighting the significance of the binding site encoder, which has not been used in similar models previously. This work expands the horizons of multi-modal protein models, paving the way for transformative applications in drug discovery, biocatalytic reaction planning, and protein engineering.", "date": "2024-11-07", "categories": []}
{"id": "2503.14571", "title": "Efficient Data Selection for Training Genomic Perturbation Models", "abstract": "Genomic studies face a vast hypothesis space, while interventions such as gene perturbations remain costly and time-consuming. To accelerate such experiments, gene perturbation models predict the transcriptional outcome of interventions. Since constructing the training set is challenging, active learning is often employed in a \"lab-in-the-loop\" process. While this strategy makes training more targeted, it is substantially slower, as it fails to exploit the inherent parallelizability of Perturb-seq experiments. Here, we focus on graph neural network-based gene perturbation models and propose a subset selection method that, unlike active learning, selects the training perturbations in one shot. Our method chooses the interventions that maximize the propagation of the supervision signal to the model. The selection criterion is defined over the input knowledge graph and is optimized with submodular maximization, ensuring a near-optimal guarantee. Experimental results across multiple datasets show that, in addition to providing months of acceleration compared to active learning, the method improves the stability of perturbation choices while maintaining competitive predictive accuracy.", "date": "2025-03-18", "categories": []}
{"id": "2510.16093", "title": "Identifying multi-omics interactions for lung cancer drug targets discovery using Kernel Machine Regression", "abstract": "Cancer exhibits diverse and complex phenotypes driven by multifaceted molecular interactions. Recent biomedical research has emphasized the comprehensive study of such diseases by integrating multi-omics datasets (genome, proteome, transcriptome, epigenome). This approach provides an efficient method for identifying genetic variants associated with cancer and offers a deeper understanding of how the disease develops and spreads. However, it is challenging to comprehend complex interactions among the features of multi-omics datasets compared to single omics. In this paper, we analyze lung cancer multi-omics datasets from The Cancer Genome Atlas (TCGA). Using four statistical methods, LIMMA, the T test, Canonical Correlation Analysis (CCA), and the Wilcoxon test, we identified differentially expressed genes across gene expression, DNA methylation, and miRNA expression data. We then integrated these multi-omics data using the Kernel Machine Regression (KMR) approach. Our findings reveal significant interactions among the three omics: gene expression, miRNA expression, and DNA methylation in lung cancer. From our data analysis, we identified 38 genes significantly associated with lung cancer. From our data analysis, we identified 38 genes significantly associated with lung cancer. Among these, eight genes of highest ranking (PDGFRB, PDGFRA, SNAI1, ID1, FGF11, TNXB, ITGB1, ZIC1) were highlighted by rigorous statistical analysis. Furthermore, in silico studies identified three top-ranked potential candidate drugs (Selinexor, Orapred, and Capmatinib) that could play a crucial role in the treatment of lung cancer. These proposed drugs are also supported by the findings of other independent studies, which underscore their potential efficacy in the fight against lung cancer.", "date": "2025-10-17", "categories": []}
{"id": "2510.16612", "title": "Accelerated Learning on Large Scale Screens using Generative Library Models", "abstract": "Biological machine learning is often bottlenecked by a lack of scaled data. One promising route to relieving data bottlenecks is through high throughput screens, which can experimentally test the activity of $10^6-10^{12}$ protein sequences in parallel. In this article, we introduce algorithms to optimize high throughput screens for data creation and model training. We focus on the large scale regime, where dataset sizes are limited by the cost of measurement and sequencing. We show that when active sequences are rare, we maximize information gain if we only collect positive examples of active sequences, i.e. $x$ with $y>0$. We can correct for the missing negative examples using a generative model of the library, producing a consistent and efficient estimate of the true $p(y | x)$. We demonstrate this approach in simulation and on a large scale screen of antibodies. Overall, co-design of experiments and inference lets us accelerate learning dramatically.", "date": "2025-10-18", "categories": []}
{"id": "2510.16674", "title": "Evaluating protein binding interfaces with PUMBA", "abstract": "Protein-protein docking tools help in studying interactions between proteins, and are essential for drug, vaccine, and therapeutic development. However, the accuracy of a docking tool depends on a robust scoring function that can reliably differentiate between native and non-native complexes. PIsToN is a state-of-the-art deep learning-based scoring function that uses Vision Transformers in its architecture. Recently, the Mamba architecture has demonstrated exceptional performance in both natural language processing and computer vision, often outperforming Transformer-based models in their domains. In this study, we introduce PUMBA (Protein-protein interface evaluation with Vision Mamba), which improves PIsToN by replacing its Vision Transformer backbone with Vision Mamba. This change allows us to leverage Mamba's efficient long-range sequence modeling for sequences of image patches. As a result, the model's ability to capture both global and local patterns in protein-protein interface features is significantly improved. Evaluation on several widely-used, large-scale public datasets demonstrates that PUMBA consistently outperforms its original Transformer-based predecessor, PIsToN.", "date": "2025-10-19", "categories": []}
{"id": "2510.17187", "title": "A Standardized Benchmark for Machine-Learned Molecular Dynamics using Weighted Ensemble Sampling", "abstract": "The rapid evolution of molecular dynamics (MD) methods, including machine-learned dynamics, has outpaced the development of standardized tools for method validation. Objective comparison between simulation approaches is often hindered by inconsistent evaluation metrics, insufficient sampling of rare conformational states, and the absence of reproducible benchmarks. To address these challenges, we introduce a modular benchmarking framework that systematically evaluates protein MD methods using enhanced sampling analysis. Our approach uses weighted ensemble (WE) sampling via The Weighted Ensemble Simulation Toolkit with Parallelization and Analysis (WESTPA), based on progress coordinates derived from Time-lagged Independent Component Analysis (TICA), enabling fast and efficient exploration of protein conformational space. The framework includes a flexible, lightweight propagator interface that supports arbitrary simulation engines, allowing both classical force fields and machine learning-based models. Additionally, the framework offers a comprehensive evaluation suite capable of computing more than 19 different metrics and visualizations across a variety of domains. We further contribute a dataset of nine diverse proteins, ranging from 10 to 224 residues, that span a variety of folding complexities and topologies. Each protein has been extensively simulated at 300K for one million MD steps per starting point (4 ns). To demonstrate the utility of our framework, we perform validation tests using classic MD simulations with implicit solvent and compare protein conformational sampling using a fully trained versus under-trained CGSchNet model. By standardizing evaluation protocols and enabling direct, reproducible comparisons across MD approaches, our open-source platform lays the groundwork for consistent, rigorous benchmarking across the molecular simulation community.", "date": "2025-10-20", "categories": []}
{"id": "2502.15646", "title": "Predicting gene essentiality and drug response from perturbation screens in preclinical cancer models with LEAP: Layered Ensemble of Autoencoders and Predictors", "abstract": "High-throughput preclinical perturbation screens, where the effects of genetic, chemical, or environmental perturbations are systematically tested on disease models, hold significant promise for machine learning-enhanced drug discovery due to their scale and causal nature. Predictive models trained on such datasets can be used to (i) infer perturbation response for previously untested disease models, and (ii) characterise the biological context that affects perturbation response. Existing predictive models suffer from limited reproducibility, generalisability and interpretability. To address these issues, we introduce a framework of Layered Ensemble of Autoencoders and Predictors (LEAP), a general and flexible ensemble strategy to aggregate predictions from multiple regressors trained using diverse gene expression representation models. LEAP consistently improves prediction performances in unscreened cell lines across modelling strategies. In particular, LEAP applied to perturbation-specific LASSO regressors (PS-LASSO) provides a favorable balance between near state-of-the-art performance and low computation time. We also propose an interpretability approach combining model distillation and stability selection to identify important biological pathways for perturbation response prediction in LEAP. Our models have the potential to accelerate the drug discovery pipeline by guiding the prioritisation of preclinical experiments and providing insights into the biological mechanisms involved in perturbation response. The code and datasets used in this work are publicly available.", "date": "2025-02-21", "categories": []}
{"id": "2510.15601", "title": "Kernel-Based Evaluation of Conditional Biological Sequence Models", "abstract": "We propose a set of kernel-based tools to evaluate the designs and tune the hyperparameters of conditional sequence models, with a focus on problems in computational biology. The backbone of our tools is a new measure of discrepancy between the true conditional distribution and the model's estimate, called the Augmented Conditional Maximum Mean Discrepancy (ACMMD). Provided that the model can be sampled from, the ACMMD can be estimated unbiasedly from data to quantify absolute model fit, integrated within hypothesis tests, and used to evaluate model reliability. We demonstrate the utility of our approach by analyzing a popular protein design model, ProteinMPNN. We are able to reject the hypothesis that ProteinMPNN fits its data for various protein families, and tune the model's temperature hyperparameter to achieve a better fit.", "date": "2025-10-17", "categories": []}
{"id": "2510.15837", "title": "Transfer Orthology Networks", "abstract": "We present Transfer Orthology Networks (TRON), a novel neural network architecture designed for cross-species transfer learning. TRON leverages orthologous relationships, represented as a bipartite graph between species, to guide knowledge transfer. Specifically, we prepend a learned species conversion layer, whose weights are masked by the biadjacency matrix of this bipartite graph, to a pre-trained feedforward neural network that predicts a phenotype from gene expression data in a source species. This allows for efficient transfer of knowledge to a target species by learning a linear transformation that maps gene expression from the source to the target species' gene space. The learned weights of this conversion layer offer a potential avenue for interpreting functional orthology, providing insights into how genes across species contribute to the phenotype of interest. TRON offers a biologically grounded and interpretable approach to cross-species transfer learning, paving the way for more effective utilization of available transcriptomic data. We are in the process of collecting cross-species transcriptomic/phenotypic data to gain experimental validation of the TRON architecture.", "date": "2025-10-17", "categories": []}
{"id": "2510.10480", "title": "Latent Retrieval Augmented Generation of Cross-Domain Protein Binders", "abstract": "Designing protein binders targeting specific sites, which requires to generate realistic and functional interaction patterns, is a fundamental challenge in drug discovery. Current structure-based generative models are limited in generating nterfaces with sufficient rationality and interpretability. In this paper, we propose Retrieval-Augmented Diffusion for Aligned interface (RADiAnce), a new framework that leverages known interfaces to guide the design of novel binders. By unifying retrieval and generation in a shared contrastive latent space, our model efficiently identifies relevant interfaces for a given binding site and seamlessly integrates them through a conditional latent diffusion generator, enabling cross-domain interface transfer. Extensive exeriments show that RADiAnce significantly outperforms baseline models across multiple metrics, including binding affinity and recovery of geometries and interactions. Additional experimental results validate cross-domain generalization, demonstrating that retrieving interfaces from diverse domains, such as peptides, antibodies, and protein fragments, enhances the generation performance of binders for other domains. Our work establishes a new paradigm for protein binder design that successfully bridges retrieval-based knowledge and generative AI, opening new possibilities for drug discovery.", "date": "2025-10-12", "categories": []}
{"id": "2510.14139", "title": "Inferred global dense residue transition graphs from primary structure sequences enable protein interaction prediction via directed graph convolutional neural networks", "abstract": "Introduction Accurate prediction of protein-protein interactions (PPIs) is crucial for understanding cellular functions and advancing drug development. Existing in-silico methods use direct sequence embeddings from Protein Language Models (PLMs). Others use Graph Neural Networks (GNNs) for 3D protein structures. This study explores less computationally intensive alternatives. We introduce a novel framework for downstream PPI prediction through link prediction. Methods We introduce a two-stage graph representation learning framework, ProtGram-DirectGCN. First, we developed ProtGram. This approach models a protein's primary structure as a hierarchy of globally inferred n-gram graphs. In these graphs, residue transition probabilities define edge weights. Each edge connects a pair of residues in a directed graph. The probabilities are aggregated from a large corpus of sequences. Second, we propose DirectGCN, a custom directed graph convolutional neural network. This model features a unique convolutional layer. It processes information through separate path-specific transformations: incoming, outgoing, and undirected. A shared transformation is also applied. These paths are combined via a learnable gating mechanism. We apply DirectGCN to ProtGram graphs to learn residue-level embeddings. These embeddings are pooled via attention to generate protein-level embeddings for prediction. Results We first established the efficacy of DirectGCN on standard node classification benchmarks. Its performance matches established methods on general datasets. The model excels at complex, directed graphs with dense, heterophilic structures. When applied to PPI prediction, the full ProtGram-DirectGCN framework delivers robust predictive power. This strong performance holds even with limited training data.", "date": "2025-10-15", "categories": []}
{"id": "2507.21938", "title": "Multi-state Protein Design with DynamicMPNN", "abstract": "Structural biology has long been dominated by the one sequence, one structure, one function paradigm, yet many critical biological processes - from enzyme catalysis to membrane transport - depend on proteins that adopt multiple conformational states. Existing multi-state design approaches rely on post-hoc aggregation of single-state predictions, achieving poor experimental success rates compared to single-state design. We introduce DynamicMPNN, an inverse folding model explicitly trained to generate sequences compatible with multiple conformations through joint learning across conformational ensembles. Trained on 46,033 conformational pairs covering 75% of CATH superfamilies and evaluated using Alphafold 3, DynamicMPNN outperforms ProteinMPNN by up to 25% on decoy-normalized RMSD and by 12% on sequence recovery across our challenging multi-state protein benchmark.", "date": "2025-07-29", "categories": []}
{"id": "2505.18966", "title": "Protein Design with Dynamic Protein Vocabulary", "abstract": "Protein design is a fundamental challenge in biotechnology, aiming to design novel sequences with specific functions within the vast space of possible proteins. Recent advances in deep generative models have enabled function-based protein design from textual descriptions, yet struggle with structural plausibility. Inspired by classical protein design methods that leverage natural protein structures, we explore whether incorporating fragments from natural proteins can enhance foldability in generative models. Our empirical results show that even random incorporation of fragments improves foldability. Building on this insight, we introduce ProDVa, a novel protein design approach that integrates a text encoder for functional descriptions, a protein language model for designing proteins, and a fragment encoder to dynamically retrieve protein fragments based on textual functional descriptions. Experimental results demonstrate that our approach effectively designs protein sequences that are both functionally aligned and structurally plausible. Compared to state-of-the-art models, ProDVa achieves comparable function alignment using less than 0.04% of the training data, while designing significantly more well-folded proteins, with the proportion of proteins having pLDDT above 70 increasing by 7.38% and those with PAE below 10 increasing by 9.6%.", "date": "2025-05-25", "categories": []}
{"id": "2510.11726", "title": "scPPDM: A Diffusion Model for Single-Cell Drug-Response Prediction", "abstract": "This paper introduces the Single-Cell Perturbation Prediction Diffusion Model (scPPDM), the first diffusion-based framework for single-cell drug-response prediction from scRNA-seq data. scPPDM couples two condition channels, pre-perturbation state and drug with dose, in a unified latent space via non-concatenative GD-Attn. During inference, factorized classifier-free guidance exposes two interpretable controls for state preservation and drug-response strength and maps dose to guidance magnitude for tunable intensity. Evaluated on the Tahoe-100M benchmark under two stringent regimes, unseen covariate combinations (UC) and unseen drugs (UD), scPPDM sets new state-of-the-art results across log fold-change recovery, delta correlations, explained variance, and DE-overlap. Representative gains include +36.11%/+34.21% on DEG logFC-Spearman/Pearson in UD over the second-best model. This control interface enables transparent what-if analyses and dose tuning, reducing experimental burden while preserving biological specificity.", "date": "2025-10-08", "categories": []}
{"id": "2510.12245", "title": "MoRA: On-the-fly Molecule-aware Low-Rank Adaptation Framework for LLM-based Multi-Modal Molecular Assistant", "abstract": "Effectively integrating molecular graph structures with Large Language Models (LLMs) is a key challenge in drug discovery. Most existing multi-modal alignment methods typically process these structures by fine-tuning the LLM or adding a static adapter simultaneously. However, these approaches have two main limitations: (1) it optimizes a shared parameter space across all molecular inputs, limiting the model's ability to capture instance-specific structural features; and (2) fine-tuning the LLM for molecular tasks can lead to catastrophic forgetting, undermining its general reasoning capabilities. In this paper, instead of static task-oriented adaptation, we propose an instance-specific parameter space alignment approach for each molecule on-the-fly. To this end, we introduce Molecule-aware Low-Rank Adaptation (MoRA) that produces a unique set of low-rank adaptation weights for each input molecular graph. These weights are then dynamically injected into a frozen LLM, allowing the model to adapt its reasoning to the structure of each molecular input, while preserving the LLM's core knowledge. Extensive experiments demonstrate that on key molecular tasks, such as chemical reaction prediction and molecular captioning, MoRA's instance-specific dynamic adaptation outperforms statically adapted baselines, including a 14.1% relative improvement in reaction prediction exact match and a 22% reduction in error for quantum property prediction. The code is available at https://github.com/jk-sounds/MoRA.", "date": "2025-10-14", "categories": []}
{"id": "2411.14833", "title": "Cell as Point: One-Stage Framework for Efficient Cell Tracking", "abstract": "Conventional multi-stage cell tracking approaches rely heavily on detection or segmentation in each frame as a prerequisite, requiring substantial resources for high-quality segmentation masks and increasing the overall prediction time. To address these limitations, we propose CAP, a novel end-to-end one-stage framework that reimagines cell tracking by treating Cell as Point. Unlike traditional methods, CAP eliminates the need for explicit detection or segmentation, instead jointly tracking cells for sequences in one stage by leveraging the inherent correlations among their trajectories. This simplification reduces both labeling requirements and pipeline complexity. However, directly processing the entire sequence in one stage poses challenges related to data imbalance in capturing cell division events and long sequence inference. To solve these challenges, CAP introduces two key innovations: (1) adaptive event-guided (AEG) sampling, which prioritizes cell division events to mitigate the occurrence imbalance of cell events, and (2) the rolling-as-window (RAW) inference strategy, which ensures continuous and stable tracking of newly emerging cells over extended sequences. By removing the dependency on segmentation-based preprocessing while addressing the challenges of imbalanced occurrence of cell events and long-sequence tracking, CAP demonstrates promising cell tracking performance and is 8 to 32 times more efficient than existing methods. The code and model checkpoints will be available soon.", "date": "2024-11-22", "categories": []}
{"id": "2506.04235", "title": "AbBiBench: A Benchmark for Antibody Binding Affinity Maturation and Design", "abstract": "We introduce AbBiBench (Antibody Binding Benchmarking), a benchmarking framework for antibody binding affinity maturation and design. Unlike previous strategies that evaluate antibodies in isolation, typically by comparing them to natural sequences with metrics such as amino acid recovery rate or structural RMSD, AbBiBench instead treats the antibody-antigen (Ab-Ag) complex as the fundamental unit. It evaluates an antibody design's binding potential by measuring how well a protein model scores the full Ab-Ag complex. We first curate, standardize, and share more than 184,500 experimental measurements of antibody mutants across 14 antibodies and 9 antigens-including influenza, lysozyme, HER2, VEGF, integrin, Ang2, and SARS-CoV-2-covering both heavy-chain and light-chain mutations. Using these datasets, we systematically compare 15 protein models including masked language models, autoregressive language models, inverse folding models, diffusion-based generative models, and geometric graph models by comparing the correlation between model likelihood and experimental affinity values. Additionally, to demonstrate AbBiBench's generative utility, we apply it to antibody F045-092 in order to introduce binding to influenza H1N1. We sample new antibody variants with the top-performing models, rank them by the structural integrity and biophysical properties of the Ab-Ag complex, and assess them with in vitro ELISA binding assays. Our findings show that structure-conditioned inverse folding models outperform others in both affinity correlation and generation tasks. Overall, AbBiBench provides a unified, biologically grounded evaluation framework to facilitate the development of more effective, function-aware antibody design models.", "date": "2025-05-23", "categories": []}
{"id": "2510.10634", "title": "ProteinAE: Protein Diffusion Autoencoders for Structure Encoding", "abstract": "Developing effective representations of protein structures is essential for advancing protein science, particularly for protein generative modeling. Current approaches often grapple with the complexities of the SE(3) manifold, rely on discrete tokenization, or the need for multiple training objectives, all of which can hinder the model optimization and generalization. We introduce ProteinAE, a novel and streamlined protein diffusion autoencoder designed to overcome these challenges by directly mapping protein backbone coordinates from E(3) into a continuous, compact latent space. ProteinAE employs a non-equivariant Diffusion Transformer with a bottleneck design for efficient compression and is trained end-to-end with a single flow matching objective, substantially simplifying the optimization pipeline. We demonstrate that ProteinAE achieves state-of-the-art reconstruction quality, outperforming existing autoencoders. The resulting latent space serves as a powerful foundation for a latent diffusion model that bypasses the need for explicit equivariance. This enables efficient, high-quality structure generation that is competitive with leading structure-based approaches and significantly outperforms prior latent-based methods. Code is available at https://github.com/OnlyLoveKFC/ProteinAE_v1.", "date": "2025-10-12", "categories": []}
{"id": "2509.24840", "title": "Cell2Text: Multimodal LLM for Generating Single-Cell Descriptions from RNA-Seq Data", "abstract": "Single-cell RNA sequencing has transformed biology by enabling the measurement of gene expression at cellular resolution, providing information for cell types, states, and disease contexts. Recently, single-cell foundation models have emerged as powerful tools for learning transferable representations directly from expression profiles, improving performance on classification and clustering tasks. However, these models are limited to discrete prediction heads, which collapse cellular complexity into predefined labels that fail to capture the richer, contextual explanations biologists need. We introduce Cell2Text, a multimodal generative framework that translates scRNA-seq profiles into structured natural language descriptions. By integrating gene-level embeddings from single-cell foundation models with pretrained large language models, Cell2Text generates coherent summaries that capture cellular identity, tissue origin, disease associations, and pathway activity, generalizing to unseen cells. Empirically, Cell2Text outperforms baselines on classification accuracy, demonstrates strong ontological consistency using PageRank-based similarity metrics, and achieves high semantic fidelity in text generation. These results demonstrate that coupling expression data with natural language offers both stronger predictive performance and inherently interpretable outputs, pointing to a scalable path for label-efficient characterization of unseen cells.", "date": "2025-09-29", "categories": []}
{"id": "2510.01298", "title": "MorphGen: Controllable and Morphologically Plausible Generative Cell-Imaging", "abstract": "Simulating in silico cellular responses to interventions is a promising direction to accelerate high-content image-based assays, critical for advancing drug discovery and gene editing. To support this, we introduce MorphGen, a state-of-the-art diffusion-based generative model for fluorescent microscopy that enables controllable generation across multiple cell types and perturbations. To capture biologically meaningful patterns consistent with known cellular morphologies, MorphGen is trained with an alignment loss to match its representations to the phenotypic embeddings of OpenPhenom, a state-of-the-art biological foundation model. Unlike prior approaches that compress multichannel stains into RGB images -- thus sacrificing organelle-specific detail -- MorphGen generates the complete set of fluorescent channels jointly, preserving per-organelle structures and enabling a fine-grained morphological analysis that is essential for biological interpretation. We demonstrate biological consistency with real images via CellProfiler features, and MorphGen attains an FID score over 35% lower than the prior state-of-the-art MorphoDiff, which only generates RGB images for a single cell type. Code is available at https://github.com/czi-ai/MorphGen.", "date": "2025-10-01", "categories": []}
{"id": "2510.08655", "title": "Knowledge Graph Sparsification for GNN-based Rare Disease Diagnosis", "abstract": "Rare genetic disease diagnosis faces critical challenges: insufficient patient data, inaccessible full genome sequencing, and the immense number of possible causative genes. These limitations cause prolonged diagnostic journeys, inappropriate treatments, and critical delays, disproportionately affecting patients in resource-limited settings where diagnostic tools are scarce. We propose RareNet, a subgraph-based Graph Neural Network that requires only patient phenotypes to identify the most likely causal gene and retrieve focused patient subgraphs for targeted clinical investigation. RareNet can function as a standalone method or serve as a pre-processing or post-processing filter for other candidate gene prioritization methods, consistently enhancing their performance while potentially enabling explainable insights. Through comprehensive evaluation on two biomedical datasets, we demonstrate competitive and robust causal gene prediction and significant performance gains when integrated with other frameworks. By requiring only phenotypic data, which is readily available in any clinical setting, RareNet democratizes access to sophisticated genetic analysis, offering particular value for underserved populations lacking advanced genomic infrastructure.", "date": "2025-10-09", "categories": []}
{"id": "2510.08703", "title": "Decoding Positive Selection in Mycobacterium tuberculosis with Phylogeny-Guided Graph Attention Models", "abstract": "Positive selection drives the emergence of adaptive mutations in Mycobacterium tuberculosis, shaping drug resistance, transmissibility, and virulence. Phylogenetic trees capture evolutionary relationships among isolates and provide a natural framework for detecting such adaptive signals. We present a phylogeny-guided graph attention network (GAT) approach, introducing a method for converting SNP-annotated phylogenetic trees into graph structures suitable for neural network analysis. Using 500 M. tuberculosis isolates from four major lineages and 249 single-nucleotide variants (84 resistance-associated and 165 neutral) across 61 drug-resistance genes, we constructed graphs where nodes represented isolates and edges reflected phylogenetic distances. Edges between isolates separated by more than seven internal nodes were pruned to emphasise local evolutionary structure. Node features encoded SNP presence or absence, and the GAT architecture included two attention layers, a residual connection, global attention pooling, and a multilayer perceptron classifier. The model achieved an accuracy of 0.88 on a held-out test set and, when applied to 146 WHO-classified \"uncertain\" variants, identified 41 candidates with convergent emergence across multiple lineages, consistent with adaptive evolution. This work demonstrates the feasibility of transforming phylogenies into GNN-compatible structures and highlights attention-based models as effective tools for detecting positive selection, aiding genomic surveillance and variant prioritisation.", "date": "2025-10-09", "categories": []}
{"id": "2510.09020", "title": "MagicDock: Toward Docking-oriented De Novo Ligand Design via Gradient Inversion", "abstract": "De novo ligand design is a fundamental task that seeks to generate protein or molecule candidates that can effectively dock with protein receptors and achieve strong binding affinity entirely from scratch. It holds paramount significance for a wide spectrum of biomedical applications. However, most existing studies are constrained by the \\textbf{Pseudo De Novo}, \\textbf{Limited Docking Modeling}, and \\textbf{Inflexible Ligand Type}. To address these issues, we propose MagicDock, a forward-looking framework grounded in the progressive pipeline and differentiable surface modeling. (1) We adopt a well-designed gradient inversion framework. To begin with, general docking knowledge of receptors and ligands is incorporated into the backbone model. Subsequently, the docking knowledge is instantiated as reverse gradient flows by binding prediction, which iteratively guide the de novo generation of ligands. (2) We emphasize differentiable surface modeling in the docking process, leveraging learnable 3D point-cloud representations to precisely capture binding details, thereby ensuring that the generated ligands preserve docking validity through direct and interpretable spatial fingerprints. (3) We introduce customized designs for different ligand types and integrate them into a unified gradient inversion framework with flexible triggers, thereby ensuring broad applicability. Moreover, we provide rigorous theoretical guarantees for each component of MagicDock. Extensive experiments across 9 scenarios demonstrate that MagicDock achieves average improvements of 27.1\\% and 11.7\\% over SOTA baselines specialized for protein or molecule ligand design, respectively.", "date": "2025-10-10", "categories": []}
{"id": "2510.09594", "title": "MODE: Learning compositional representations of complex systems with Mixtures Of Dynamical Experts", "abstract": "Dynamical systems in the life sciences are often composed of complex mixtures of overlapping behavioral regimes. Cellular subpopulations may shift from cycling to equilibrium dynamics or branch towards different developmental fates. The transitions between these regimes can appear noisy and irregular, posing a serious challenge to traditional, flow-based modeling techniques which assume locally smooth dynamics. To address this challenge, we propose MODE (Mixture Of Dynamical Experts), a graphical modeling framework whose neural gating mechanism decomposes complex dynamics into sparse, interpretable components, enabling both the unsupervised discovery of behavioral regimes and accurate long-term forecasting across regime transitions. Crucially, because agents in our framework can jump to different governing laws, MODE is especially tailored to the aforementioned noisy transitions. We evaluate our method on a battery of synthetic and real datasets from computational biology. First, we systematically benchmark MODE on an unsupervised classification task using synthetic dynamical snapshot data, including in noisy, few-sample settings. Next, we show how MODE succeeds on challenging forecasting tasks which simulate key cycling and branching processes in cell biology. Finally, we deploy our method on human, single-cell RNA sequencing data and show that it can not only distinguish proliferation from differentiation dynamics but also predict when cells will commit to their ultimate fate, a key outstanding challenge in computational biology.", "date": "2025-10-10", "categories": []}
{"id": "2409.04922", "title": "Nearest Neighbor CCP-Based Molecular Sequence Analysis", "abstract": "Molecular sequence analysis is crucial for comprehending several biological processes, including protein-protein interactions, functional annotation, and disease classification. The large number of sequences and the inherently complicated nature of protein structures make it challenging to analyze such data. Finding patterns and enhancing subsequent research requires the use of dimensionality reduction and feature selection approaches. Recently, a method called Correlated Clustering and Projection (CCP) has been proposed as an effective method for biological sequencing data. The CCP technique is still costly to compute even though it is effective for sequence visualization. Furthermore, its utility for classifying molecular sequences is still uncertain. To solve these two problems, we present a Nearest Neighbor Correlated Clustering and Projection (CCP-NN)-based technique for efficiently preprocessing molecular sequence data. To group related molecular sequences and produce representative supersequences, CCP makes use of sequence-to-sequence correlations. As opposed to conventional methods, CCP doesn't rely on matrix diagonalization, therefore it can be applied to a range of machine-learning problems. We estimate the density map and compute the correlation using a nearest-neighbor search technique. We performed molecular sequence classification using CCP and CCP-NN representations to assess the efficacy of our proposed approach. Our findings show that CCP-NN considerably improves classification task accuracy as well as significantly outperforms CCP in terms of computational runtime.", "date": "2024-09-07", "categories": []}
{"id": "2503.00096", "title": "BixBench: a Comprehensive Benchmark for LLM-based Agents in Computational Biology", "abstract": "Large Language Models (LLMs) and LLM-based agents show great promise in accelerating scientific research. Existing benchmarks for measuring this potential and guiding future development continue to evolve from pure recall and rote knowledge tasks, towards more practical work such as literature review and experimental planning. Bioinformatics is a domain where fully autonomous AI-driven discovery may be near, but no extensive benchmarks for measuring progress have been introduced to date. We therefore present the Bioinformatics Benchmark (BixBench), a dataset comprising over 50 real-world scenarios of practical biological data analysis with nearly 300 associated open-answer questions designed to measure the ability of LLM-based agents to explore biological datasets, perform long, multi-step analytical trajectories, and interpret the nuanced results of those analyses. We evaluate the performance of two frontier LLMs (GPT-4o and Claude 3.5 Sonnet) using a custom agent framework we open source. We find that even the latest frontier models only achieve 17% accuracy in the open-answer regime, and no better than random in a multiple-choice setting. By exposing the current limitations of frontier models, we hope BixBench can spur the development of agents capable of conducting rigorous bioinformatic analysis and accelerate scientific discovery.", "date": "2025-02-28", "categories": []}
{"id": "2510.07477", "title": "HEMERA: A Human-Explainable Transformer Model for Estimating Lung Cancer Risk using GWAS Data", "abstract": "Lung cancer (LC) is the third most common cancer and the leading cause of cancer deaths in the US. Although smoking is the primary risk factor, the occurrence of LC in never-smokers and familial aggregation studies highlight a genetic component. Genetic biomarkers identified through genome-wide association studies (GWAS) are promising tools for assessing LC risk. We introduce HEMERA (Human-Explainable Transformer Model for Estimating Lung Cancer Risk using GWAS Data), a new framework that applies explainable transformer-based deep learning to GWAS data of single nucleotide polymorphisms (SNPs) for predicting LC risk. Unlike prior approaches, HEMERA directly processes raw genotype data without clinical covariates, introducing additive positional encodings, neural genotype embeddings, and refined variant filtering. A post hoc explainability module based on Layer-wise Integrated Gradients enables attribution of model predictions to specific SNPs, aligning strongly with known LC risk loci. Trained on data from 27,254 Million Veteran Program participants, HEMERA achieved >99% AUC (area under receiver characteristics) score. These findings support transparent, hypothesis-generating models for personalized LC risk assessment and early intervention.", "date": "2025-10-08", "categories": []}
{"id": "2510.07706", "title": "Large Language Models Meet Virtual Cell: A Survey", "abstract": "Large language models (LLMs) are transforming cellular biology by enabling the development of \"virtual cells\"--computational systems that represent, predict, and reason about cellular states and behaviors. This work provides a comprehensive review of LLMs for virtual cell modeling. We propose a unified taxonomy that organizes existing methods into two paradigms: LLMs as Oracles, for direct cellular modeling, and LLMs as Agents, for orchestrating complex scientific tasks. We identify three core tasks--cellular representation, perturbation prediction, and gene regulation inference--and review their associated models, datasets, evaluation benchmarks, as well as the critical challenges in scalability, generalizability, and interpretability.", "date": "2025-10-09", "categories": []}
{"id": "2510.07964", "title": "PRESCRIBE: Predicting Single-Cell Responses with Bayesian Estimation", "abstract": "In single-cell perturbation prediction, a central task is to forecast the effects of perturbing a gene unseen in the training data. The efficacy of such predictions depends on two factors: (1) the similarity of the target gene to those covered in the training data, which informs model (epistemic) uncertainty, and (2) the quality of the corresponding training data, which reflects data (aleatoric) uncertainty. Both factors are critical for determining the reliability of a prediction, particularly as gene perturbation is an inherently stochastic biochemical process. In this paper, we propose PRESCRIBE (PREdicting Single-Cell Response wIth Bayesian Estimation), a multivariate deep evidential regression framework designed to measure both sources of uncertainty jointly. Our analysis demonstrates that PRESCRIBE effectively estimates a confidence score for each prediction, which strongly correlates with its empirical accuracy. This capability enables the filtering of untrustworthy results, and in our experiments, it achieves steady accuracy improvements of over 3% compared to comparable baselines.", "date": "2025-10-09", "categories": []}
{"id": "2509.03487", "title": "SafeProtein: Red-Teaming Framework and Benchmark for Protein Foundation Models", "abstract": "Proteins play crucial roles in almost all biological processes. The advancement of deep learning has greatly accelerated the development of protein foundation models, leading to significant successes in protein understanding and design. However, the lack of systematic red-teaming for these models has raised serious concerns about their potential misuse, such as generating proteins with biological safety risks. This paper introduces SafeProtein, the first red-teaming framework designed for protein foundation models to the best of our knowledge. SafeProtein combines multimodal prompt engineering and heuristic beam search to systematically design red-teaming methods and conduct tests on protein foundation models. We also curated SafeProtein-Bench, which includes a manually constructed red-teaming benchmark dataset and a comprehensive evaluation protocol. SafeProtein achieved continuous jailbreaks on state-of-the-art protein foundation models (up to 70% attack success rate for ESM3), revealing potential biological safety risks in current protein foundation models and providing insights for the development of robust security protection technologies for frontier models. The codes will be made publicly available at https://github.com/jigang-fan/SafeProtein.", "date": "2025-09-03", "categories": []}
{"id": "2510.05886", "title": "acia-workflows: Automated Single-cell Imaging Analysis for Scalable and Deep Learning-based Live-cell Imaging Analysis Workflows", "abstract": "Live-cell imaging (LCI) technology enables the detailed spatio-temporal characterization of living cells at the single-cell level, which is critical for advancing research in the life sciences, from biomedical applications to bioprocessing. High-throughput setups with tens to hundreds of parallel cell cultivations offer the potential for robust and reproducible insights. However, these insights are obscured by the large amount of LCI data recorded per experiment. Recent advances in state-of-the-art deep learning methods for cell segmentation and tracking now enable the automated analysis of such large data volumes, offering unprecedented opportunities to systematically study single-cell dynamics. The next key challenge lies in integrating these powerful tools into accessible, flexible, and user-friendly workflows that support routine application in biological research. In this work, we present acia-workflows, a platform that combines three key components: (1) the Automated live-Cell Imaging Analysis (acia) Python library, which supports the modular design of image analysis pipelines offering eight deep learning segmentation and tracking approaches; (2) workflows that assemble the image analysis pipeline, its software dependencies, documentation, and visualizations into a single Jupyter Notebook, leading to accessible, reproducible and scalable analysis workflows; and (3) a collection of application workflows showcasing the analysis and customization capabilities in real-world applications. Specifically, we present three workflows to investigate various types of microfluidic LCI experiments ranging from growth rate comparisons to precise, minute-resolution quantitative analyses of individual dynamic cells responses to changing oxygen conditions. Our collection of more than ten application workflows is open source and publicly available at https://github.com/JuBiotech/acia-workflows.", "date": "2025-10-07", "categories": []}
{"id": "2510.06396", "title": "Adaptive Protein Design Protocols and Middleware", "abstract": "Computational protein design is experiencing a transformation driven by AI/ML. However, the range of potential protein sequences and structures is astronomically vast, even for moderately sized proteins. Hence, achieving convergence between generated and predicted structures demands substantial computational resources for sampling. The Integrated Machine-learning for Protein Structures at Scale (IMPRESS) offers methods and advanced computing systems for coupling AI to high-performance computing tasks, enabling the ability to evaluate the effectiveness of protein designs as they are developed, as well as the models and simulations used to generate data and train models. This paper introduces IMPRESS and demonstrates the development and implementation of an adaptive protein design protocol and its supporting computing infrastructure. This leads to increased consistency in the quality of protein design and enhanced throughput of protein design due to dynamic resource allocation and asynchronous workload execution.", "date": "2025-10-07", "categories": []}
{"id": "2410.00709", "title": "Binding Affinity Prediction: From Conventional to Machine Learning-Based Approaches", "abstract": "Protein-ligand binding is the process by which a small molecule (drug or inhibitor) attaches to a target protein. Binding affinity, which characterizes the strength of biomolecular interactions, is essential for tackling diverse challenges in life sciences, including therapeutic design, protein engineering, enzyme optimization, and elucidating biological mechanisms. Much work has been devoted to predicting binding affinity over the past decades. Here, we review recent significant works, with a focus on methods, evaluation strategies, and benchmark datasets. We note growing use of both traditional machine learning and deep learning models for predicting binding affinity, accompanied by an increasing amount of data on proteins and small drug-like molecules. With improved predictive performance and the FDA's phasing out of animal testing, AI-driven in silico models, such as AI virtual cells (AIVCs), are poised to advance binding affinity prediction; reciprocally, progress in building binding affinity predictors can refine AIVCs. Future efforts in binding affinity prediction and AI-driven in silico models can enhance the simulation of temporal dynamics, cell-type specificity, and multi-omics integration to support more accurate and personalized outcomes.", "date": "2024-09-30", "categories": []}
{"id": "2510.02476", "title": "Uncertainty-Guided Model Selection for Tabular Foundation Models in Biomolecule Efficacy Prediction", "abstract": "In-context learners like TabPFN are promising for biomolecule efficacy prediction, where established molecular feature sets and relevant experimental results can serve as powerful contextual examples. However, their performance is highly sensitive to the provided context, making strategies like post-hoc ensembling of models trained on different data subsets a viable approach. An open question is how to select the best models for the ensemble without access to ground truth labels. In this study, we investigate an uncertainty-guided strategy for model selection. We demonstrate on an siRNA knockdown efficacy task that a TabPFN model using straightforward sequence-based features can surpass specialized state-of-the-art predictors. We also show that the model's predicted inter-quantile range (IQR), a measure of its uncertainty, has a negative correlation with true prediction error. We developed the OligoICP method, which selects and averages an ensemble of models with the lowest mean IQR for siRNA efficacy prediction, achieving superior performance compared to naive ensembling or using a single model trained on all available data. This finding highlights model uncertainty as a powerful, label-free heuristic for optimizing biomolecule efficacy predictions.", "date": "2025-10-02", "categories": []}
{"id": "2510.03287", "title": "SoC-DT: Standard-of-Care Aligned Digital Twins for Patient-Specific Tumor Dynamics", "abstract": "Accurate prediction of tumor trajectories under standard-of-care (SoC) therapies remains a major unmet need in oncology. This capability is essential for optimizing treatment planning and anticipating disease progression. Conventional reaction-diffusion models are limited in scope, as they fail to capture tumor dynamics under heterogeneous therapeutic paradigms. There is hence a critical need for computational frameworks that can realistically simulate SoC interventions while accounting for inter-patient variability in genomics, demographics, and treatment regimens. We introduce Standard-of-Care Digital Twin (SoC-DT), a differentiable framework that unifies reaction-diffusion tumor growth models, discrete SoC interventions (surgery, chemotherapy, radiotherapy) along with genomic and demographic personalization to predict post-treatment tumor structure on imaging. An implicit-explicit exponential time-differencing solver, IMEX-SoC, is also proposed, which ensures stability, positivity, and scalability in SoC treatment situations. Evaluated on both synthetic data and real world glioma data, SoC-DT consistently outperforms classical PDE baselines and purely data-driven neural models in predicting tumor dynamics. By bridging mechanistic interpretability with modern differentiable solvers, SoC-DT establishes a principled foundation for patient-specific digital twins in oncology, enabling biologically consistent tumor dynamics estimation. Code will be made available upon acceptance.", "date": "2025-09-29", "categories": []}
{"id": "2510.03326", "title": "NS-Pep: De novo Peptide Design with Non-Standard Amino Acids", "abstract": "Peptide drugs incorporating non-standard amino acids (NSAAs) offer improved binding affinity and improved pharmacological properties. However, existing peptide design methods are limited to standard amino acids, leaving NSAA-aware design largely unexplored. We introduce NS-Pep, a unified framework for co-designing peptide sequences and structures with NSAAs. The main challenge is that NSAAs are extremely underrepresented-even the most frequent one, SEP, accounts for less than 0.4% of residues-resulting in a severe long-tailed distribution. To improve generalization to rare amino acids, we propose Residue Frequency-Guided Modification (RFGM), which mitigates over-penalization through frequency-aware logit calibration, supported by both theoretical and empirical analysis. Furthermore, we identify that insufficient side-chain modeling limits geometric representation of NSAAs. To address this, we introduce Progressive Side-chain Perception (PSP) for coarse-to-fine torsion and location prediction, and Interaction-Aware Weighting (IAW) to emphasize pocket-proximal residues. Moreover, NS-Pep generalizes naturally to the peptide folding task with NSAAs, addressing a major limitation of current tools. Experiments show that NS-Pep improves sequence recovery rate and binding affinity by 6.23% and 5.12%, respectively, and outperforms AlphaFold3 by 17.76% in peptide folding success rate.", "date": "2025-10-01", "categories": []}
{"id": "2510.03811", "title": "Curriculum-Augmented GFlowNets For mRNA Sequence Generation", "abstract": "Designing mRNA sequences is a major challenge in developing next-generation therapeutics, since it involves exploring a vast space of possible nucleotide combinations while optimizing sequence properties like stability, translation efficiency, and protein expression. While Generative Flow Networks are promising for this task, their training is hindered by sparse, long-horizon rewards and multi-objective trade-offs. We propose Curriculum-Augmented GFlowNets (CAGFN), which integrate curriculum learning with multi-objective GFlowNets to generate de novo mRNA sequences. CAGFN integrates a length-based curriculum that progressively adapts the maximum sequence length guiding exploration from easier to harder subproblems. We also provide a new mRNA design environment for GFlowNets which, given a target protein sequence and a combination of biological objectives, allows for the training of models that generate plausible mRNA candidates. This provides a biologically motivated setting for applying and advancing GFlowNets in therapeutic sequence design. On different mRNA design tasks, CAGFN improves Pareto performance and biological plausibility, while maintaining diversity. Moreover, CAGFN reaches higher-quality solutions faster than a GFlowNet trained with random sequence sampling (no curriculum), and enables generalization to out-of-distribution sequences.", "date": "2025-10-04", "categories": []}
{"id": "2510.04126", "title": "Attending on Multilevel Structure of Proteins enables Accurate Prediction of Cold-Start Drug-Target Interactions", "abstract": "Cold-start drug-target interaction (DTI) prediction focuses on interaction between novel drugs and proteins. Previous methods typically learn transferable interaction patterns between structures of drug and proteins to tackle it. However, insight from proteomics suggest that protein have multi-level structures and they all influence the DTI. Existing works usually represent protein with only primary structures, limiting their ability to capture interactions involving higher-level structures. Inspired by this insight, we propose ColdDTI, a framework attending on protein multi-level structure for cold-start DTI prediction. We employ hierarchical attention mechanism to mine interaction between multi-level protein structures (from primary to quaternary) and drug structures at both local and global granularities. Then, we leverage mined interactions to fuse structure representations of different levels for final prediction. Our design captures biologically transferable priors, avoiding the risk of overfitting caused by excessive reliance on representation learning. Experiments on benchmark datasets demonstrate that ColdDTI consistently outperforms previous methods in cold-start settings.", "date": "2025-10-05", "categories": []}
{"id": "2510.04315", "title": "GenAR: Next-Scale Autoregressive Generation for Spatial Gene Expression Prediction", "abstract": "Spatial Transcriptomics (ST) offers spatially resolved gene expression but remains costly. Predicting expression directly from widely available Hematoxylin and Eosin (H&E) stained images presents a cost-effective alternative. However, most computational approaches (i) predict each gene independently, overlooking co-expression structure, and (ii) cast the task as continuous regression despite expression being discrete counts. This mismatch can yield biologically implausible outputs and complicate downstream analyses. We introduce GenAR, a multi-scale autoregressive framework that refines predictions from coarse to fine. GenAR clusters genes into hierarchical groups to expose cross-gene dependencies, models expression as codebook-free discrete token generation to directly predict raw counts, and conditions decoding on fused histological and spatial embeddings. From an information-theoretic perspective, the discrete formulation avoids log-induced biases and the coarse-to-fine factorization aligns with a principled conditional decomposition. Extensive experimental results on four Spatial Transcriptomics datasets across different tissue types demonstrate that GenAR achieves state-of-the-art performance, offering potential implications for precision medicine and cost-effective molecular profiling. Code is publicly available at https://github.com/oyjr/genar.", "date": "2025-10-05", "categories": []}
{"id": "2510.04776", "title": "MetaMP: Seamless Metadata Enrichment and AI Application Framework for Enhanced Membrane Protein Visualization and Analysis", "abstract": "Structural biology has made significant progress in determining membrane proteins, leading to a remarkable increase in the number of available structures in dedicated databases. The inherent complexity of membrane protein structures, coupled with challenges such as missing data, inconsistencies, and computational barriers from disparate sources, underscores the need for improved database integration. To address this gap, we present MetaMP, a framework that unifies membrane-protein databases within a web application and uses machine learning for classification. MetaMP improves data quality by enriching metadata, offering a user-friendly interface, and providing eight interactive views for streamlined exploration. MetaMP was effective across tasks of varying difficulty, demonstrating advantages across different levels without compromising speed or accuracy, according to user evaluations. Moreover, MetaMP supports essential functions such as structure classification and outlier detection.\n  We present three practical applications of Artificial Intelligence (AI) in membrane protein research: predicting transmembrane segments, reconciling legacy databases, and classifying structures with explainable AI support. In a validation focused on statistics, MetaMP resolved 77% of data discrepancies and accurately predicted the class of newly identified membrane proteins 98% of the time and overtook expert curation. Altogether, MetaMP is a much-needed resource that harmonizes current knowledge and empowers AI-driven exploration of membrane-protein architecture.", "date": "2025-10-06", "categories": []}
{"id": "2412.05788", "title": "On diffusion posterior sampling via sequential Monte Carlo for zero-shot scaffolding of protein motifs", "abstract": "With the advent of diffusion models, new proteins can be generated at an unprecedented rate. The motif scaffolding problem requires steering this generative process to yield proteins with a desirable functional substructure called a motif. While models have been trained to take the motif as conditional input, recent techniques in diffusion posterior sampling can be leveraged as zero-shot alternatives whose approximations can be corrected with sequential Monte Carlo (SMC) algorithms. In this work, we introduce a new set of guidance potentials for describing scaffolding tasks and solve them by adapting SMC-aided diffusion posterior samplers with an unconditional model, Genie, as a prior. In single motif problems, we find that (i) the proposed potentials perform comparably, if not better, than the conventional masking approach, (ii) samplers based on reconstruction guidance outperform their replacement method counterparts, and (iii) measurement tilted proposals and twisted targets improve performance substantially. Furthermore, as a demonstration, we provide solutions to two multi-motif problems by pairing reconstruction guidance with an SE(3)-invariant potential. We also produce designable internally symmetric monomers with a guidance potential for point symmetry constraints. Our code is available at: https://github.com/matsagad/mres-project.", "date": "2024-12-08", "categories": []}
{"id": "2510.00352", "title": "AReUReDi: Annealed Rectified Updates for Refining Discrete Flows with Multi-Objective Guidance", "abstract": "Designing sequences that satisfy multiple, often conflicting, objectives is a central challenge in therapeutic and biomolecular engineering. Existing generative frameworks largely operate in continuous spaces with single-objective guidance, while discrete approaches lack guarantees for multi-objective Pareto optimality. We introduce AReUReDi (Annealed Rectified Updates for Refining Discrete Flows), a discrete optimization algorithm with theoretical guarantees of convergence to the Pareto front. Building on Rectified Discrete Flows (ReDi), AReUReDi combines Tchebycheff scalarization, locally balanced proposals, and annealed Metropolis-Hastings updates to bias sampling toward Pareto-optimal states while preserving distributional invariance. Applied to peptide and SMILES sequence design, AReUReDi simultaneously optimizes up to five therapeutic properties (including affinity, solubility, hemolysis, half-life, and non-fouling) and outperforms both evolutionary and diffusion-based baselines. These results establish AReUReDi as a powerful, sequence-based framework for multi-property biomolecule generation.", "date": "2025-09-30", "categories": []}
{"id": "2510.02416", "title": "Cross-Platform DNA Methylation Classifier for the Eight Molecular Subtypes of Group 3 & 4 Medulloblastoma", "abstract": "Medulloblastoma is a malignant pediatric brain cancer, and the discovery of molecular subgroups is enabling personalized treatment strategies. In 2019, a consensus identified eight novel subtypes within Groups 3 and 4, each displaying heterogeneous characteristics. Classifiers are essential for translating these findings into clinical practice by supporting clinical trials, personalized therapy development and application, and patient monitoring. This study presents a DNA methylation-based, cross-platform machine learning classifier capable of distinguishing these subtypes on both HM450 and EPIC methylation array samples. Across two independent test sets, the model achieved weighted F1 = 0.95 and balanced accuracy = 0.957, consistent across platforms. As the first cross-platform solution, it provides backward compatibility while extending applicability to a newer platform, also enhancing accessibility. It also has the potential to become the first publicly available classifier for these subtypes once deployed through a web application, as planned in the future. This work overall takes steps in the direction of advancing precision medicine and improving clinical outcomes for patients within the majority prevalence medulloblastoma subgroups, groups 3 and 4.", "date": "2025-10-02", "categories": []}
{"id": "2505.12626", "title": "scSiameseClu: A Siamese Clustering Framework for Interpreting single-cell RNA Sequencing Data", "abstract": "Single-cell RNA sequencing (scRNA-seq) reveals cell heterogeneity, with cell clustering playing a key role in identifying cell types and marker genes. Recent advances, especially graph neural networks (GNNs)-based methods, have significantly improved clustering performance. However, the analysis of scRNA-seq data remains challenging due to noise, sparsity, and high dimensionality. Compounding these challenges, GNNs often suffer from over-smoothing, limiting their ability to capture complex biological information. In response, we propose scSiameseClu, a novel Siamese Clustering framework for interpreting single-cell RNA-seq data, comprising of 3 key steps: (1) Dual Augmentation Module, which applies biologically informed perturbations to the gene expression matrix and cell graph relationships to enhance representation robustness; (2) Siamese Fusion Module, which combines cross-correlation refinement and adaptive information fusion to capture complex cellular relationships while mitigating over-smoothing; and (3) Optimal Transport Clustering, which utilizes Sinkhorn distance to efficiently align cluster assignments with predefined proportions while maintaining balance. Comprehensive evaluations on seven real-world datasets demonstrate that scSiameseClu outperforms state-of-the-art methods in single-cell clustering, cell type annotation, and cell type classification, providing a powerful tool for scRNA-seq data interpretation.", "date": "2025-05-19", "categories": []}
{"id": "2510.01428", "title": "BioVERSE: Representation Alignment of Biomedical Modalities to LLMs for Multi-Modal Reasoning", "abstract": "Recent advances in large language models (LLMs) and biomedical foundation models (BioFMs) have achieved strong results in biological text reasoning, molecular modeling, and single-cell analysis, yet they remain siloed in disjoint embedding spaces, limiting cross-modal reasoning. We present BIOVERSE (Biomedical Vector Embedding Realignment for Semantic Engagement), a two-stage approach that adapts pretrained BioFMs as modality encoders and aligns them with LLMs through lightweight, modality-specific projection layers. The approach first aligns each modality to a shared LLM space through independently trained projections, allowing them to interoperate naturally, and then applies standard instruction tuning with multi-modal data to bring them together for downstream reasoning. By unifying raw biomedical data with knowledge embedded in LLMs, the approach enables zero-shot annotation, cross-modal question answering, and interactive, explainable dialogue. Across tasks spanning cell-type annotation, molecular description, and protein function reasoning, compact BIOVERSE configurations surpass larger LLM baselines while enabling richer, generative outputs than existing BioFMs, establishing a foundation for principled multi-modal biomedical reasoning.", "date": "2025-10-01", "categories": []}
{"id": "2510.02139", "title": "BioinfoMCP: A Unified Platform Enabling MCP Interfaces in Agentic Bioinformatics", "abstract": "Bioinformatics tools are essential for complex computational biology tasks, yet their integration with emerging AI-agent frameworks is hindered by incompatible interfaces, heterogeneous input-output formats, and inconsistent parameter conventions. The Model Context Protocol (MCP) provides a standardized framework for tool-AI communication, but manually converting hundreds of existing and rapidly growing specialized bioinformatics tools into MCP-compliant servers is labor-intensive and unsustainable. Here, we present BioinfoMCP, a unified platform comprising two components: BioinfoMCP Converter, which automatically generates robust MCP servers from tool documentation using large language models, and BioinfoMCP Benchmark, which systematically validates the reliability and versatility of converted tools across diverse computational tasks. We present a platform of 38 MCP-converted bioinformatics tools, extensively validated to show that 94.7% successfully executed complex workflows across three widely used AI-agent platforms. By removing technical barriers to AI automation, BioinfoMCP enables natural-language interaction with sophisticated bioinformatics analyses without requiring extensive programming expertise, offering a scalable path to intelligent, interoperable computational biology.", "date": "2025-10-02", "categories": []}
{"id": "2510.00086", "title": "Behavioural Classification in C. elegans: a Spatio-Temporal Analysis of Locomotion", "abstract": "The 1mm roundworm C. elegans is a model organism used in many sub-areas of biology to investigate different types of biological processes. In order to complement the n-vivo analysis with computer-based investigations, several methods have been proposed to simulate the worm behaviour. These methods extract discrete behavioural units from the flow of the worm movements using different types of tracking techniques. Nevertheless, these techniques require a clear view of the entire worm body, which is not always achievable. For example, this happens in high density worm conditions, which are particularly informative to understand the influence of the social context on the single worm behaviour. In this paper, we illustrate and evaluate a method to extract behavioural units from recordings of C. elegans movements which do not necessarily require a clear view of the entire worm body. Moreover, the behavioural units are defined by an unsupervised automatic pipeline which frees the process from predefined assumptions that inevitably bias the behavioural analysis. The behavioural units resulting from the automatic method are interpreted by comparing them with hand-designed behavioural units. The effectiveness of the automatic method is evaluated by measuring the extent to which the movement of a simulated worm, with an agent-based model, matches the movement of a natural worm. Our results indicate that spatio-temporal locomotory patterns emerge even from single point worm tracking. Moreover, we show that such patterns represent a fundamental aspect of the behavioural classification process.", "date": "2025-09-30", "categories": []}
{"id": "2510.00392", "title": "A Deep Learning Pipeline for Epilepsy Genomic Analysis Using GPT-2 XL and NVIDIA H100", "abstract": "Epilepsy is a chronic neurological condition characterized by recurrent seizures, with global prevalence estimated at 50 million people worldwide. While progress in high-throughput sequencing has allowed for broad-based transcriptomic profiling of brain tissues, the deciphering of these highly complex datasets remains one of the challenges. To address this issue, in this paper we propose a new analysis pipeline that integrates the power of deep learning strategies with GPU-acceleration computation for investigating Gene expression patterns in epilepsy. Specifically, our proposed approach employs GPT-2 XL, a transformer-based Large Language Model (LLM) with 1.5 billion parameters for genomic sequence analysis over the latest NVIDIA H100 Tensor Core GPUs based on Hopper architecture. Our proposed method enables efficient preprocessing of RNA sequence data, gene sequence encoding, and subsequent pattern identification. We conducted experiments on two epilepsy datasets including GEO accession GSE264537 and GSE275235. The obtained results reveal several significant transcriptomic modifications, including reduced hippocampal astrogliosis after ketogenic diet treatment as well as restored excitatory-inhibitory signaling equilibrium in zebrafish epilepsy model. Moreover, our results highlight the effectiveness of leveraging LLMs in combination with advanced hardware acceleration for transcriptomic characterization in neurological diseases.", "date": "2025-10-01", "categories": []}
{"id": "2510.00907", "title": "BoMGene: Integrating Boruta-mRMR feature selection for enhanced Gene expression classification", "abstract": "Feature selection is a crucial step in analyzing gene expression data, enhancing classification performance, and reducing computational costs for high-dimensional datasets. This paper proposes BoMGene, a hybrid feature selection method that effectively integrates two popular techniques: Boruta and Minimum Redundancy Maximum Relevance (mRMR). The method aims to optimize the feature space and enhance classification accuracy. Experiments were conducted on 25 publicly available gene expression datasets, employing widely used classifiers such as Support Vector Machine (SVM), Random Forest, XGBoost (XGB), and Gradient Boosting Machine (GBM). The results show that using the Boruta-mRMR combination cuts down the number of features chosen compared to just using mRMR, which helps to speed up training time while keeping or even improving classification accuracy compared to using individual feature selection methods. The proposed approach demonstrates clear advantages in accuracy, stability, and practical applicability for multi-class gene expression data analysis", "date": "2025-10-01", "categories": []}
{"id": "2510.01118", "title": "Breaking the Euclidean Barrier: Hyperboloid-Based Biological Sequence Analysis", "abstract": "Genomic sequence analysis plays a crucial role in various scientific and medical domains. Traditional machine-learning approaches often struggle to capture the complex relationships and hierarchical structures of sequence data when working in high-dimensional Euclidean spaces. This limitation hinders accurate sequence classification and similarity measurement. To address these challenges, this research proposes a method to transform the feature representation of biological sequences into the hyperboloid space. By applying a transformation, the sequences are mapped onto the hyperboloid, preserving their inherent structural information. Once the sequences are represented in the hyperboloid space, a kernel matrix is computed based on the hyperboloid features. The kernel matrix captures the pairwise similarities between sequences, enabling more effective analysis of biological sequence relationships. This approach leverages the inner product of the hyperboloid feature vectors to measure the similarity between pairs of sequences. The experimental evaluation of the proposed approach demonstrates its efficacy in capturing important sequence correlations and improving classification accuracy.", "date": "2025-10-01", "categories": []}
{"id": "2404.06167", "title": "scCDCG: Efficient Deep Structural Clustering for single-cell RNA-seq via Deep Cut-informed Graph Embedding", "abstract": "Single-cell RNA sequencing (scRNA-seq) is essential for unraveling cellular heterogeneity and diversity, offering invaluable insights for bioinformatics advancements. Despite its potential, traditional clustering methods in scRNA-seq data analysis often neglect the structural information embedded in gene expression profiles, crucial for understanding cellular correlations and dependencies. Existing strategies, including graph neural networks, face challenges in handling the inefficiency due to scRNA-seq data's intrinsic high-dimension and high-sparsity. Addressing these limitations, we introduce scCDCG (single-cell RNA-seq Clustering via Deep Cut-informed Graph), a novel framework designed for efficient and accurate clustering of scRNA-seq data that simultaneously utilizes intercellular high-order structural information. scCDCG comprises three main components: (i) A graph embedding module utilizing deep cut-informed techniques, which effectively captures intercellular high-order structural information, overcoming the over-smoothing and inefficiency issues prevalent in prior graph neural network methods. (ii) A self-supervised learning module guided by optimal transport, tailored to accommodate the unique complexities of scRNA-seq data, specifically its high-dimension and high-sparsity. (iii) An autoencoder-based feature learning module that simplifies model complexity through effective dimension reduction and feature extraction. Our extensive experiments on 6 datasets demonstrate scCDCG's superior performance and efficiency compared to 7 established models, underscoring scCDCG's potential as a transformative tool in scRNA-seq data analysis. Our code is available at: https://github.com/XPgogogo/scCDCG.", "date": "2024-04-09", "categories": []}
{"id": "2508.07043", "title": "K-Dense Analyst: Towards Fully Automated Scientific Analysis", "abstract": "The complexity of modern bioinformatics analysis has created a critical gap between data generation and developing scientific insights. While large language models (LLMs) have shown promise in scientific reasoning, they remain fundamentally limited when dealing with real-world analytical workflows that demand iterative computation, tool integration and rigorous validation. We introduce K-Dense Analyst, a hierarchical multi-agent system that achieves autonomous bioinformatics analysis through a dual-loop architecture. K-Dense Analyst, part of the broader K-Dense platform, couples planning with validated execution using specialized agents to decompose complex objectives into executable, verifiable tasks within secure computational environments. On BixBench, a comprehensive benchmark for open-ended biological analysis, K-Dense Analyst achieves 29.2% accuracy, surpassing the best-performing language model (GPT-5) by 6.3 percentage points, representing nearly 27% improvement over what is widely considered the most powerful LLM available. Remarkably, K-Dense Analyst achieves this performance using Gemini 2.5 Pro, which attains only 18.3% accuracy when used directly, demonstrating that our architectural innovations unlock capabilities far beyond the underlying model's baseline performance. Our insights demonstrate that autonomous scientific reasoning requires more than enhanced language models, it demands purpose-built systems that can bridge the gap between high-level scientific objectives and low-level computational execution. These results represent a significant advance toward fully autonomous computational biologists capable of accelerating discovery across the life sciences.", "date": "2025-08-09", "categories": []}
{"id": "2509.25274", "title": "DNABERT-2: Fine-Tuning a Genomic Language Model for Colorectal Gene Enhancer Classification", "abstract": "Gene enhancers control when and where genes switch on, yet their sequence diversity and tissue specificity make them hard to pinpoint in colorectal cancer. We take a sequence-only route and fine-tune DNABERT-2, a transformer genomic language model that uses byte-pair encoding to learn variable-length tokens from DNA. Using assays curated via the Johnston Cancer Research Centre at Queen's University Belfast, we assembled a balanced corpus of 2.34 million 1 kb enhancer sequences, applied summit-centered extraction and rigorous de-duplication including reverse-complement collapse, and split the data stratified by class. With a 4096-term vocabulary and a 232-token context chosen empirically, the DNABERT-2-117M classifier was trained with Optuna-tuned hyperparameters and evaluated on 350742 held-out sequences. The model reached PR-AUC 0.759, ROC-AUC 0.743, and best F1 0.704 at an optimized threshold (0.359), with recall 0.835 and precision 0.609. Against a CNN-based EnhancerNet trained on the same data, DNABERT-2 delivered stronger threshold-independent ranking and higher recall, although point accuracy was lower. To our knowledge, this is the first study to apply a second-generation genomic language model with BPE tokenization to enhancer classification in colorectal cancer, demonstrating the feasibility of capturing tumor-associated regulatory signals directly from DNA sequence alone. Overall, our results show that transformer-based genomic models can move beyond motif-level encodings toward holistic classification of regulatory elements, offering a novel path for cancer genomics. Next steps will focus on improving precision, exploring hybrid CNN-transformer designs, and validating across independent datasets to strengthen real-world utility.", "date": "2025-09-28", "categories": []}
{"id": "2509.26377", "title": "MC-GNNAS-Dock: Multi-criteria GNN-based Algorithm Selection for Molecular Docking", "abstract": "Molecular docking is a core tool in drug discovery for predicting ligand-target interactions. Despite the availability of diverse search-based and machine learning approaches, no single docking algorithm consistently dominates, as performance varies by context. To overcome this challenge, algorithm selection frameworks such as GNNAS-Dock, built on graph neural networks, have been proposed. This study introduces an enhanced system, MC-GNNAS-Dock, with three key advances. First, a multi-criteria evaluation integrates binding-pose accuracy (RMSD) with validity checks from PoseBusters, offering a more rigorous assessment. Second, architectural refinements by inclusion of residual connections strengthen predictive robustness. Third, rank-aware loss functions are incorporated to sharpen rank learning. Extensive experiments are performed on a curated dataset containing approximately 3200 protein-ligand complexes from PDBBind. MC-GNNAS-Dock demonstrates consistently superior performance, achieving up to 5.4% (3.4%) gains under composite criteria of RMSD below 1\\AA{} (2\\AA{}) with PoseBuster-validity compared to the single best solver (SBS) Uni-Mol Docking V2.", "date": "2025-09-30", "categories": []}
{"id": "2406.01651", "title": "FusionDTI: Fine-grained Binding Discovery with Token-level Fusion for Drug-Target Interaction", "abstract": "Predicting drug-target interaction (DTI) is critical in the drug discovery process. Despite remarkable advances in recent DTI models through the integration of representations from diverse drug and target encoders, such models often struggle to capture the fine-grained interactions between drugs and protein, i.e. the binding of specific drug atoms (or substructures) and key amino acids of proteins, which is crucial for understanding the binding mechanisms and optimising drug design. To address this issue, this paper introduces a novel model, called FusionDTI, which uses a token-level Fusion module to effectively learn fine-grained information for Drug-Target Interaction. In particular, our FusionDTI model uses the SELFIES representation of drugs to mitigate sequence fragment invalidation and incorporates the structure-aware (SA) vocabulary of target proteins to address the limitation of amino acid sequences in structural information, additionally leveraging pre-trained language models extensively trained on large-scale biomedical datasets as encoders to capture the complex information of drugs and targets. Experiments on three well-known benchmark datasets show that our proposed FusionDTI model achieves the best performance in DTI prediction compared with seven existing state-of-the-art baselines. Furthermore, our case study indicates that FusionDTI could highlight the potential binding sites, enhancing the explainability of the DTI prediction.", "date": "2024-06-03", "categories": []}
{"id": "2505.20346", "title": "PDFBench: A Benchmark for De novo Protein Design from Function", "abstract": "Function-guided protein design is a crucial task with significant applications in drug discovery and enzyme engineering. However, the field lacks a unified and comprehensive evaluation framework. Current models are assessed using inconsistent and limited subsets of metrics, which prevents fair comparison and a clear understanding of the relationships between different evaluation criteria. To address this gap, we introduce PDFBench, the first comprehensive benchmark for function-guided denovo protein design. Our benchmark systematically evaluates eight state-of-the-art models on 16 metrics across two key settings: description-guided design, for which we repurpose the Mol-Instructions dataset, originally lacking quantitative benchmarking, and keyword-guided design, for which we introduce a new test set, SwissTest, created with a strict datetime cutoff to ensure data integrity. By benchmarking across a wide array of metrics and analyzing their correlations, PDFBench enables more reliable model comparisons and provides key insights to guide future research.", "date": "2025-05-25", "categories": []}
{"id": "2509.22853", "title": "Patient-specific Biomolecular Instruction Tuning", "abstract": "Proteomics data is essential to pathogenic understanding of a disease phenotype. In cancer, analysis of molecular signatures enables precision medicine through the identification of biological processes that drive individualized tumor progression, therapeutic resistance, and clinical heterogeneity. Recent advances in multimodal large language models (LLMs) have shown remarkable capacity to integrate and reason across heterogeneous data modalities. However, performing multi-modal language modeling for molecular understanding of patient-specific proteomics remains a significant challenge due to two barriers: (1) the lack of instruction-tuning datasets that enable clinical interpretation from proteomics data, and (2) the absence of language modeling architectures designed to capture the rich heterogeneity of molecular data. In this work, we introduce CPTAC-PROTSTRUCT, the first instruction tuning dataset for molecular understanding of oncology, comprising over 400k open-ended examples derived from individualized proteomic profiles curated from the largest national proteomics cancer study (CPTAC). Additionally, we propose KRONOS (Knowledge Representation of patient Omics Networks in Oncology via Structured tuning), a novel graph-LLM framework that leverages molecular interaction topology with proteomics to learn patient-specific graph representations for enhanced clinical reasoning. We show that KRONOS achieves competitive performance across benchmark clinical tasks, including molecular classification, temporal trajectory modeling, and tumor stage prediction from proteomics data. Ultimately, this approach empowers LLMs to understand patient-level pathogenesis, advancing precision medicine through more accurate diagnosis, prognosis, and treatment stratification.", "date": "2025-09-26", "categories": []}
{"id": "2509.23254", "title": "ABConformer: Physics-inspired Sliding Attention for Antibody-Antigen Interface Prediction", "abstract": "Accurate prediction of antibody-antigen (Ab-Ag) interfaces is critical for vaccine design, immunodiagnostics, and therapeutic antibody development. However, achieving reliable predictions from sequences alone remains a challenge. In this paper, we present ABCONFORMER, a model based on the Conformer backbone that captures both local and global features of a biosequence. To accurately capture Ab-Ag interactions, we introduced the physics-inspired sliding attention, enabling residue-level contact recovery without relying on three-dimensional structural data. ABConformer can accurately predict paratopes and epitopes given the antibody and antigen sequence, and predict pan-epitopes on the antigen without antibody information. In comparison experiments, ABCONFORMER achieves state-of-the-art performance on a recent SARS-CoV-2 Ab-Ag dataset, and surpasses widely used sequence-based methods for antibody-agnostic epitope prediction. Ablation studies further quantify the contribution of each component, demonstrating that, compared to conventional cross-attention, sliding attention significantly enhances the precision of epitope prediction. To facilitate reproducibility, we will release the code under an open-source license upon acceptance.", "date": "2025-09-27", "categories": []}
{"id": "2509.24933", "title": "Is Sequence Information All You Need for Bayesian Optimization of Antibodies?", "abstract": "Bayesian optimization is a natural candidate for the engineering of antibody therapeutic properties, which is often iterative and expensive. However, finding the optimal choice of surrogate model for optimization over the highly structured antibody space is difficult, and may differ depending on the property being optimized. Moreover, to the best of our knowledge, no prior works have attempted to incorporate structural information into antibody Bayesian optimization. In this work, we explore different approaches to incorporating structural information into Bayesian optimization, and compare them to a variety of sequence-only approaches on two different antibody properties, binding affinity and stability. In addition, we propose the use of a protein language model-based ``soft constraint,'' which helps guide the optimization to promising regions of the space. We find that certain types of structural information improve data efficiency in early optimization rounds for stability, but have equivalent peak performance. Moreover, when incorporating the protein language model soft constraint we find that the data efficiency gap is diminished for affinity and eliminated for stability, resulting in sequence-only methods that match the performance of structure-based methods, raising questions about the necessity of structure in Bayesian optimization for antibodies.", "date": "2025-09-29", "categories": []}
{"id": "2509.21424", "title": "PhenoMoler: Phenotype-Guided Molecular Optimization via Chemistry Large Language Model", "abstract": "Current molecular generative models primarily focus on improving drug-target binding affinity and specificity, often neglecting the system-level phenotypic effects elicited by compounds. Transcriptional profiles, as molecule-level readouts of drug-induced phenotypic shifts, offer a powerful opportunity to guide molecular design in a phenotype-aware manner. We present PhenoMoler, a phenotype-guided molecular generation framework that integrates a chemistry large language model with expression profiles to enable biologically informed drug design. By conditioning the generation on drug-induced differential expression signatures, PhenoMoler explicitly links transcriptional responses to chemical structure. By selectively masking and reconstructing specific substructures-scaffolds, side chains, or linkers-PhenoMoler supports fine-grained, controllable molecular optimization. Extensive experiments demonstrate that PhenoMoler generates chemically valid, novel, and diverse molecules aligned with desired phenotypic profiles. Compared to FDA-approved drugs, the generated compounds exhibit comparable or enhanced drug-likeness (QED), optimized physicochemical properties, and superior binding affinity to key cancer targets. These findings highlight PhenoMoler's potential for phenotype-guided and structure-controllable molecular optimization.", "date": "2025-09-25", "categories": []}
{"id": "2509.21689", "title": "SpecMER: Fast Protein Generation with K-mer Guided Speculative Decoding", "abstract": "Autoregressive models have transformed protein engineering by enabling the generation of novel protein sequences beyond those found in nature. However, their sequential inference introduces significant latency, limiting their utility in high-throughput protein screening. Speculative decoding accelerates generation by employing a lightweight draft model to sample tokens, which a larger target model then verifies and refines. Yet, in protein sequence generation, draft models are typically agnostic to the structural and functional constraints of the target protein, leading to biologically implausible outputs and a shift in the likelihood distribution of generated sequences. We introduce SpecMER (Speculative Decoding via k-mer Guidance), a novel framework that incorporates biological, structural, and functional priors using k-mer motifs extracted from multiple sequence alignments. By scoring candidate sequences in parallel and selecting those most consistent with known biological patterns, SpecMER significantly improves sequence plausibility while retaining the efficiency of speculative decoding. SpecMER achieves 24-32% speedup over standard autoregressive decoding, along with higher acceptance rates and improved sequence likelihoods.", "date": "2025-09-25", "categories": []}
{"id": "2509.20842", "title": "Robust Multi-Omics Integration from Incomplete Modalities Significantly Improves Prediction of Alzheimer's Disease", "abstract": "Multi-omics data capture complex biomolecular interactions and provide insights into metabolism and disease. However, missing modalities hinder integrative analysis across heterogeneous omics. To address this, we present MOIRA (Multi-Omics Integration with Robustness to Absent modalities), an early integration method enabling robust learning from incomplete omics data via representation alignment and adaptive aggregation. MOIRA leverages all samples, including those with missing modalities, by projecting each omics dataset onto a shared embedding space where a learnable weighting mechanism fuses them. Evaluated on the Religious Order Study and Memory and Aging Project (ROSMAP) dataset for Alzheimer's Disease (AD), MOIRA outperformed existing approaches, and further ablation studies confirmed modality-wise contributions. Feature importance analysis revealed AD-related biomarkers consistent with prior literature, highlighting the biological relevance of our approach.", "date": "2025-09-25", "categories": []}
{"id": "2506.06290", "title": "CellCLIP -- Learning Perturbation Effects in Cell Painting via Text-Guided Contrastive Learning", "abstract": "High-content screening (HCS) assays based on high-throughput microscopy techniques such as Cell Painting have enabled the interrogation of cells' morphological responses to perturbations at an unprecedented scale. The collection of such data promises to facilitate a better understanding of the relationships between different perturbations and their effects on cellular state. Towards achieving this goal, recent advances in cross-modal contrastive learning could, in theory, be leveraged to learn a unified latent space that aligns perturbations with their corresponding morphological effects. However, the application of such methods to HCS data is not straightforward due to substantial differences in the semantics of Cell Painting images compared to natural images, and the difficulty of representing different classes of perturbations (e.g., small molecule vs CRISPR gene knockout) in a single latent space. In response to these challenges, here we introduce CellCLIP, a cross-modal contrastive learning framework for HCS data. CellCLIP leverages pre-trained image encoders coupled with a novel channel encoding scheme to better capture relationships between different microscopy channels in image embeddings, along with natural language encoders for representing perturbations. Our framework outperforms current open-source models, demonstrating the best performance in both cross-modal retrieval and biologically meaningful downstream tasks while also achieving significant reductions in computation time.", "date": "2025-05-16", "categories": []}
{"id": "2509.18408", "title": "Explicit Path CGR: Maintaining Sequence Fidelity in Geometric Representations", "abstract": "We present a novel information-preserving Chaos Game Representation (CGR) method, also called Reverse-CGR (R-CGR), for biological sequence analysis that addresses the fundamental limitation of traditional CGR approaches - the loss of sequence information during geometric mapping. Our method introduces complete sequence recovery through explicit path encoding combined with rational arithmetic precision control, enabling perfect sequence reconstruction from stored geometric traces. Unlike purely geometric approaches, our reversibility is achieved through comprehensive path storage that maintains both positional and character information at each step. We demonstrate the effectiveness of R-CGR on biological sequence classification tasks, achieving competitive performance compared to traditional sequence-based methods while providing interpretable geometric visualizations. The approach generates feature-rich images suitable for deep learning while maintaining complete sequence information through explicit encoding, opening new avenues for interpretable bioinformatics analysis where both accuracy and sequence recovery are essential.", "date": "2025-09-22", "categories": []}
{"id": "2509.19566", "title": "Nano Bio-Agents (NBA): Small Language Model Agents for Genomics", "abstract": "We investigate the application of Small Language Models (<10 billion parameters) for genomics question answering via agentic framework to address hallucination issues and computational cost challenges. The Nano Bio-Agent (NBA) framework we implemented incorporates task decomposition, tool orchestration, and API access into well-established systems such as NCBI and AlphaGenome. Results show that SLMs combined with such agentic framework can achieve comparable and in many cases superior performance versus existing approaches utilising larger models, with our best model-agent combination achieving 98% accuracy on the GeneTuring benchmark. Notably, small 3-10B parameter models consistently achieve 85-97% accuracy while requiring much lower computational resources than conventional approaches. This demonstrates promising potential for efficiency gains, cost savings, and democratization of ML-powered genomics tools while retaining highly robust and accurate performance.", "date": "2025-09-23", "categories": []}
{"id": "2509.19604", "title": "Improved Therapeutic Antibody Reformatting through Multimodal Machine Learning", "abstract": "Modern therapeutic antibody design often involves composing multi-part assemblages of individual functional domains, each of which may be derived from a different source or engineered independently. While these complex formats can expand disease applicability and improve safety, they present a significant engineering challenge: the function and stability of individual domains are not guaranteed in the novel format, and the entire molecule may no longer be synthesizable. To address these challenges, we develop a machine learning framework to predict \"reformatting success\" -- whether converting an antibody from one format to another will succeed or not. Our framework incorporates both antibody sequence and structural context, incorporating an evaluation protocol that reflects realistic deployment scenarios. In experiments on a real-world antibody reformatting dataset, we find the surprising result that large pretrained protein language models (PLMs) fail to outperform simple, domain-tailored, multimodal representations. This is particularly evident in the most difficult evaluation setting, where we test model generalization to a new starting antibody. In this challenging \"new antibody, no data\" scenario, our best multimodal model achieves high predictive accuracy, enabling prioritization of promising candidates and reducing wasted experimental effort.", "date": "2025-09-23", "categories": []}
{"id": "2509.19766", "title": "Dynamicasome: a molecular dynamics-guided and AI-driven pathogenicity prediction catalogue for all genetic mutations", "abstract": "Advances in genomic medicine accelerate the identi cation of mutations in disease-associated genes, but the pathogenicity of many mutations remains unknown, hindering their use in diagnostics and clinical decision-making. Predictive AI models are generated to combat this issue, but current tools display low accuracy when tested against functionally validated datasets. We show that integrating detailed conformational data extracted from molecular dynamics simulations (MDS) into advanced AI-based models increases their predictive power. We carry out an exhaustive mutational analysis of the disease gene PMM2 and subject structural models of each variant to MDS. AI models trained on this dataset outperform existing tools when predicting the known pathogenicity of mutations. Our best performing model, a neuronal networks model, also predicts the pathogenicity of several PMM2 mutations currently considered of unknown signi cance. We believe this model helps alleviate the burden of unknown variants in genomic medicine.", "date": "2025-09-23", "categories": []}
{"id": "2509.19896", "title": "Efficient Cell Painting Image Representation Learning via Cross-Well Aligned Masked Siamese Network", "abstract": "Computational models that predict cellular phenotypic responses to chemical and genetic perturbations can accelerate drug discovery by prioritizing therapeutic hypotheses and reducing costly wet-lab iteration. However, extracting biologically meaningful and batch-robust cell painting representations remains challenging. Conventional self-supervised and contrastive learning approaches often require a large-scale model and/or a huge amount of carefully curated data, still struggling with batch effects. We present Cross-Well Aligned Masked Siamese Network (CWA-MSN), a novel representation learning framework that aligns embeddings of cells subjected to the same perturbation across different wells, enforcing semantic consistency despite batch effects. Integrated into a masked siamese architecture, this alignment yields features that capture fine-grained morphology while remaining data- and parameter-efficient. For instance, in a gene-gene relationship retrieval benchmark, CWA-MSN outperforms the state-of-the-art publicly available self-supervised (OpenPhenom) and contrastive learning (CellCLIP) methods, improving the benchmark scores by +29\\% and +9\\%, respectively, while training on substantially fewer data (e.g., 0.2M images for CWA-MSN vs. 2.2M images for OpenPhenom) or smaller model size (e.g., 22M parameters for CWA-MSN vs. 1.48B parameters for CellCLIP). Extensive experiments demonstrate that CWA-MSN is a simple and effective way to learn cell image representation, enabling efficient phenotype modeling even under limited data and parameter budgets.", "date": "2025-09-24", "categories": []}
{"id": "2509.20240", "title": "A HyperGraphMamba-Based Multichannel Adaptive Model for ncRNA Classification", "abstract": "Non-coding RNAs (ncRNAs) play pivotal roles in gene expression regulation and the pathogenesis of various diseases. Accurate classification of ncRNAs is essential for functional annotation and disease diagnosis. To address existing limitations in feature extraction depth and multimodal fusion, we propose HGMamba-ncRNA, a HyperGraphMamba-based multichannel adaptive model, which integrates sequence, secondary structure, and optionally available expression features of ncRNAs to enhance classification performance. Specifically, the sequence of ncRNA is modeled using a parallel Multi-scale Convolution and LSTM architecture (MKC-L) to capture both local patterns and long-range dependencies of nucleotides. The structure modality employs a multi-scale graph transformer (MSGraphTransformer) to represent the multi-level topological characteristics of ncRNA secondary structures. The expression modality utilizes a Chebyshev Polynomial-based Kolmogorov-Arnold Network (CPKAN) to effectively model and interpret high-dimensional expression profiles. Finally, by incorporating virtual nodes to facilitate efficient and comprehensive multimodal interaction, HyperGraphMamba is proposed to adaptively align and integrate multichannel heterogeneous modality features. Experiments conducted on three public datasets demonstrate that HGMamba-ncRNA consistently outperforms state-of-the-art methods in terms of accuracy and other metrics. Extensive empirical studies further confirm the model's robustness, effectiveness, and strong transferability, offering a novel and reliable strategy for complex ncRNA functional classification. Code and datasets are available at https://anonymous.4open.science/r/HGMamba-ncRNA-94D0.", "date": "2025-09-24", "categories": []}
{"id": "2509.20290", "title": "PGCLODA: Prompt-Guided Graph Contrastive Learning for Oligopeptide-Infectious Disease Association Prediction", "abstract": "Infectious diseases continue to pose a serious threat to public health, underscoring the urgent need for effective computational approaches to screen novel anti-infective agents. Oligopeptides have emerged as promising candidates in antimicrobial research due to their structural simplicity, high bioavailability, and low susceptibility to resistance. Despite their potential, computational models specifically designed to predict associations between oligopeptides and infectious diseases remain scarce. This study introduces a prompt-guided graph-based contrastive learning framework (PGCLODA) to uncover potential associations. A tripartite graph is constructed with oligopeptides, microbes, and diseases as nodes, incorporating both structural and semantic information. To preserve critical regions during contrastive learning, a prompt-guided graph augmentation strategy is employed to generate meaningful paired views. A dual encoder architecture, integrating Graph Convolutional Network (GCN) and Transformer, is used to jointly capture local and global features. The fused embeddings are subsequently input into a multilayer perceptron (MLP) classifier for final prediction. Experimental results on a benchmark dataset indicate that PGCLODA consistently outperforms state-of-the-art models in AUROC, AUPRC, and accuracy. Ablation and hyperparameter studies confirm the contribution of each module. Case studies further validate the generalization ability of PGCLODA and its potential to uncover novel, biologically relevant associations. These findings offer valuable insights for mechanism-driven discovery and oligopeptide-based drug development. The source code of PGCLODA is available online at https://github.com/jjnlcode/PGCLODA.", "date": "2025-09-24", "categories": []}
{"id": "2412.19191", "title": "Biology-Instructions: A Dataset and Benchmark for Multi-Omics Sequence Understanding Capability of Large Language Models", "abstract": "Large language models (LLMs) have shown remarkable capabilities in general domains, but their application to multi-omics biology remains underexplored. To address this gap, we introduce Biology-Instructions, the first large-scale instruction-tuning dataset for multi-omics biological sequences, including DNA, RNA, proteins, and multi-molecules. This dataset bridges LLMs and complex biological sequence-related tasks, enhancing their versatility and reasoning while maintaining conversational fluency. We also highlight significant limitations of current state-of-the-art LLMs on multi-omics tasks without specialized training. To overcome this, we propose ChatMultiOmics, a strong baseline with a novel three-stage training pipeline, demonstrating superior biological understanding through Biology-Instructions. Both resources are publicly available, paving the way for better integration of LLMs in multi-omics analysis. The Biology-Instructions is publicly available at: https://github.com/hhnqqq/Biology-Instructions.", "date": "2024-12-26", "categories": []}
{"id": "2509.18140", "title": "A Machine Learning Framework for Pathway-Driven Therapeutic Target Discovery in Metabolic Disorders", "abstract": "Metabolic disorders, particularly type 2 diabetes mellitus (T2DM), represent a significant global health burden, disproportionately impacting genetically predisposed populations such as the Pima Indians (a Native American tribe from south central Arizona). This study introduces a novel machine learning (ML) framework that integrates predictive modeling with gene-agnostic pathway mapping to identify high-risk individuals and uncover potential therapeutic targets. Using the Pima Indian dataset, logistic regression and t-tests were applied to identify key predictors of T2DM, yielding an overall model accuracy of 78.43%. To bridge predictive analytics with biological relevance, we developed a pathway mapping strategy that links identified predictors to critical signaling networks, including insulin signaling, AMPK, and PPAR pathways. This approach provides mechanistic insights without requiring direct molecular data. Building upon these connections, we propose therapeutic strategies such as dual GLP-1/GIP receptor agonists, AMPK activators, SIRT1 modulators, and phytochemical, further validated through pathway enrichment analyses. Overall, this framework advances precision medicine by offering interpretable and scalable solutions for early detection and targeted intervention in metabolic disorders. The key contributions of this work are: (1) development of an ML framework combining logistic regression and principal component analysis (PCA) for T2DM risk prediction; (2) introduction of a gene-agnostic pathway mapping approach to generate mechanistic insights; and (3) identification of novel therapeutic strategies tailored for high-risk populations.", "date": "2025-09-14", "categories": []}
{"id": "2509.18153", "title": "A deep reinforcement learning platform for antibiotic discovery", "abstract": "Antimicrobial resistance (AMR) is projected to cause up to 10 million deaths annually by 2050, underscoring the urgent need for new antibiotics. Here we present ApexAmphion, a deep-learning framework for de novo design of antibiotics that couples a 6.4-billion-parameter protein language model with reinforcement learning. The model is first fine-tuned on curated peptide data to capture antimicrobial sequence regularities, then optimised with proximal policy optimization against a composite reward that combines predictions from a learned minimum inhibitory concentration (MIC) classifier with differentiable physicochemical objectives. In vitro evaluation of 100 designed peptides showed low MIC values (nanomolar range in some cases) for all candidates (100% hit rate). Moreover, 99 our of 100 compounds exhibited broad-spectrum antimicrobial activity against at least two clinically relevant bacteria. The lead molecules killed bacteria primarily by potently targeting the cytoplasmic membrane. By unifying generation, scoring and multi-objective optimization with deep reinforcement learning in a single pipeline, our approach rapidly produces diverse, potent candidates, offering a scalable route to peptide antibiotics and a platform for iterative steering toward potency and developability within hours.", "date": "2025-09-16", "categories": []}
{"id": "2509.18529", "title": "Reverse-Complement Consistency for DNA Language Models", "abstract": "A fundamental property of DNA is that the reverse complement (RC) of a sequence often carries identical biological meaning. However, state-of-the-art DNA language models frequently fail to capture this symmetry, producing inconsistent predictions for a sequence and its RC counterpart, which undermines their reliability. In this work, we introduce Reverse-Complement Consistency Regularization (RCCR), a simple and model-agnostic fine-tuning objective that directly penalizes the divergence between a model's prediction on a sequence and the aligned prediction on its reverse complement. We evaluate RCCR across three diverse backbones (Nucleotide Transformer, HyenaDNA, DNABERT-2) on a wide range of genomic tasks, including sequence classification, scalar regression, and profile prediction. Our experiments show that RCCR substantially improves RC robustness by dramatically reducing prediction flips and errors, all while maintaining or improving task accuracy compared to baselines such as RC data augmentation and test-time averaging. By integrating a key biological prior directly into the learning process, RCCR produces a single, intrinsically robust, and computationally efficient model fine-tuning recipe for diverse biology tasks.", "date": "2025-09-23", "categories": []}
{"id": "2509.18998", "title": "Bayesian Calibration and Model Assessment of Cell Migration Dynamics with Surrogate Model Integration", "abstract": "Computational models provide crucial insights into complex biological processes such as cancer evolution, but their mechanistic nature often makes them nonlinear and parameter-rich, complicating calibration. We systematically evaluate parameter probability distributions in cell migration models using Bayesian calibration across four complementary strategies: parametric and surrogate models, each with and without explicit model discrepancy. This approach enables joint analysis of parameter uncertainty, predictive performance, and interpretability. Applied to a real data experiment of glioblastoma progression in microfluidic devices, surrogate models achieve higher computational efficiency and predictive accuracy, whereas parametric models yield more reliable parameter estimates due to their mechanistic grounding. Incorporating model discrepancy exposes structural limitations, clarifying where model refinement is necessary. Together, these comparisons offer practical guidance for calibrating and improving computational models of complex biological systems.", "date": "2025-09-23", "categories": []}
{"id": "2509.17224", "title": "AI-based Methods for Simulating, Sampling, and Predicting Protein Ensembles", "abstract": "Advances in deep learning have opened an era of abundant and accurate predicted protein structures; however, similar progress in protein ensembles has remained elusive. This review highlights several recent research directions towards AI-based predictions of protein ensembles, including coarse-grained force fields, generative models, multiple sequence alignment perturbation methods, and modeling of ensemble descriptors. An emphasis is placed on realistic assessments of the technological maturity of current methods, the strengths and weaknesses of broad families of techniques, and promising machine learning frameworks at an early stage of development. We advocate for \"closing the loop\" between model training, simulation, and inference to overcome challenges in training data availability and to enable the next generation of models.", "date": "2025-09-21", "categories": []}
{"id": "2509.14723", "title": "Transcoder-based Circuit Analysis for Interpretable Single-Cell Foundation Models", "abstract": "Single-cell foundation models (scFMs) have demonstrated state-of-the-art performance on various tasks, such as cell-type annotation and perturbation response prediction, by learning gene regulatory networks from large-scale transcriptome data. However, a significant challenge remains: the decision-making processes of these models are less interpretable compared to traditional methods like differential gene expression analysis. Recently, transcoders have emerged as a promising approach for extracting interpretable decision circuits from large language models (LLMs). In this work, we train a transcoder on the cell2sentence (C2S) model, a state-of-the-art scFM. By leveraging the trained transcoder, we extract internal decision-making circuits from the C2S model. We demonstrate that the discovered circuits correspond to real-world biological mechanisms, confirming the potential of transcoders to uncover biologically plausible pathways within complex single-cell models.", "date": "2025-09-18", "categories": []}
{"id": "2402.17131", "title": "Predicting O-GlcNAcylation Sites in Mammalian Proteins with Transformers and RNNs Trained with a New Loss Function", "abstract": "O-GlcNAcylation, a subtype of glycosylation, has the potential to be an important target for therapeutics, but methods to reliably predict O-GlcNAcylation sites had not been available until 2023; a 2021 review correctly noted that published models were insufficient and failed to generalize. Moreover, many are no longer usable. In 2023, a considerably better recurrent neural network (RNN) model was published. This article creates improved models by using a new loss function, which we call the weighted focal differentiable MCC. RNN models trained with this new loss display superior performance to models trained using the weighted cross-entropy loss; this new function can also be used to fine-tune trained models. An RNN trained with this loss achieves state-of-the-art performance in O-GlcNAcylation site prediction with an F$_1$ score of 38.88% and an MCC of 38.20% on an independent test set from the largest dataset available.", "date": "2024-02-27", "categories": []}
{"id": "2507.13912", "title": "Self-supervised learning on gene expression data", "abstract": "Predicting phenotypes from gene expression data is a crucial task in biomedical research, enabling insights into disease mechanisms, drug responses, and personalized medicine. Traditional machine learning and deep learning rely on supervised learning, which requires large quantities of labeled data that are costly and time-consuming to obtain in the case of gene expression data. Self-supervised learning has recently emerged as a promising approach to overcome these limitations by extracting information directly from the structure of unlabeled data. In this study, we investigate the application of state-of-the-art self-supervised learning methods to bulk gene expression data for phenotype prediction. We selected three self-supervised methods, based on different approaches, to assess their ability to exploit the inherent structure of the data and to generate qualitative representations which can be used for downstream predictive tasks. By using several publicly available gene expression datasets, we demonstrate how the selected methods can effectively capture complex information and improve phenotype prediction accuracy. The results obtained show that self-supervised learning methods can outperform traditional supervised models besides offering significant advantage by reducing the dependency on annotated data. We provide a comprehensive analysis of the performance of each method by highlighting their strengths and limitations. We also provide recommendations for using these methods depending on the case under study. Finally, we outline future research directions to enhance the application of self-supervised learning in the field of gene expression data analysis. This study is the first work that deals with bulk RNA-Seq data and self-supervised learning.", "date": "2025-07-18", "categories": []}
{"id": "2509.13376", "title": "Unleashing the power of computational insights in revealing the complexity of biological systems in the new era of spatial multi-omics", "abstract": "Recent advances in spatial omics technologies have revolutionized our ability to study biological systems with unprecedented resolution. By preserving the spatial context of molecular measurements, these methods enable comprehensive mapping of cellular heterogeneity, tissue architecture, and dynamic biological processes in developmental biology, neuroscience, oncology, and evolutionary studies. This review highlights a systematic overview of the continuous advancements in both technology and computational algorithms that are paving the way for a deeper, more systematic comprehension of the structure and mechanisms of mammalian tissues and organs by using spatial multi-omics. Our viewpoint demonstrates how advanced machine learning algorithms and multi-omics integrative modeling can decode complex biological processes, including the spatial organization and topological relationships of cells during organ development, as well as key molecular signatures and regulatory networks underlying tumorigenesis and metastasis. Finally, we outline future directions for technological innovation and modeling insights of spatial omics in precision medicine.", "date": "2025-09-16", "categories": []}
{"id": "2509.14029", "title": "Deep Learning-Driven Peptide Classification in Biological Nanopores", "abstract": "A device capable of performing real time classification of proteins in a clinical setting would allow for inexpensive and rapid disease diagnosis. One such candidate for this technology are nanopore devices. These devices work by measuring a current signal that arises when a protein or peptide enters a nanometer-length-scale pore. Should this current be uniquely related to the structure of the peptide and its interactions with the pore, the signals can be used to perform identification. While such a method would allow for real time identification of peptides and proteins in a clinical setting, to date, the complexities of these signals limit their accuracy. In this work, we tackle the issue of classification by converting the current signals into scaleogram images via wavelet transforms, capturing amplitude, frequency, and time information in a modality well-suited to machine learning algorithms. When tested on 42 peptides, our method achieved a classification accuracy of ~$81\\,\\%$, setting a new state-of-the-art in the field and taking a step toward practical peptide/protein diagnostics at the point of care. In addition, we demonstrate model transfer techniques that will be critical when deploying these models into real hardware, paving the way to a new method for real-time disease diagnosis.", "date": "2025-09-17", "categories": []}
{"id": "2509.12732", "title": "A Novel Recurrent Neural Network Framework for Prediction and Treatment of Oncogenic Mutation Progression", "abstract": "Despite significant medical advancements, cancer remains the second leading cause of death, with over 600,000 deaths per year in the US. One emerging field, pathway analysis, is promising but still relies on manually derived wet lab data, which is time-consuming to acquire. This work proposes an efficient, effective end-to-end framework for Artificial Intelligence (AI) based pathway analysis that predicts both cancer severity and mutation progression, thus recommending possible treatments. The proposed technique involves a novel combination of time-series machine learning models and pathway analysis. First, mutation sequences were isolated from The Cancer Genome Atlas (TCGA) Database. Then, a novel preprocessing algorithm was used to filter key mutations by mutation frequency. This data was fed into a Recurrent Neural Network (RNN) that predicted cancer severity. Then, the model probabilistically used the RNN predictions, information from the preprocessing algorithm, and multiple drug-target databases to predict future mutations and recommend possible treatments. This framework achieved robust results and Receiver Operating Characteristic (ROC) curves (a key statistical metric) with accuracies greater than 60%, similar to existing cancer diagnostics. In addition, preprocessing played an instrumental role in isolating important mutations, demonstrating that each cancer stage studied may contain on the order of a few-hundred key driver mutations, consistent with current research. Heatmaps based on predicted gene frequency were also generated, highlighting key mutations in each cancer. Overall, this work is the first to propose an efficient, cost-effective end-to-end framework for projecting cancer progression and providing possible treatments without relying on expensive, time-consuming wet lab work.", "date": "2025-09-16", "categories": []}
{"id": "2509.12976", "title": "SHREC 2025: Protein surface shape retrieval including electrostatic potential", "abstract": "This SHREC 2025 track dedicated to protein surface shape retrieval involved 9 participating teams. We evaluated the performance in retrieval of 15 proposed methods on a large dataset of 11,555 protein surfaces with calculated electrostatic potential (a key molecular surface descriptor). The performance in retrieval of the proposed methods was evaluated through different metrics (Accuracy, Balanced accuracy, F1 score, Precision and Recall). The best retrieval performance was achieved by the proposed methods that used the electrostatic potential complementary to molecular surface shape. This observation was also valid for classes with limited data which highlights the importance of taking into account additional molecular surface descriptors.", "date": "2025-09-16", "categories": []}
{"id": "2509.13216", "title": "Flow-Based Fragment Identification via Binding Site-Specific Latent Representations", "abstract": "Fragment-based drug design is a promising strategy leveraging the binding of small chemical moieties that can efficiently guide drug discovery. The initial step of fragment identification remains challenging, as fragments often bind weakly and non-specifically. We developed a protein-fragment encoder that relies on a contrastive learning approach to map both molecular fragments and protein surfaces in a shared latent space. The encoder captures interaction-relevant features and allows to perform virtual screening as well as generative design with our new method LatentFrag. In LatentFrag, fragment embeddings and positions are generated conditioned on the protein surface while being chemically realistic by construction. Our expressive fragment and protein representations allow location of protein-fragment interaction sites with high sensitivity and we observe state-of-the-art fragment recovery rates when sampling from the learned distribution of latent fragment embeddings. Our generative method outperforms common methods such as virtual screening at a fraction of its computational cost providing a valuable starting point for fragment hit discovery. We further show the practical utility of LatentFrag and extend the workflow to full ligand design tasks. Together, these approaches contribute to advancing fragment identification and provide valuable tools for fragment-based drug discovery.", "date": "2025-09-16", "categories": []}
{"id": "2509.10527", "title": "An Interpretable Ensemble Framework for Multi-Omics Dementia Biomarker Discovery Under HDLSS Conditions", "abstract": "Biomarker discovery in neurodegenerative diseases requires robust, interpretable frameworks capable of integrating high-dimensional multi-omics data under low-sample conditions. We propose a novel ensemble approach combining Graph Attention Networks (GAT), MultiOmics Variational AutoEncoder (MOVE), Elastic-net sparse regression, and Storey's False Discovery Rate (FDR). This framework is benchmarked against state-of-the-art methods including DIABLO, MOCAT, AMOGEL, and MOMLIN. We evaluate performance using both simulated multi-omics data and the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. Our method demonstrates superior predictive accuracy, feature selection precision, and biological relevance. Biomarker gene maps derived from both datasets are visualized and interpreted, offering insights into latent molecular mechanisms underlying dementia.", "date": "2025-09-04", "categories": []}
{"id": "2509.10575", "title": "Gene-R1: Reasoning with Data-Augmented Lightweight LLMs for Gene Set Analysis", "abstract": "The gene set analysis (GSA) is a foundational approach for uncovering the molecular functions associated with a group of genes. Recently, LLM-powered methods have emerged to annotate gene sets with biological functions together with coherent explanatory insights. However, existing studies primarily focus on proprietary models, which have been shown to outperform their open-source counterparts despite concerns over cost and data privacy. Furthermore, no research has investigated the application of advanced reasoning strategies to the GSA task. To address this gap, we introduce Gene-R1, a data-augmented learning framework that equips lightweight and open-source LLMs with step-by-step reasoning capabilities tailored to GSA. Experiments on 1,508 in-distribution gene sets demonstrate that Gene-R1 achieves substantial performance gains, matching commercial LLMs. On 106 out-of-distribution gene sets, Gene-R1 performs comparably to both commercial and large-scale LLMs, exhibiting robust generalizability across diverse gene sources.", "date": "2025-09-11", "categories": []}
{"id": "2509.11046", "title": "Hybrid Quantum Neural Networks for Efficient Protein-Ligand Binding Affinity Prediction", "abstract": "Protein-ligand binding affinity is critical in drug discovery, but experimentally determining it is time-consuming and expensive. Artificial intelligence (AI) has been used to predict binding affinity, significantly accelerating this process. However, the high-performance requirements and vast datasets involved in affinity prediction demand increasingly large AI models, requiring substantial computational resources and training time. Quantum machine learning has emerged as a promising solution to these challenges. In particular, hybrid quantum-classical models can reduce the number of parameters while maintaining or improving performance compared to classical counterparts. Despite these advantages, challenges persist: why hybrid quantum models achieve these benefits, whether quantum neural networks (QNNs) can replace classical neural networks, and whether such models are feasible on noisy intermediate-scale quantum (NISQ) devices. This study addresses these challenges by proposing a hybrid quantum neural network (HQNN) that empirically demonstrates the capability to approximate non-linear functions in the latent feature space derived from classical embedding. The primary goal of this study is to achieve a parameter-efficient model in binding affinity prediction while ensuring feasibility on NISQ devices. Numerical results indicate that HQNN achieves comparable or superior performance and parameter efficiency compared to classical neural networks, underscoring its potential as a viable replacement. This study highlights the potential of hybrid QML in computational drug discovery, offering insights into its applicability and advantages in addressing the computational challenges of protein-ligand binding affinity prediction.", "date": "2025-09-14", "categories": []}
{"id": "2509.11131", "title": "Neural cellular automata: applications to biology and beyond classical AI", "abstract": "Neural Cellular Automata (NCA) represent a powerful framework for modeling biological self-organization, extending classical rule-based systems with trainable, differentiable (or evolvable) update rules that capture the adaptive self-regulatory dynamics of living matter. By embedding Artificial Neural Networks (ANNs) as local decision-making centers and interaction rules between localized agents, NCA can simulate processes across molecular, cellular, tissue, and system-level scales, offering a multiscale competency architecture perspective on evolution, development, regeneration, aging, morphogenesis, and robotic control. These models not only reproduce biologically inspired target patterns but also generalize to novel conditions, demonstrating robustness to perturbations and the capacity for open-ended adaptation and reasoning. Given their immense success in recent developments, we here review current literature of NCAs that are relevant primarily for biological or bioengineering applications. Moreover, we emphasize that beyond biology, NCAs display robust and generalizing goal-directed dynamics without centralized control, e.g., in controlling or regenerating composite robotic morphologies or even on cutting-edge reasoning tasks such as ARC-AGI-1. In addition, the same principles of iterative state-refinement is reminiscent to modern generative Artificial Intelligence (AI), such as probabilistic diffusion models. Their governing self-regulatory behavior is constraint to fully localized interactions, yet their collective behavior scales into coordinated system-level outcomes. We thus argue that NCAs constitute a unifying computationally lean paradigm that not only bridges fundamental insights from multiscale biology with modern generative AI, but have the potential to design truly bio-inspired collective intelligence capable of hierarchical reasoning and control.", "date": "2025-09-14", "categories": []}
{"id": "2509.11345", "title": "BiLSTM-VHP: BiLSTM-Powered Network for Viral Host Prediction", "abstract": "Recorded history shows the long coexistence of humans and animals, suggesting it began much earlier. Despite some beneficial interdependence, many animals carry viral diseases that can spread to humans. These diseases are known as zoonotic diseases. Recent outbreaks of SARS-CoV-2, Monkeypox and swine flu viruses have shown how these viruses can disrupt human life and cause death. Fast and accurate predictions of the host from which the virus spreads can help prevent these diseases from spreading. This work presents BiLSTM-VHP, a lightweight bidirectional long short-term memory (LSTM)-based architecture that can predict the host from the nucleotide sequence of orthohantavirus, rabies lyssavirus, and rotavirus A with high accuracy. The proposed model works with nucleotide sequences of 400 bases in length and achieved a prediction accuracy of 89.62% for orthohantavirus, 96.58% for rotavirus A, and 77.22% for rabies lyssavirus outperforming previous studies. Moreover, performance of the model is assessed using the confusion matrix, F-1 score, precision, recall, microaverage AUC. In addition, we introduce three curated datasets of orthohantavirus, rotavirus A, and rabies lyssavirus containing 8,575, 95,197, and 22,052 nucleotide sequences divided into 9, 12, and 29 host classes, respectively. The codes and dataset are available at https://doi.org/10.17605/OSF.IO/ANFKR", "date": "2025-09-14", "categories": []}
{"id": "2509.11782", "title": "Multimodal Regression for Enzyme Turnover Rates Prediction", "abstract": "The enzyme turnover rate is a fundamental parameter in enzyme kinetics, reflecting the catalytic efficiency of enzymes. However, enzyme turnover rates remain scarce across most organisms due to the high cost and complexity of experimental measurements. To address this gap, we propose a multimodal framework for predicting the enzyme turnover rate by integrating enzyme sequences, substrate structures, and environmental factors. Our model combines a pre-trained language model and a convolutional neural network to extract features from protein sequences, while a graph neural network captures informative representations from substrate molecules. An attention mechanism is incorporated to enhance interactions between enzyme and substrate representations. Furthermore, we leverage symbolic regression via Kolmogorov-Arnold Networks to explicitly learn mathematical formulas that govern the enzyme turnover rate, enabling interpretable and accurate predictions. Extensive experiments demonstrate that our framework outperforms both traditional and state-of-the-art deep learning approaches. This work provides a robust tool for studying enzyme kinetics and holds promise for applications in enzyme engineering, biotechnology, and industrial biocatalysis.", "date": "2025-09-15", "categories": []}
{"id": "2502.16446", "title": "Auxiliary Discrminator Sequence Generative Adversarial Networks (ADSeqGAN) for Few Sample Molecule Generation", "abstract": "In this work, we introduce Auxiliary Discriminator Sequence Generative Adversarial Networks (ADSeqGAN), a novel approach for molecular generation in small-sample datasets. Traditional generative models often struggle with limited training data, particularly in drug discovery, where molecular datasets for specific therapeutic targets, such as nucleic acids binders and central nervous system (CNS) drugs, are scarce. ADSeqGAN addresses this challenge by integrating an auxiliary random forest classifier as an additional discriminator into the GAN framework, significantly improves molecular generation quality and class specificity. Our method incorporates pretrained generator and Wasserstein distance to enhance training stability and diversity. We evaluate ADSeqGAN across three representative cases. First, on nucleic acid- and protein-targeting molecules, ADSeqGAN shows superior capability in generating nucleic acid binders compared to baseline models. Second, through oversampling, it markedly improves CNS drug generation, achieving higher yields than traditional de novo models. Third, in cannabinoid receptor type 1 (CB1) ligand design, ADSeqGAN generates novel druglike molecules, with 32.8\\% predicted actives surpassing hit rates of CB1-focused and general-purpose libraries when assessed by a target-specific LRIP-SF scoring function. Overall, ADSeqGAN offers a versatile framework for molecular design in data-scarce scenarios, with demonstrated applications in nucleic acid binders, CNS drugs, and CB1 ligands.", "date": "2025-02-23", "categories": []}
{"id": "2509.07983", "title": "Steering Protein Language Models", "abstract": "Protein Language Models (PLMs), pre-trained on extensive evolutionary data from natural proteins, have emerged as indispensable tools for protein design. While powerful, PLMs often struggle to produce proteins with precisely specified functionalities or properties due to inherent challenges in controlling their outputs. In this work, we investigate the potential of Activation Steering, a technique originally developed for controlling text generation in Large Language Models (LLMs), to direct PLMs toward generating protein sequences with targeted properties. We propose a simple yet effective method that employs activation editing to steer PLM outputs, and extend this approach to protein optimization through a novel editing site identification module. Through comprehensive experiments on lysozyme-like sequence generation and optimization, we demonstrate that our methods can be seamlessly integrated into both auto-encoding and autoregressive PLMs without requiring additional training. These results highlight a promising direction for precise protein engineering using foundation models.", "date": "2025-07-01", "categories": []}
{"id": "2501.04718", "title": "Knowledge-Guided Biomarker Identification for Label-Free Single-Cell RNA-Seq Data: A Reinforcement Learning Perspective", "abstract": "Gene panel selection aims to identify the most informative genomic biomarkers in label-free genomic datasets. Traditional approaches, which rely on domain expertise, embedded machine learning models, or heuristic-based iterative optimization, often introduce biases and inefficiencies, potentially obscuring critical biological signals. To address these challenges, we present an iterative gene panel selection strategy that harnesses ensemble knowledge from existing gene selection algorithms to establish preliminary boundaries or prior knowledge, which guide the initial search space. Subsequently, we incorporate reinforcement learning through a reward function shaped by expert behavior, enabling dynamic refinement and targeted selection of gene panels. This integration mitigates biases stemming from initial boundaries while capitalizing on RL's stochastic adaptability. Comprehensive comparative experiments, case studies, and downstream analyses demonstrate the effectiveness of our method, highlighting its improved precision and efficiency for label-free biomarker discovery. Our results underscore the potential of this approach to advance single-cell genomics data analysis.", "date": "2025-01-02", "categories": []}
{"id": "2509.06465", "title": "CAME-AB: Cross-Modality Attention with Mixture-of-Experts for Antibody Binding Site Prediction", "abstract": "Antibody binding site prediction plays a pivotal role in computational immunology and therapeutic antibody design. Existing sequence or structure methods rely on single-view features and fail to identify antibody-specific binding sites on the antigens. In this paper, we propose \\textbf{CAME-AB}, a novel Cross-modality Attention framework with a Mixture-of-Experts (MoE) backbone for robust antibody binding site prediction. CAME-AB integrates five biologically grounded modalities, including raw amino acid encodings, BLOSUM substitution profiles, pretrained language model embeddings, structure-aware features, and GCN-refined biochemical graphs, into a unified multimodal representation. To enhance adaptive cross-modal reasoning, we propose an \\emph{adaptive modality fusion} module that learns to dynamically weight each modality based on its global relevance and input-specific contribution. A Transformer encoder combined with an MoE module further promotes feature specialization and capacity expansion. We additionally incorporate a supervised contrastive learning objective to explicitly shape the latent space geometry, encouraging intra-class compactness and inter-class separability. To improve optimization stability and generalization, we apply stochastic weight averaging during training. Extensive experiments on benchmark antibody-antigen datasets demonstrate that CAME-AB consistently outperforms strong baselines on multiple metrics, including Precision, Recall, F1-score, AUC-ROC, and MCC. Ablation studies further validate the effectiveness of each architectural component and the benefit of multimodal feature integration. The model implementation details and the codes are available on https://anonymous.4open.science/r/CAME-AB-C525", "date": "2025-09-08", "categories": []}
{"id": "2411.10548", "title": "BioNeMo Framework: a modular, high-performance library for AI model development in drug discovery", "abstract": "Artificial Intelligence models encoding biology and chemistry are opening new routes to high-throughput and high-quality in-silico drug development. However, their training increasingly relies on computational scale, with recent protein language models (pLM) training on hundreds of graphical processing units (GPUs). We introduce the BioNeMo Framework to facilitate the training of computational biology and chemistry AI models across hundreds of GPUs. Its modular design allows the integration of individual components, such as data loaders, into existing workflows and is open to community contributions. We detail technical features of the BioNeMo Framework through use cases such as pLM pre-training and fine-tuning. On 256 NVIDIA A100s, BioNeMo Framework trains a three billion parameter BERT-based pLM on over one trillion tokens in 4.2 days. The BioNeMo Framework is open-source and free for everyone to use.", "date": "2024-11-15", "categories": []}
{"id": "2509.06986", "title": "CellPainTR: Generalizable Representation Learning for Cross-Dataset Cell Painting Analysis", "abstract": "Large-scale biological discovery requires integrating massive, heterogeneous datasets like those from the JUMP Cell Painting consortium, but technical batch effects and a lack of generalizable models remain critical roadblocks. To address this, we introduce CellPainTR, a Transformer-based architecture designed to learn foundational representations of cellular morphology that are robust to batch effects. Unlike traditional methods that require retraining on new data, CellPainTR's design, featuring source-specific context tokens, allows for effective out-of-distribution (OOD) generalization to entirely unseen datasets without fine-tuning. We validate CellPainTR on the large-scale JUMP dataset, where it outperforms established methods like ComBat and Harmony in both batch integration and biological signal preservation. Critically, we demonstrate its robustness through a challenging OOD task on the unseen Bray et al. dataset, where it maintains high performance despite significant domain and feature shifts. Our work represents a significant step towards creating truly foundational models for image-based profiling, enabling more reliable and scalable cross-study biological analysis.", "date": "2025-09-02", "categories": []}
{"id": "2402.12391", "title": "Toward a Team of AI-made Scientists for Scientific Discovery from Gene Expression Data", "abstract": "Machine learning has emerged as a powerful tool for scientific discovery, enabling researchers to extract meaningful insights from complex datasets. For instance, it has facilitated the identification of disease-predictive genes from gene expression data, significantly advancing healthcare. However, the traditional process for analyzing such datasets demands substantial human effort and expertise for the data selection, processing, and analysis. To address this challenge, we introduce a novel framework, a Team of AI-made Scientists (TAIS), designed to streamline the scientific discovery pipeline. TAIS comprises simulated roles, including a project manager, data engineer, and domain expert, each represented by a Large Language Model (LLM). These roles collaborate to replicate the tasks typically performed by data scientists, with a specific focus on identifying disease-predictive genes. Furthermore, we have curated a benchmark dataset to assess TAIS's effectiveness in gene identification, demonstrating our system's potential to significantly enhance the efficiency and scope of scientific exploration. Our findings represent a solid step towards automating scientific discovery through large language models.", "date": "2024-02-15", "categories": []}
{"id": "2403.03726", "title": "Diffusion on language model encodings for protein sequence generation", "abstract": "Protein sequence design has seen significant advances through discrete diffusion and autoregressive approaches, yet the potential of continuous diffusion remains underexplored. Here, we present DiMA, a latent diffusion framework that operates on protein language model representations. Through systematic exploration of architectural choices and diffusion components, we develop a robust methodology that generalizes across multiple protein encoders ranging from 8M to 3B parameters. We demonstrate that our framework achieves consistently high performance across sequence-only (ESM-2, ESMc), dual-decodable (CHEAP), and multimodal (SaProt) representations using the same architecture and training approach. We extensively evaluate existing methods alongside DiMA using multiple metrics across two protein modalities, covering quality, diversity, novelty, and distribution matching of generated proteins. DiMA consistently produces novel, high-quality and diverse protein sequences and achieves strong results compared to baselines such as autoregressive, discrete diffusion and flow matching language models. The model demonstrates versatile functionality, supporting conditional generation tasks including protein family-generation, motif scaffolding and infilling, and fold-specific sequence design. This work provides a universal continuous diffusion framework for protein sequence generation, offering both architectural insights and practical applicability across various protein design scenarios. Code is released at \\href{https://github.com/MeshchaninovViacheslav/DiMA}{GitHub}.", "date": "2024-03-06", "categories": []}
{"id": "2509.05302", "title": "Sesame: Opening the door to protein pockets", "abstract": "Molecular docking is a cornerstone of drug discovery, relying on high-resolution ligand-bound structures to achieve accurate predictions. However, obtaining these structures is often costly and time-intensive, limiting their availability. In contrast, ligand-free structures are more accessible but suffer from reduced docking performance due to pocket geometries being less suited for ligand accommodation in apo structures. Traditional methods for artificially inducing these conformations, such as molecular dynamics simulations, are computationally expensive. In this work, we introduce Sesame, a generative model designed to predict this conformational change efficiently. By generating geometries better suited for ligand accommodation at a fraction of the computational cost, Sesame aims to provide a scalable solution for improving virtual screening workflows.", "date": "2025-08-21", "categories": []}
{"id": "2509.05388", "title": "Augmented Structure Preserving Neural Networks for cell biomechanics", "abstract": "Cell biomechanics involve a great number of complex phenomena that are fundamental to the evolution of life itself and other associated processes, ranging from the very early stages of embryo-genesis to the maintenance of damaged structures or the growth of tumors. Given the importance of such phenomena, increasing research has been dedicated to their understanding, but the many interactions between them and their influence on the decisions of cells as a collective network or cluster remain unclear. We present a new approach that combines Structure Preserving Neural Networks, which study cell movements as a purely mechanical system, with other Machine Learning tools (Artificial Neural Networks), which allow taking into consideration environmental factors that can be directly deduced from an experiment with Computer Vision techniques. This new model, tested on simulated and real cell migration cases, predicts complete cell trajectories following a roll-out policy with a high level of accuracy. This work also includes a mitosis event prediction model based on Neural Networks architectures which makes use of the same observed features.", "date": "2025-09-05", "categories": []}
{"id": "2503.00648", "title": "T-cell receptor specificity landscape revealed through de novo peptide design", "abstract": "T-cells play a key role in adaptive immunity by mounting specific responses against diverse pathogens. An effective binding between T-cell receptors (TCRs) and pathogen-derived peptides presented on Major Histocompatibility Complexes (MHCs) mediate an immune response. However, predicting these interactions remains challenging due to limited functional data on T-cell reactivities. Here, we introduce a computational approach to predict TCR interactions with peptides presented on MHC class I alleles, and to design novel immunogenic peptides for specified TCR-MHC complexes. Our method leverages HERMES, a structure-based, physics-guided machine learning model trained on the protein universe to predict amino acid preferences based on local structural environments. Despite no direct training on TCR-pMHC data, the implicit physical reasoning in HERMES enables us to make accurate predictions of both TCR-pMHC binding affinities and T-cell activities across diverse viral epitopes and cancer neoantigens, achieving up to 0.72 correlation with experimental data. Leveraging our TCR recognition model, we develop a computational protocol for de novo design of immunogenic peptides. Through experimental validation in three TCR-MHC systems targeting viral and cancer peptides, we demonstrate that our designs -- with up to five substitutions from the native sequence -- activate T-cells at success rates of up to 50%. Lastly, we use our generative framework to quantify the diversity of the peptide recognition landscape for various TCR-MHC complexes, offering key insights into T-cell specificity in both humans and mice. Our approach provides a platform for immunogenic peptide and neoantigen design, as well as for evaluating TCR specificity, offering a computational framework to inform design of engineered T-cell therapies and vaccines.", "date": "2025-03-01", "categories": []}
{"id": "2505.02980", "title": "Completing Spatial Transcriptomics Data for Gene Expression Prediction Benchmarking", "abstract": "Spatial Transcriptomics is a groundbreaking technology that integrates histology images with spatially resolved gene expression profiles. Among the various Spatial Transcriptomics techniques available, Visium has emerged as the most widely adopted. However, its accessibility is limited by high costs, the need for specialized expertise, and slow clinical integration. Additionally, gene capture inefficiencies lead to significant dropout, corrupting acquired data. To address these challenges, the deep learning community has explored the gene expression prediction task directly from histology images. Yet, inconsistencies in datasets, preprocessing, and training protocols hinder fair comparisons between models. To bridge this gap, we introduce SpaRED, a systematically curated database comprising 26 public datasets, providing a standardized resource for model evaluation. We further propose SpaCKLE, a state-of-the-art transformer-based gene expression completion model that reduces mean squared error by over 82.5% compared to existing approaches. Finally, we establish the SpaRED benchmark, evaluating eight state-of-the-art prediction models on both raw and SpaCKLE-completed data, demonstrating SpaCKLE substantially improves the results across all the gene expression prediction models. Altogether, our contributions constitute the most comprehensive benchmark of gene expression prediction from histology images to date and a stepping stone for future research on Spatial Transcriptomics.", "date": "2025-05-05", "categories": []}
{"id": "2509.03551", "title": "Predicting Antimicrobial Resistance (AMR) in Campylobacter, a Foodborne Pathogen, and Cost Burden Analysis Using Machine Learning", "abstract": "Antimicrobial resistance (AMR) poses a significant public health and economic challenge, increasing treatment costs and reducing antibiotic effectiveness. This study employs machine learning to analyze genomic and epidemiological data from the public databases for molecular typing and microbial genome diversity (PubMLST), incorporating data from UK government-supported AMR surveillance by the Food Standards Agency and Food Standards Scotland. We identify AMR patterns in Campylobacter jejuni and Campylobacter coli isolates collected in the UK from 2001 to 2017. The research integrates whole-genome sequencing (WGS) data, epidemiological metadata, and economic projections to identify key resistance determinants and forecast future resistance trends and healthcare costs. We investigate gyrA mutations for fluoroquinolone resistance and the tet(O) gene for tetracycline resistance, training a Random Forest model validated with bootstrap resampling (1,000 samples, 95% confidence intervals), achieving 74% accuracy in predicting AMR phenotypes. Time-series forecasting models (SARIMA, SIR, and Prophet) predict a rise in campylobacteriosis cases, potentially exceeding 130 cases per 100,000 people by 2050, with an economic burden projected to surpass 1.9 billion GBP annually if left unchecked. An enhanced Random Forest system, analyzing 6,683 isolates, refines predictions by incorporating temporal patterns, uncertainty estimation, and resistance trend modeling, indicating sustained high beta-lactam resistance, increasing fluoroquinolone resistance, and fluctuating tetracycline resistance.", "date": "2025-09-03", "categories": []}
{"id": "2509.03885", "title": "Topotein: Topological Deep Learning for Protein Representation Learning", "abstract": "Protein representation learning (PRL) is crucial for understanding structure-function relationships, yet current sequence- and graph-based methods fail to capture the hierarchical organization inherent in protein structures. We introduce Topotein, a comprehensive framework that applies topological deep learning to PRL through the novel Protein Combinatorial Complex (PCC) and Topology-Complete Perceptron Network (TCPNet). Our PCC represents proteins at multiple hierarchical levels -- from residues to secondary structures to complete proteins -- while preserving geometric information at each level. TCPNet employs SE(3)-equivariant message passing across these hierarchical structures, enabling more effective capture of multi-scale structural patterns. Through extensive experiments on four PRL tasks, TCPNet consistently outperforms state-of-the-art geometric graph neural networks. Our approach demonstrates particular strength in tasks such as fold classification which require understanding of secondary structure arrangements, validating the importance of hierarchical topological features for protein analysis.", "date": "2025-09-04", "categories": []}
{"id": "2509.02610", "title": "Resilient Biosecurity in the Era of AI-Enabled Bioweapons", "abstract": "Recent advances in generative biology have enabled the design of novel proteins, creating significant opportunities for drug discovery while also introducing new risks, including the potential development of synthetic bioweapons. Existing biosafety measures primarily rely on inference-time filters such as sequence alignment and protein-protein interaction (PPI) prediction to detect dangerous outputs. In this study, we evaluate the performance of three leading PPI prediction tools: AlphaFold 3, AF3Complex, and SpatialPPIv2. These models were tested on well-characterized viral-host interactions, such as those involving Hepatitis B and SARS-CoV-2. Despite being trained on many of the same viruses, the models fail to detect a substantial number of known interactions. Strikingly, none of the tools successfully identify any of the four experimentally validated SARS-CoV-2 mutants with confirmed binding. These findings suggest that current predictive filters are inadequate for reliably flagging even known biological threats and are even more unlikely to detect novel ones. We argue for a shift toward response-oriented infrastructure, including rapid experimental validation, adaptable biomanufacturing, and regulatory frameworks capable of operating at the speed of AI-driven developments.", "date": "2025-08-30", "categories": []}
{"id": "2509.02639", "title": "Enhanced Single-Cell RNA-seq Embedding through Gene Expression and Data-Driven Gene-Gene Interaction Integration", "abstract": "Single-cell RNA sequencing (scRNA-seq) provides unprecedented insights into cellular heterogeneity, enabling detailed analysis of complex biological systems at single-cell resolution. However, the high dimensionality and technical noise inherent in scRNA-seq data pose significant analytical challenges. While current embedding methods focus primarily on gene expression levels, they often overlook crucial gene-gene interactions that govern cellular identity and function. To address this limitation, we present a novel embedding approach that integrates both gene expression profiles and data-driven gene-gene interactions. Our method first constructs a Cell-Leaf Graph (CLG) using random forest models to capture regulatory relationships between genes, while simultaneously building a K-Nearest Neighbor Graph (KNNG) to represent expression similarities between cells. These graphs are then combined into an Enriched Cell-Leaf Graph (ECLG), which serves as input for a graph neural network to compute cell embeddings. By incorporating both expression levels and gene-gene interactions, our approach provides a more comprehensive representation of cellular states. Extensive evaluation across multiple datasets demonstrates that our method enhances the detection of rare cell populations and improves downstream analyses such as visualization, clustering, and trajectory inference. This integrated approach represents a significant advance in single-cell data analysis, offering a more complete framework for understanding cellular diversity and dynamics.", "date": "2025-09-01", "categories": []}
{"id": "2509.02642", "title": "BioMD: All-atom Generative Model for Biomolecular Dynamics Simulation", "abstract": "Molecular dynamics (MD) simulations are essential tools in computational chemistry and drug discovery, offering crucial insights into dynamic molecular behavior. However, their utility is significantly limited by substantial computational costs, which severely restrict accessible timescales for many biologically relevant processes. Despite the encouraging performance of existing machine learning (ML) methods, they struggle to generate extended biomolecular system trajectories, primarily due to the lack of MD datasets and the large computational demands of modeling long historical trajectories. Here, we introduce BioMD, the first all-atom generative model to simulate long-timescale protein-ligand dynamics using a hierarchical framework of forecasting and interpolation. We demonstrate the effectiveness and versatility of BioMD on the DD-13M (ligand unbinding) and MISATO datasets. For both datasets, BioMD generates highly realistic conformations, showing high physical plausibility and low reconstruction errors. Besides, BioMD successfully generates ligand unbinding paths for 97.1% of the protein-ligand systems within ten attempts, demonstrating its ability to explore critical unbinding pathways. Collectively, these results establish BioMD as a tool for simulating complex biomolecular processes, offering broad applicability for computational chemistry and drug discovery.", "date": "2025-09-02", "categories": []}
{"id": "2509.03425", "title": "LINKER: Learning Interactions Between Functional Groups and Residues With Chemical Knowledge-Enhanced Reasoning and Explainability", "abstract": "Accurate identification of interactions between protein residues and ligand functional groups is essential to understand molecular recognition and guide rational drug design. Existing deep learning approaches for protein-ligand interpretability often rely on 3D structural input or use distance-based contact labels, limiting both their applicability and biological relevance. We introduce LINKER, the first sequence-based model to predict residue-functional group interactions in terms of biologically defined interaction types, using only protein sequences and the ligand SMILES as input. LINKER is trained with structure-supervised attention, where interaction labels are derived from 3D protein-ligand complexes via functional group-based motif extraction. By abstracting ligand structures into functional groups, the model focuses on chemically meaningful substructures while predicting interaction types rather than mere spatial proximity. Crucially, LINKER requires only sequence-level input at inference time, enabling large-scale application in settings where structural data is unavailable. Experiments on the LP-PDBBind benchmark demonstrate that structure-informed supervision over functional group abstractions yields interaction predictions closely aligned with ground-truth biochemical annotations.", "date": "2025-09-03", "categories": []}
{"id": "2508.09926", "title": "Towards Comprehensive Cellular Characterisation of H&E slides", "abstract": "Cell detection, segmentation and classification are essential for analyzing tumor microenvironments (TME) on hematoxylin and eosin (H&E) slides. Existing methods suffer from poor performance on understudied cell types (rare or not present in public datasets) and limited cross-domain generalization. To address these shortcomings, we introduce HistoPLUS, a state-of-the-art model for cell analysis, trained on a novel curated pan-cancer dataset of 108,722 nuclei covering 13 cell types. In external validation across 4 independent cohorts, HistoPLUS outperforms current state-of-the-art models in detection quality by 5.2% and overall F1 classification score by 23.7%, while using 5x fewer parameters. Notably, HistoPLUS unlocks the study of 7 understudied cell types and brings significant improvements on 8 of 13 cell types. Moreover, we show that HistoPLUS robustly transfers to two oncology indications unseen during training. To support broader TME biomarker research, we release the model weights and inference code at https://github.com/owkin/histoplus/.", "date": "2025-08-13", "categories": []}
{"id": "2508.21484", "title": "Data-driven Discovery of Digital Twins in Biomedical Research", "abstract": "Recent technological advances have expanded the availability of high-throughput biological datasets, enabling the reliable design of digital twins of biomedical systems or patients. Such computational tools represent key reaction networks driving perturbation or drug response and can guide drug discovery and personalized therapeutics. Yet, their development still relies on laborious data integration by the human modeler, so that automated approaches are critically needed. The success of data-driven system discovery in Physics, rooted in clean datasets and well-defined governing laws, has fueled interest in applying similar techniques in Biology, which presents unique challenges. Here, we reviewed methodologies for automatically inferring digital twins from biological time series, which mostly involve symbolic or sparse regression. We evaluate algorithms according to eight biological and methodological challenges, associated to noisy/incomplete data, multiple conditions, prior knowledge integration, latent variables, high dimensionality, unobserved variable derivatives, candidate library design, and uncertainty quantification. Upon these criteria, sparse regression generally outperformed symbolic regression, particularly when using Bayesian frameworks. We further highlight the emerging role of deep learning and large language models, which enable innovative prior knowledge integration, though the reliability and consistency of such approaches must be improved. While no single method addresses all challenges, we argue that progress in learning digital twins will come from hybrid and modular frameworks combining chemical reaction network-based mechanistic grounding, Bayesian uncertainty quantification, and the generative and knowledge integration capacities of deep learning. To support their development, we further propose a benchmarking framework to evaluate methods across all challenges.", "date": "2025-08-29", "categories": []}
{"id": "2509.00123", "title": "Friend or Foe", "abstract": "A fundamental challenge in microbial ecology is determining whether bacteria compete or cooperate in different environmental conditions. With recent advances in genome-scale metabolic models, we are now capable of simulating interactions between thousands of pairs of bacteria in thousands of different environmental settings at a scale infeasible experimentally. These approaches can generate tremendous amounts of data that can be exploited by state-of-the-art machine learning algorithms to uncover the mechanisms driving interactions. Here, we present Friend or Foe, a compendium of 64 tabular environmental datasets, consisting of more than 26M shared environments for more than 10K pairs of bacteria sampled from two of the largest collections of metabolic models. The Friend or Foe datasets are curated for a wide range of machine learning tasks -- supervised, unsupervised, and generative -- to address specific questions underlying bacterial interactions. We benchmarked a selection of the most recent models for each of these tasks and our results indicate that machine learning can be successful in this application to microbial ecology. Going beyond, analyses of the Friend or Foe compendium can shed light on the predictability of bacterial interactions and highlight novel research directions into how bacteria infer and navigate their relationships.", "date": "2025-08-29", "categories": []}
{"id": "2509.00524", "title": "Biological Pathway Informed Models with Graph Attention Networks (GATs)", "abstract": "Biological pathways map gene-gene interactions that govern all human processes. Despite their importance, most ML models treat genes as unstructured tokens, discarding known pathway structure. The latest pathway-informed models capture pathway-pathway interactions, but still treat each pathway as a \"bag of genes\" via MLPs, discarding its topology and gene-gene interactions. We propose a Graph Attention Network (GAT) framework that models pathways at the gene level. We show that GATs generalize much better than MLPs, achieving an 81% reduction in MSE when predicting pathway dynamics under unseen treatment conditions. We further validate the correctness of our biological prior by encoding drug mechanisms via edge interventions, boosting model robustness. Finally, we show that our GAT model is able to correctly rediscover all five gene-gene interactions in the canonical TP53-MDM2-MDM4 feedback loop from raw time-series mRNA data, demonstrating potential to generate novel biological hypotheses directly from experimental data.", "date": "2025-08-30", "categories": []}
{"id": "2509.00684", "title": "Valid Property-Enhanced Contrastive Learning for Targeted Optimization & Resampling for Novel Drug Design", "abstract": "Efficiently steering generative models toward pharmacologically relevant regions of chemical space remains a major obstacle in molecular drug discovery under low-data regimes. We present VECTOR+: Valid-property-Enhanced Contrastive Learning for Targeted Optimization and Resampling, a framework that couples property-guided representation learning with controllable molecule generation. VECTOR+ applies to both regression and classification tasks and enables interpretable, data-efficient exploration of functional chemical space. We evaluate on two datasets: a curated PD-L1 inhibitor set (296 compounds with experimental $IC_{50}$ values) and a receptor kinase inhibitor set (2,056 molecules by binding mode). Despite limited training data, VECTOR+ generates novel, synthetically tractable candidates. Against PD-L1 (PDB 5J89), 100 of 8,374 generated molecules surpass a docking threshold of $-15.0$ kcal/mol, with the best scoring $-17.6$ kcal/mol compared to the top reference inhibitor ($-15.4$ kcal/mol). The best-performing molecules retain the conserved biphenyl pharmacophore while introducing novel motifs. Molecular dynamics (250 ns) confirm binding stability (ligand RMSD < $2.5$ angstroms). VECTOR+ generalizes to kinase inhibitors, producing compounds with stronger docking scores than established drugs such as brigatinib and sorafenib. Benchmarking against JT-VAE and MolGPT across docking, novelty, uniqueness, and Tanimoto similarity highlights the superior performance of our method. These results position our work as a robust, extensible approach for property-conditioned molecular design in low-data settings, bridging contrastive learning and generative modeling for reproducible, AI-accelerated discovery.", "date": "2025-08-31", "categories": []}
{"id": "2509.01526", "title": "Prediction, Generation of WWTPs microbiome community structures and Clustering of WWTPs various feature attributes using DE-BP model, SiTime-GAN model and DPNG-EPMC ensemble clustering algorithm with modulation of microbial ecosystem health", "abstract": "Microbiomes not only underpin Earth's biogeochemical cycles but also play crucial roles in both engineered and natural ecosystems, such as the soil, wastewater treatment, and the human gut. However, microbiome engineering faces significant obstacles to surmount to deliver the desired improvements in microbiome control. Here, we use the backpropagation neural network (BPNN), optimized through differential evolution (DE-BP), to predict the microbial composition of activated sludge (AS) systems collected from wastewater treatment plants (WWTPs) located worldwide. Furthermore, we introduce a novel clustering algorithm termed Directional Position Nonlinear Emotional Preference Migration Behavior Clustering (DPNG-EPMC). This method is applied to conduct a clustering analysis of WWTPs across various feature attributes. Finally, we employ the Similar Time Generative Adversarial Networks (SiTime-GAN), to synthesize novel microbial compositions and feature attributes data. As a result, we demonstrate that the DE-BP model can provide superior predictions of the microbial composition. Additionally, we show that the DPNG-EPMC can be applied to the analysis of WWTPs under various feature attributes. Finally, we demonstrate that the SiTime-GAN model can generate valuable incremental synthetic data. Our results, obtained through predicting the microbial community and conducting analysis of WWTPs under various feature attributes, develop an understanding of the factors influencing AS communities.", "date": "2025-09-01", "categories": []}
{"id": "2509.02069", "title": "Data-Dependent Smoothing for Protein Discovery with Walk-Jump Sampling", "abstract": "Diffusion models have emerged as a powerful class of generative models by learning to iteratively reverse the noising process. Their ability to generate high-quality samples has extended beyond high-dimensional image data to other complex domains such as proteins, where data distributions are typically sparse and unevenly spread. Importantly, the sparsity itself is uneven. Empirically, we observed that while a small fraction of samples lie in dense clusters, the majority occupy regions of varying sparsity across the data space. Existing approaches largely ignore this data-dependent variability. In this work, we introduce a Data-Dependent Smoothing Walk-Jump framework that employs kernel density estimation (KDE) as a preprocessing step to estimate the noise scale $\\sigma$ for each data point, followed by training a score model with these data-dependent $\\sigma$ values. By incorporating local data geometry into the denoising process, our method accounts for the heterogeneous distribution of protein data. Empirical evaluations demonstrate that our approach yields consistent improvements across multiple metrics, highlighting the importance of data-aware sigma prediction for generative modeling in sparse, high-dimensional settings.", "date": "2025-09-02", "categories": []}
{"id": "2508.21076", "title": "Pep2Prob Benchmark: Predicting Fragment Ion Probability for MS$^2$-based Proteomics", "abstract": "Proteins perform nearly all cellular functions and constitute most drug targets, making their analysis fundamental to understanding human biology in health and disease. Tandem mass spectrometry (MS$^2$) is the major analytical technique in proteomics that identifies peptides by ionizing them, fragmenting them, and using the resulting mass spectra to identify and quantify proteins in biological samples. In MS$^2$ analysis, peptide fragment ion probability prediction plays a critical role, enhancing the accuracy of peptide identification from mass spectra as a complement to the intensity information. Current approaches rely on global statistics of fragmentation, which assumes that a fragment's probability is uniform across all peptides. Nevertheless, this assumption is oversimplified from a biochemical principle point of view and limits accurate prediction. To address this gap, we present Pep2Prob, the first comprehensive dataset and benchmark designed for peptide-specific fragment ion probability prediction. The proposed dataset contains fragment ion probability statistics for 608,780 unique precursors (each precursor is a pair of peptide sequence and charge state), summarized from more than 183 million high-quality, high-resolution, HCD MS$^2$ spectra with validated peptide assignments and fragmentation annotations. We establish baseline performance using simple statistical rules and learning-based methods, and find that models leveraging peptide-specific information significantly outperform previous methods using only global fragmentation statistics. Furthermore, performance across benchmark models with increasing capacities suggests that the peptide-fragmentation relationship exhibits complex nonlinearities requiring sophisticated machine learning approaches.", "date": "2025-08-12", "categories": []}
{"id": "2508.21082", "title": "ImmunoAI: Accelerated Antibody Discovery Using Gradient-Boosted Machine Learning with Thermodynamic-Hydrodynamic Descriptors and 3D Geometric Interface Topology", "abstract": "Human metapneumovirus (hMPV) poses serious risks to pediatric, elderly, and immunocompromised populations. Traditional antibody discovery pipelines require 10-12 months, limiting their applicability for rapid outbreak response. This project introduces ImmunoAI, a machine learning framework that accelerates antibody discovery by predicting high-affinity candidates using gradient-boosted models trained on thermodynamic, hydrodynamic, and 3D topological interface descriptors. A dataset of 213 antibody-antigen complexes was curated to extract geometric and physicochemical features, and a LightGBM regressor was trained to predict binding affinity with high precision. The model reduced the antibody candidate search space by 89%, and fine-tuning on 117 SARS-CoV-2 binding pairs further reduced Root Mean Square Error (RMSE) from 1.70 to 0.92. In the absence of an experimental structure for the hMPV A2.2 variant, AlphaFold2 was used to predict its 3D structure. The fine-tuned model identified two optimal antibodies with predicted picomolar affinities targeting key mutation sites (G42V and E96K), making them excellent candidates for experimental testing. In summary, ImmunoAI shortens design cycles and enables faster, structure-informed responses to viral outbreaks.", "date": "2025-08-25", "categories": []}
{"id": "2508.21468", "title": "Controllable 3D Molecular Generation for Structure-Based Drug Design Through Bayesian Flow Networks and Gradient Integration", "abstract": "Recent advances in Structure-based Drug Design (SBDD) have leveraged generative models for 3D molecular generation, predominantly evaluating model performance by binding affinity to target proteins. However, practical drug discovery necessitates high binding affinity along with synthetic feasibility and selectivity, critical properties that were largely neglected in previous evaluations. To address this gap, we identify fundamental limitations of conventional diffusion-based generative models in effectively guiding molecule generation toward these diverse pharmacological properties. We propose CByG, a novel framework extending Bayesian Flow Network into a gradient-based conditional generative model that robustly integrates property-specific guidance. Additionally, we introduce a comprehensive evaluation scheme incorporating practical benchmarks for binding affinity, synthetic feasibility, and selectivity, overcoming the limitations of conventional evaluation methods. Extensive experiments demonstrate that our proposed CByG framework significantly outperforms baseline models across multiple essential evaluation criteria, highlighting its effectiveness and practicality for real-world drug discovery applications.", "date": "2025-08-29", "categories": []}
{"id": "2505.02033", "title": "Quantum-Enhanced Classification of Brain Tumors Using DNA Microarray Gene Expression Profiles", "abstract": "DNA microarray technology enables the simultaneous measurement of expression levels of thousands of genes, thereby facilitating the understanding of the molecular mechanisms underlying complex diseases such as brain tumors and the identification of diagnostic genetic signatures. To derive meaningful biological insights from the high-dimensional and complex gene features obtained through this technology and to analyze gene properties in detail, classical AI-based approaches such as machine learning and deep learning are widely employed. However, these methods face various limitations in managing high-dimensional vector spaces and modeling the intricate relationships among genes. In particular, challenges such as hyperparameter tuning, computational costs, and high processing power requirements can hinder their efficiency. To overcome these limitations, quantum computing and quantum AI approaches are gaining increasing attention. Leveraging quantum properties such as superposition and entanglement, quantum methods enable more efficient parallel processing of high-dimensional data and offer faster and more effective solutions to problems that are computationally demanding for classical methods. In this study, a novel model called \"Deep VQC\" is proposed, based on the Variational Quantum Classifier approach. Developed using microarray data containing 54,676 gene features, the model successfully classified four different types of brain tumors-ependymoma, glioblastoma, medulloblastoma, and pilocytic astrocytoma-alongside healthy samples with high accuracy. Furthermore, compared to classical ML algorithms, our model demonstrated either superior or comparable classification performance. These results highlight the potential of quantum AI methods as an effective and promising approach for the analysis and classification of complex structures such as brain tumors based on gene expression features.", "date": "2025-05-04", "categories": []}
{"id": "2508.19424", "title": "MS-ConTab: Multi-Scale Contrastive Learning of Mutation Signatures for Pan Cancer Representation and Stratification", "abstract": "Motivation. Understanding the pan-cancer mutational landscape offers critical insights into the molecular mechanisms underlying tumorigenesis. While patient-level machine learning techniques have been widely employed to identify tumor subtypes, cohort-level clustering, where entire cancer types are grouped based on shared molecular features, has largely relied on classical statistical methods.\n  Results. In this study, we introduce a novel unsupervised contrastive learning framework to cluster 43 cancer types based on coding mutation data derived from the COSMIC database. For each cancer type, we construct two complementary mutation signatures: a gene-level profile capturing nucleotide substitution patterns across the most frequently mutated genes, and a chromosome-level profile representing normalized substitution frequencies across chromosomes. These dual views are encoded using TabNet encoders and optimized via a multi-scale contrastive learning objective (NT-Xent loss) to learn unified cancer-type embeddings. We demonstrate that the resulting latent representations yield biologically meaningful clusters of cancer types, aligning with known mutational processes and tissue origins. Our work represents the first application of contrastive learning to cohort-level cancer clustering, offering a scalable and interpretable framework for mutation-driven cancer subtyping.", "date": "2025-08-26", "categories": []}
{"id": "2508.20130", "title": "Artificial Intelligence for CRISPR Guide RNA Design: Explainable Models and Off-Target Safety", "abstract": "CRISPR-based genome editing has revolutionized biotechnology, yet optimizing guide RNA (gRNA) design for efficiency and safety remains a critical challenge. Recent advances (2020--2025, updated to reflect current year if needed) demonstrate that artificial intelligence (AI), especially deep learning, can markedly improve the prediction of gRNA on-target activity and identify off-target risks. In parallel, emerging explainable AI (XAI) techniques are beginning to illuminate the black-box nature of these models, offering insights into sequence features and genomic contexts that drive Cas enzyme performance. Here we review how state-of-the-art machine learning models are enhancing gRNA design for CRISPR systems, highlight strategies for interpreting model predictions, and discuss new developments in off-target prediction and safety assessment. We emphasize breakthroughs from top-tier journals that underscore an interdisciplinary convergence of AI and genome editing to enable more efficient, specific, and clinically viable CRISPR applications.", "date": "2025-08-26", "categories": []}
{"id": "2508.01490", "title": "A Large-Scale Benchmark of Cross-Modal Learning for Histology and Gene Expression in Spatial Transcriptomics", "abstract": "Spatial transcriptomics enables simultaneous measurement of gene expression and tissue morphology, offering unprecedented insights into cellular organization and disease mechanisms. However, the field lacks comprehensive benchmarks for evaluating multimodal learning methods that leverage both histology images and gene expression data. Here, we present HESCAPE, a large-scale benchmark for cross-modal contrastive pretraining in spatial transcriptomics, built on a curated pan-organ dataset spanning 6 different gene panels and 54 donors. We systematically evaluated state-of-the-art image and gene expression encoders across multiple pretraining strategies and assessed their effectiveness on two downstream tasks: gene mutation classification and gene expression prediction. Our benchmark demonstrates that gene expression encoders are the primary determinant of strong representational alignment, and that gene models pretrained on spatial transcriptomics data outperform both those trained without spatial data and simple baseline approaches. However, downstream task evaluation reveals a striking contradiction: while contrastive pretraining consistently improves gene mutation classification performance, it degrades direct gene expression prediction compared to baseline encoders trained without cross-modal objectives. We identify batch effects as a key factor that interferes with effective cross-modal alignment. Our findings highlight the critical need for batch-robust multimodal learning approaches in spatial transcriptomics. To accelerate progress in this direction, we release HESCAPE, providing standardized datasets, evaluation protocols, and benchmarking tools for the community", "date": "2025-08-02", "categories": []}
{"id": "2505.02766", "title": "Giving Simulated Cells a Voice: Evolving Prompt-to-Intervention Models for Cellular Control", "abstract": "Guiding biological systems toward desired states, such as morphogenetic outcomes, remains a fundamental challenge with far-reaching implications for medicine and synthetic biology. While large language models (LLMs) have enabled natural language as an interface for interpretable control in AI systems, their use as mediators for steering biological or cellular dynamics remains largely unexplored.\n  In this work, we present a functional pipeline that translates natural language prompts into spatial vector fields capable of directing simulated cellular collectives. Our approach combines a large language model with an evolvable neural controller (Prompt-to-Intervention, or P2I), optimized via evolutionary strategies to generate behaviors such as clustering or scattering in a simulated 2D environment.\n  We demonstrate that even with constrained vocabulary and simplified cell models, evolved P2I networks can successfully align cellular dynamics with user-defined goals expressed in plain language. This work offers a complete loop from language input to simulated bioelectric-like intervention to behavioral output, providing a foundation for future systems capable of natural language-driven cellular control.", "date": "2025-05-05", "categories": []}
{"id": "2506.10271", "title": "Evaluating DNA function understanding in genomic language models using evolutionarily implausible sequences", "abstract": "Genomic language models (gLMs) hold promise for generating novel, functional DNA sequences for synthetic biology. However, realizing this potential requires models to go beyond evolutionary plausibility and understand how DNA sequence encodes gene expression and regulation. We introduce a benchmark called Nullsettes, which assesses how well models can predict in silico loss-of-function (LOF) mutations, in synthetic expression cassettes with little evolutionary precedent. Testing 12 state-of-the-art gLMs, we find that most fail to consistently detect these strong LOF mutations. All models show a sharp drop in predictive accuracy as the likelihood assigned to the original (nonmutant) sequence decreases, suggesting that gLMs rely heavily on pattern-matching to their evolutionary prior rather than on any mechanistic understanding of gene expression. Our findings highlight fundamental limitations in how gLMs generalize to engineered, non-natural sequences, and underscore the need for benchmarks and modeling strategies that prioritize functional understanding.", "date": "2025-06-12", "categories": []}
{"id": "2508.18304", "title": "scI2CL: Effectively Integrating Single-cell Multi-omics by Intra- and Inter-omics Contrastive Learning", "abstract": "Single-cell multi-omics data contain huge information of cellular states, and analyzing these data can reveal valuable insights into cellular heterogeneity, diseases, and biological processes. However, as cell differentiation \\& development is a continuous and dynamic process, it remains challenging to computationally model and infer cell interaction patterns based on single-cell multi-omics data. This paper presents scI2CL, a new single-cell multi-omics fusion framework based on intra- and inter-omics contrastive learning, to learn comprehensive and discriminative cellular representations from complementary multi-omics data for various downstream tasks. Extensive experiments of four downstream tasks validate the effectiveness of scI2CL and its superiority over existing peers. Concretely, in cell clustering, scI2CL surpasses eight state-of-the-art methods on four widely-used real-world datasets. In cell subtyping, scI2CL effectively distinguishes three latent monocyte cell subpopulations, which are not discovered by existing methods. Simultaneously, scI2CL is the only method that correctly constructs the cell developmental trajectory from hematopoietic stem and progenitor cells to Memory B cells. In addition, scI2CL resolves the misclassification of cell types between two subpopulations of CD4+ T cells, while existing methods fail to precisely distinguish the mixed cells. In summary, scI2CL can accurately characterize cross-omics relationships among cells, thus effectively fuses multi-omics data and learns discriminative cellular representations to support various downstream analysis tasks.", "date": "2025-08-23", "categories": []}
{"id": "2508.18446", "title": "From Prediction to Simulation: AlphaFold 3 as a Differentiable Framework for Structural Biology", "abstract": "AlphaFold 3 represents a transformative advancement in computational biology, enhancing protein structure prediction through novel multi-scale transformer architectures, biologically informed cross-attention mechanisms, and geometry-aware optimization strategies. These innovations dramatically improve predictive accuracy and generalization across diverse protein families, surpassing previous methods. Crucially, AlphaFold 3 embodies a paradigm shift toward differentiable simulation, bridging traditional static structural modeling with dynamic molecular simulations. By reframing protein folding predictions as a differentiable process, AlphaFold 3 serves as a foundational framework for integrating deep learning with physics-based molecular", "date": "2025-08-25", "categories": []}
{"id": "2508.18567", "title": "Sparse Autoencoders for Low-$N$ Protein Function Prediction and Design", "abstract": "Predicting protein function from amino acid sequence remains a central challenge in data-scarce (low-$N$) regimes, limiting machine learning-guided protein design when only small amounts of assay-labeled sequence-function data are available. Protein language models (pLMs) have advanced the field by providing evolutionary-informed embeddings and sparse autoencoders (SAEs) have enabled decomposition of these embeddings into interpretable latent variables that capture structural and functional features. However, the effectiveness of SAEs for low-$N$ function prediction and protein design has not been systematically studied. Herein, we evaluate SAEs trained on fine-tuned ESM2 embeddings across diverse fitness extrapolation and protein engineering tasks. We show that SAEs, with as few as 24 sequences, consistently outperform or compete with their ESM2 baselines in fitness prediction, indicating that their sparse latent space encodes compact and biologically meaningful representations that generalize more effectively from limited data. Moreover, steering predictive latents exploits biological motifs in pLM representations, yielding top-fitness variants in 83% of cases compared to designing with ESM2 alone.", "date": "2025-08-25", "categories": []}
{"id": "2508.18638", "title": "Biologically Disentangled Multi-Omic Modeling Reveals Mechanistic Insights into Pan-Cancer Immunotherapy Resistance", "abstract": "Immune checkpoint inhibitors (ICIs) have transformed cancer treatment, yet patient responses remain highly variable, and the biological mechanisms underlying resistance are poorly understood. While machine learning models hold promise for predicting responses to ICIs, most existing methods lack interpretability and do not effectively leverage the biological structure inherent to multi-omics data. Here, we introduce the Biologically Disentangled Variational Autoencoder (BDVAE), a deep generative model that integrates transcriptomic and genomic data through modality- and pathway-specific encoders. Unlike existing rigid, pathway-informed models, BDVAE employs a modular encoder architecture combined with variational inference to learn biologically meaningful latent features associated with immune, genomic, and metabolic processes. Applied to a pan-cancer cohort of 366 patients across four cancer types treated with ICIs, BDVAE accurately predicts treatment response (AUC-ROC = 0.94 on unseen test data) and uncovers critical resistance mechanisms, including immune suppression, metabolic shifts, and neuronal signaling. Importantly, BDVAE reveals that resistance spans a continuous biological spectrum rather than strictly binary states, reflecting gradations of tumor dysfunction. Several latent features correlate with survival outcomes and known clinical subtypes, demonstrating BDVAE's capability to generate interpretable, clinically relevant insights. These findings underscore the value of biologically structured machine learning in elucidating complex resistance patterns and guiding precision immunotherapy strategies.", "date": "2025-08-26", "categories": []}
{"id": "2504.00020", "title": "Celler:A Genomic Language Model for Long-Tailed Single-Cell Annotation", "abstract": "Recent breakthroughs in single-cell technology have ushered in unparalleled opportunities to decode the molecular intricacy of intricate biological systems, especially those linked to diseases unique to humans. However, these progressions have also ushered in novel obstacles-specifically, the efficient annotation of extensive, long-tailed single-cell data pertaining to disease conditions. To effectively surmount this challenge, we introduce Celler, a state-of-the-art generative pre-training model crafted specifically for the annotation of single-cell data. Celler incorporates two groundbreaking elements: First, we introduced the Gaussian Inflation (GInf) Loss function. By dynamically adjusting sample weights, GInf Loss significantly enhances the model's ability to learn from rare categories while reducing the risk of overfitting for common categories. Secondly, we introduce an innovative Hard Data Mining (HDM) strategy into the training process, specifically targeting the challenging-to-learn minority data samples, which significantly improved the model's predictive accuracy. Additionally, to further advance research in this field, we have constructed a large-scale single-cell dataset: Celler-75, which encompasses 40 million cells distributed across 80 human tissues and 75 specific diseases. This dataset provides critical support for comprehensively exploring the potential of single-cell technology in disease research. Our code is available at https://github.com/AI4science-ym/HiCeller.", "date": "2025-03-28", "categories": []}
{"id": "2508.16587", "title": "HemePLM-Diffuse: A Scalable Generative Framework for Protein-Ligand Dynamics in Large Biomolecular System", "abstract": "Comprehending the long-timescale dynamics of protein-ligand complexes is very important for drug discovery and structural biology, but it continues to be computationally challenging for large biomolecular systems. We introduce HemePLM-Diffuse, an innovative generative transformer model that is designed for accurate simulation of protein-ligand trajectories, inpaints the missing ligand fragments, and sample transition paths in systems with more than 10,000 atoms. HemePLM-Diffuse has features of SE(3)-Invariant tokenization approach for proteins and ligands, that utilizes time-aware cross-attentional diffusion to effectively capture atomic motion. We also demonstrate its capabilities using the 3CQV HEME system, showing enhanced accuracy and scalability compared to leading models such as TorchMD-Net, MDGEN, and Uni-Mol.", "date": "2025-08-07", "categories": []}
{"id": "2508.17389", "title": "Neural Proteomics Fields for Super-resolved Spatial Proteomics Prediction", "abstract": "Spatial proteomics maps protein distributions in tissues, providing transformative insights for life sciences. However, current sequencing-based technologies suffer from low spatial resolution, and substantial inter-tissue variability in protein expression further compromises the performance of existing molecular data prediction methods. In this work, we introduce the novel task of spatial super-resolution for sequencing-based spatial proteomics (seq-SP) and, to the best of our knowledge, propose the first deep learning model for this task--Neural Proteomics Fields (NPF). NPF formulates seq-SP as a protein reconstruction problem in continuous space by training a dedicated network for each tissue. The model comprises a Spatial Modeling Module, which learns tissue-specific protein spatial distributions, and a Morphology Modeling Module, which extracts tissue-specific morphological features. Furthermore, to facilitate rigorous evaluation, we establish an open-source benchmark dataset, Pseudo-Visium SP, for this task. Experimental results demonstrate that NPF achieves state-of-the-art performance with fewer learnable parameters, underscoring its potential for advancing spatial proteomics research. Our code and dataset are publicly available at https://github.com/Bokai-Zhao/NPF.", "date": "2025-08-24", "categories": []}
{"id": "2502.15346", "title": "Drug-Target Interaction/Affinity Prediction: Deep Learning Models and Advances Review", "abstract": "Drug discovery remains a slow and expensive process that involves many steps, from detecting the target structure to obtaining approval from the Food and Drug Administration (FDA), and is often riddled with safety concerns. Accurate prediction of how drugs interact with their targets and the development of new drugs by using better methods and technologies have immense potential to speed up this process, ultimately leading to faster delivery of life-saving medications. Traditional methods used for drug-target interaction prediction show limitations, particularly in capturing complex relationships between drugs and their targets. As an outcome, deep learning models have been presented to overcome the challenges of interaction prediction through their precise and efficient end results. By outlining promising research avenues and models, each with a different solution but similar to the problem, this paper aims to give researchers a better idea of methods for even more accurate and efficient prediction of drug-target interaction, ultimately accelerating the development of more effective drugs. A total of 180 prediction methods for drug-target interactions were analyzed throughout the period spanning 2016 to 2025 using different frameworks based on machine learning, mainly deep learning and graph neural networks. Additionally, this paper discusses the novelty, architecture, and input representation of these models.", "date": "2025-02-21", "categories": []}
{"id": "2507.08902", "title": "Generation of structure-guided pMHC-I libraries using Diffusion Models", "abstract": "Personalized vaccines and T-cell immunotherapies depend critically on identifying peptide-MHC class I (pMHC-I) interactions capable of eliciting potent immune responses. However, current benchmarks and models inherit biases present in mass-spectrometry and binding-assay datasets, limiting discovery of novel peptide ligands. To address this issue, we introduce a structure-guided benchmark of pMHC-I peptides designed using diffusion models conditioned on crystal structure interaction distances. Spanning twenty high-priority HLA alleles, this benchmark is independent of previously characterized peptides yet reproduces canonical anchor residue preferences, indicating structural generalization without experimental dataset bias. Using this resource, we demonstrate that state-of-the-art sequence-based predictors perform poorly at recognizing the binding potential of these structurally stable designs, indicating allele-specific limitations invisible in conventional evaluations. Our geometry-aware design pipeline yields peptides with high predicted structural integrity and higher residue diversity than existing datasets, representing a key resource for unbiased model training and evaluation. Our code, and data are available at: https://github.com/sermare/struct-mhc-dev.", "date": "2025-07-11", "categories": []}
{"id": "2508.14924", "title": "A U-Statistic-based random forest approach for genetic interaction study", "abstract": "Variations in complex traits are influenced by multiple genetic variants, environmental risk factors, and their interactions. Though substantial progress has been made in identifying single genetic variants associated with complex traits, detecting the gene-gene and gene-environment interactions remains a great challenge. When a large number of genetic variants and environmental risk factors are involved, searching for interactions is limited to pair-wise interactions due to the exponentially increased feature space and computational intensity. Alternatively, recursive partitioning approaches, such as random forests, have gained popularity in high-dimensional genetic association studies. In this article, we propose a U-Statistic-based random forest approach, referred to as Forest U-Test, for genetic association studies with quantitative traits. Through simulation studies, we showed that the Forest U-Test outperformed existing methods. The proposed method was also applied to study Cannabis Dependence CD, using three independent datasets from the Study of Addiction: Genetics and Environment. A significant joint association was detected with an empirical p-value less than 0.001. The finding was also replicated in two independent datasets with p-values of 5.93e-19 and 4.70e-17, respectively.", "date": "2025-08-19", "categories": []}
{"id": "2508.14934", "title": "AGP: A Novel Arabidopsis thaliana Genomics-Phenomics Dataset and its HyperGraph Baseline Benchmarking", "abstract": "Understanding which genes control which traits in an organism remains one of the central challenges in biology. Despite significant advances in data collection technology, our ability to map genes to traits is still limited. This genome-to-phenome (G2P) challenge spans several problem domains, including plant breeding, and requires models capable of reasoning over high-dimensional, heterogeneous, and biologically structured data. Currently, however, many datasets solely capture genetic information or solely capture phenotype information. Additionally, phenotype data is very heterogeneous, which many datasets do not fully capture. The critical drawback is that these datasets are not integrated, that is, they do not link with each other to describe the same biological specimens. This limits machine learning models' ability to be informed on the various aspects of these specimens, impacting the breadth of correlations learned, and therefore their ability to make more accurate predictions. To address this gap, we present the Arabidopsis Genomics-Phenomics (AGP) Dataset, a curated multi-modal dataset linking gene expression profiles with phenotypic trait measurements in Arabidopsis thaliana, a model organism in plant biology. AGP supports tasks such as phenotype prediction and interpretable graph learning. In addition, we benchmark conventional regression and explanatory baselines, including a biologically-informed hypergraph baseline, to validate gene-trait associations. To the best of our knowledge, this is the first dataset that provides multi-modal gene information and heterogeneous trait or phenotype data for the same Arabidopsis thaliana specimens. With AGP, we aim to foster the research community towards accurately understanding the connection between genotypes and phenotypes using gene information, higher-order gene pairings, and trait data from several sources.", "date": "2025-08-19", "categories": []}
{"id": "2508.15103", "title": "Equi-mRNA: Protein Translation Equivariant Encoding for mRNA Language Models", "abstract": "The growing importance of mRNA therapeutics and synthetic biology highlights the need for models that capture the latent structure of synonymous codon (different triplets encoding the same amino acid) usage, which subtly modulates translation efficiency and gene expression. While recent efforts incorporate codon-level inductive biases through auxiliary objectives, they often fall short of explicitly modeling the structured relationships that arise from the genetic code's inherent symmetries. We introduce Equi-mRNA, the first codon-level equivariant mRNA language model that explicitly encodes synonymous codon symmetries as cyclic subgroups of 2D Special Orthogonal matrix (SO(2)). By combining group-theoretic priors with an auxiliary equivariance loss and symmetry-aware pooling, Equi-mRNA learns biologically grounded representations that outperform vanilla baselines across multiple axes. On downstream property-prediction tasks including expression, stability, and riboswitch switching Equi-mRNA delivers up to approximately 10% improvements in accuracy. In sequence generation, it produces mRNA constructs that are up to approximately 4x more realistic under Frechet BioDistance metrics and approximately 28% better preserve functional properties compared to vanilla baseline. Interpretability analyses further reveal that learned codon-rotation distributions recapitulate known GC-content biases and tRNA abundance patterns, offering novel insights into codon usage. Equi-mRNA establishes a new biologically principled paradigm for mRNA modeling, with significant implications for the design of next-generation therapeutics.", "date": "2025-08-20", "categories": []}
{"id": "2412.06805", "title": "HiCat: A Semi-Supervised Approach for Cell Type Annotation", "abstract": "We introduce HiCat (Hybrid Cell Annotation using Transformative embeddings), a novel semi-supervised pipeline for annotating cell types from single-cell RNA sequencing data. HiCat fuses the strengths of supervised learning for known cell types with unsupervised learning to identify novel types. This hybrid approach incorporates both reference and query genomic data for feature engineering, enhancing the embedding learning process, increasing the effective sample size for unsupervised techniques, and improving the transferability of the supervised model trained on reference data when applied to query datasets. The pipeline follows six key steps: (1) removing batch effects using Harmony to generate a 50-dimensional principal component embedding; (2) applying UMAP for dimensionality reduction to two dimensions to capture crucial data patterns; (3) conducting unsupervised clustering of cells with DBSCAN, yielding a one-dimensional cluster membership vector; (4) merging the multi-resolution results of the previous steps into a 53-dimensional feature space that encompasses both reference and query data; (5) training a CatBoost model on the reference dataset to predict cell types in the query dataset; and (6) resolving inconsistencies between the supervised predictions and unsupervised cluster labels. When benchmarked on 10 publicly available genomic datasets, HiCat surpasses other methods, particularly in differentiating and identifying multiple new cell types. Its capacity to accurately classify novel cell types showcases its robustness and adaptability within intricate biological datasets.", "date": "2024-11-25", "categories": []}
{"id": "2508.14393", "title": "Img2ST-Net: Efficient High-Resolution Spatial Omics Prediction from Whole Slide Histology Images via Fully Convolutional Image-to-Image Learning", "abstract": "Recent advances in multi-modal AI have demonstrated promising potential for generating the currently expensive spatial transcriptomics (ST) data directly from routine histology images, offering a means to reduce the high cost and time-intensive nature of ST data acquisition. However, the increasing resolution of ST, particularly with platforms such as Visium HD achieving 8um or finer, introduces significant computational and modeling challenges. Conventional spot-by-spot sequential regression frameworks become inefficient and unstable at this scale, while the inherent extreme sparsity and low expression levels of high-resolution ST further complicate both prediction and evaluation. To address these limitations, we propose Img2ST-Net, a novel histology-to-ST generation framework for efficient and parallel high-resolution ST prediction. Unlike conventional spot-by-spot inference methods, Img2ST-Net employs a fully convolutional architecture to generate dense, HD gene expression maps in a parallelized manner. By modeling HD ST data as super-pixel representations, the task is reformulated from image-to-omics inference into a super-content image generation problem with hundreds or thousands of output channels. This design not only improves computational efficiency but also better preserves the spatial organization intrinsic to spatial omics data. To enhance robustness under sparse expression patterns, we further introduce SSIM-ST, a structural-similarity-based evaluation metric tailored for high-resolution ST analysis. We present a scalable, biologically coherent framework for high-resolution ST prediction. Img2ST-Net offers a principled solution for efficient and accurate ST inference at scale. Our contributions lay the groundwork for next-generation ST modeling that is robust and resolution-aware. The source code has been made publicly available at https://github.com/hrlblab/Img2ST-Net.", "date": "2025-08-20", "categories": []}
{"id": "2508.14844", "title": "Multimodal Quantum Vision Transformer for Enzyme Commission Classification from Biochemical Representations", "abstract": "Accurately predicting enzyme functionality remains one of the major challenges in computational biology, particularly for enzymes with limited structural annotations or sequence homology. We present a novel multimodal Quantum Machine Learning (QML) framework that enhances Enzyme Commission (EC) classification by integrating four complementary biochemical modalities: protein sequence embeddings, quantum-derived electronic descriptors, molecular graph structures, and 2D molecular image representations. Quantum Vision Transformer (QVT) backbone equipped with modality-specific encoders and a unified cross-attention fusion module. By integrating graph features and spatial patterns, our method captures key stereoelectronic interactions behind enzyme function. Experimental results demonstrate that our multimodal QVT model achieves a top-1 accuracy of 85.1%, outperforming sequence-only baselines by a substantial margin and achieving better performance results compared to other QML models.", "date": "2025-08-20", "categories": []}
{"id": "2407.06295", "title": "Engineering morphogenesis of cell clusters with differentiable programming", "abstract": "Understanding the rules underlying organismal development is a major unsolved problem in biology. Each cell in a developing organism responds to signals in its local environment by dividing, excreting, consuming, or reorganizing, yet how these individual actions coordinate over a macroscopic number of cells to grow complex structures with exquisite functionality is unknown. Here we use recent advances in automatic differentiation to discover local interaction rules and genetic networks that yield emergent, systems-level characteristics in a model of development. We consider a growing tissue with cellular interactions mediated by morphogen diffusion, cell adhesion and mechanical stress. Each cell has an internal genetic network that is used to make decisions based on the cell's local environment. We show that one can learn the parameters governing cell interactions in the form of interpretable genetic networks for complex developmental scenarios, including directed axial elongation, cell type homeostasis via chemical signaling and homogenization of growth via mechanical stress. When combined with recent experimental advances measuring spatio-temporal dynamics and gene expression of cells in a growing tissue, the methodology outlined here offers a promising path to unraveling the cellular bases of development.", "date": "2024-07-08", "categories": []}
{"id": "2508.13208", "title": "Modeling GRNs with a Probabilistic Categorical Framework", "abstract": "Understanding the complex and stochastic nature of Gene Regulatory Networks (GRNs) remains a central challenge in systems biology. Existing modeling paradigms often struggle to effectively capture the intricate, multi-factor regulatory logic and to rigorously manage the dual uncertainties of network structure and kinetic parameters. In response, this work introduces the Probabilistic Categorical GRN(PC-GRN) framework. It is a novel theoretical approach founded on the synergistic integration of three core methodologies. Firstly, category theory provides a formal language for the modularity and composition of regulatory pathways. Secondly, Bayesian Typed Petri Nets (BTPNs) serve as an interpretable,mechanistic substrate for modeling stochastic cellular processes, with kinetic parameters themselves represented as probability distributions. The central innovation of PC-GRN is its end-to-end generative Bayesian inference engine, which learns a full posterior distribution over BTPN models (P (G, {\\Theta}|D)) directly from data. This is achieved by the novel interplay of a GFlowNet, which learns a policy to sample network topologies, and a HyperNetwork, which performs amortized inference to predict their corresponding parameter distributions. The resulting framework provides a mathematically rigorous, biologically interpretable, and uncertainty-aware representation of GRNs, advancing predictive modeling and systems-level analysis.", "date": "2025-08-16", "categories": []}
{"id": "2508.13552", "title": "Collapsing ROC approach for risk prediction research on both common and rare variants", "abstract": "Risk prediction that capitalizes on emerging genetic findings holds great promise for improving public health and clinical care. However, recent risk prediction research has shown that predictive tests formed on existing common genetic loci, including those from genome-wide association studies, have lacked sufficient accuracy for clinical use. Because most rare variants on the genome have not yet been studied for their role in risk prediction, future disease prediction discoveries should shift toward a more comprehensive risk prediction strategy that takes into account both common and rare variants. We are proposing a collapsing receiver operating characteristic CROC approach for risk prediction research on both common and rare variants. The new approach is an extension of a previously developed forward ROC FROC approach, with additional procedures for handling rare variants. The approach was evaluated through the use of 533 single-nucleotide polymorphisms SNPs in 37 candidate genes from the Genetic Analysis Workshop 17 mini-exome data set. We found that a prediction model built on all SNPs gained more accuracy AUC = 0.605 than one built on common variants alone AUC = 0.585. We further evaluated the performance of two approaches by gradually reducing the number of common variants in the analysis. We found that the CROC method attained more accuracy than the FROC method when the number of common variants in the data decreased. In an extreme scenario, when there are only rare variants in the data, the CROC reached an AUC value of 0.603, whereas the FROC had an AUC value of 0.524.", "date": "2025-08-19", "categories": []}
{"id": "2508.02834", "title": "Learning from B Cell Evolution: Adaptive Multi-Expert Diffusion for Antibody Design via Online Optimization", "abstract": "Recent advances in diffusion models have shown remarkable potential for antibody design, yet existing approaches apply uniform generation strategies that cannot adapt to each antigen's unique requirements. Inspired by B cell affinity maturation, where antibodies evolve through multi-objective optimization balancing affinity, stability, and self-avoidance, we propose the first biologically-motivated framework that leverages physics-based domain knowledge within an online meta-learning system. Our method employs multiple specialized experts (van der Waals, molecular recognition, energy balance, and interface geometry) whose parameters evolve during generation based on iterative feedback, mimicking natural antibody refinement cycles. Instead of fixed protocols, this adaptive guidance discovers personalized optimization strategies for each target. Our experiments demonstrate that this approach: (1) discovers optimal SE(3)-equivariant guidance strategies for different antigen classes without pre-training, preserving molecular symmetries throughout optimization; (2) significantly enhances hotspot coverage and interface quality through target-specific adaptation, achieving balanced multi-objective optimization characteristic of therapeutic antibodies; (3) establishes a paradigm for iterative refinement where each antibody-antigen system learns its unique optimization profile through online evaluation; (4) generalizes effectively across diverse design challenges, from small epitopes to large protein interfaces, enabling precision-focused campaigns for individual targets.", "date": "2025-07-25", "categories": []}
{"id": "2508.12617", "title": "A Generalized Genetic Random Field Method for the Genetic Association Analysis of Sequencing Data", "abstract": "With the advance of high-throughput sequencing technologies, it has become feasible to investigate the influence of the entire spectrum of sequencing variations on complex human diseases. Although association studies utilizing the new sequencing technologies hold great promise to unravel novel genetic variants, especially rare genetic variants that contribute to human diseases, the statistical analysis of high-dimensional sequencing data remains a challenge. Advanced analytical methods are in great need to facilitate high-dimensional sequencing data analyses. In this article, we propose a generalized genetic random field (GGRF) method for association analyses of sequencing data. Like other similarity-based methods (e.g., SIMreg and SKAT), the new method has the advantages of avoiding the need to specify thresholds for rare variants and allowing for testing multiple variants acting in different directions and magnitude of effects. The method is built on the generalized estimating equation framework and thus accommodates a variety of disease phenotypes (e.g., quantitative and binary phenotypes). Moreover, it has a nice asymptotic property, and can be applied to small-scale sequencing data without need for small-sample adjustment. Through simulations, we demonstrate that the proposed GGRF attains an improved or comparable power over a commonly used method, SKAT, under various disease scenarios, especially when rare variants play a significant role in disease etiology. We further illustrate GGRF with an application to a real dataset from the Dallas Heart Study. By using GGRF, we were able to detect the association of two candidate genes, ANGPTL3 and ANGPTL4, with serum triglyceride.", "date": "2025-08-18", "categories": []}
{"id": "2502.06173", "title": "Uncertainty-Aware Adaptation of Large Language Models for Protein-Protein Interaction Analysis", "abstract": "Identification of protein-protein interactions (PPIs) helps derive cellular mechanistic understanding, particularly in the context of complex conditions such as neurodegenerative disorders, metabolic syndromes, and cancer. Large Language Models (LLMs) have demonstrated remarkable potential in predicting protein structures and interactions via automated mining of vast biomedical literature; yet their inherent uncertainty remains a key challenge for deriving reproducible findings, critical for biomedical applications. In this study, we present an uncertainty-aware adaptation of LLMs for PPI analysis, leveraging fine-tuned LLaMA-3 and BioMedGPT models. To enhance prediction reliability, we integrate LoRA ensembles and Bayesian LoRA models for uncertainty quantification (UQ), ensuring confidence-calibrated insights into protein behavior. Our approach achieves competitive performance in PPI identification across diverse disease contexts while addressing model uncertainty, thereby enhancing trustworthiness and reproducibility in computational biology. These findings underscore the potential of uncertainty-aware LLM adaptation for advancing precision medicine and biomedical research.", "date": "2025-02-10", "categories": []}
{"id": "2508.10541", "title": "Driving Accurate Allergen Prediction with Protein Language Models and Generalization-Focused Evaluation", "abstract": "Allergens, typically proteins capable of triggering adverse immune responses, represent a significant public health challenge. To accurately identify allergen proteins, we introduce Applm (Allergen Prediction with Protein Language Models), a computational framework that leverages the 100-billion parameter xTrimoPGLM protein language model. We show that Applm consistently outperforms seven state-of-the-art methods in a diverse set of tasks that closely resemble difficult real-world scenarios. These include identifying novel allergens that lack similar examples in the training set, differentiating between allergens and non-allergens among homologs with high sequence similarity, and assessing functional consequences of mutations that create few changes to the protein sequences. Our analysis confirms that xTrimoPGLM, originally trained on one trillion tokens to capture general protein sequence characteristics, is crucial for Applm's performance by detecting important differences among protein sequences. In addition to providing Applm as open-source software, we also provide our carefully curated benchmark datasets to facilitate future research.", "date": "2025-08-14", "categories": []}
{"id": "2508.11069", "title": "Functional Analysis of Variance for Association Studies", "abstract": "While progress has been made in identifying common genetic variants associated with human diseases, for most of common complex diseases, the identified genetic variants only account for a small proportion of heritability. Challenges remain in finding additional unknown genetic variants predisposing to complex diseases. With the advance in next-generation sequencing technologies, sequencing studies have become commonplace in genetic research. The ongoing exome-sequencing and whole-genome-sequencing studies generate a massive amount of sequencing variants and allow researchers to comprehensively investigate their role in human diseases. The discovery of new disease-associated variants can be enhanced by utilizing powerful and computationally efficient statistical methods. In this paper, we propose a functional analysis of variance (FANOVA) method for testing an association of sequence variants in a genomic region with a qualitative trait. The FANOVA has a number of advantages: (1) it tests for a joint effect of gene variants, including both common and rare; (2) it fully utilizes linkage disequilibrium and genetic position information; and (3) allows for either protective or risk-increasing causal variants. Through simulations, we show that FANOVA outperform two popularly used methods - SKAT and a previously proposed method based on functional linear models (FLM), - especially if a sample size of a study is small and/or sequence variants have low to moderate effects. We conduct an empirical study by applying three methods (FANOVA, SKAT and FLM) to sequencing data from Dallas Heart Study. While SKAT and FLM respectively detected ANGPTL 4 and ANGPTL 3 associated with obesity, FANOVA was able to identify both genes associated with obesity.", "date": "2025-08-14", "categories": []}
{"id": "2508.11075", "title": "Abundance-Aware Set Transformer for Microbiome Sample Embedding", "abstract": "Microbiome sample representation to input into LLMs is essential for downstream tasks such as phenotype prediction and environmental classification. While prior studies have explored embedding-based representations of each microbiome sample, most rely on simple averaging over sequence embeddings, often overlooking the biological importance of taxa abundance. In this work, we propose an abundance-aware variant of the Set Transformer to construct fixed-size sample-level embeddings by weighting sequence embeddings according to their relative abundance. Without modifying the model architecture, we replicate embedding vectors proportional to their abundance and apply self-attention-based aggregation. Our method outperforms average pooling and unweighted Set Transformers on real-world microbiome classification tasks, achieving perfect performance in some cases. These results demonstrate the utility of abundance-aware aggregation for robust and biologically informed microbiome representation. To the best of our knowledge, this is one of the first approaches to integrate sequence-level abundance into Transformer-based sample embeddings.", "date": "2025-08-14", "categories": []}
{"id": "2508.11424", "title": "Generative Co-Design of Antibody Sequences and Structures via Black-Box Guidance in a Shared Latent Space", "abstract": "Advancements in deep generative models have enabled the joint modeling of antibody sequence and structure, given the antigen-antibody complex as context. However, existing approaches for optimizing complementarity-determining regions (CDRs) to improve developability properties operate in the raw data space, leading to excessively costly evaluations due to the inefficient search process. To address this, we propose LatEnt blAck-box Design (LEAD), a sequence-structure co-design framework that optimizes both sequence and structure within their shared latent space. Optimizing shared latent codes can not only break through the limitations of existing methods, but also ensure synchronization of different modality designs. Particularly, we design a black-box guidance strategy to accommodate real-world scenarios where many property evaluators are non-differentiable. Experimental results demonstrate that our LEAD achieves superior optimization performance for both single and multi-property objectives. Notably, LEAD reduces query consumption by a half while surpassing baseline methods in property optimization. The code is available at https://github.com/EvaFlower/LatEnt-blAck-box-Design.", "date": "2025-08-15", "categories": []}
{"id": "2507.08162", "title": "AmpLyze: A Deep Learning Model for Predicting the Hemolytic Concentration", "abstract": "Red-blood-cell lysis (HC50) is the principal safety barrier for antimicrobial-peptide (AMP) therapeutics, yet existing models only say \"toxic\" or \"non-toxic.\" AmpLyze closes this gap by predicting the actual HC50 value from sequence alone and explaining the residues that drive toxicity. The model couples residue-level ProtT5/ESM2 embeddings with sequence-level descriptors in dual local and global branches, aligned by a cross-attention module and trained with log-cosh loss for robustness to assay noise. The optimal AmpLyze model reaches a PCC of 0.756 and an MSE of 0.987, outperforming classical regressors and the state-of-the-art. Ablations confirm that both branches are essential, and cross-attention adds a further 1% PCC and 3% MSE improvement. Expected-Gradients attributions reveal known toxicity hotspots and suggest safer substitutions. By turning hemolysis assessment into a quantitative, sequence-based, and interpretable prediction, AmpLyze facilitates AMP design and offers a practical tool for early-stage toxicity screening.", "date": "2025-07-10", "categories": []}
{"id": "2508.09659", "title": "Thermal Tracks: A Gaussian process-based framework for universal melting curve analysis enabling unconstrained hit identification in thermal proteome profiling experiments", "abstract": "Thermal Tracks is a Python-based statistical framework for analyzing protein thermal stability data that overcomes key limitations of existing thermal proteome profiling (TPP) work-flows. Unlike standard approaches that assume sigmoidal melting curves and are constrained by empirical null distributions (limiting significant hits to approximately 5 % of data), Thermal Tracks uses Gaussian Process (GP) models with squared-exponential kernels to flexibly model any melting curve shape while generating unbiased null distributions through kernel priors. This framework is particularly valuable for analyzing proteome-wide perturbations that significantly alter protein thermal stability, such as pathway inhibitions, genetic modifications, or environmental stresses, where conventional TPP methods may miss biologically relevant changes due to their statistical constraints. Furthermore, Thermal Tracks excels at analyzing proteins with un-conventional melting profiles, including phase-separating proteins and membrane proteins, which often exhibit complex, non-sigmoidal thermal stability behaviors. Thermal Tracks is freely available from GitHub and is implemented in Python, providing an accessible and flexible tool for proteome-wide thermal profiling studies.", "date": "2025-08-13", "categories": []}
{"id": "2508.10232", "title": "CellSymphony: Deciphering the molecular and phenotypic orchestration of cells with single-cell pathomics", "abstract": "Xenium, a new spatial transcriptomics platform, enables subcellular-resolution profiling of complex tumor tissues. Despite the rich morphological information in histology images, extracting robust cell-level features and integrating them with spatial transcriptomics data remains a critical challenge. We introduce CellSymphony, a flexible multimodal framework that leverages foundation model-derived embeddings from both Xenium transcriptomic profiles and histology images at true single-cell resolution. By learning joint representations that fuse spatial gene expression with morphological context, CellSymphony achieves accurate cell type annotation and uncovers distinct microenvironmental niches across three cancer types. This work highlights the potential of foundation models and multimodal fusion for deciphering the physiological and phenotypic orchestration of cells within complex tissue ecosystems.", "date": "2025-08-13", "categories": []}
{"id": "2508.10629", "title": "Energy-Based Models for Predicting Mutational Effects on Proteins", "abstract": "Predicting changes in binding free energy ($\\Delta\\Delta G$) is a vital task in protein engineering and protein-protein interaction (PPI) engineering for drug discovery. Previous works have observed a high correlation between $\\Delta\\Delta G$ and entropy, using probabilities of biologically important objects such as side chain angles and residue identities to estimate $\\Delta\\Delta G$. However, estimating the full conformational distribution of a protein complex is generally considered intractable. In this work, we propose a new approach to $\\Delta\\Delta G$ prediction that avoids this issue by instead leveraging energy-based models for estimating the probability of a complex's conformation. Specifically, we novelly decompose $\\Delta\\Delta G$ into a sequence-based component estimated by an inverse folding model and a structure-based component estimated by an energy model. This decomposition is made tractable by assuming equilibrium between the bound and unbound states, allowing us to simplify the estimation of degeneracies associated with each state. Unlike previous deep learning-based methods, our method incorporates an energy-based physical inductive bias by connecting the often-used sequence log-odds ratio-based approach to $\\Delta\\Delta G$ prediction with a new $\\Delta\\Delta E$ term grounded in statistical mechanics. We demonstrate superiority over existing state-of-the-art structure and sequence-based deep learning methods in $\\Delta\\Delta G$ prediction and antibody optimization against SARS-CoV-2.", "date": "2025-08-14", "categories": []}
{"id": "2508.10775", "title": "IBEX: Information-Bottleneck-EXplored Coarse-to-Fine Molecular Generation under Limited Data", "abstract": "Three-dimensional generative models increasingly drive structure-based drug discovery, yet it remains constrained by the scarce publicly available protein-ligand complexes. Under such data scarcity, almost all existing pipelines struggle to learn transferable geometric priors and consequently overfit to training-set biases. As such, we present IBEX, an Information-Bottleneck-EXplored coarse-to-fine pipeline to tackle the chronic shortage of protein-ligand complex data in structure-based drug design. Specifically, we use PAC-Bayesian information-bottleneck theory to quantify the information density of each sample. This analysis reveals how different masking strategies affect generalization and indicates that, compared with conventional de novo generation, the constrained Scaffold Hopping task endows the model with greater effective capacity and improved transfer performance. IBEX retains the original TargetDiff architecture and hyperparameters for training to generate molecules compatible with the binding pocket; it then applies an L-BFGS optimization step to finely refine each conformation by optimizing five physics-based terms and adjusting six translational and rotational degrees of freedom in under one second. With only these modifications, IBEX raises the zero-shot docking success rate on CBGBench CrossDocked2020-based from 53% to 64%, improves the mean Vina score from $-7.41 kcal mol^{-1}$ to $-8.07 kcal mol^{-1}$, and achieves the best median Vina energy in 57 of 100 pockets versus 3 for the original TargetDiff. IBEX also increases the QED by 25%, achieves state-of-the-art validity and diversity, and markedly reduces extrapolation error.", "date": "2025-08-14", "categories": []}
{"id": "2508.10841", "title": "Performance of universal machine-learned potentials with explicit long-range interactions in biomolecular simulations", "abstract": "Universal machine-learned potentials promise transferable accuracy across compositional and vibrational degrees of freedom, yet their application to biomolecular simulations remains underexplored. This work systematically evaluates equivariant message-passing architectures trained on the SPICE-v2 dataset with and without explicit long-range dispersion and electrostatics. We assess the impact of model size, training data composition, and electrostatic treatment across in- and out-of-distribution benchmark datasets, as well as molecular simulations of bulk liquid water, aqueous NaCl solutions, and biomolecules, including alanine tripeptide, the mini-protein Trp-cage, and Crambin. While larger models improve accuracy on benchmark datasets, this trend does not consistently extend to properties obtained from simulations. Predicted properties also depend on the composition of the training dataset. Long-range electrostatics show no systematic impact across systems. However, for Trp-cage, their inclusion yields increased conformational variability. Our results suggest that imbalanced datasets and immature evaluation practices currently challenge the applicability of universal machine-learned potentials to biomolecular simulations.", "date": "2025-08-14", "categories": []}
{"id": "2506.17857", "title": "AbRank: A Benchmark Dataset and Metric-Learning Framework for Antibody-Antigen Affinity Ranking", "abstract": "Accurate prediction of antibody-antigen (Ab-Ag) binding affinity is essential for therapeutic design and vaccine development, yet the performance of current models is limited by noisy experimental labels, heterogeneous assay conditions, and poor generalization across the vast antibody and antigen sequence space. We introduce AbRank, a large-scale benchmark and evaluation framework that reframes affinity prediction as a pairwise ranking problem. AbRank aggregates over 380,000 binding assays from nine heterogeneous sources, spanning diverse antibodies, antigens, and experimental conditions, and introduces standardized data splits that systematically increase distribution shift, from local perturbations such as point mutations to broad generalization across novel antigens and antibodies. To ensure robust supervision, AbRank defines an m-confident ranking framework by filtering out comparisons with marginal affinity differences, focusing training on pairs with at least an m-fold difference in measured binding strength. As a baseline for the benchmark, we introduce WALLE-Affinity, a graph-based approach that integrates protein language model embeddings with structural information to predict pairwise binding preferences. Our benchmarks reveal significant limitations in current methods under realistic generalization settings and demonstrate that ranking-based training improves robustness and transferability. In summary, AbRank offers a robust foundation for machine learning models to generalize across the antibody-antigen space, with direct relevance for scalable, structure-aware antibody therapeutic design.", "date": "2025-06-21", "categories": []}
{"id": "2508.09499", "title": "CWFBind: Geometry-Awareness for Fast and Accurate Protein-Ligand Docking", "abstract": "Accurately predicting the binding conformation of small-molecule ligands to protein targets is a critical step in rational drug design. Although recent deep learning-based docking surpasses traditional methods in speed and accuracy, many approaches rely on graph representations and language model-inspired encoders while neglecting critical geometric information, resulting in inaccurate pocket localization and unrealistic binding conformations. In this study, we introduce CWFBind, a weighted, fast, and accurate docking method based on local curvature features. Specifically, we integrate local curvature descriptors during the feature extraction phase to enrich the geometric representation of both proteins and ligands, complementing existing chemical, sequence, and structural features. Furthermore, we embed degree-aware weighting mechanisms into the message passing process, enhancing the model's ability to capture spatial structural distinctions and interaction strengths. To address the class imbalance challenge in pocket prediction, CWFBind employs a ligand-aware dynamic radius strategy alongside an enhanced loss function, facilitating more precise identification of binding regions and key residues. Comprehensive experimental evaluations demonstrate that CWFBind achieves competitive performance across multiple docking benchmarks, offering a balanced trade-off between accuracy and efficiency.", "date": "2025-08-13", "categories": []}
{"id": "2406.13839", "title": "RNA-FrameFlow: Flow Matching for de novo 3D RNA Backbone Design", "abstract": "We introduce RNA-FrameFlow, the first generative model for 3D RNA backbone design. We build upon SE(3) flow matching for protein backbone generation and establish protocols for data preparation and evaluation to address unique challenges posed by RNA modeling. We formulate RNA structures as a set of rigid-body frames and associated loss functions which account for larger, more conformationally flexible RNA backbones (13 atoms per nucleotide) vs. proteins (4 atoms per residue). Toward tackling the lack of diversity in 3D RNA datasets, we explore training with structural clustering and cropping augmentations. Additionally, we define a suite of evaluation metrics to measure whether the generated RNA structures are globally self-consistent (via inverse folding followed by forward folding) and locally recover RNA-specific structural descriptors. The most performant version of RNA-FrameFlow generates locally realistic RNA backbones of 40-150 nucleotides, over 40% of which pass our validity criteria as measured by a self-consistency TM-score >= 0.45, at which two RNAs have the same global fold. Open-source code: https://github.com/rish-16/rna-backbone-design", "date": "2024-06-19", "categories": []}
{"id": "2503.10708", "title": "Exploration of Hepatitis B Virus Infection Dynamics through Physics-Informed Deep Learning Approach", "abstract": "Accurate forecasting of viral disease outbreaks is crucial for guiding public health responses and preventing widespread loss of life. In recent years, Physics-Informed Neural Networks (PINNs) have emerged as a promising framework that can capture the intricate dynamics of viral infection and reliably predict its future progression. However, despite notable advances, the application of PINNs in disease modeling remains limited. Standard PINNs are effective in simulating disease dynamics through forward modeling but often face challenges in estimating key biological parameters from sparse or noisy experimental data when applied in an inverse framework. To overcome these limitations, a recent extension known as Disease Informed Neural Networks (DINNs) has emerged, offering a more robust approach to parameter estimation tasks. In this work, we apply this DINNs technique on a recently proposed hepatitis B virus (HBV) infection dynamics model to predict infection transmission within the liver. This model consists of four compartments: uninfected and infected hepatocytes, rcDNA-containing capsids, and free viruses. Leveraging the power of DINNs, we study the impacts of (i) variations in parameter range, (ii) experimental noise in data, (iii) sample sizes, (iv) network architecture and (v) learning rate. We employ this methodology in experimental data collected from nine HBV-infected chimpanzees and observe that it reliably estimates the model parameters. DINNs can capture infection dynamics and predict their future progression even when data of some compartments of the system are missing. Additionally, it identifies the influential model parameters that determine whether the HBV infection is cleared or persists within the host.", "date": "2025-03-12", "categories": []}
{"id": "2508.07127", "title": "How Effectively Can Large Language Models Connect SNP Variants and ECG Phenotypes for Cardiovascular Risk Prediction?", "abstract": "Cardiovascular disease (CVD) prediction remains a tremendous challenge due to its multifactorial etiology and global burden of morbidity and mortality. Despite the growing availability of genomic and electrophysiological data, extracting biologically meaningful insights from such high-dimensional, noisy, and sparsely annotated datasets remains a non-trivial task. Recently, LLMs has been applied effectively to predict structural variations in biological sequences. In this work, we explore the potential of fine-tuned LLMs to predict cardiac diseases and SNPs potentially leading to CVD risk using genetic markers derived from high-throughput genomic profiling. We investigate the effect of genetic patterns associated with cardiac conditions and evaluate how LLMs can learn latent biological relationships from structured and semi-structured genomic data obtained by mapping genetic aspects that are inherited from the family tree. By framing the problem as a Chain of Thought (CoT) reasoning task, the models are prompted to generate disease labels and articulate informed clinical deductions across diverse patient profiles and phenotypes. The findings highlight the promise of LLMs in contributing to early detection, risk assessment, and ultimately, the advancement of personalized medicine in cardiac care.", "date": "2025-08-10", "categories": []}
{"id": "2508.07345", "title": "ProteoKnight: Convolution-based phage virion protein classification and uncertainty analysis", "abstract": "\\textbf{Introduction:} Accurate prediction of Phage Virion Proteins (PVP) is essential for genomic studies due to their crucial role as structural elements in bacteriophages. Computational tools, particularly machine learning, have emerged for annotating phage protein sequences from high-throughput sequencing. However, effective annotation requires specialized sequence encodings. Our paper introduces ProteoKnight, a new image-based encoding method that addresses spatial constraints in existing techniques, yielding competitive performance in PVP classification using pre-trained convolutional neural networks. Additionally, our study evaluates prediction uncertainty in binary PVP classification through Monte Carlo Dropout (MCD). \\textbf{Methods:} ProteoKnight adapts the classical DNA-Walk algorithm for protein sequences, incorporating pixel colors and adjusting walk distances to capture intricate protein features. Encoded sequences were classified using multiple pre-trained CNNs. Variance and entropy measures assessed prediction uncertainty across proteins of various classes and lengths. \\textbf{Results:} Our experiments achieved 90.8% accuracy in binary classification, comparable to state-of-the-art methods. Multi-class classification accuracy remains suboptimal. Our uncertainty analysis unveils variability in prediction confidence influenced by protein class and sequence length. \\textbf{Conclusions:} Our study surpasses frequency chaos game representation (FCGR) by introducing novel image encoding that mitigates spatial information loss limitations. Our classification technique yields accurate and robust PVP predictions while identifying low-confidence predictions.", "date": "2025-08-10", "categories": []}
{"id": "2508.05006", "title": "The Docking Game: Loop Self-Play for Fast, Dynamic, and Accurate Prediction of Flexible Protein-Ligand Binding", "abstract": "Molecular docking is a crucial aspect of drug discovery, as it predicts the binding interactions between small-molecule ligands and protein pockets. However, current multi-task learning models for docking often show inferior performance in ligand docking compared to protein pocket docking. This disparity arises largely due to the distinct structural complexities of ligands and proteins. To address this issue, we propose a novel game-theoretic framework that models the protein-ligand interaction as a two-player game called the Docking Game, with the ligand docking module acting as the ligand player and the protein pocket docking module as the protein player. To solve this game, we develop a novel Loop Self-Play (LoopPlay) algorithm, which alternately trains these players through a two-level loop. In the outer loop, the players exchange predicted poses, allowing each to incorporate the other's structural predictions, which fosters mutual adaptation over multiple iterations. In the inner loop, each player dynamically refines its predictions by incorporating its own predicted ligand or pocket poses back into its model. We theoretically show the convergence of LoopPlay, ensuring stable optimization. Extensive experiments conducted on public benchmark datasets demonstrate that LoopPlay achieves approximately a 10\\% improvement in predicting accurate binding modes compared to previous state-of-the-art methods. This highlights its potential to enhance the accuracy of molecular docking in drug discovery.", "date": "2025-08-07", "categories": []}
{"id": "2502.20632", "title": "Lattice Protein Folding with Variational Annealing", "abstract": "Understanding the principles of protein folding is a cornerstone of computational biology, with implications for drug design, bioengineering, and the understanding of fundamental biological processes. Lattice protein folding models offer a simplified yet powerful framework for studying the complexities of protein folding, enabling the exploration of energetically optimal folds under constrained conditions. However, finding these optimal folds is a computationally challenging combinatorial optimization problem. In this work, we introduce a novel upper-bound training scheme that employs masking to identify the lowest-energy folds in two-dimensional Hydrophobic-Polar (HP) lattice protein folding. By leveraging Dilated Recurrent Neural Networks (RNNs) integrated with an annealing process driven by temperature-like fluctuations, our method accurately predicts optimal folds for benchmark systems of up to 60 beads. Our approach also effectively masks invalid folds from being sampled without compromising the autoregressive sampling properties of RNNs. This scheme is generalizable to three spatial dimensions and can be extended to lattice protein models with larger alphabets. Our findings emphasize the potential of advanced machine learning techniques in tackling complex protein folding problems and a broader class of constrained combinatorial optimization challenges.", "date": "2025-02-28", "categories": []}
{"id": "2508.04739", "title": "CodonMoE: DNA Language Models for mRNA Analyses", "abstract": "Genomic language models (gLMs) face a fundamental efficiency challenge: either maintain separate specialized models for each biological modality (DNA and RNA) or develop large multi-modal architectures. Both approaches impose significant computational burdens - modality-specific models require redundant infrastructure despite inherent biological connections, while multi-modal architectures demand massive parameter counts and extensive cross-modality pretraining. To address this limitation, we introduce CodonMoE (Adaptive Mixture of Codon Reformative Experts), a lightweight adapter that transforms DNA language models into effective RNA analyzers without RNA-specific pretraining. Our theoretical analysis establishes CodonMoE as a universal approximator at the codon level, capable of mapping arbitrary functions from codon sequences to RNA properties given sufficient expert capacity. Across four RNA prediction tasks spanning stability, expression, and regulation, DNA models augmented with CodonMoE significantly outperform their unmodified counterparts, with HyenaDNA+CodonMoE series achieving state-of-the-art results using 80% fewer parameters than specialized RNA models. By maintaining sub-quadratic complexity while achieving superior performance, our approach provides a principled path toward unifying genomic language modeling, leveraging more abundant DNA data and reducing computational overhead while preserving modality-specific performance advantages.", "date": "2025-08-06", "categories": []}
{"id": "2508.04742", "title": "Discovery of Disease Relationships via Transcriptomic Signature Analysis Powered by Agentic AI", "abstract": "Modern disease classification often overlooks molecular commonalities hidden beneath divergent clinical presentations. This study introduces a transcriptomics-driven framework for discovering disease relationships by analyzing over 1300 disease-condition pairs using GenoMAS, a fully automated agentic AI system. Beyond identifying robust gene-level overlaps, we develop a novel pathway-based similarity framework that integrates multi-database enrichment analysis to quantify functional convergence across diseases. The resulting disease similarity network reveals both known comorbidities and previously undocumented cross-category links. By examining shared biological pathways, we explore potential molecular mechanisms underlying these connections-offering functional hypotheses that go beyond symptom-based taxonomies. We further show how background conditions such as obesity and hypertension modulate transcriptomic similarity, and identify therapeutic repurposing opportunities for rare diseases like autism spectrum disorder based on their molecular proximity to better-characterized conditions. In addition, this work demonstrates how biologically grounded agentic AI can scale transcriptomic analysis while enabling mechanistic interpretation across complex disease landscapes. All results are publicly accessible at github.com/KeeeeChen/Pathway_Similarity_Network.", "date": "2025-08-06", "categories": []}
{"id": "2508.04743", "title": "Alz-QNet: A Quantum Regression Network for Studying Alzheimer's Gene Interactions", "abstract": "Understanding the molecular-level mechanisms underpinning Alzheimer's disease (AD) by studying crucial genes associated with the disease remains a challenge. Alzheimer's, being a multifactorial disease, requires understanding the gene-gene interactions underlying it for theranostics and progress. In this article, a novel attempt has been made using a quantum regression to decode how some crucial genes in the AD Amyloid Beta Precursor Protein ($APP$), Sterol regulatory element binding transcription factor 14 ($FGF14$), Yin Yang 1 ($YY1$), and Phospholipase D Family Member 3 ($PLD3$) etc. become influenced by other prominent switching genes during disease progression, which may help in gene expression-based therapy for AD. Our proposed Quantum Regression Network (Alz-QNet) introduces a pioneering approach with insights from the state-of-the-art Quantum Gene Regulatory Networks (QGRN) to unravel the gene interactions involved in AD pathology, particularly within the Entorhinal Cortex (EC), where early pathological changes occur. Using the proposed Alz-QNet framework, we explore the interactions between key genes ($APP$, $FGF14$, $YY1$, $EGR1$, $GAS7$, $AKT3$, $SREBF2$, and $PLD3$) within the CE microenvironment of AD patients, studying genetic samples from the database $GSE138852$, all of which are believed to play a crucial role in the progression of AD. Our investigation uncovers intricate gene-gene interactions, shedding light on the potential regulatory mechanisms that underlie the pathogenesis of AD, which help us to find potential gene inhibitors or regulators for theranostics.", "date": "2025-08-06", "categories": []}
{"id": "2508.04757", "title": "Embedding Is (Almost) All You Need: Retrieval-Augmented Inference for Generalizable Genomic Prediction Tasks", "abstract": "Large pre-trained DNA language models such as DNABERT-2, Nucleotide Transformer, and HyenaDNA have demonstrated strong performance on various genomic benchmarks. However, most applications rely on expensive fine-tuning, which works best when the training and test data share a similar distribution. In this work, we investigate whether task-specific fine-tuning is always necessary. We show that simple embedding-based pipelines that extract fixed representations from these models and feed them into lightweight classifiers can achieve competitive performance. In evaluation settings with different data distributions, embedding-based methods often outperform fine-tuning while reducing inference time by 10x to 20x. Our results suggest that embedding extraction is not only a strong baseline but also a more generalizable and efficient alternative to fine-tuning, especially for deployment in diverse or unseen genomic contexts. For example, in enhancer classification, HyenaDNA embeddings combined with zCurve achieve 0.68 accuracy (vs. 0.58 for fine-tuning), with an 88% reduction in inference time and over 8x lower carbon emissions (0.02 kg vs. 0.17 kg CO2). In non-TATA promoter classification, DNABERT-2 embeddings with zCurve or GC content reach 0.85 accuracy (vs. 0.89 with fine-tuning) with a 22x lower carbon footprint (0.02 kg vs. 0.44 kg CO2). These results show that embedding-based pipelines offer over 10x better carbon efficiency while maintaining strong predictive performance. The code is available here: https://github.com/NIRJHOR-DATTA/EMBEDDING-IS-ALMOST-ALL-YOU-NEED.", "date": "2025-08-06", "categories": []}
{"id": "2508.02743", "title": "A Novel cVAE-Augmented Deep Learning Framework for Pan-Cancer RNA-Seq Classification", "abstract": "Pan-cancer classification using transcriptomic (RNA-Seq) data can inform tumor subtyping and therapy selection, but is challenging due to extremely high dimensionality and limited sample sizes. In this study, we propose a novel deep learning framework that uses a class-conditional variational autoencoder (cVAE) to augment training data for pan-cancer gene expression classification. Using 801 tumor RNA-Seq samples spanning 5 cancer types from The Cancer Genome Atlas (TCGA), we first perform feature selection to reduce 20,531 gene expression features to the 500 most variably expressed genes. A cVAE is then trained on this data to learn a latent representation of gene expression conditioned on cancer type, enabling the generation of synthetic gene expression samples for each tumor class. We augment the training set with these cVAE-generated samples (doubling the dataset size) to mitigate overfitting and class imbalance. A two-layer multilayer perceptron (MLP) classifier is subsequently trained on the augmented dataset to predict tumor type. The augmented framework achieves high classification accuracy (~98%) on a held-out test set, substantially outperforming a classifier trained on the original data alone. We present detailed experimental results, including VAE training curves, classifier performance metrics (ROC curves and confusion matrix), and architecture diagrams to illustrate the approach. The results demonstrate that cVAE-based synthetic augmentation can significantly improve pan-cancer prediction performance, especially for underrepresented cancer classes.", "date": "2025-08-02", "categories": []}
{"id": "2501.06608", "title": "Dual-Modality Representation Learning for Molecular Property Prediction", "abstract": "Molecular property prediction has attracted substantial attention recently. Accurate prediction of drug properties relies heavily on effective molecular representations. The structures of chemical compounds are commonly represented as graphs or SMILES sequences. Recent advances in learning drug properties commonly employ Graph Neural Networks (GNNs) based on the graph representation. For the SMILES representation, Transformer-based architectures have been adopted by treating each SMILES string as a sequence of tokens. Because each representation has its own advantages and disadvantages, combining both representations in learning drug properties is a promising direction. We propose a method named Dual-Modality Cross-Attention (DMCA) that can effectively combine the strengths of two representations by employing the cross-attention mechanism. DMCA was evaluated across eight datasets including both classification and regression tasks. Results show that our method achieves the best overall performance, highlighting its effectiveness in leveraging the complementary information from both graph and SMILES modalities.", "date": "2025-01-11", "categories": []}
{"id": "2504.19034", "title": "On learning functions over biological sequence space: relating Gaussian process priors, regularization, and gauge fixing", "abstract": "Mappings from biological sequences (DNA, RNA, protein) to quantitative measures of sequence functionality play an important role in contemporary biology. We are interested in the related tasks of (i) inferring predictive sequence-to-function maps and (ii) decomposing sequence-function maps to elucidate the contributions of individual subsequences. Because each sequence-function map can be written as a weighted sum over subsequences in multiple ways, meaningfully interpreting these weights requires ``gauge-fixing,'' i.e., defining a unique representation for each map. Recent work has established that most existing gauge-fixed representations arise as the unique solutions to $L_2$-regularized regression in an overparameterized ``weight space'' where the choice of regularizer defines the gauge. Here, we establish the relationship between regularized regression in overparameterized weight space and Gaussian process approaches that operate in ``function space,'' i.e.~the space of all real-valued functions on a finite set of sequences. We disentangle how weight space regularizers both impose an implicit prior on the learned function and restrict the optimal weights to a particular gauge. We show how to construct regularizers that correspond to arbitrary explicit Gaussian process priors combined with a wide variety of gauges and characterize the implicit function space priors associated with the most common weight space regularizers. Finally, we derive the posterior distribution of a broad class of sequence-to-function statistics, including gauge-fixed weights and multiple systems for expressing higher-order epistatic coefficients. We show that such distributions can be efficiently computed for product-kernel priors using a kernel trick.", "date": "2025-04-26", "categories": []}
{"id": "2507.02724", "title": "Hierarchical Multi-Label Contrastive Learning for Protein-Protein Interaction Prediction Across Organisms", "abstract": "Recent advances in AI for science have highlighted the power of contrastive learning in bridging heterogeneous biological data modalities. Building on this paradigm, we propose HIPPO (HIerarchical Protein-Protein interaction prediction across Organisms), a hierarchical contrastive framework for protein-protein interaction(PPI) prediction, where protein sequences and their hierarchical attributes are aligned through multi-tiered biological representation matching. The proposed approach incorporates hierarchical contrastive loss functions that emulate the structured relationship among functional classes of proteins. The framework adaptively incorporates domain and family knowledge through a data-driven penalty mechanism, enforcing consistency between the learned embedding space and the intrinsic hierarchy of protein functions. Experiments on benchmark datasets demonstrate that HIPPO achieves state-of-the-art performance, outperforming existing methods and showing robustness in low-data regimes. Notably, the model demonstrates strong zero-shot transferability to other species without retraining, enabling reliable PPI prediction and functional inference even in less characterized or rare organisms where experimental data are limited. Further analysis reveals that hierarchical feature fusion is critical for capturing conserved interaction determinants, such as binding motifs and functional annotations. This work advances cross-species PPI prediction and provides a unified framework for interaction prediction in scenarios with sparse or imbalanced multi-species data.", "date": "2025-07-03", "categories": []}
{"id": "2410.07501", "title": "Inferring biological processes with intrinsic noise from cross-sectional data", "abstract": "Inferring dynamical models from data continues to be a significant challenge in computational biology, especially given the stochastic nature of many biological processes. We explore a common scenario in omics, where statistically independent cross-sectional samples are available at a few time points, and the goal is to infer the underlying diffusion process that generated the data. Existing inference approaches often simplify or ignore noise intrinsic to the system, compromising accuracy for the sake of optimization ease. We circumvent this compromise by inferring the phase-space probability flow that shares the same time-dependent marginal distributions as the underlying stochastic process. Our approach, probability flow inference (PFI), disentangles force from intrinsic stochasticity while retaining the algorithmic ease of ODE inference. Analytically, we prove that for Ornstein-Uhlenbeck processes the regularized PFI formalism yields a unique solution in the limit of well-sampled distributions. In practical applications, we show that PFI enables accurate parameter and force estimation in high-dimensional stochastic reaction networks, and that it allows inference of cell differentiation dynamics with molecular noise, outperforming state-of-the-art approaches.", "date": "2024-10-10", "categories": []}
{"id": "2507.22587", "title": "Deep learning of geometrical cell division rules", "abstract": "The positioning of new cellular walls during cell division plays a key role in shaping plant tissue organization. The influence of cell geometry on the positioning of division planes has been previously captured into various geometrical rules. Accordingly, linking cell shape to division orientation has relied on the comparison between observed division patterns and predictions under specific rules. The need to define a priori the tested rules is a fundamental limitation of this hypothesis-driven approach. As an alternative, we introduce a data-based approach to investigate the relation between cell geometry and division plane positioning, exploiting the ability of deep neural network to learn complex relationships across multidimensional spaces. Adopting an image-based cell representation, we show how division patterns can be learned and predicted from mother cell geometry using a UNet architecture modified to operate on cell masks. Using synthetic data and A. thaliana embryo cells, we evaluate the model performances on a wide range of diverse cell shapes and division patterns. We find that the trained model accounted for embryo division patterns that were previously irreconcilable under existing geometrical rules. Our work shows the potential of deep networks to understand cell division patterns and to generate new hypotheses on the control of cell division positioning.", "date": "2025-07-30", "categories": []}
{"id": "2507.21706", "title": "EnTao-GPM: DNA Foundation Model for Predicting the Germline Pathogenic Mutations", "abstract": "Distinguishing pathogenic mutations from benign polymorphisms remains a critical challenge in precision medicine. EnTao-GPM, developed by Fudan University and BioMap, addresses this through three innovations: (1) Cross-species targeted pre-training on disease-relevant mammalian genomes (human, pig, mouse), leveraging evolutionary conservation to enhance interpretation of pathogenic motifs, particularly in non-coding regions; (2) Germline mutation specialization via fine-tuning on ClinVar and HGMD, improving accuracy for both SNVs and non-SNVs; (3) Interpretable clinical framework integrating DNA sequence embeddings with LLM-based statistical explanations to provide actionable insights. Validated against ClinVar, EnTao-GPM demonstrates superior accuracy in mutation classification. It revolutionizes genetic testing by enabling faster, more accurate, and accessible interpretation for clinical diagnostics (e.g., variant assessment, risk identification, personalized treatment) and research, advancing personalized medicine.", "date": "2025-07-29", "categories": []}
{"id": "2411.11513", "title": "A Modular Open Source Framework for Genomic Variant Calling", "abstract": "Variant calling is a fundamental task in genomic research, essential for detecting genetic variations such as single nucleotide polymorphisms (SNPs) and insertions or deletions (indels). This paper presents an enhancement to DeepChem, a widely used open-source drug discovery framework, through the integration of DeepVariant. In particular, we introduce a variant calling pipeline that leverages DeepVariant's convolutional neural network (CNN) architecture to improve the accuracy and reliability of variant detection. The implemented pipeline includes stages for realignment of sequencing reads, candidate variant detection, and pileup image generation, followed by variant classification using a modified Inception v3 model. Our work adds a modular and extensible variant calling framework to the DeepChem framework and enables future work integrating DeepChem's drug discovery infrastructure more tightly with bioinformatics pipelines.", "date": "2024-11-18", "categories": []}
{"id": "2505.02022", "title": "NbBench: Benchmarking Language Models for Comprehensive Nanobody Tasks", "abstract": "Nanobodies -- single-domain antibody fragments derived from camelid heavy-chain-only antibodies -- exhibit unique advantages such as compact size, high stability, and strong binding affinity, making them valuable tools in therapeutics and diagnostics. While recent advances in pretrained protein and antibody language models (PPLMs and PALMs) have greatly enhanced biomolecular understanding, nanobody-specific modeling remains underexplored and lacks a unified benchmark. To address this gap, we introduce NbBench, the first comprehensive benchmark suite for nanobody representation learning. Spanning eight biologically meaningful tasks across nine curated datasets, NbBench encompasses structure annotation, binding prediction, and developability assessment. We systematically evaluate eleven representative models -- including general-purpose protein LMs, antibody-specific LMs, and nanobody-specific LMs -- in a frozen setting. Our analysis reveals that antibody language models excel in antigen-related tasks, while performance on regression tasks such as thermostability and affinity remains challenging across all models. Notably, no single model consistently outperforms others across all tasks. By standardizing datasets, task definitions, and evaluation protocols, NbBench offers a reproducible foundation for assessing and advancing nanobody modeling.", "date": "2025-05-04", "categories": []}
{"id": "2507.19565", "title": "Review of Deep Learning Applications to Structural Proteomics Enabled by Cryogenic Electron Microscopy and Tomography", "abstract": "The past decade's \"cryoEM revolution\" has produced exponential growth in high-resolution structural data through advances in cryogenic electron microscopy (cryoEM) and tomography (cryoET). Deep learning integration into structural proteomics workflows addresses longstanding challenges including low signal-to-noise ratios, preferred orientation artifacts, and missing-wedge problems that historically limited efficiency and scalability. This review examines AI applications across the entire cryoEM pipeline, from automated particle picking using convolutional neural networks (Topaz, crYOLO, CryoSegNet) to computational solutions for preferred orientation bias (spIsoNet, cryoPROS) and advanced denoising algorithms (Topaz-Denoise). In cryoET, tools like IsoNet employ U-Net architectures for simultaneous missing-wedge correction and noise reduction, while TomoNet streamlines subtomogram averaging through AI-driven particle detection. The workflow culminates with automated atomic model building using sophisticated tools like ModelAngelo, DeepTracer, and CryoREAD that translate density maps into interpretable biological structures. These AI-enhanced approaches have achieved near-atomic resolution reconstructions with minimal manual intervention, resolved previously intractable datasets suffering from severe orientation bias, and enabled successful application to diverse biological systems from HIV virus-like particles to in situ ribosomal complexes. As deep learning evolves, particularly with large language models and vision transformers, the future promises sophisticated automation and accessibility in structural biology, potentially revolutionizing our understanding of macromolecular architecture and function.", "date": "2025-07-25", "categories": []}
{"id": "2507.19755", "title": "Modeling enzyme temperature stability from sequence segment perspective", "abstract": "Developing enzymes with desired thermal properties is crucial for a wide range of industrial and research applications, and determining temperature stability is an essential step in this process. Experimental determination of thermal parameters is labor-intensive, time-consuming, and costly. Moreover, existing computational approaches are often hindered by limited data availability and imbalanced distributions. To address these challenges, we introduce a curated temperature stability dataset designed for model development and benchmarking in enzyme thermal modeling. Leveraging this dataset, we present the \\textit{Segment Transformer}, a novel deep learning framework that enables efficient and accurate prediction of enzyme temperature stability. The model achieves state-of-the-art performance with an RMSE of 24.03, MAE of 18.09, and Pearson and Spearman correlations of 0.33, respectively. These results highlight the effectiveness of incorporating segment-level representations, grounded in the biological observation that different regions of a protein sequence contribute unequally to thermal behavior. As a proof of concept, we applied the Segment Transformer to guide the engineering of a cutinase enzyme. Experimental validation demonstrated a 1.64-fold improvement in relative activity following heat treatment, achieved through only 17 mutations and without compromising catalytic function.", "date": "2025-07-26", "categories": []}
{"id": "2507.19805", "title": "Sequence-based protein-protein interaction prediction and its applications in drug discovery", "abstract": "Aberrant protein-protein interactions (PPIs) underpin a plethora of human diseases, and disruption of these harmful interactions constitute a compelling treatment avenue. Advances in computational approaches to PPI prediction have closely followed progress in deep learning and natural language processing. In this review, we outline the state-of the-art for sequence-based PPI prediction methods and explore their impact on target identification and drug discovery. We begin with an overview of commonly used training data sources and techniques used to curate these data to enhance the quality of the training set. Subsequently, we survey various PPI predictor types, including traditional similarity-based approaches, and deep learning-based approaches with a particular emphasis on the transformer architecture. Finally, we provide examples of PPI prediction in systems-level proteomics analyses, target identification, and design of therapeutic peptides and antibodies. We also take the opportunity to showcase the potential of PPI-aware drug discovery models in accelerating therapeutic development.", "date": "2025-07-26", "categories": []}
{"id": "2507.20130", "title": "Generative molecule evolution using 3D pharmacophore for efficient Structure-Based Drug Design", "abstract": "Recent advances in generative models, particularly diffusion and auto-regressive models, have revolutionized fields like computer vision and natural language processing. However, their application to structure-based drug design (SBDD) remains limited due to critical data constraints. To address the limitation of training data for models targeting SBDD tasks, we propose an evolutionary framework named MEVO, which bridges the gap between billion-scale small molecule dataset and the scarce protein-ligand complex dataset, and effectively increase the abundance of training data for generative SBDD models. MEVO is composed of three key components: a high-fidelity VQ-VAE for molecule representation in latent space, a diffusion model for pharmacophore-guided molecule generation, and a pocket-aware evolutionary strategy for molecule optimization with physics-based scoring function. This framework efficiently generate high-affinity binders for various protein targets, validated with predicted binding affinities using free energy perturbation (FEP) methods. In addition, we showcase the capability of MEVO in designing potent inhibitors to KRAS$^{\\textrm{G12D}}$, a challenging target in cancer therapeutics, with similar affinity to the known highly active inhibitor evaluated by FEP calculations. With high versatility and generalizability, MEVO offers an effective and data-efficient model for various tasks in structure-based ligand design.", "date": "2025-07-27", "categories": []}
{"id": "2507.20243", "title": "Protein-SE(3): Benchmarking SE(3)-based Generative Models for Protein Structure Design", "abstract": "SE(3)-based generative models have shown great promise in protein geometry modeling and effective structure design. However, the field currently lacks a modularized benchmark to enable comprehensive investigation and fair comparison of different methods. In this paper, we propose Protein-SE(3), a new benchmark based on a unified training framework, which comprises protein scaffolding tasks, integrated generative models, high-level mathematical abstraction, and diverse evaluation metrics. Recent advanced generative models designed for protein scaffolding, from multiple perspectives like DDPM (Genie1 and Genie2), Score Matching (FrameDiff and RfDiffusion) and Flow Matching (FoldFlow and FrameFlow) are integrated into our framework. All integrated methods are fairly investigated with the same training dataset and evaluation metrics. Furthermore, we provide a high-level abstraction of the mathematical foundations behind the generative models, enabling fast prototyping of future algorithms without reliance on explicit protein structures. Accordingly, we release the first comprehensive benchmark built upon unified training framework for SE(3)-based protein structure design, which is publicly accessible at https://github.com/BruthYU/protein-se3.", "date": "2025-07-27", "categories": []}
{"id": "2507.20644", "title": "Deep Generative Models of Evolution: SNP-level Population Adaptation by Genomic Linkage Incorporation", "abstract": "The investigation of allele frequency trajectories in populations evolving under controlled environmental pressures has become a popular approach to study evolutionary processes on the molecular level. Statistical models based on well-defined evolutionary concepts can be used to validate different hypotheses about empirical observations. Despite their popularity, classic statistical models like the Wright-Fisher model suffer from simplified assumptions such as the independence of selected loci along a chromosome and uncertainty about the parameters. Deep generative neural networks offer a powerful alternative known for the integration of multivariate dependencies and noise reduction. Due to their high data demands and challenging interpretability they have, so far, not been widely considered in the area of population genomics. To address the challenges in the area of Evolve and Resequencing experiments (E&R) based on pooled sequencing (Pool-Seq) data, we introduce a deep generative neural network that aims to model a concept of evolution based on empirical observations over time. The proposed model estimates the distribution of allele frequency trajectories by embedding the observations from single nucleotide polymorphisms (SNPs) with information from neighboring loci. Evaluation on simulated E&R experiments demonstrates the model's ability to capture the distribution of allele frequency trajectories and illustrates the representational power of deep generative models on the example of linkage disequilibrium (LD) estimation. Inspecting the internally learned representations enables estimating pairwise LD, which is typically inaccessible in Pool-Seq data. Our model provides competitive LD estimation in Pool-Seq data high degree of LD when compared to existing methods.", "date": "2025-07-28", "categories": []}
{"id": "2507.20925", "title": "Zero-Shot Learning with Subsequence Reordering Pretraining for Compound-Protein Interaction", "abstract": "Given the vastness of chemical space and the ongoing emergence of previously uncharacterized proteins, zero-shot compound-protein interaction (CPI) prediction better reflects the practical challenges and requirements of real-world drug development. Although existing methods perform adequately during certain CPI tasks, they still face the following challenges: (1) Representation learning from local or complete protein sequences often overlooks the complex interdependencies between subsequences, which are essential for predicting spatial structures and binding properties. (2) Dependence on large-scale or scarce multimodal protein datasets demands significant training data and computational resources, limiting scalability and efficiency. To address these challenges, we propose a novel approach that pretrains protein representations for CPI prediction tasks using subsequence reordering, explicitly capturing the dependencies between protein subsequences. Furthermore, we apply length-variable protein augmentation to ensure excellent pretraining performance on small training datasets. To evaluate the model's effectiveness and zero-shot learning ability, we combine it with various baseline methods. The results demonstrate that our approach can improve the baseline model's performance on the CPI task, especially in the challenging zero-shot scenario. Compared to existing pre-training models, our model demonstrates superior performance, particularly in data-scarce scenarios where training samples are limited. Our implementation is available at https://github.com/Hoch-Zhang/PSRP-CPI.", "date": "2025-07-28", "categories": []}
{"id": "2501.06158", "title": "GenMol: A Drug Discovery Generalist with Discrete Diffusion", "abstract": "Drug discovery is a complex process that involves multiple stages and tasks. However, existing molecular generative models can only tackle some of these tasks. We present Generalist Molecular generative model (GenMol), a versatile framework that uses only a single discrete diffusion model to handle diverse drug discovery scenarios. GenMol generates Sequential Attachment-based Fragment Embedding (SAFE) sequences through non-autoregressive bidirectional parallel decoding, thereby allowing the utilization of a molecular context that does not rely on the specific token ordering while having better sampling efficiency. GenMol uses fragments as basic building blocks for molecules and introduces fragment remasking, a strategy that optimizes molecules by regenerating masked fragments, enabling effective exploration of chemical space. We further propose molecular context guidance (MCG), a guidance method tailored for masked discrete diffusion of GenMol. GenMol significantly outperforms the previous GPT-based model in de novo generation and fragment-constrained generation, and achieves state-of-the-art performance in goal-directed hit generation and lead optimization. These results demonstrate that GenMol can tackle a wide range of drug discovery tasks, providing a unified and versatile approach for molecular design. Our code is available at https://github.com/NVIDIA-Digital-Bio/genmol.", "date": "2025-01-10", "categories": []}
{"id": "2507.11588", "title": "SToFM: a Multi-scale Foundation Model for Spatial Transcriptomics", "abstract": "Spatial Transcriptomics (ST) technologies provide biologists with rich insights into single-cell biology by preserving spatial context of cells. Building foundational models for ST can significantly enhance the analysis of vast and complex data sources, unlocking new perspectives on the intricacies of biological tissues. However, modeling ST data is inherently challenging due to the need to extract multi-scale information from tissue slices containing vast numbers of cells. This process requires integrating macro-scale tissue morphology, micro-scale cellular microenvironment, and gene-scale gene expression profile. To address this challenge, we propose SToFM, a multi-scale Spatial Transcriptomics Foundation Model. SToFM first performs multi-scale information extraction on each ST slice, to construct a set of ST sub-slices that aggregate macro-, micro- and gene-scale information. Then an SE(2) Transformer is used to obtain high-quality cell representations from the sub-slices. Additionally, we construct \\textbf{SToCorpus-88M}, the largest high-resolution spatial transcriptomics corpus for pretraining. SToFM achieves outstanding performance on a variety of downstream tasks, such as tissue region semantic segmentation and cell type annotation, demonstrating its comprehensive understanding of ST data through capturing and integrating multi-scale information.", "date": "2025-07-15", "categories": []}
{"id": "2507.16978", "title": "Fast and Scalable Gene Embedding Search: A Comparative Study of FAISS and ScaNN", "abstract": "The exponential growth of DNA sequencing data has outpaced traditional heuristic-based methods, which struggle to scale effectively. Efficient computational approaches are urgently needed to support large-scale similarity search, a foundational task in bioinformatics for detecting homology, functional similarity, and novelty among genomic and proteomic sequences. Although tools like BLAST have been widely used and remain effective in many scenarios, they suffer from limitations such as high computational cost and poor performance on divergent sequences.\n  In this work, we explore embedding-based similarity search methods that learn latent representations capturing deeper structural and functional patterns beyond raw sequence alignment. We systematically evaluate two state-of-the-art vector search libraries, FAISS and ScaNN, on biologically meaningful gene embeddings. Unlike prior studies, our analysis focuses on bioinformatics-specific embeddings and benchmarks their utility for detecting novel sequences, including those from uncharacterized taxa or genes lacking known homologs. Our results highlight both computational advantages (in memory and runtime efficiency) and improved retrieval quality, offering a promising alternative to traditional alignment-heavy tools.", "date": "2025-07-22", "categories": []}
{"id": "2507.14156", "title": "All-atom inverse protein folding through discrete flow matching", "abstract": "The recent breakthrough of AlphaFold3 in modeling complex biomolecular interactions, including those between proteins and ligands, nucleotides, or metal ions, creates new opportunities for protein design. In so-called inverse protein folding, the objective is to find a sequence of amino acids that adopts a target protein structure. Many inverse folding methods struggle to predict sequences for complexes that contain non-protein components, and perform poorly with complexes that adopt multiple structural states. To address these challenges, we present ADFLIP (All-atom Discrete FLow matching Inverse Protein folding), a generative model based on discrete flow-matching for designing protein sequences conditioned on all-atom structural contexts. ADFLIP progressively incorporates predicted amino acid side chains as structural context during sequence generation and enables the design of dynamic protein complexes through ensemble sampling across multiple structural states. Furthermore, ADFLIP implements training-free classifier guidance sampling, which allows the incorporation of arbitrary pre-trained models to optimise the designed sequence for desired protein properties. We evaluated the performance of ADFLIP on protein complexes with small-molecule ligands, nucleotides, or metal ions, including dynamic complexes for which structure ensembles were determined by nuclear magnetic resonance (NMR). Our model achieves state-of-the-art performance in single-structure and multi-structure inverse folding tasks, demonstrating excellent potential for all-atom protein design. The code is available at https://github.com/ykiiiiii/ADFLIP.", "date": "2025-07-04", "categories": []}
{"id": "2507.14639", "title": "KinForm: Kinetics Informed Feature Optimised Representation Models for Enzyme $k_{cat}$ and $K_{M}$ Prediction", "abstract": "Kinetic parameters such as the turnover number ($k_{cat}$) and Michaelis constant ($K_{\\mathrm{M}}$) are essential for modelling enzymatic activity but experimental data remains limited in scale and diversity. Previous methods for predicting enzyme kinetics typically use mean-pooled residue embeddings from a single protein language model to represent the protein. We present KinForm, a machine learning framework designed to improve predictive accuracy and generalisation for kinetic parameters by optimising protein feature representations. KinForm combines several residue-level embeddings (Evolutionary Scale Modeling Cambrian, Evolutionary Scale Modeling 2, and ProtT5-XL-UniRef50), taken from empirically selected intermediate transformer layers and applies weighted pooling based on per-residue binding-site probability. To counter the resulting high dimensionality, we apply dimensionality reduction using principal--component analysis (PCA) on concatenated protein features, and rebalance the training data via a similarity-based oversampling strategy. KinForm outperforms baseline methods on two benchmark datasets. Improvements are most pronounced in low sequence similarity bins. We observe improvements from binding-site probability pooling, intermediate-layer selection, PCA, and oversampling of low-identity proteins. We also find that removing sequence overlap between folds provides a more realistic evaluation of generalisation and should be the standard over random splitting when benchmarking kinetic prediction models.", "date": "2025-07-19", "categories": []}
{"id": "2507.15087", "title": "Evaluation of Coding Schemes for Transformer-based Gene Sequence Modeling", "abstract": "Currently, many studies view DNA sequences as a special type of language and utilize Transformers to model them. These studies use fixed-length k-mer segmentation and BPE subword tokenization but lack a systematic evaluation to determine which is superior. We compare k-mer segmentation with k=1,3,4,5,6, a 4,096-token BPE vocabulary, and three positional encoding methods-sinusoidal, AliBi, and RoPE. Each configuration is trained from scratch in 3, 6, 12, and 24-layer Transformer encoders and evaluated on GUE benchmark dataset. In general, BPE delivers higher and more stable performance across tasks by compressing frequent motifs into variable-length tokens, reducing sequence length, and improving model generalization. RoPE excels at capturing periodic motifs and extrapolating to long sequences, while AliBi also performs well on tasks driven by local dependencies. In terms of depth, we observe significant gains when increasing layers from 3 to 12, with only marginal improvements or slight overfitting at 24 layers. This study provides practical guidance for designing tokenization and positional encoding in DNA Transformer models.", "date": "2025-07-20", "categories": []}
{"id": "2507.09754", "title": "Explainable AI in Genomics: Transcription Factor Binding Site Prediction with Mixture of Experts", "abstract": "Transcription Factor Binding Site (TFBS) prediction is crucial for understanding gene regulation and various biological processes. This study introduces a novel Mixture of Experts (MoE) approach for TFBS prediction, integrating multiple pre-trained Convolutional Neural Network (CNN) models, each specializing in different TFBS patterns. We evaluate the performance of our MoE model against individual expert models on both in-distribution and out-of-distribution (OOD) datasets, using six randomly selected transcription factors (TFs) for OOD testing. Our results demonstrate that the MoE model achieves competitive or superior performance across diverse TF binding sites, particularly excelling in OOD scenarios. The Analysis of Variance (ANOVA) statistical test confirms the significance of these performance differences. Additionally, we introduce ShiftSmooth, a novel attribution mapping technique that provides more robust model interpretability by considering small shifts in input sequences. Through comprehensive explainability analysis, we show that ShiftSmooth offers superior attribution for motif discovery and localization compared to traditional Vanilla Gradient methods. Our work presents an efficient, generalizable, and interpretable solution for TFBS prediction, potentially enabling new discoveries in genome biology and advancing our understanding of transcriptional regulation.", "date": "2025-07-13", "categories": []}
{"id": "2507.13950", "title": "MoDyGAN: Combining Molecular Dynamics With GANs to Investigate Protein Conformational Space", "abstract": "Extensively exploring protein conformational landscapes remains a major challenge in computational biology due to the high computational cost involved in dynamic physics-based simulations. In this work, we propose a novel pipeline, MoDyGAN, that leverages molecular dynamics (MD) simulations and generative adversarial networks (GANs) to explore protein conformational spaces. MoDyGAN contains a generator that maps Gaussian distributions into MD-derived protein trajectories, and a refinement module that combines ensemble learning with a dual-discriminator to further improve the plausibility of generated conformations. Central to our approach is an innovative representation technique that reversibly transforms 3D protein structures into 2D matrices, enabling the use of advanced image-based GAN architectures. We use three rigid proteins to demonstrate that MoDyGAN can generate plausible new conformations. We also use deca-alanine as a case study to show that interpolations within the latent space closely align with trajectories obtained from steered molecular dynamics (SMD) simulations. Our results suggest that representing proteins as image-like data unlocks new possibilities for applying advanced deep learning techniques to biomolecular simulation, leading to an efficient sampling of conformational states. Additionally, the proposed framework holds strong potential for extension to other complex 3D structures.", "date": "2025-07-18", "categories": []}
{"id": "2507.11660", "title": "STAGED: A Multi-Agent Neural Network for Learning Cellular Interaction Dynamics", "abstract": "The advent of single-cell technology has significantly improved our understanding of cellular states and subpopulations in various tissues under normal and diseased conditions by employing data-driven approaches such as clustering and trajectory inference. However, these methods consider cells as independent data points of population distributions. With spatial transcriptomics, we can represent cellular organization, along with dynamic cell-cell interactions that lead to changes in cell state. Still, key computational advances are necessary to enable the data-driven learning of such complex interactive cellular dynamics. While agent-based modeling (ABM) provides a powerful framework, traditional approaches rely on handcrafted rules derived from domain knowledge rather than data-driven approaches. To address this, we introduce Spatio Temporal Agent-Based Graph Evolution Dynamics(STAGED) integrating ABM with deep learning to model intercellular communication, and its effect on the intracellular gene regulatory network. Using graph ODE networks (GDEs) with shared weights per cell type, our approach represents genes as vertices and interactions as directed edges, dynamically learning their strengths through a designed attention mechanism. Trained to match continuous trajectories of simulated as well as inferred trajectories from spatial transcriptomics data, the model captures both intercellular and intracellular interactions, enabling a more adaptive and accurate representation of cellular dynamics.", "date": "2025-07-15", "categories": []}
{"id": "2411.02572", "title": "ViTally Consistent: Scaling Biological Representation Learning for Cell Microscopy", "abstract": "Large-scale cell microscopy screens are used in drug discovery and molecular biology research to study the effects of millions of chemical and genetic perturbations on cells. To use these images in downstream analysis, we need models that can map each image into a feature space that represents diverse biological phenotypes consistently, in the sense that perturbations with similar biological effects have similar representations. In this work, we present the largest foundation model for cell microscopy data to date, a new 1.9 billion-parameter ViT-G/8 MAE trained on over 8 billion microscopy image crops. Compared to a previous published ViT-L/8 MAE, our new model achieves a 60% improvement in linear separability of genetic perturbations and obtains the best overall performance on whole-genome biological relationship recall and replicate consistency benchmarks. Beyond scaling, we developed two key methods that improve performance: (1) training on a curated and diverse dataset; and, (2) using biologically motivated linear probing tasks to search across each transformer block for the best candidate representation of whole-genome screens. We find that many self-supervised vision transformers, pretrained on either natural or microscopy images, yield significantly more biologically meaningful representations of microscopy images in their intermediate blocks than in their typically used final blocks. More broadly, our approach and results provide insights toward a general strategy for successfully building foundation models for large-scale biological data.", "date": "2024-11-04", "categories": []}
{"id": "2501.17965", "title": "Variational Combinatorial Sequential Monte Carlo for Bayesian Phylogenetics in Hyperbolic Space", "abstract": "Hyperbolic space naturally encodes hierarchical structures such as phylogenies (binary trees), where inward-bending geodesics reflect paths through least common ancestors, and the exponential growth of neighborhoods mirrors the super-exponential scaling of topologies. This scaling challenge limits the efficiency of Euclidean-based approximate inference methods. Motivated by the geometric connections between trees and hyperbolic space, we develop novel hyperbolic extensions of two sequential search algorithms: Combinatorial and Nested Combinatorial Sequential Monte Carlo (\\textsc{Csmc} and \\textsc{Ncsmc}). Our approach introduces consistent and unbiased estimators, along with variational inference methods (\\textsc{H-Vcsmc} and \\textsc{H-Vncsmc}), which outperform their Euclidean counterparts. Empirical results demonstrate improved speed, scalability and performance in high-dimensional phylogenetic inference tasks.", "date": "2025-01-29", "categories": []}
{"id": "2507.10502", "title": "Benchmarking and Evaluation of AI Models in Biology: Outcomes and Recommendations from the CZI Virtual Cells Workshop", "abstract": "Artificial intelligence holds immense promise for transforming biology, yet a lack of standardized, cross domain, benchmarks undermines our ability to build robust, trustworthy models. Here, we present insights from a recent workshop that convened machine learning and computational biology experts across imaging, transcriptomics, proteomics, and genomics to tackle this gap. We identify major technical and systemic bottlenecks such as data heterogeneity and noise, reproducibility challenges, biases, and the fragmented ecosystem of publicly available resources and propose a set of recommendations for building benchmarking frameworks that can efficiently compare ML models of biological systems across tasks and data modalities. By promoting high quality data curation, standardized tooling, comprehensive evaluation metrics, and open, collaborative platforms, we aim to accelerate the development of robust benchmarks for AI driven Virtual Cells. These benchmarks are crucial for ensuring rigor, reproducibility, and biological relevance, and will ultimately advance the field toward integrated models that drive new discoveries, therapeutic insights, and a deeper understanding of cellular systems.", "date": "2025-07-14", "categories": []}
{"id": "2507.11789", "title": "Enforcing Latent Euclidean Geometry in Single-Cell VAEs for Manifold Interpolation", "abstract": "Latent space interpolations are a powerful tool for navigating deep generative models in applied settings. An example is single-cell RNA sequencing, where existing methods model cellular state transitions as latent space interpolations with variational autoencoders, often assuming linear shifts and Euclidean geometry. However, unless explicitly enforced, linear interpolations in the latent space may not correspond to geodesic paths on the data manifold, limiting methods that assume Euclidean geometry in the data representations. We introduce FlatVI, a novel training framework that regularises the latent manifold of discrete-likelihood variational autoencoders towards Euclidean geometry, specifically tailored for modelling single-cell count data. By encouraging straight lines in the latent space to approximate geodesic interpolations on the decoded single-cell manifold, FlatVI enhances compatibility with downstream approaches that assume Euclidean latent geometry. Experiments on synthetic data support the theoretical soundness of our approach, while applications to time-resolved single-cell RNA sequencing data demonstrate improved trajectory reconstruction and manifold interpolation.", "date": "2025-07-15", "categories": []}
{"id": "2507.11839", "title": "Protenix-Mini: Efficient Structure Predictor via Compact Architecture, Few-Step Diffusion and Switchable pLM", "abstract": "Lightweight inference is critical for biomolecular structure prediction and other downstream tasks, enabling efficient real-world deployment and inference-time scaling for large-scale applications. In this work, we address the challenge of balancing model efficiency and prediction accuracy by making several key modifications, 1) Multi-step AF3 sampler is replaced by a few-step ODE sampler, significantly reducing computational overhead for the diffusion module part during inference; 2) In the open-source Protenix framework, a subset of pairformer or diffusion transformer blocks doesn't make contributions to the final structure prediction, presenting opportunities for architectural pruning and lightweight redesign; 3) A model incorporating an ESM module is trained to substitute the conventional MSA module, reducing MSA preprocessing time. Building on these key insights, we present Protenix-Mini, a compact and optimized model designed for efficient protein structure prediction. This streamlined version incorporates a more efficient architectural design with a two-step Ordinary Differential Equation (ODE) sampling strategy. By eliminating redundant Transformer components and refining the sampling process, Protenix-Mini significantly reduces model complexity with slight accuracy drop. Evaluations on benchmark datasets demonstrate that it achieves high-fidelity predictions, with only a negligible 1 to 5 percent decrease in performance on benchmark datasets compared to its full-scale counterpart. This makes Protenix-Mini an ideal choice for applications where computational resources are limited but accurate structure prediction remains crucial.", "date": "2025-07-16", "categories": []}
{"id": "2507.11950", "title": "RNAMunin: A Deep Machine Learning Model for Non-coding RNA Discovery", "abstract": "Functional annotation of microbial genomes is often biased toward protein-coding genes, leaving a vast, unexplored landscape of non-coding RNAs (ncRNAs) that are critical for regulating bacterial and archaeal physiology, stress response and metabolism. Identifying ncRNAs directly from genomic sequence is a paramount challenge in bioinformatics and biology, essential for understanding the complete regulatory potential of an organism. This paper presents RNAMunin, a machine learning (ML) model that is capable of finding ncRNAs using genomic sequence alone. It is also computationally viable for large sequence datasets such as long read metagenomic assemblies with contigs totaling multiple Gbp. RNAMunin is trained on Rfam sequences extracted from approximately 60 Gbp of long read metagenomes from 16 San Francisco Estuary samples. We know of no other model that can detect ncRNAs based solely on genomic sequence at this scale. Since RNAMunin only requires genomic sequence as input, we do not need for an ncRNA to be transcribed to find it, i.e., we do not need transcriptomics data. We wrote this manuscript in a narrative style in order to best convey how RNAMunin was developed and how it works in detail. Unlike almost all current ML models, at approximately 1M parameters, RNAMunin is very small and very fast.", "date": "2025-07-16", "categories": []}
{"id": "2507.10877", "title": "BioScore: A Foundational Scoring Function For Diverse Biomolecular Complexes", "abstract": "Structural assessment of biomolecular complexes is vital for translating molecular models into functional insights, shaping our understanding of biology and aiding drug discovery. However, current structure-based scoring functions often lack generalizability across diverse biomolecular systems. We present BioScore, a foundational scoring function that addresses key challenges -- data sparsity, cross-system representation, and task compatibility -- through a dual-scale geometric graph learning framework with tailored modules for structure assessment and affinity prediction. BioScore supports a wide range of tasks, including affinity prediction, conformation ranking, and structure-based virtual screening. Evaluated on 16 benchmarks spanning proteins, nucleic acids, small molecules, and carbohydrates, BioScore consistently outperforms or matches 70 traditional and deep learning methods. Our newly proposed PPI Benchmark further enables comprehensive evaluation of protein-protein complex scoring. BioScore demonstrates broad applicability: (1) pretraining on mixed-structure data boosts protein-protein affinity prediction by up to 40% and antigen-antibody binding correlation by over 90%; (2) cross-system generalizability enables zero- and few-shot prediction with up to 71% correlation gain; and (3) its unified representation captures chemically challenging systems such as cyclic peptides, improving affinity prediction by over 60%. BioScore establishes a robust and generalizable framework for structural assessment across complex biomolecular landscapes.", "date": "2025-07-15", "categories": []}
{"id": "2507.10923", "title": "Enhancing Safe and Controllable Protein Generation via Knowledge Preference Optimization", "abstract": "Protein language models have emerged as powerful tools for sequence generation, offering substantial advantages in functional optimization and denovo design. However, these models also present significant risks of generating harmful protein sequences, such as those that enhance viral transmissibility or evade immune responses. These concerns underscore critical biosafety and ethical challenges. To address these issues, we propose a Knowledge-guided Preference Optimization (KPO) framework that integrates prior knowledge via a Protein Safety Knowledge Graph. This framework utilizes an efficient graph pruning strategy to identify preferred sequences and employs reinforcement learning to minimize the risk of generating harmful proteins. Experimental results demonstrate that KPO effectively reduces the likelihood of producing hazardous sequences while maintaining high functionality, offering a robust safety assurance framework for applying generative models in biotechnology.", "date": "2025-07-15", "categories": []}
{"id": "2403.00043", "title": "RiNALMo: General-Purpose RNA Language Models Can Generalize Well on Structure Prediction Tasks", "abstract": "While RNA has recently been recognized as an interesting small-molecule drug target, many challenges remain to be addressed before we take full advantage of it. This emphasizes the necessity to improve our understanding of its structures and functions. Over the years, sequencing technologies have produced an enormous amount of unlabeled RNA data, which hides a huge potential. Motivated by the successes of protein language models, we introduce RiboNucleic Acid Language Model (RiNALMo) to unveil the hidden code of RNA. RiNALMo is the largest RNA language model to date, with 650M parameters pre-trained on 36M non-coding RNA sequences from several databases. It can extract hidden knowledge and capture the underlying structure information implicitly embedded within the RNA sequences. RiNALMo achieves state-of-the-art results on several downstream tasks. Notably, we show that its generalization capabilities overcome the inability of other deep learning methods for secondary structure prediction to generalize on unseen RNA families.", "date": "2024-02-29", "categories": []}
{"id": "2507.04225", "title": "Zero-Shot Cyclic Peptide Design via Composable Geometric Constraints", "abstract": "Cyclic peptides, characterized by geometric constraints absent in linear peptides, offer enhanced biochemical properties, presenting new opportunities to address unmet medical needs. However, designing target-specific cyclic peptides remains underexplored due to limited training data. To bridge the gap, we propose CP-Composer, a novel generative framework that enables zero-shot cyclic peptide generation via composable geometric constraints. Our approach decomposes complex cyclization patterns into unit constraints, which are incorporated into a diffusion model through geometric conditioning on nodes and edges. During training, the model learns from unit constraints and their random combinations in linear peptides, while at inference, novel constraint combinations required for cyclization are imposed as input. Experiments show that our model, despite trained with linear peptides, is capable of generating diverse target-binding cyclic peptides, reaching success rates from 38% to 84% on different cyclization strategies.", "date": "2025-07-06", "categories": []}
{"id": "2507.08966", "title": "ToxBench: A Binding Affinity Prediction Benchmark with AB-FEP-Calculated Labels for Human Estrogen Receptor Alpha", "abstract": "Protein-ligand binding affinity prediction is essential for drug discovery and toxicity assessment. While machine learning (ML) promises fast and accurate predictions, its progress is constrained by the availability of reliable data. In contrast, physics-based methods such as absolute binding free energy perturbation (AB-FEP) deliver high accuracy but are computationally prohibitive for high-throughput applications. To bridge this gap, we introduce ToxBench, the first large-scale AB-FEP dataset designed for ML development and focused on a single pharmaceutically critical target, Human Estrogen Receptor Alpha (ER$\\alpha$). ToxBench contains 8,770 ER$\\alpha$-ligand complex structures with binding free energies computed via AB-FEP with a subset validated against experimental affinities at 1.75 kcal/mol RMSE, along with non-overlapping ligand splits to assess model generalizability. Using ToxBench, we further benchmark state-of-the-art ML methods, and notably, our proposed DualBind model, which employs a dual-loss framework to effectively learn the binding energy function. The benchmark results demonstrate the superior performance of DualBind and the potential of ML to approximate AB-FEP at a fraction of the computational cost.", "date": "2025-07-11", "categories": []}
{"id": "2507.09054", "title": "Conformation-Aware Structure Prediction of Antigen-Recognizing Immune Proteins", "abstract": "We introduce Ibex, a pan-immunoglobulin structure prediction model that achieves state-of-the-art accuracy in modeling the variable domains of antibodies, nanobodies, and T-cell receptors. Unlike previous approaches, Ibex explicitly distinguishes between bound and unbound protein conformations by training on labeled apo and holo structural pairs, enabling accurate prediction of both states at inference time. Using a comprehensive private dataset of high-resolution antibody structures, we demonstrate superior out-of-distribution performance compared to existing specialized and general protein structure prediction tools. Ibex combines the accuracy of cutting-edge models with significantly reduced computational requirements, providing a robust foundation for accelerating large molecule design and therapeutic development.", "date": "2025-07-11", "categories": []}
{"id": "2507.08542", "title": "CircFormerMoE: An End-to-End Deep Learning Framework for Circular RNA Splice Site Detection and Pairing in Plant Genomes", "abstract": "Circular RNAs (circRNAs) are important components of the non-coding RNA regulatory network. Previous circRNA identification primarily relies on high-throughput RNA sequencing (RNA-seq) data combined with alignment-based algorithms that detect back-splicing signals. However, these methods face several limitations: they can't predict circRNAs directly from genomic DNA sequences and relies heavily on RNA experimental data; they involve high computational costs due to complex alignment and filtering steps; and they are inefficient for large-scale or genome-wide circRNA prediction. The challenge is even greater in plants, where plant circRNA splice sites often lack the canonical GT-AG motif seen in human mRNA splicing, and no efficient deep learning model with strong generalization capability currently exists. Furthermore, the number of currently identified plant circRNAs is likely far lower than their true abundance. In this paper, we propose a deep learning framework named CircFormerMoE based on transformers and mixture-of experts for predicting circRNAs directly from plant genomic DNA. Our framework consists of two subtasks known as splicing site detection (SSD) and splicing site pairing (SSP). The model's effectiveness has been validated on gene data of 10 plant species. Trained on known circRNA instances, it is also capable of discovering previously unannotated circRNAs. In addition, we performed interpretability analyses on the trained model to investigate the sequence patterns contributing to its predictions. Our framework provides a fast and accurate computational method and tool for large-scale circRNA discovery in plants, laying a foundation for future research in plant functional genomics and non-coding RNA annotation.", "date": "2025-07-11", "categories": []}
{"id": "2507.07663", "title": "MolCLIP: A Molecular-Auxiliary CLIP Framework for Identifying Drug Mechanism of Action Based on Time-Lapsed Mitochondrial Images", "abstract": "Drug Mechanism of Action (MoA) mainly investigates how drug molecules interact with cells, which is crucial for drug discovery and clinical application. Recently, deep learning models have been used to recognize MoA by relying on high-content and fluorescence images of cells exposed to various drugs. However, these methods focus on spatial characteristics while overlooking the temporal dynamics of live cells. Time-lapse imaging is more suitable for observing the cell response to drugs. Additionally, drug molecules can trigger cellular dynamic variations related to specific MoA. This indicates that the drug molecule modality may complement the image counterpart. This paper proposes MolCLIP, the first visual language model to combine microscopic cell video- and molecule-modalities. MolCLIP designs a molecule-auxiliary CLIP framework to guide video features in learning the distribution of the molecular latent space. Furthermore, we integrate a metric learning strategy with MolCLIP to optimize the aggregation of video features. Experimental results on the MitoDataset demonstrate that MolCLIP achieves improvements of 51.2% and 20.5% in mAP for drug identification and MoA recognition, respectively.", "date": "2025-07-10", "categories": []}
{"id": "2507.07862", "title": "Predicting and generating antibiotics against future pathogens with ApexOracle", "abstract": "Antimicrobial resistance (AMR) is escalating and outpacing current antibiotic development. Thus, discovering antibiotics effective against emerging pathogens is becoming increasingly critical. However, existing approaches cannot rapidly identify effective molecules against novel pathogens or emerging drug-resistant strains. Here, we introduce ApexOracle, an artificial intelligence (AI) model that both predicts the antibacterial potency of existing compounds and designs de novo molecules active against strains it has never encountered. Departing from models that rely solely on molecular features, ApexOracle incorporates pathogen-specific context through the integration of molecular features captured via a foundational discrete diffusion language model and a dual-embedding framework that combines genomic- and literature-derived strain representations. Across diverse bacterial species and chemical modalities, ApexOracle consistently outperformed state-of-the-art approaches in activity prediction and demonstrated reliable transferability to novel pathogens with little or no antimicrobial data. Its unified representation-generation architecture further enables the in silico creation of \"new-to-nature\" molecules with high predicted efficacy against priority threats. By pairing rapid activity prediction with targeted molecular generation, ApexOracle offers a scalable strategy for countering AMR and preparing for future infectious-disease outbreaks.", "date": "2025-07-10", "categories": []}
{"id": "2507.07887", "title": "Automating MD simulations for Proteins using Large language Models: NAMD-Agent", "abstract": "Molecular dynamics simulations are an essential tool in understanding protein structure, dynamics, and function at the atomic level. However, preparing high quality input files for MD simulations can be a time consuming and error prone process. In this work, we introduce an automated pipeline that leverages Large Language Models (LLMs), specifically Gemini 2.0 Flash, in conjunction with python scripting and Selenium based web automation to streamline the generation of MD input files. The pipeline exploits CHARMM GUI's comprehensive web-based interface for preparing simulation-ready inputs for NAMD. By integrating Gemini's code generation and iterative refinement capabilities, simulation scripts are automatically written, executed, and revised to navigate CHARMM GUI, extract appropriate parameters, and produce the required NAMD input files. Post processing is performed using additional software to further refine the simulation outputs, thereby enabling a complete and largely hands free workflow. Our results demonstrate that this approach reduces setup time, minimizes manual errors, and offers a scalable solution for handling multiple protein systems in parallel. This automated framework paves the way for broader application of LLMs in computational structural biology, offering a robust and adaptable platform for future developments in simulation automation.", "date": "2025-07-10", "categories": []}
{"id": "2503.09312", "title": "Terrier: A Deep Learning Repeat Classifier", "abstract": "Repetitive DNA sequences underpin genome architecture and evolutionary processes, yet they remain challenging to classify accurately. Terrier is a deep learning model designed to overcome these challenges by classifying repetitive DNA sequences using a publicly available, curated repeat sequence library trained under the RepeatMasker schema. Poor representation of taxa within repeat databases often limits the classification accuracy and reproducibility of current repeat annotation methods, limiting our understanding of repeat evolution and function. Terrier overcomes these challenges by leveraging deep learning for improved accuracy. Trained on Repbase, which includes over 100,000 repeat families -- four times more than Dfam -- Terrier maps 97.1% of Repbase sequences to RepeatMasker categories, offering the most comprehensive classification system available. When benchmarked against DeepTE, TERL, and TEclass2 in model organisms (rice, fruit flies, humans, and mice), Terrier achieved superior accuracy while classifying a broader range of sequences. Further validation in non-model amphibian, flatworm and Northern krill genomes highlights its effectiveness in improving classification in non-model species, facilitating research on repeat-driven evolution, genomic instability, and phenotypic variation.", "date": "2025-03-12", "categories": []}
{"id": "2507.06366", "title": "DecoyDB: A Dataset for Graph Contrastive Learning in Protein-Ligand Binding Affinity Prediction", "abstract": "Predicting the binding affinity of protein-ligand complexes plays a vital role in drug discovery. Unfortunately, progress has been hindered by the lack of large-scale and high-quality binding affinity labels. The widely used PDBbind dataset has fewer than 20K labeled complexes. Self-supervised learning, especially graph contrastive learning (GCL), provides a unique opportunity to break the barrier by pre-training graph neural network models based on vast unlabeled complexes and fine-tuning the models on much fewer labeled complexes. However, the problem faces unique challenges, including a lack of a comprehensive unlabeled dataset with well-defined positive/negative complex pairs and the need to design GCL algorithms that incorporate the unique characteristics of such data. To fill the gap, we propose DecoyDB, a large-scale, structure-aware dataset specifically designed for self-supervised GCL on protein-ligand complexes. DecoyDB consists of high-resolution ground truth complexes (less than 2.5 Angstrom) and diverse decoy structures with computationally generated binding poses that range from realistic to suboptimal (negative pairs). Each decoy is annotated with a Root Mean Squared Deviation (RMSD) from the native pose. We further design a customized GCL framework to pre-train graph neural networks based on DecoyDB and fine-tune the models with labels from PDBbind. Extensive experiments confirm that models pre-trained with DecoyDB achieve superior accuracy, label efficiency, and generalizability.", "date": "2025-07-08", "categories": []}
{"id": "2507.06418", "title": "PAST: A multimodal single-cell foundation model for histopathology and spatial transcriptomics in cancer", "abstract": "While pathology foundation models have transformed cancer image analysis, they often lack integration with molecular data at single-cell resolution, limiting their utility for precision oncology. Here, we present PAST, a pan-cancer single-cell foundation model trained on 20 million paired histopathology images and single-cell transcriptomes spanning multiple tumor types and tissue contexts. By jointly encoding cellular morphology and gene expression, PAST learns unified cross-modal representations that capture both spatial and molecular heterogeneity at the cellular level. This approach enables accurate prediction of single-cell gene expression, virtual molecular staining, and multimodal survival analysis directly from routine pathology slides. Across diverse cancers and downstream tasks, PAST consistently exceeds the performance of existing approaches, demonstrating robust generalizability and scalability. Our work establishes a new paradigm for pathology foundation models, providing a versatile tool for high-resolution spatial omics, mechanistic discovery, and precision cancer research.", "date": "2025-07-08", "categories": []}
{"id": "2507.05265", "title": "BMFM-DNA: A SNP-aware DNA foundation model to capture variant effects", "abstract": "Large language models (LLMs) trained on text demonstrated remarkable results on natural language processing (NLP) tasks. These models have been adapted to decipher the language of DNA, where sequences of nucleotides act as \"words\" that encode genomic functions. However, the genome differs fundamentally from natural language, as it lacks clearly defined words or a consistent grammar. Although DNA language models (DNALMs) such as DNABERT, GENA-LM have achieved high level of performance on genome-related biological tasks, these models do not encode biological functions in the presence of sequence variations. To address this problem, we pre-train foundation models that effectively integrate sequence variations, in particular Single Nucleotide Polymorphisms (SNPs), as they underlie important biological functions. Specifically, we use ModernBERT to pre-train two different Biomedical Foundation Models (BMFM), namely, BMFM-DNA-REF in which the model is trained with sequences of varying lengths along with their reverse complements derived from the reference genome and BMFM-DNA-SNP in which the model is trained with sequences created using a novel representation scheme that encodes sequence variations. Our findings indicate that integrating sequence variations into DNALMs helps capture the biological functions as seen in improvements on all fine-tuning tasks. To explore the model's practical utility, we experimented with various strategies for SNP imputation on promoter detection task introduced in DNABERT-2. However, we acknowledge that the current benchmarks are limited in their ability to fully evaluate these models. To enable more comprehensive assessment in the future and encourage community contributions, we release our models through HuggingFace and the code to reproduce the results at https://github.com/BiomedSciAI/biomed-multi-omic", "date": "2025-06-26", "categories": []}
{"id": "2507.05502", "title": "Predicting mutational effects on protein binding from folding energy", "abstract": "Accurate estimation of mutational effects on protein-protein binding energies is an open problem with applications in structural biology and therapeutic design. Several deep learning predictors for this task have been proposed, but, presumably due to the scarcity of binding data, these methods underperform computationally expensive estimates based on empirical force fields. In response, we propose a transfer-learning approach that leverages advances in protein sequence modeling and folding stability prediction for this task. The key idea is to parameterize the binding energy as the difference between the folding energy of the protein complex and the sum of the folding energies of its binding partners. We show that using a pre-trained inverse-folding model as a proxy for folding energy provides strong zero-shot performance, and can be fine-tuned with (1) copious folding energy measurements and (2) more limited binding energy measurements. The resulting predictor, StaB-ddG, is the first deep learning predictor to match the accuracy of the state-of-the-art empirical force-field method FoldX, while offering an over 1,000x speed-up.", "date": "2025-07-07", "categories": []}
{"id": "2507.02902", "title": "Controllable diffusion-based generation for multi-channel biological data", "abstract": "Spatial profiling technologies in biology, such as imaging mass cytometry (IMC) and spatial transcriptomics (ST), generate high-dimensional, multi-channel data with strong spatial alignment and complex inter-channel relationships. Generative modeling of such data requires jointly capturing intra- and inter-channel structure, while also generalizing across arbitrary combinations of observed and missing channels for practical application. Existing diffusion-based models generally assume low-dimensional inputs (e.g., RGB images) and rely on simple conditioning mechanisms that break spatial correspondence and ignore inter-channel dependencies. This work proposes a unified diffusion framework for controllable generation over structured and spatial biological data. Our model contains two key innovations: (1) a hierarchical feature injection mechanism that enables multi-resolution conditioning on spatially aligned channels, and (2) a combination of latent-space and output-space channel-wise attention to capture inter-channel relationships. To support flexible conditioning and generalization to arbitrary subsets of observed channels, we train the model using a random masking strategy, enabling it to reconstruct missing channels from any combination of inputs. We demonstrate state-of-the-art performance across both spatial and non-spatial prediction tasks, including protein imputation in IMC and gene-to-protein prediction in single-cell datasets, and show strong generalization to unseen conditional configurations.", "date": "2025-06-24", "categories": []}
{"id": "2507.02980", "title": "Modeling Gene Expression Distributional Shifts for Unseen Genetic Perturbations", "abstract": "We train a neural network to predict distributional responses in gene expression following genetic perturbations. This is an essential task in early-stage drug discovery, where such responses can offer insights into gene function and inform target identification. Existing methods only predict changes in the mean expression, overlooking stochasticity inherent in single-cell data. In contrast, we offer a more realistic view of cellular responses by modeling expression distributions. Our model predicts gene-level histograms conditioned on perturbations and outperforms baselines in capturing higher-order statistics, such as variance, skewness, and kurtosis, at a fraction of the training cost. To generalize to unseen perturbations, we incorporate prior knowledge via gene embeddings from large language models (LLMs). While modeling a richer output space, the method remains competitive in predicting mean expression changes. This work offers a practical step towards more expressive and biologically informative models of perturbation effects.", "date": "2025-07-01", "categories": []}
{"id": "2507.03656", "title": "MLASDO: a software tool to detect and explain clinical and omics inconsistencies applied to the Parkinson's Progression Markers Initiative cohort", "abstract": "Inconsistencies between clinical and omics data may arise within medical cohorts. The identification, annotation and explanation of anomalous omics-based patients or individuals may become crucial to better reshape the disease, e.g., by detecting early onsets signaled by the omics and undetectable from observable symptoms. Here, we developed MLASDO (Machine Learning based Anomalous Sample Detection on Omics), a new method and software tool to identify, characterize and automatically describe anomalous samples based on omics data. Its workflow is based on three steps: (1) classification of healthy and cases individuals using a support vector machine algorithm; (2) detection of anomalous samples within groups; (3) explanation of anomalous individuals based on clinical data and expert knowledge. We showcase MLASDO using transcriptomics data of 317 healthy controls (HC) and 465 Parkinson's disease (PD) cases from the Parkinson's Progression Markers Initiative. In this cohort, MLASDO detected 15 anomalous HC with a PD-like transcriptomic signature and PD-like clinical features, including a lower proportion of CD4/CD8 naive T-cells and CD4 memory T-cells compared to HC (P<3.5*10^-3). MLASDO also identified 22 anomalous PD cases with a transcriptomic signature more similar to that of HC and some clinical features more similar to HC, including a lower proportion of mature neutrophils compared to PD cases (P<6*10^-3). In summary, MLASDO is a powerful tool that can help the clinician to detect and explain anomalous HC and cases of interest to be followed up. MLASDO is an open-source R package available at: https://github.com/JoseAdrian3/MLASDO.", "date": "2025-07-04", "categories": []}
{"id": "2507.00884", "title": "A Scalable and Quantum-Accurate Foundation Model for Biomolecular Force Field via Linearly Tensorized Quadrangle Attention", "abstract": "Accurate atomistic biomolecular simulations are vital for disease mechanism understanding, drug discovery, and biomaterial design, but existing simulation methods exhibit significant limitations. Classical force fields are efficient but lack accuracy for transition states and fine conformational details critical in many chemical and biological processes. Quantum Mechanics (QM) methods are highly accurate but computationally infeasible for large-scale or long-time simulations. AI-based force fields (AIFFs) aim to achieve QM-level accuracy with efficiency but struggle to balance many-body modeling complexity, accuracy, and speed, often constrained by limited training data and insufficient validation for generalizability. To overcome these challenges, we introduce LiTEN, a novel equivariant neural network with Tensorized Quadrangle Attention (TQA). TQA efficiently models three- and four-body interactions with linear complexity by reparameterizing high-order tensor features via vector operations, avoiding costly spherical harmonics. Building on LiTEN, LiTEN-FF is a robust AIFF foundation model, pre-trained on the extensive nablaDFT dataset for broad chemical generalization and fine-tuned on SPICE for accurate solvated system simulations. LiTEN achieves state-of-the-art (SOTA) performance across most evaluation subsets of rMD17, MD22, and Chignolin, outperforming leading models such as MACE, NequIP, and EquiFormer. LiTEN-FF enables the most comprehensive suite of downstream biomolecular modeling tasks to date, including QM-level conformer searches, geometry optimization, and free energy surface construction, while offering 10x faster inference than MACE-OFF for large biomolecules (~1000 atoms). In summary, we present a physically grounded, highly efficient framework that advances complex biomolecular modeling, providing a versatile foundation for drug discovery and related applications.", "date": "2025-07-01", "categories": []}
{"id": "2507.02345", "title": "HelixDesign-Antibody: A Scalable Production-Grade Platform for Antibody Design Built on HelixFold3", "abstract": "Antibody engineering is essential for developing therapeutics and advancing biomedical research. Traditional discovery methods often rely on time-consuming and resource-intensive experimental screening. To enhance and streamline this process, we introduce a production-grade, high-throughput platform built on HelixFold3, HelixDesign-Antibody, which utilizes the high-accuracy structure prediction model, HelixFold3. The platform facilitates the large-scale generation of antibody candidate sequences and evaluates their interaction with antigens. Integrated high-performance computing (HPC) support enables high-throughput screening, addressing challenges such as fragmented toolchains and high computational demands. Validation on multiple antigens showcases the platform's ability to generate diverse and high-quality antibodies, confirming a scaling law where exploring larger sequence spaces increases the likelihood of identifying optimal binders. This platform provides a seamless, accessible solution for large-scale antibody design and is available via the antibody design page of PaddleHelix platform.", "date": "2025-07-03", "categories": []}
{"id": "2507.02624", "title": "A Matrix Variational Auto-Encoder for Variant Effect Prediction in Pharmacogenes", "abstract": "Variant effect predictors (VEPs) aim to assess the functional impact of protein variants, traditionally relying on multiple sequence alignments (MSAs). This approach assumes that naturally occurring variants are fit, an assumption challenged by pharmacogenomics, where some pharmacogenes experience low evolutionary pressure. Deep mutational scanning (DMS) datasets provide an alternative by offering quantitative fitness scores for variants. In this work, we propose a transformer-based matrix variational auto-encoder (matVAE) with a structured prior and evaluate its performance on 33 DMS datasets corresponding to 26 drug target and ADME proteins from the ProteinGym benchmark. Our model trained on MSAs (matVAE-MSA) outperforms the state-of-the-art DeepSequence model in zero-shot prediction on DMS datasets, despite using an order of magnitude fewer parameters and requiring less computation at inference time. We also compare matVAE-MSA to matENC-DMS, a model of similar capacity trained on DMS data, and find that the latter performs better on supervised prediction tasks. Additionally, incorporating AlphaFold-generated structures into our transformer model further improves performance, achieving results comparable to DeepSequence trained on MSAs and finetuned on DMS. These findings highlight the potential of DMS datasets to replace MSAs without significant loss in predictive performance, motivating further development of DMS datasets and exploration of their relationships to enhance variant effect prediction.", "date": "2025-07-03", "categories": []}
{"id": "2507.02670", "title": "Guided Generation for Developable Antibodies", "abstract": "Therapeutic antibodies require not only high-affinity target engagement, but also favorable manufacturability, stability, and safety profiles for clinical effectiveness. These properties are collectively called `developability'. To enable a computational framework for optimizing antibody sequences for favorable developability, we introduce a guided discrete diffusion model trained on natural paired heavy- and light-chain sequences from the Observed Antibody Space (OAS) and quantitative developability measurements for 246 clinical-stage antibodies. To steer generation toward biophysically viable candidates, we integrate a Soft Value-based Decoding in Diffusion (SVDD) Module that biases sampling without compromising naturalness. In unconstrained sampling, our model reproduces global features of both the natural repertoire and approved therapeutics, and under SVDD guidance we achieve significant enrichment in predicted developability scores over unguided baselines. When combined with high-throughput developability assays, this framework enables an iterative, ML-driven pipeline for designing antibodies that satisfy binding and biophysical criteria in tandem.", "date": "2025-07-03", "categories": []}
{"id": "2507.01163", "title": "cp_measure: API-first feature extraction for image-based profiling workflows", "abstract": "Biological image analysis has traditionally focused on measuring specific visual properties of interest for cells or other entities. A complementary paradigm gaining increasing traction is image-based profiling - quantifying many distinct visual features to form comprehensive profiles which may reveal hidden patterns in cellular states, drug responses, and disease mechanisms. While current tools like CellProfiler can generate these feature sets, they pose significant barriers to automated and reproducible analyses, hindering machine learning workflows. Here we introduce cp_measure, a Python library that extracts CellProfiler's core measurement capabilities into a modular, API-first tool designed for programmatic feature extraction. We demonstrate that cp_measure features retain high fidelity with CellProfiler features while enabling seamless integration with the scientific Python ecosystem. Through applications to 3D astrocyte imaging and spatial transcriptomics, we showcase how cp_measure enables reproducible, automated image-based profiling pipelines that scale effectively for machine learning applications in computational biology.", "date": "2025-07-01", "categories": []}
{"id": "2410.11281", "title": "DynaCLR: Contrastive Learning of Cellular Dynamics with Temporal Regularization", "abstract": "We report DynaCLR, a self-supervised method for embedding cell and organelle Dynamics via Contrastive Learning of Representations of time-lapse images. DynaCLR integrates single-cell tracking and time-aware contrastive sampling to learn robust, temporally regularized representations of cell dynamics. DynaCLR embeddings generalize effectively to in-distribution and out-of-distribution datasets, and can be used for several downstream tasks with sparse human annotations. We demonstrate efficient annotations of cell states with a human-in-the-loop using fluorescence and label-free imaging channels. DynaCLR method enables diverse downstream biological analyses: classification of cell division and infection, clustering heterogeneous cell migration patterns, cross-modal distillation of cell states from fluorescence to label-free channel, alignment of asynchronous cellular responses and broken cell tracks, and discovering organelle response due to infection. DynaCLR is a flexible method for comparative analyses of dynamic cellular responses to pharmacological, microbial, and genetic perturbations. We provide PyTorch-based implementations of the model training and inference pipeline (https://github.com/mehta-lab/viscy) and a GUI (https://github.com/czbiohub-sf/napari-iohub) for the visualization and annotation of trajectories of cells in the real space and the embedding space.", "date": "2024-10-15", "categories": []}
{"id": "2411.19906", "title": "A Graph-Based Classical and Quantum Approach to Deterministic L-System Inference", "abstract": "L-systems can be made to model and create simulations of many biological processes, such as plant development. Finding an L-system for a given process is typically solved by hand, by experts, in a massively time-consuming process. It would be significant if this could be done automatically from data, such as from sequences of images. In this paper, we are interested in inferring a particular type of L-system, deterministic context-free L-system (D0L-system) from a sequence of strings. We introduce the characteristic graph of a sequence of strings, which we then utilize to translate our problem (inferring D0L-systems) in polynomial time into the maximum independent set problem (MIS) and the SAT problem. After that, we offer a classical exact algorithm and an approximate quantum algorithm for the problem.", "date": "2024-11-29", "categories": []}
{"id": "2507.00087", "title": "pUniFind: a unified large pre-trained deep learning model pushing the limit of mass spectra interpretation", "abstract": "Deep learning has advanced mass spectrometry data interpretation, yet most models remain feature extractors rather than unified scoring frameworks. We present pUniFind, the first large-scale multimodal pre-trained model in proteomics that integrates end-to-end peptide-spectrum scoring with open, zero-shot de novo sequencing. Trained on over 100 million open search-derived spectra, pUniFind aligns spectral and peptide modalities via cross modality prediction and outperforms traditional engines across diverse datasets, particularly achieving a 42.6 percent increase in the number of identified peptides in immunopeptidomics. Supporting over 1,300 modifications, pUniFind identifies 60 percent more PSMs than existing de novo methods despite a 300-fold larger search space. A deep learning based quality control module further recovers 38.5 percent additional peptides including 1,891 mapped to the genome but absent from reference proteomes while preserving full fragment ion coverage. These results establish a unified, scalable deep learning framework for proteomic analysis, offering improved sensitivity, modification coverage, and interpretability.", "date": "2025-06-30", "categories": []}
{"id": "2506.19598", "title": "Training Flexible Models of Genetic Variant Effects from Functional Annotations using Accelerated Linear Algebra", "abstract": "To understand how genetic variants in human genomes manifest in phenotypes -- traits like height or diseases like asthma -- geneticists have sequenced and measured hundreds of thousands of individuals. Geneticists use this data to build models that predict how a genetic variant impacts phenotype given genomic features of the variant, like DNA accessibility or the presence of nearby DNA-bound proteins. As more data and features become available, one might expect predictive models to improve. Unfortunately, training these models is bottlenecked by the need to solve expensive linear algebra problems because variants in the genome are correlated with nearby variants, requiring inversion of large matrices. Previous methods have therefore been restricted to fitting small models, and fitting simplified summary statistics, rather than the full likelihood of the statistical model. In this paper, we leverage modern fast linear algebra techniques to develop DeepWAS (Deep genome Wide Association Studies), a method to train large and flexible neural network predictive models to optimize likelihood. Notably, we find that larger models only improve performance when using our full likelihood approach; when trained by fitting traditional summary statistics, larger models perform no better than small ones. We find larger models trained on more features make better predictions, potentially improving disease predictions and therapeutic target identification.", "date": "2025-06-24", "categories": []}
{"id": "2506.22963", "title": "CN-SBM: Categorical Block Modelling For Primary and Residual Copy Number Variation", "abstract": "Cancer is a genetic disorder whose clonal evolution can be monitored by tracking noisy genome-wide copy number variants. We introduce the Copy Number Stochastic Block Model (CN-SBM), a probabilistic framework that jointly clusters samples and genomic regions based on discrete copy number states using a bipartite categorical block model. Unlike models relying on Gaussian or Poisson assumptions, CN-SBM respects the discrete nature of CNV calls and captures subpopulation-specific patterns through block-wise structure. Using a two-stage approach, CN-SBM decomposes CNV data into primary and residual components, enabling detection of both large-scale chromosomal alterations and finer aberrations. We derive a scalable variational inference algorithm for application to large cohorts and high-resolution data. Benchmarks on simulated and real datasets show improved model fit over existing methods. Applied to TCGA low-grade glioma data, CN-SBM reveals clinically relevant subtypes and structured residual variation, aiding patient stratification in survival analysis. These results establish CN-SBM as an interpretable, scalable framework for CNV analysis with direct relevance for tumor heterogeneity and prognosis.", "date": "2025-06-28", "categories": []}
{"id": "2506.23827", "title": "Spatially Gene Expression Prediction using Dual-Scale Contrastive Learning", "abstract": "Spatial transcriptomics (ST) provides crucial insights into tissue micro-environments, but is limited to its high cost and complexity. As an alternative, predicting gene expression from pathology whole slide images (WSI) is gaining increasing attention. However, existing methods typically rely on single patches or a single pathology modality, neglecting the complex spatial and molecular interactions between target and neighboring information (e.g., gene co-expression). This leads to a failure in establishing connections among adjacent regions and capturing intricate cross-modal relationships. To address these issues, we propose NH2ST, a framework that integrates spatial context and both pathology and gene modalities for gene expression prediction. Our model comprises a query branch and a neighbor branch to process paired target patch and gene data and their neighboring regions, where cross-attention and contrastive learning are employed to capture intrinsic associations and ensure alignments between pathology and gene expression. Extensive experiments on six datasets demonstrate that our model consistently outperforms existing methods, achieving over 20% in PCC metrics. Codes are available at https://github.com/MCPathology/NH2ST", "date": "2025-06-30", "categories": []}
{"id": "2403.15011", "title": "Cell Tracking according to Biological Needs -- Strong Mitosis-aware Multi-Hypothesis Tracker with Aleatoric Uncertainty", "abstract": "Cell tracking and segmentation assist biologists in extracting insights from large-scale microscopy time-lapse data. Driven by local accuracy metrics, current tracking approaches often suffer from a lack of long-term consistency and the ability to reconstruct lineage trees correctly. To address this issue, we introduce an uncertainty estimation technique for motion estimation frameworks and extend the multi-hypothesis tracking framework. Our uncertainty estimation lifts motion representations into probabilistic spatial densities using problem-specific test-time augmentations. Moreover, we introduce a novel mitosis-aware assignment problem formulation that allows multi-hypothesis trackers to model cell splits and to resolve false associations and mitosis detections based on long-term conflicts. In our framework, explicit biological knowledge is modeled in assignment costs. We evaluate our approach on nine competitive datasets and demonstrate that we outperform the current state-of-the-art on biologically inspired metrics substantially, achieving improvements by a factor of approximately 6 and uncover new insights into the behavior of motion estimation uncertainty.", "date": "2024-03-22", "categories": []}
{"id": "2506.22228", "title": "Uncovering smooth structures in single-cell data with PCS-guided neighbor embeddings", "abstract": "Single-cell sequencing is revolutionizing biology by enabling detailed investigations of cell-state transitions. Many biological processes unfold along continuous trajectories, yet it remains challenging to extract smooth, low-dimensional representations from inherently noisy, high-dimensional single-cell data. Neighbor embedding (NE) algorithms, such as t-SNE and UMAP, are widely used to embed high-dimensional single-cell data into low dimensions. But they often introduce undesirable distortions, resulting in misleading interpretations. Existing evaluation methods for NE algorithms primarily focus on separating discrete cell types rather than capturing continuous cell-state transitions, while dynamic modeling approaches rely on strong assumptions about cellular processes and specialized data. To address these challenges, we build on the Predictability-Computability-Stability (PCS) framework for reliable and reproducible data-driven discoveries. First, we systematically evaluate popular NE algorithms through empirical analysis, simulation, and theory, and reveal their key shortcomings, such as artifacts and instability. We then introduce NESS, a principled and interpretable machine learning approach to improve NE representations by leveraging algorithmic stability and to enable robust inference of smooth biological structures. NESS offers useful concepts, quantitative stability metrics, and efficient computational workflows to uncover developmental trajectories and cell-state transitions in single-cell data. Finally, we apply NESS to six single-cell datasets, spanning pluripotent stem cell differentiation, organoid development, and multiple tissue-specific lineage trajectories. Across these diverse contexts, NESS consistently yields useful biological insights, such as identification of transitional and stable cell states and quantification of transcriptional dynamics during development.", "date": "2025-06-27", "categories": []}
{"id": "2506.20693", "title": "E-ABIN: an Explainable module for Anomaly detection in BIological Networks", "abstract": "The increasing availability of large-scale omics data calls for robust analytical frameworks capable of handling complex gene expression datasets while offering interpretable results. Recent advances in artificial intelligence have enabled the identification of aberrant molecular patterns distinguishing disease states from healthy controls. Coupled with improvements in model interpretability, these tools now support the identification of genes potentially driving disease phenotypes. However, current approaches to gene anomaly detection often remain limited to single datasets and lack accessible graphical interfaces. Here, we introduce E-ABIN, a general-purpose, explainable framework for Anomaly detection in Biological Networks. E-ABIN combines classical machine learning and graph-based deep learning techniques within a unified, user-friendly platform, enabling the detection and interpretation of anomalies from gene expression or methylation-derived networks. By integrating algorithms such as Support Vector Machines, Random Forests, Graph Autoencoders (GAEs), and Graph Adversarial Attributed Networks (GAANs), E-ABIN ensures a high predictive accuracy while maintaining interpretability. We demonstrate the utility of E-ABIN through case studies of bladder cancer and coeliac disease, where it effectively uncovers biologically relevant anomalies and offers insights into disease mechanisms.", "date": "2025-06-25", "categories": []}
{"id": "2506.20694", "title": "Evaluating PDE discovery methods for multiscale modeling of biological signals", "abstract": "Biological systems are non-linear, include unobserved variables and the physical principles that govern their dynamics are partly unknown. This makes the characterization of their behavior very challenging. Notably, their activity occurs on multiple interdependent spatial and temporal scales that require linking mechanisms across scales. To address the challenge of bridging gaps between scales, we leverage partial differential equations (PDE) discovery. PDE discovery suggests meso-scale dynamics characteristics from micro-scale data. In this article, we present our framework combining particle-based simulations and PDE discovery and conduct preliminary experiments to assess equation discovery in controlled settings. We evaluate five state-of-the-art PDE discovery methods on particle-based simulations of calcium diffusion in astrocytes. The performances of the methods are evaluated on both the form of the discovered equation and the forecasted temporal variations of calcium concentration. Our results show that several methods accurately recover the diffusion term, highlighting the potential of PDE discovery for capturing macroscopic dynamics in biological systems from microscopic data.", "date": "2025-06-25", "categories": []}
{"id": "2506.20697", "title": "scMamba: A Scalable Foundation Model for Single-Cell Multi-Omics Integration Beyond Highly Variable Feature Selection", "abstract": "The advent of single-cell multi-omics technologies has enabled the simultaneous profiling of diverse omics layers within individual cells. Integrating such multimodal data provides unprecedented insights into cellular identity, regulatory processes, and disease mechanisms. However, it remains challenging, as current methods often rely on selecting highly variable genes or peaks during preprocessing, which may inadvertently discard crucial biological information. Here, we present scMamba, a foundation model designed to integrate single-cell multi-omics data without the need for prior feature selection while preserving genomic positional information. scMamba introduces a patch-based cell tokenization strategy that treats genomics regions as words (tokens) and cells as sentences. Building upon the concept of state space duality, scMamba distills rich biological insights from high-dimensional, sparse single-cell multi-omics data. Additionally, our novel contrastive learning approach, enhanced with cosine similarity regularization, enables superior alignment across omics layers compared to traditional methods. Systematic benchmarking across multiple datasets demonstrates that scMamba significantly outperforms state-of-the-art methods in preserving biological variation, aligning omics layers, and enhancing key downstream tasks such as clustering, cell type annotation, and trajectory inference. Our findings position scMamba as a powerful tool for large-scale single-cell multi-omics integration, capable of handling large-scale atlases and advancing biological discovery.", "date": "2025-06-25", "categories": []}
{"id": "2506.20957", "title": "Antibody Design and Optimization with Multi-scale Equivariant Graph Diffusion Models for Accurate Complex Antigen Binding", "abstract": "Antibody design remains a critical challenge in therapeutic and diagnostic development, particularly for complex antigens with diverse binding interfaces. Current computational methods face two main limitations: (1) capturing geometric features while preserving symmetries, and (2) generalizing novel antigen interfaces. Despite recent advancements, these methods often fail to accurately capture molecular interactions and maintain structural integrity. To address these challenges, we propose \\textbf{AbMEGD}, an end-to-end framework integrating \\textbf{M}ulti-scale \\textbf{E}quivariant \\textbf{G}raph \\textbf{D}iffusion for antibody sequence and structure co-design. Leveraging advanced geometric deep learning, AbMEGD combines atomic-level geometric features with residue-level embeddings, capturing local atomic details and global sequence-structure interactions. Its E(3)-equivariant diffusion method ensures geometric precision, computational efficiency, and robust generalizability for complex antigens. Furthermore, experiments using the SAbDab database demonstrate a 10.13\\% increase in amino acid recovery, 3.32\\% rise in improvement percentage, and a 0.062~\\AA\\ reduction in root mean square deviation within the critical CDR-H3 region compared to DiffAb, a leading antibody design model. These results highlight AbMEGD's ability to balance structural integrity with improved functionality, establishing a new benchmark for sequence-structure co-design and affinity optimization. The code is available at: https://github.com/Patrick221215/AbMEGD.", "date": "2025-06-26", "categories": []}
{"id": "2503.00089", "title": "Protein Structure Tokenization: Benchmarking and New Recipe", "abstract": "Recent years have witnessed a surge in the development of protein structural tokenization methods, which chunk protein 3D structures into discrete or continuous representations. Structure tokenization enables the direct application of powerful techniques like language modeling for protein structures, and large multimodal models to integrate structures with protein sequences and functional texts. Despite the progress, the capabilities and limitations of these methods remain poorly understood due to the lack of a unified evaluation framework. We first introduce StructTokenBench, a framework that comprehensively evaluates the quality and efficiency of structure tokenizers, focusing on fine-grained local substructures rather than global structures, as typical in existing benchmarks. Our evaluations reveal that no single model dominates all benchmarking perspectives. Observations of codebook under-utilization led us to develop AminoAseed, a simple yet effective strategy that enhances codebook gradient updates and optimally balances codebook size and dimension for improved tokenizer utilization and quality. Compared to the leading model ESM3, our method achieves an average of 6.31% performance improvement across 24 supervised tasks, with sensitivity and utilization rates increased by 12.83% and 124.03%, respectively. Source code and model weights are available at https://github.com/KatarinaYuan/StructTokenBench", "date": "2025-02-28", "categories": []}
{"id": "2506.18940", "title": "eccDNAMamba: A Pre-Trained Model for Ultra-Long eccDNA Sequence Analysis", "abstract": "Extrachromosomal circular DNA (eccDNA) plays key regulatory roles and contributes to oncogene overexpression in cancer through high-copy amplification and long-range interactions. Despite advances in modeling, no pre-trained models currently support full-length circular eccDNA for downstream analysis. Existing genomic models are either limited to single-nucleotide resolution or hindered by the inefficiency of the quadratic attention mechanism. Here, we introduce eccDNAMamba, the first bidirectional state-space encoder tailored for circular DNA sequences. It combines forward and reverse passes for full-context representation learning with linear-time complexity, and preserves circular structure through a novel augmentation strategy. Tested on two real-world datasets, eccDNAMamba achieves strong classification performance and scales to sequences up to 200 Kbp, offering a robust and efficient framework for modeling circular genomes. Our codes are available at https://github.com/zzq1zh/GenAI-Lab.", "date": "2025-06-22", "categories": []}
{"id": "2506.19820", "title": "ProxelGen: Generating Proteins as 3D Densities", "abstract": "We develop ProxelGen, a protein structure generative model that operates on 3D densities as opposed to the prevailing 3D point cloud representations. Representing proteins as voxelized densities, or proxels, enables new tasks and conditioning capabilities. We generate proteins encoded as proxels via a 3D CNN-based VAE in conjunction with a diffusion model operating on its latent space. Compared to state-of-the-art models, ProxelGen's samples achieve higher novelty, better FID scores, and the same level of designability as the training set. ProxelGen's advantages are demonstrated in a standard motif scaffolding benchmark, and we show how 3D density-based generation allows for more flexible shape conditioning.", "date": "2025-06-24", "categories": []}
{"id": "2412.13722", "title": "Data-driven Discovery of Biophysical T Cell Receptor Co-specificity Rules", "abstract": "The biophysical interactions between the T cell receptor (TCR) and its ligands determine the specificity of the cellular immune response. However, the immense diversity of receptors and ligands has made it challenging to discover generalizable rules across the distinct binding affinity landscapes created by different ligands. Here, we present an optimization framework for discovering biophysical rules that predict whether TCRs share specificity to a ligand. Applying this framework to TCRs associated with a collection of SARS-CoV-2 peptides we systematically characterize how co-specificity depends on the type and position of amino-acid differences between receptors. We also demonstrate that the inferred rules generalize to ligands highly dissimilar to any seen during training. Our analysis reveals that matching of steric properties between substituted amino acids is more important for receptor co-specificity than the hydrophobic properties that prominently determine evolutionary substitutability. Our analysis also quantifies the substantial importance of positions not in direct contact with the peptide for specificity. These findings highlight the potential for data-driven approaches to uncover the molecular mechanisms underpinning the specificity of adaptive immune responses.", "date": "2024-12-18", "categories": []}
{"id": "2412.17240", "title": "Rethinking Cancer Gene Identification through Graph Anomaly Analysis", "abstract": "Graph neural networks (GNNs) have shown promise in integrating protein-protein interaction (PPI) networks for identifying cancer genes in recent studies. However, due to the insufficient modeling of the biological information in PPI networks, more faithfully depiction of complex protein interaction patterns for cancer genes within the graph structure remains largely unexplored. This study takes a pioneering step toward bridging biological anomalies in protein interactions caused by cancer genes to statistical graph anomaly. We find a unique graph anomaly exhibited by cancer genes, namely weight heterogeneity, which manifests as significantly higher variance in edge weights of cancer gene nodes within the graph. Additionally, from the spectral perspective, we demonstrate that the weight heterogeneity could lead to the \"flattening out\" of spectral energy, with a concentration towards the extremes of the spectrum. Building on these insights, we propose the HIerarchical-Perspective Graph Neural Network (HIPGNN) that not only determines spectral energy distribution variations on the spectral perspective, but also perceives detailed protein interaction context on the spatial perspective. Extensive experiments are conducted on two reprocessed datasets STRINGdb and CPDB, and the experimental results demonstrate the superiority of HIPGNN.", "date": "2024-12-23", "categories": []}
{"id": "2506.17963", "title": "OmniESI: A unified framework for enzyme-substrate interaction prediction with progressive conditional deep learning", "abstract": "Understanding and modeling enzyme-substrate interactions is crucial for catalytic mechanism research, enzyme engineering, and metabolic engineering. Although a large number of predictive methods have emerged, they do not incorporate prior knowledge of enzyme catalysis to rationally modulate general protein-molecule features that are misaligned with catalytic patterns. To address this issue, we introduce a two-stage progressive framework, OmniESI, for enzyme-substrate interaction prediction through conditional deep learning. By decomposing the modeling of enzyme-substrate interactions into a two-stage progressive process, OmniESI incorporates two conditional networks that respectively emphasize enzymatic reaction specificity and crucial catalysis-related interactions, facilitating a gradual feature modulation in the latent space from general protein-molecule domain to catalysis-aware domain. On top of this unified architecture, OmniESI can adapt to a variety of downstream tasks, including enzyme kinetic parameter prediction, enzyme-substrate pairing prediction, enzyme mutational effect prediction, and enzymatic active site annotation. Under the multi-perspective performance evaluation of in-distribution and out-of-distribution settings, OmniESI consistently delivered superior performance than state-of-the-art specialized methods across seven benchmarks. More importantly, the proposed conditional networks were shown to internalize the fundamental patterns of catalytic efficiency while significantly improving prediction performance, with only negligible parameter increases (0.16%), as demonstrated by ablation studies on key components. Overall, OmniESI represents a unified predictive approach for enzyme-substrate interactions, providing an effective tool for catalytic mechanism cracking and enzyme engineering with strong generalization and broad applicability.", "date": "2025-06-22", "categories": []}
{"id": "2506.18797", "title": "A Multi-view Divergence-Convergence Feature Augmentation Framework for Drug-related Microbes Prediction", "abstract": "In the study of drug function and precision medicine, identifying new drug-microbe associations is crucial. However, current methods isolate association and similarity analysis of drug and microbe, lacking effective inter-view optimization and coordinated multi-view feature fusion. In our study, a multi-view Divergence-Convergence Feature Augmentation framework for Drug-related Microbes Prediction (DCFA_DMP) is proposed, to better learn and integrate association information and similarity information. In the divergence phase, DCFA_DMP strengthens the complementarity and diversity between heterogeneous information and similarity information by performing Adversarial Learning method between the association network view and different similarity views, optimizing the feature space. In the convergence phase, a novel Bidirectional Synergistic Attention Mechanism is proposed to deeply synergize the complementary features between different views, achieving a deep fusion of the feature space. Moreover, Transformer graph learning is alternately applied on the drug-microbe heterogeneous graph, enabling each drug or microbe node to focus on the most relevant nodes. Numerous experiments demonstrate DCFA_DMP's significant performance in predicting drug-microbe associations. It also proves effectiveness in predicting associations for new drugs and microbes in cold start experiments, further confirming its stability and reliability in predicting potential drug-microbe associations.", "date": "2025-06-23", "categories": []}
{"id": "2302.11430", "title": "Differentiable Rotamer Sampling with Molecular Force Fields", "abstract": "Molecular dynamics is the primary computational method by which modern structural biology explores macromolecule structure and function. Boltzmann generators have been proposed as an alternative to molecular dynamics, by replacing the integration of molecular systems over time with the training of generative neural networks. This neural network approach to MD samples rare events at a higher rate than traditional MD, however critical gaps in the theory and computational feasibility of Boltzmann generators significantly reduce their usability. Here, we develop a mathematical foundation to overcome these barriers; we demonstrate that the Boltzmann generator approach is sufficiently rapid to replace traditional MD for complex macromolecules, such as proteins in specific applications, and we provide a comprehensive toolkit for the exploration of molecular energy landscapes with neural networks.", "date": "2023-02-22", "categories": []}
{"id": "2506.15761", "title": "Advancing Digital Precision Medicine for Chronic Fatigue Syndrome through Longitudinal Large-Scale Multi-Modal Biological Omics Modeling with Machine Learning and Artificial Intelligence", "abstract": "We studied a generalized question: chronic diseases like ME/CFS and long COVID exhibit high heterogeneity with multifactorial etiology and progression, complicating diagnosis and treatment. To address this, we developed BioMapAI, an explainable Deep Learning framework using the richest longitudinal multi-omics dataset for ME/CFS to date. This dataset includes gut metagenomics, plasma metabolome, immune profiling, blood labs, and clinical symptoms. By connecting multi-omics to a symptom matrix, BioMapAI identified both disease- and symptom-specific biomarkers, reconstructed symptoms, and achieved state-of-the-art precision in disease classification. We also created the first connectivity map of these omics in both healthy and disease states and revealed how microbiome-immune-metabolome crosstalk shifted from healthy to ME/CFS.", "date": "2025-06-18", "categories": []}
{"id": "2506.14796", "title": "PFMBench: Protein Foundation Model Benchmark", "abstract": "This study investigates the current landscape and future directions of protein foundation model research. While recent advancements have transformed protein science and engineering, the field lacks a comprehensive benchmark for fair evaluation and in-depth understanding. Since ESM-1B, numerous protein foundation models have emerged, each with unique datasets and methodologies. However, evaluations often focus on limited tasks tailored to specific models, hindering insights into broader generalization and limitations. Specifically, researchers struggle to understand the relationships between tasks, assess how well current models perform across them, and determine the criteria in developing new foundation models. To fill this gap, we present PFMBench, a comprehensive benchmark evaluating protein foundation models across 38 tasks spanning 8 key areas of protein science. Through hundreds of experiments on 17 state-of-the-art models across 38 tasks, PFMBench reveals the inherent correlations between tasks, identifies top-performing models, and provides a streamlined evaluation protocol. Code is available at \\href{https://github.com/biomap-research/PFMBench}{\\textcolor{blue}{GitHub}}.", "date": "2025-06-01", "categories": []}
{"id": "2506.14853", "title": "DisProtEdit: Exploring Disentangled Representations for Multi-Attribute Protein Editing", "abstract": "We introduce DisProtEdit, a controllable protein editing framework that leverages dual-channel natural language supervision to learn disentangled representations of structural and functional properties. Unlike prior approaches that rely on joint holistic embeddings, DisProtEdit explicitly separates semantic factors, enabling modular and interpretable control. To support this, we construct SwissProtDis, a large-scale multimodal dataset where each protein sequence is paired with two textual descriptions, one for structure and one for function, automatically decomposed using a large language model. DisProtEdit aligns protein and text embeddings using alignment and uniformity objectives, while a disentanglement loss promotes independence between structural and functional semantics. At inference time, protein editing is performed by modifying one or both text inputs and decoding from the updated latent representation. Experiments on protein editing and representation learning benchmarks demonstrate that DisProtEdit performs competitively with existing methods while providing improved interpretability and controllability. On a newly constructed multi-attribute editing benchmark, the model achieves a both-hit success rate of up to 61.7%, highlighting its effectiveness in coordinating simultaneous structural and functional edits.", "date": "2025-06-17", "categories": []}
{"id": "2410.04461", "title": "Improved Off-policy Reinforcement Learning in Biological Sequence Design", "abstract": "Designing biological sequences with desired properties is challenging due to vast search spaces and limited evaluation budgets. Although reinforcement learning methods use proxy models for rapid reward evaluation, insufficient training data can cause proxy misspecification on out-of-distribution inputs. To address this, we propose a novel off-policy search, $\\delta$-Conservative Search, that enhances robustness by restricting policy exploration to reliable regions. Starting from high-score offline sequences, we inject noise by randomly masking tokens with probability $\\delta$, then denoise them using our policy. We further adapt $\\delta$ based on proxy uncertainty on each data point, aligning the level of conservativeness with model confidence. Experimental results show that our conservative search consistently enhances the off-policy training, outperforming existing machine learning methods in discovering high-score sequences across diverse tasks, including DNA, RNA, protein, and peptide design.", "date": "2024-10-06", "categories": []}
{"id": "2506.14488", "title": "Reimagining Target-Aware Molecular Generation through Retrieval-Enhanced Aligned Diffusion", "abstract": "Breakthroughs in high-accuracy protein structure prediction, such as AlphaFold, have established receptor-based molecule design as a critical driver for rapid early-phase drug discovery. However, most approaches still struggle to balance pocket-specific geometric fit with strict valence and synthetic constraints. To resolve this trade-off, a Retrieval-Enhanced Aligned Diffusion termed READ is introduced, which is the first to merge molecular Retrieval-Augmented Generation with an SE(3)-equivariant diffusion model. Specifically, a contrastively pre-trained encoder aligns atom-level representations during training, then retrieves graph embeddings of pocket-matched scaffolds to guide each reverse-diffusion step at inference. This single mechanism can inject real-world chemical priors exactly where needed, producing valid, diverse, and shape-complementary ligands. Experimental results demonstrate that READ can achieve very competitive performance in CBGBench, surpassing state-of-the-art generative models and even native ligands. That suggests retrieval and diffusion can be co-optimized for faster, more reliable structure-based drug design.", "date": "2025-06-17", "categories": []}
{"id": "2211.15669", "title": "PINNet: a deep neural network with pathway prior knowledge for Alzheimer's disease", "abstract": "Identification of Alzheimer's Disease (AD)-related transcriptomic signatures from blood is important for early diagnosis of the disease. Deep learning techniques are potent classifiers for AD diagnosis, but most have been unable to identify biomarkers because of their lack of interpretability. To address these challenges, we propose a pathway information-based neural network (PINNet) to predict AD patients and analyze blood and brain transcriptomic signatures using an interpretable deep learning model. PINNet is a deep neural network (DNN) model with pathway prior knowledge from either the Gene Ontology or Kyoto Encyclopedia of Genes and Genomes databases. Then, a backpropagation-based model interpretation method was applied to reveal essential pathways and genes for predicting AD. We compared the performance of PINNet with a DNN model without a pathway. Performances of PINNet outperformed or were similar to those of DNN without a pathway using blood and brain gene expressions, respectively. Moreover, PINNet considers more AD-related genes as essential features than DNN without a pathway in the learning process. Pathway analysis of protein-protein interaction modules of highly contributed genes showed that AD-related genes in blood were enriched with cell migration, PI3K-Akt, MAPK signaling, and apoptosis in blood. The pathways enriched in the brain module included cell migration, PI3K-Akt, MAPK signaling, apoptosis, protein ubiquitination, and t-cell activation. Collectively, with prior knowledge about pathways, PINNet reveals essential pathways related to AD.", "date": "2022-11-27", "categories": []}
{"id": "2409.02143", "title": "MLOmics: Cancer Multi-Omics Database for Machine Learning", "abstract": "Framing the investigation of diverse cancers as a machine learning problem has recently shown significant potential in multi-omics analysis and cancer research. Empowering these successful machine learning models are the high-quality training datasets with sufficient data volume and adequate preprocessing. However, while there exist several public data portals, including The Cancer Genome Atlas (TCGA) multi-omics initiative or open-bases such as the LinkedOmics, these databases are not off-the-shelf for existing machine learning models. In this paper, we introduce MLOmics, an open cancer multi-omics database aiming at serving better the development and evaluation of bioinformatics and machine learning models. MLOmics contains 8,314 patient samples covering all 32 cancer types with four omics types, stratified features, and extensive baselines. Complementary support for downstream analysis and bio-knowledge linking are also included to support interdisciplinary analysis.", "date": "2024-09-02", "categories": []}
{"id": "2409.02335", "title": "What Do You See in Common? Learning Hierarchical Prototypes over Tree-of-Life to Discover Evolutionary Traits", "abstract": "A grand challenge in biology is to discover evolutionary traits - features of organisms common to a group of species with a shared ancestor in the tree of life (also referred to as phylogenetic tree). With the growing availability of image repositories in biology, there is a tremendous opportunity to discover evolutionary traits directly from images in the form of a hierarchy of prototypes. However, current prototype-based methods are mostly designed to operate over a flat structure of classes and face several challenges in discovering hierarchical prototypes, including the issue of learning over-specific prototypes at internal nodes. To overcome these challenges, we introduce the framework of Hierarchy aligned Commonality through Prototypical Networks (HComP-Net). The key novelties in HComP-Net include a novel over-specificity loss to avoid learning over-specific prototypes, a novel discriminative loss to ensure prototypes at an internal node are absent in the contrasting set of species with different ancestry, and a novel masking module to allow for the exclusion of over-specific prototypes at higher levels of the tree without hampering classification performance. We empirically show that HComP-Net learns prototypes that are accurate, semantically consistent, and generalizable to unseen species in comparison to baselines.", "date": "2024-09-03", "categories": []}
{"id": "2504.03847", "title": "Interpretable Multimodal Learning for Tumor Protein-Metal Binding: Progress, Challenges, and Perspectives", "abstract": "In cancer therapeutics, protein-metal binding mechanisms critically govern the pharmacokinetics and targeting efficacy of drugs, thereby fundamentally shaping the rational design of anticancer metallodrugs. While conventional laboratory methods used to study such mechanisms are often costly, low throughput, and limited in capturing dynamic biological processes, machine learning (ML) has emerged as a promising alternative. Despite increasing efforts to develop protein-metal binding datasets and ML algorithms, the application of ML in tumor protein-metal binding remains limited. Key challenges include a shortage of high-quality, tumor-specific datasets, insufficient consideration of multiple data modalities, and the complexity of interpreting results due to the ''black box'' nature of complex ML models. This paper summarizes recent progress and ongoing challenges in using ML to predict tumor protein-metal binding, focusing on data, modeling, and interpretability. We present multimodal protein-metal binding datasets and outline strategies for acquiring, curating, and preprocessing them for training ML models. Moreover, we explore the complementary value provided by different data modalities and examine methods for their integration. We also review approaches for improving model interpretability to support more trustworthy decisions in cancer research. Finally, we offer our perspective on research opportunities and propose strategies to address the scarcity of tumor protein data and the limited number of predictive models for tumor protein-metal binding. We also highlight two promising directions for effective metal-based drug design: integrating protein-protein interaction data to provide structural insights into metal-binding events and predicting structural changes in tumor proteins after metal binding.", "date": "2025-04-04", "categories": []}
{"id": "2506.12185", "title": "Artificial Intelligence and Machine Learning in the Development of Vaccines and Immunotherapeutics Yesterday, Today, and Tomorrow", "abstract": "In the past, the development of vaccines and immunotherapeutics relied heavily on trial-and-error experimentation and extensive in vivo testing, often requiring years of pre-clinical and clinical trials. Today, artificial intelligence (AI) and deep learning (DL) are actively transforming vaccine and immunotherapeutic design, by (i) offering predictive frameworks that support rapid, data-driven decision-making; (ii) increasingly being implemented as time- and resource-efficient strategies that integrate computational models, systems vaccinology, and multi-omics data to better phenotype, differentiate, and classify patient diseases and cancers; predict patients' immune responses; and identify the factors contributing to optimal vaccine and immunotherapeutic protective efficacy; (iii) refining the selection of B- and T-cell antigen/epitope targets to enhance efficacy and durability of immune protection; and (iv) enabling a deeper understanding of immune regulation, immune evasion, immune checkpoints, and regulatory pathways. The future of AI and DL points toward (i) replacing animal preclinical testing of drugs, vaccines, and immunotherapeutics with computational-based models, as recently proposed by the United States FDA; and (ii) enabling real-time in vivo modeling for immunobridging and prediction of protection in clinical trials. This may result in a fast and transformative shift for the development of personal vaccines and immunotherapeutics against infectious pathogens and cancers.", "date": "2025-06-13", "categories": []}
{"id": "2506.12821", "title": "PDCNet: a benchmark and general deep learning framework for activity prediction of peptide-drug conjugates", "abstract": "Peptide-drug conjugates (PDCs) represent a promising therapeutic avenue for human diseases, particularly in cancer treatment. Systematic elucidation of structure-activity relationships (SARs) and accurate prediction of the activity of PDCs are critical for the rational design and optimization of these conjugates. To this end, we carefully design and construct a benchmark PDCs dataset compiled from literature-derived collections and PDCdb database, and then develop PDCNet, the first unified deep learning framework for forecasting the activity of PDCs. The architecture systematically captures the complex factors underlying anticancer decisions of PDCs in real-word scenarios through a multi-level feature fusion framework that collaboratively characterizes and learns the features of peptides, linkers, and payloads. Leveraging a curated PDCs benchmark dataset, comprehensive evaluation results show that PDCNet demonstrates superior predictive capability, with the highest AUC, F1, MCC and BA scores of 0.9213, 0.7656, 0.7071 and 0.8388 for the test set, outperforming eight established traditional machine learning models. Multi-level validations, including 5-fold cross-validation, threshold testing, ablation studies, model interpretability analysis and external independent testing, further confirm the superiority, robustness, and usability of the PDCNet architecture. We anticipate that PDCNet represents a novel paradigm, incorporating both a benchmark dataset and advanced models, which can accelerate the design and discovery of new PDC-based therapeutic agents.", "date": "2025-06-15", "categories": []}
{"id": "2506.13006", "title": "Antibody Foundational Model : Ab-RoBERTa", "abstract": "With the growing prominence of antibody-based therapeutics, antibody engineering has gained increasing attention as a critical area of research and development. Recent progress in transformer-based protein large language models (LLMs) has demonstrated promising applications in protein sequence design and structural prediction. Moreover, the availability of large-scale antibody datasets such as the Observed Antibody Space (OAS) database has opened new avenues for the development of LLMs specialized for processing antibody sequences. Among these, RoBERTa has demonstrated improved performance relative to BERT, while maintaining a smaller parameter count (125M) compared to the BERT-based protein model, ProtBERT (420M). This reduced model size enables more efficient deployment in antibody-related applications. However, despite the numerous advantages of the RoBERTa architecture, antibody-specific foundational models built upon it have remained inaccessible to the research community. In this study, we introduce Ab-RoBERTa, a RoBERTa-based antibody-specific LLM, which is publicly available at https://huggingface.co/mogam-ai/Ab-RoBERTa. This resource is intended to support a wide range of antibody-related research applications including paratope prediction or humanness assessment.", "date": "2025-06-16", "categories": []}
{"id": "2506.13119", "title": "PhenoKG: Knowledge Graph-Driven Gene Discovery and Patient Insights from Phenotypes Alone", "abstract": "Identifying causative genes from patient phenotypes remains a significant challenge in precision medicine, with important implications for the diagnosis and treatment of genetic disorders. We propose a novel graph-based approach for predicting causative genes from patient phenotypes, with or without an available list of candidate genes, by integrating a rare disease knowledge graph (KG). Our model, combining graph neural networks and transformers, achieves substantial improvements over the current state-of-the-art. On the real-world MyGene2 dataset, it attains a mean reciprocal rank (MRR) of 24.64\\% and nDCG@100 of 33.64\\%, surpassing the best baseline (SHEPHERD) at 19.02\\% MRR and 30.54\\% nDCG@100. We perform extensive ablation studies to validate the contribution of each model component. Notably, the approach generalizes to cases where only phenotypic data are available, addressing key challenges in clinical decision support when genomic information is incomplete.", "date": "2025-06-16", "categories": []}
{"id": "2506.13344", "title": "LapDDPM: A Conditional Graph Diffusion Model for scRNA-seq Generation with Spectral Adversarial Perturbations", "abstract": "Generating high-fidelity and biologically plausible synthetic single-cell RNA sequencing (scRNA-seq) data, especially with conditional control, is challenging due to its high dimensionality, sparsity, and complex biological variations. Existing generative models often struggle to capture these unique characteristics and ensure robustness to structural noise in cellular networks. We introduce LapDDPM, a novel conditional Graph Diffusion Probabilistic Model for robust and high-fidelity scRNA-seq generation. LapDDPM uniquely integrates graph-based representations with a score-based diffusion model, enhanced by a novel spectral adversarial perturbation mechanism on graph edge weights. Our contributions are threefold: we leverage Laplacian Positional Encodings (LPEs) to enrich the latent space with crucial cellular relationship information; we develop a conditional score-based diffusion model for effective learning and generation from complex scRNA-seq distributions; and we employ a unique spectral adversarial training scheme on graph edge weights, boosting robustness against structural variations. Extensive experiments on diverse scRNA-seq datasets demonstrate LapDDPM's superior performance, achieving high fidelity and generating biologically-plausible, cell-type-specific samples. LapDDPM sets a new benchmark for conditional scRNA-seq data generation, offering a robust tool for various downstream biological applications.", "date": "2025-06-16", "categories": []}
{"id": "2506.13485", "title": "Curriculum Learning for Biological Sequence Prediction: The Case of De Novo Peptide Sequencing", "abstract": "Peptide sequencing-the process of identifying amino acid sequences from mass spectrometry data-is a fundamental task in proteomics. Non-Autoregressive Transformers (NATs) have proven highly effective for this task, outperforming traditional methods. Unlike autoregressive models, which generate tokens sequentially, NATs predict all positions simultaneously, leveraging bidirectional context through unmasked self-attention. However, existing NAT approaches often rely on Connectionist Temporal Classification (CTC) loss, which presents significant optimization challenges due to CTC's complexity and increases the risk of training failures. To address these issues, we propose an improved non-autoregressive peptide sequencing model that incorporates a structured protein sequence curriculum learning strategy. This approach adjusts protein's learning difficulty based on the model's estimated protein generational capabilities through a sampling process, progressively learning peptide generation from simple to complex sequences. Additionally, we introduce a self-refining inference-time module that iteratively enhances predictions using learned NAT token embeddings, improving sequence accuracy at a fine-grained level. Our curriculum learning strategy reduces NAT training failures frequency by more than 90% based on sampled training over various data distributions. Evaluations on nine benchmark species demonstrate that our approach outperforms all previous methods across multiple metrics and species.", "date": "2025-06-16", "categories": []}
{"id": "2302.04611", "title": "A Text-guided Protein Design Framework", "abstract": "Current AI-assisted protein design mainly utilizes protein sequential and structural information. Meanwhile, there exists tremendous knowledge curated by humans in the text format describing proteins' high-level functionalities. Yet, whether the incorporation of such text data can help protein design tasks has not been explored. To bridge this gap, we propose ProteinDT, a multi-modal framework that leverages textual descriptions for protein design. ProteinDT consists of three subsequent steps: ProteinCLAP which aligns the representation of two modalities, a facilitator that generates the protein representation from the text modality, and a decoder that creates the protein sequences from the representation. To train ProteinDT, we construct a large dataset, SwissProtCLAP, with 441K text and protein pairs. We quantitatively verify the effectiveness of ProteinDT on three challenging tasks: (1) over 90% accuracy for text-guided protein generation; (2) best hit ratio on 12 zero-shot text-guided protein editing tasks; (3) superior performance on four out of six protein property prediction benchmarks.", "date": "2023-02-09", "categories": []}
{"id": "2406.12022", "title": "Constructing Ancestral Recombination Graphs through Reinforcement Learning", "abstract": "Over the years, many approaches have been proposed to build ancestral recombination graphs (ARGs), graphs used to represent the genetic relationship between individuals. Among these methods, many rely on the assumption that the most likely graph is among the shortest ones. In this paper, we propose a new approach to build short ARGs: Reinforcement Learning (RL). We exploit the similarities between finding the shortest path between a set of genetic sequences and their most recent common ancestor and finding the shortest path between the entrance and exit of a maze, a classic RL problem. In the maze problem, the learner, called the agent, must learn the directions to take in order to escape as quickly as possible, whereas in our problem, the agent must learn the actions to take between coalescence, mutation, and recombination in order to reach the most recent common ancestor as quickly as possible. Our results show that RL can be used to build ARGs as short as those built with a heuristic algorithm optimized to build short ARGs, and sometimes even shorter. Moreover, our method allows to build a distribution of short ARGs for a given sample, and can also generalize learning to new samples not used during the learning process.", "date": "2024-06-17", "categories": []}
{"id": "2506.11158", "title": "Brain-wide interpolation and conditioning of gene expression in the human brain using Implicit Neural Representations", "abstract": "In this paper, we study the efficacy and utility of recent advances in non-local, non-linear image interpolation and extrapolation algorithms, specifically, ideas based on Implicit Neural Representations (INR), as a tool for analysis of spatial transcriptomics data. We seek to utilize the microarray gene expression data sparsely sampled in the healthy human brain, and produce fully resolved spatial maps of any given gene across the whole brain at a voxel-level resolution. To do so, we first obtained the 100 top AD risk genes, whose baseline spatial transcriptional profiles were obtained from the Allen Human Brain Atlas (AHBA). We adapted Implicit Neural Representation models so that the pipeline can produce robust voxel-resolution quantitative maps of all genes. We present a variety of experiments using interpolations obtained from Abagen as a baseline/reference.", "date": "2025-06-11", "categories": []}
{"id": "2506.11182", "title": "Multimodal Modeling of CRISPR-Cas12 Activity Using Foundation Models and Chromatin Accessibility Data", "abstract": "Predicting guide RNA (gRNA) activity is critical for effective CRISPR-Cas12 genome editing but remains challenging due to limited data, variation across protospacer adjacent motifs (PAMs-short sequence requirements for Cas binding), and reliance on large-scale training. We investigate whether pre-trained biological foundation model originally trained on transcriptomic data can improve gRNA activity estimation even without domain-specific pre-training. Using embeddings from existing RNA foundation model as input to lightweight regressor, we show substantial gains over traditional baselines. We also integrate chromatin accessibility data to capture regulatory context, improving performance further. Our results highlight the effectiveness of pre-trained foundation models and chromatin accessibility data for gRNA activity prediction.", "date": "2025-06-12", "categories": []}
{"id": "2406.15669", "title": "CARE: a Benchmark Suite for the Classification and Retrieval of Enzymes", "abstract": "Enzymes are important proteins that catalyze chemical reactions. In recent years, machine learning methods have emerged to predict enzyme function from sequence; however, there are no standardized benchmarks to evaluate these methods. We introduce CARE, a benchmark and dataset suite for the Classification And Retrieval of Enzymes (CARE). CARE centers on two tasks: (1) classification of a protein sequence by its enzyme commission (EC) number and (2) retrieval of an EC number given a chemical reaction. For each task, we design train-test splits to evaluate different kinds of out-of-distribution generalization that are relevant to real use cases. For the classification task, we provide baselines for state-of-the-art methods. Because the retrieval task has not been previously formalized, we propose a method called Contrastive Reaction-EnzymE Pretraining (CREEP) as one of the first baselines for this task and compare it to the recent method, CLIPZyme. CARE is available at https://github.com/jsunn-y/CARE/.", "date": "2024-06-21", "categories": []}
{"id": "2410.03634", "title": "Function-Guided Conditional Generation Using Protein Language Models with Adapters", "abstract": "The conditional generation of proteins with desired functions is a key goal for generative models. Existing methods based on prompting of protein language models (PLMs) can generate proteins conditioned on a target functionality, such as a desired enzyme family. However, these methods are limited to simple, tokenized conditioning and have not been shown to generalize to unseen functions. In this study, we propose ProCALM (Protein Conditionally Adapted Language Model), an approach for the conditional generation of proteins using adapters to PLMs. While previous methods have used adapters for structure-conditioned generation from PLMs, our implementation of ProCALM involves finetuning ProGen2 to condition generation based on versatile representations of protein function-e.g. enzyme family, taxonomy, or natural language descriptions. ProCALM matches or exceeds the performance of existing methods at conditional sequence generation from target functions. Impressively, it can also generalize to rare and unseen functions. Overall, ProCALM is a flexible and computationally efficient approach, and we expect that it can be extended to a wide range of generative language models.", "date": "2024-10-04", "categories": []}
{"id": "2506.10015", "title": "Identifying critical residues of a protein using meaningfully-thresholded Random Geometric Graphs", "abstract": "Identification of critical residues of a protein is actively pursued, since such residues are essential for protein function. We present three ways of recognising critical residues of an example protein, the evolution of which is tracked via molecular dynamical simulations. Our methods are based on learning a Random Geometric Graph (RGG) variable, where the state variable of each of 156 residues, is attached to a node of this graph, with the RGG learnt using the matrix of correlations between state variables of each residue-pair. Given the categorical nature of the state variable, correlation between a residue pair is computed using Cramer's V. We advance an organic thresholding to learn an RGG, and compare results against extant thresholding techniques, when parametrising criticality as the nodal degree in the learnt RGG. Secondly, we develop a criticality measure by ranking the computed differences between the posterior probability of the full graph variable defined on all 156 residues, and that of the graph with all but one residue omitted. A third parametrisation of criticality informs on the dynamical variation of nodal degrees as the protein evolves during the simulation. Finally, we compare results obtained with the three distinct criticality parameters, against experimentally-ascertained critical residues.", "date": "2025-05-28", "categories": []}
{"id": "2409.06694", "title": "DANCE: Deep Learning-Assisted Analysis of Protein Sequences Using Chaos Enhanced Kaleidoscopic Images", "abstract": "Cancer is a complex disease characterized by uncontrolled cell growth. T cell receptors (TCRs), crucial proteins in the immune system, play a key role in recognizing antigens, including those associated with cancer. Recent advancements in sequencing technologies have facilitated comprehensive profiling of TCR repertoires, uncovering TCRs with potent anti-cancer activity and enabling TCR-based immunotherapies. However, analyzing these intricate biomolecules necessitates efficient representations that capture their structural and functional information. T-cell protein sequences pose unique challenges due to their relatively smaller lengths compared to other biomolecules. An image-based representation approach becomes a preferred choice for efficient embeddings, allowing for the preservation of essential details and enabling comprehensive analysis of T-cell protein sequences. In this paper, we propose to generate images from the protein sequences using the idea of Chaos Game Representation (CGR) using the Kaleidoscopic images approach. This Deep Learning Assisted Analysis of Protein Sequences Using Chaos Enhanced Kaleidoscopic Images (called DANCE) provides a unique way to visualize protein sequences by recursively applying chaos game rules around a central seed point. we perform the classification of the T cell receptors (TCRs) protein sequences in terms of their respective target cancer cells, as TCRs are known for their immune response against cancer disease. The TCR sequences are converted into images using the DANCE method. We employ deep-learning vision models to perform the classification to obtain insights into the relationship between the visual patterns observed in the generated kaleidoscopic images and the underlying protein properties. By combining CGR-based image generation with deep learning classification, this study opens novel possibilities in the protein analysis domain.", "date": "2024-09-10", "categories": []}
{"id": "2503.16563", "title": "Chem42: a Family of chemical Language Models for Target-aware Ligand Generation", "abstract": "Revolutionizing drug discovery demands more than just understanding molecular interactions - it requires generative models that can design novel ligands tailored to specific biological targets. While chemical Language Models (cLMs) have made strides in learning molecular properties, most fail to incorporate target-specific insights, restricting their ability to drive de-novo ligand generation. Chem42, a cutting-edge family of generative chemical Language Models, is designed to bridge this gap. By integrating atomic-level interactions with multimodal inputs from Prot42, a complementary protein Language Model, Chem42 achieves a sophisticated cross-modal representation of molecular structures, interactions, and binding patterns. This innovative framework enables the creation of structurally valid, synthetically accessible ligands with enhanced target specificity. Evaluations across diverse protein targets confirm that Chem42 surpasses existing approaches in chemical validity, target-aware design, and predicted binding affinity. By reducing the search space of viable drug candidates, Chem42 could accelerate the drug discovery pipeline, offering a powerful generative AI tool for precision medicine. Our Chem42 models set a new benchmark in molecule property prediction, conditional molecule generation, and target-aware ligand design. The models are publicly available at huggingface.co/inceptionai.", "date": "2025-03-20", "categories": []}
{"id": "2506.09052", "title": "Llama-Affinity: A Predictive Antibody Antigen Binding Model Integrating Antibody Sequences with Llama3 Backbone Architecture", "abstract": "Antibody-facilitated immune responses are central to the body's defense against pathogens, viruses, and other foreign invaders. The ability of antibodies to specifically bind and neutralize antigens is vital for maintaining immunity. Over the past few decades, bioengineering advancements have significantly accelerated therapeutic antibody development. These antibody-derived drugs have shown remarkable efficacy, particularly in treating cancer, SARS-CoV-2, autoimmune disorders, and infectious diseases. Traditionally, experimental methods for affinity measurement have been time-consuming and expensive. With the advent of artificial intelligence, in silico medicine has been revolutionized; recent developments in machine learning, particularly the use of large language models (LLMs) for representing antibodies, have opened up new avenues for AI-based design and improved affinity prediction. Herein, we present an advanced antibody-antigen binding affinity prediction model (LlamaAffinity), leveraging an open-source Llama 3 backbone and antibody sequence data sourced from the Observed Antibody Space (OAS) database. The proposed approach shows significant improvement over existing state-of-the-art (SOTA) methods (AntiFormer, AntiBERTa, AntiBERTy) across multiple evaluation metrics. Specifically, the model achieved an accuracy of 0.9640, an F1-score of 0.9643, a precision of 0.9702, a recall of 0.9586, and an AUC-ROC of 0.9936. Moreover, this strategy unveiled higher computational efficiency, with a five-fold average cumulative training time of only 0.46 hours, significantly lower than in previous studies.", "date": "2025-05-17", "categories": []}
{"id": "2506.09063", "title": "Reconstructing Heterogeneous Biomolecules via Hierarchical Gaussian Mixtures and Part Discovery", "abstract": "Cryo-EM is a transformational paradigm in molecular biology where computational methods are used to infer 3D molecular structure at atomic resolution from extremely noisy 2D electron microscope images. At the forefront of research is how to model the structure when the imaged particles exhibit non-rigid conformational flexibility and compositional variation where parts are sometimes missing. We introduce a novel 3D reconstruction framework with a hierarchical Gaussian mixture model, inspired in part by Gaussian Splatting for 4D scene reconstruction. In particular, the structure of the model is grounded in an initial process that infers a part-based segmentation of the particle, providing essential inductive bias in order to handle both conformational and compositional variability. The framework, called CryoSPIRE, is shown to reveal biologically meaningful structures on complex experimental datasets, and establishes a new state-of-the-art on CryoBench, a benchmark for cryo-EM heterogeneity methods.", "date": "2025-06-06", "categories": []}
{"id": "2506.09496", "title": "EnerBridge-DPO: Energy-Guided Protein Inverse Folding with Markov Bridges and Direct Preference Optimization", "abstract": "Designing protein sequences with optimal energetic stability is a key challenge in protein inverse folding, as current deep learning methods are primarily trained by maximizing sequence recovery rates, often neglecting the energy of the generated sequences. This work aims to overcome this limitation by developing a model that directly generates low-energy, stable protein sequences. We propose EnerBridge-DPO, a novel inverse folding framework focused on generating low-energy, high-stability protein sequences. Our core innovation lies in: First, integrating Markov Bridges with Direct Preference Optimization (DPO), where energy-based preferences are used to fine-tune the Markov Bridge model. The Markov Bridge initiates optimization from an information-rich prior sequence, providing DPO with a pool of structurally plausible sequence candidates. Second, an explicit energy constraint loss is introduced, which enhances the energy-driven nature of DPO based on prior sequences, enabling the model to effectively learn energy representations from a wealth of prior knowledge and directly predict sequence energy values, thereby capturing quantitative features of the energy landscape. Our evaluations demonstrate that EnerBridge-DPO can design protein complex sequences with lower energy while maintaining sequence recovery rates comparable to state-of-the-art models, and accurately predicts $\\Delta \\Delta G$ values between various sequences.", "date": "2025-06-11", "categories": []}
{"id": "2503.01718", "title": "Learning surrogate equations for the analysis of an agent-based cancer model", "abstract": "In this paper, we adapt a two-species agent-based cancer model that describes the interaction between cancer cells and healthy cells on a uniform grid to include the interaction with a third species -- namely immune cells. We run six different scenarios to explore the competition between cancer and immune cells and the initial concentration of the immune cells on cancer dynamics. We then use coupled equation learning to construct a population-based reaction model for each scenario. We show how they can be unified into a single surrogate population-based reaction model, whose underlying three coupled ordinary differential equations are much easier to analyse than the original agent-based model. As an example, by finding the single steady state of the cancer concentration, we are able to find a linear relationship between this concentration and the initial concentration of the immune cells. This then enables us to estimate suitable values for the competition and initial concentration to reduce the cancer substantially without performing additional complex and expensive simulations from an agent-based stochastic model.", "date": "2025-03-03", "categories": []}
{"id": "2506.08365", "title": "AlphaFold Database Debiasing for Robust Inverse Folding", "abstract": "The AlphaFold Protein Structure Database (AFDB) offers unparalleled structural coverage at near-experimental accuracy, positioning it as a valuable resource for data-driven protein design. However, its direct use in training deep models that are sensitive to fine-grained atomic geometry, such as inverse folding, exposes a critical limitation. Comparative analysis of structural feature distributions reveals that AFDB structures exhibit distinct statistical regularities, reflecting a systematic geometric bias that deviates from the conformational diversity found in experimentally determined structures from the Protein Data Bank (PDB). While AFDB structures are cleaner and more idealized, PDB structures capture the intrinsic variability and physical realism essential for generalization in downstream tasks. To address this discrepancy, we introduce a Debiasing Structure AutoEncoder (DeSAE) that learns to reconstruct native-like conformations from intentionally corrupted backbone geometries. By training the model to recover plausible structural states, DeSAE implicitly captures a more robust and natural structural manifold. At inference, applying DeSAE to AFDB structures produces debiased structures that significantly improve inverse folding performance across multiple benchmarks. This work highlights the critical impact of subtle systematic biases in predicted structures and presents a principled framework for debiasing, significantly boosting the performance of structure-based learning tasks like inverse folding.", "date": "2025-06-10", "categories": []}
{"id": "2506.08854", "title": "Spatial Transcriptomics Expression Prediction from Histopathology Based on Cross-Modal Mask Reconstruction and Contrastive Learning", "abstract": "Spatial transcriptomics is a technology that captures gene expression levels at different spatial locations, widely used in tumor microenvironment analysis and molecular profiling of histopathology, providing valuable insights into resolving gene expression and clinical diagnosis of cancer. Due to the high cost of data acquisition, large-scale spatial transcriptomics data remain challenging to obtain. In this study, we develop a contrastive learning-based deep learning method to predict spatially resolved gene expression from whole-slide images. Evaluation across six different disease datasets demonstrates that, compared to existing studies, our method improves Pearson Correlation Coefficient (PCC) in the prediction of highly expressed genes, highly variable genes, and marker genes by 6.27%, 6.11%, and 11.26% respectively. Further analysis indicates that our method preserves gene-gene correlations and applies to datasets with limited samples. Additionally, our method exhibits potential in cancer tissue localization based on biomarker expression.", "date": "2025-06-10", "categories": []}
{"id": "2506.08954", "title": "Protriever: End-to-End Differentiable Protein Homology Search for Fitness Prediction", "abstract": "Retrieving homologous protein sequences is essential for a broad range of protein modeling tasks such as fitness prediction, protein design, structure modeling, and protein-protein interactions. Traditional workflows have relied on a two-step process: first retrieving homologs via Multiple Sequence Alignments (MSA), then training models on one or more of these alignments. However, MSA-based retrieval is computationally expensive, struggles with highly divergent sequences or complex insertions & deletions patterns, and operates independently of the downstream modeling objective. We introduce Protriever, an end-to-end differentiable framework that learns to retrieve relevant homologs while simultaneously training for the target task. When applied to protein fitness prediction, Protriever achieves state-of-the-art performance compared to sequence-based models that rely on MSA-based homolog retrieval, while being two orders of magnitude faster through efficient vector search. Protriever is both architecture- and task-agnostic, and can flexibly adapt to different retrieval strategies and protein databases at inference time -- offering a scalable alternative to alignment-centric approaches.", "date": "2025-06-10", "categories": []}
{"id": "2411.14464", "title": "JESTR: Joint Embedding Space Technique for Ranking Candidate Molecules for the Annotation of Untargeted Metabolomics Data", "abstract": "Motivation: A major challenge in metabolomics is annotation: assigning molecular structures to mass spectral fragmentation patterns. Despite recent advances in molecule-to-spectra and in spectra-to-molecular fingerprint prediction (FP), annotation rates remain low. Results: We introduce in this paper a novel paradigm (JESTR) for annotation. Unlike prior approaches that explicitly construct molecular fingerprints or spectra, JESTR leverages the insight that molecules and their corresponding spectra are views of the same data and effectively embeds their representations in a joint space. Candidate structures are ranked based on cosine similarity between the embeddings of query spectrum and each candidate. We evaluate JESTR against mol-to-spec and spec-to-FP annotation tools on three datasets. On average, for rank@[1-5], JESTR outperforms other tools by 23.6%-71.6%. We further demonstrate the strong value of regularization with candidate molecules during training, boosting rank@1 performance by 11.4% and enhancing the model's ability to discern between target and candidate molecules. When comparing JESTR's performance against that of publicly available pretrained models of SIRIUS and CFM-ID on appropriate subsets of MassSpecGym benchmark dataset, JESTR outperforms these tools by 31% and 238%, respectively. Through JESTR, we offer a novel promising avenue towards accurate annotation, therefore unlocking valuable insights into the metabolome.", "date": "2024-11-18", "categories": []}
{"id": "2506.07035", "title": "AnnoDPO: Protein Functional Annotation Learning with Direct Preference Optimization", "abstract": "Deciphering protein function remains a fundamental challenge in protein representation learning. The task presents significant difficulties for protein language models (PLMs) due to the sheer volume of functional annotation categories and the highly imbalanced distribution of annotated instances across biological ontologies. Inspired by the remarkable success of reinforcement learning from human feedback (RLHF) in large language model (LLM) alignment, we propose AnnoDPO, a novel multi-modal framework for protein function prediction that leverages Direct Preference Optimization (DPO) to enhance annotation learning. Our methodology addresses the dual challenges of annotation scarcity and category imbalance through preference-aligned training objectives, establishing a new paradigm for biological knowledge integration in protein representation learning.", "date": "2025-06-08", "categories": []}
{"id": "2506.07591", "title": "Automating Exploratory Multiomics Research via Language Models", "abstract": "This paper introduces PROTEUS, a fully automated system that produces data-driven hypotheses from raw data files. We apply PROTEUS to clinical proteogenomics, a field where effective downstream data analysis and hypothesis proposal is crucial for producing novel discoveries. PROTEUS uses separate modules to simulate different stages of the scientific process, from open-ended data exploration to specific statistical analysis and hypothesis proposal. It formulates research directions, tools, and results in terms of relationships between biological entities, using unified graph structures to manage complex research processes. We applied PROTEUS to 10 clinical multiomics datasets from published research, arriving at 360 total hypotheses. Results were evaluated through external data validation and automatic open-ended scoring. Through exploratory and iterative research, the system can navigate high-throughput and heterogeneous multiomics data to arrive at hypotheses that balance reliability and novelty. In addition to accelerating multiomic analysis, PROTEUS represents a path towards tailoring general autonomous systems to specialized scientific domains to achieve open-ended hypothesis generation from data.", "date": "2025-06-09", "categories": []}
{"id": "2410.21283", "title": "pLDDT-Predictor: High-speed Protein Screening Using Transformer and ESM2", "abstract": "Recent advancements in protein structure prediction, particularly AlphaFold2, have revolutionized structural biology by achieving near-experimental accuracy ($\\text{average RMSD} < 1.5\\text{\\AA}$). However, the computational demands of these models (approximately 30 minutes per protein on an RTX 4090) significantly limit their application in high-throughput protein screening. While large language models like ESM (Evolutionary Scale Modeling) have shown promise in extracting structural information directly from protein sequences, rapid assessment of protein structure quality for large-scale analyses remains a major challenge.\n  We introduce pLDDT-Predictor, a high-speed protein screening tool that achieves a $250,000\\times$ speedup compared to AlphaFold2 by leveraging pre-trained ESM2 protein embeddings and a Transformer architecture. Our model predicts AlphaFold2's pLDDT (predicted Local Distance Difference Test) scores with a Pearson correlation of 0.7891 and processes proteins in just 0.007 seconds on average. Using a comprehensive dataset of 1.5 million diverse protein sequences (ranging from 50 to 2048 amino acids), we demonstrate that pLDDT-Predictor accurately classifies high-confidence structures (pLDDT $>$ 70) with 91.2\\% accuracy and achieves an MSE of 84.8142 compared to AlphaFold2's predictions.\n  The source code and pre-trained models are freely available at https://github.com/jw-chae/pLDDT_Predictor, enabling the research community to perform rapid, large-scale protein structure quality assessments.", "date": "2024-10-11", "categories": []}
{"id": "2501.05644", "title": "Interpretable Enzyme Function Prediction via Residue-Level Detection", "abstract": "Predicting multiple functions labeled with Enzyme Commission (EC) numbers from the enzyme sequence is of great significance but remains a challenge due to its sparse multi-label classification nature, i.e., each enzyme is typically associated with only a few labels out of more than 6000 possible EC numbers. However, existing machine learning algorithms generally learn a fixed global representation for each enzyme to classify all functions, thereby they lack interpretability and the fine-grained information of some function-specific local residue fragments may be overwhelmed. Here we present an attention-based framework, namely ProtDETR (Protein Detection Transformer), by casting enzyme function prediction as a detection problem. It uses a set of learnable functional queries to adaptatively extract different local representations from the sequence of residue-level features for predicting different EC numbers. ProtDETR not only significantly outperforms existing deep learning-based enzyme function prediction methods, but also provides a new interpretable perspective on automatically detecting different local regions for identifying different functions through cross-attentions between queries and residue-level features. Code is available at https://github.com/yangzhao1230/ProtDETR.", "date": "2025-01-10", "categories": []}
{"id": "2506.05361", "title": "Scalable Generation of Spatial Transcriptomics from Histology Images via Whole-Slide Flow Matching", "abstract": "Spatial transcriptomics (ST) has emerged as a powerful technology for bridging histology imaging with gene expression profiling. However, its application has been limited by low throughput and the need for specialized experimental facilities. Prior works sought to predict ST from whole-slide histology images to accelerate this process, but they suffer from two major limitations. First, they do not explicitly model cell-cell interaction as they factorize the joint distribution of whole-slide ST data and predict the gene expression of each spot independently. Second, their encoders struggle with memory constraints due to the large number of spots (often exceeding 10,000) in typical ST datasets. Herein, we propose STFlow, a flow matching generative model that considers cell-cell interaction by modeling the joint distribution of gene expression of an entire slide. It also employs an efficient slide-level encoder with local spatial attention, enabling whole-slide processing without excessive memory overhead. On the recently curated HEST-1k and STImage-1K4M benchmarks, STFlow substantially outperforms state-of-the-art baselines and achieves over 18% relative improvements over the pathology foundation models.", "date": "2025-05-25", "categories": []}
{"id": "2506.05443", "title": "UniPTMs: The First Unified Multi-type PTM Site Prediction Model via Master-Slave Architecture-Based Multi-Stage Fusion Strategy and Hierarchical Contrastive Loss", "abstract": "As a core mechanism of epigenetic regulation in eukaryotes, protein post-translational modifications (PTMs) require precise prediction to decipher dynamic life activity networks. To address the limitations of existing deep learning models in cross-modal feature fusion, domain generalization, and architectural optimization, this study proposes UniPTMs: the first unified framework for multi-type PTM prediction. The framework innovatively establishes a \"Master-Slave\" dual-path collaborative architecture: The master path dynamically integrates high-dimensional representations of protein sequences, structures, and evolutionary information through a Bidirectional Gated Cross-Attention (BGCA) module, while the slave path optimizes feature discrepancies and recalibration between structural and traditional features using a Low-Dimensional Fusion Network (LDFN). Complemented by a Multi-scale Adaptive convolutional Pyramid (MACP) for capturing local feature patterns and a Bidirectional Hierarchical Gated Fusion Network (BHGFN) enabling multi-level feature integration across paths, the framework employs a Hierarchical Dynamic Weighting Fusion (HDWF) mechanism to intelligently aggregate multimodal features. Enhanced by a novel Hierarchical Contrastive loss function for feature consistency optimization, UniPTMs demonstrates significant performance improvements (3.2%-11.4% MCC and 4.2%-14.3% AP increases) over state-of-the-art models across five modification types and transcends the Single-Type Prediction Paradigm. To strike a balance between model complexity and performance, we have also developed a lightweight variant named UniPTMs-mini.", "date": "2025-06-05", "categories": []}
{"id": "2505.07286", "title": "Piloting Structure-Based Drug Design via Modality-Specific Optimal Schedule", "abstract": "Structure-Based Drug Design (SBDD) is crucial for identifying bioactive molecules. Recent deep generative models are faced with challenges in geometric structure modeling. A major bottleneck lies in the twisted probability path of multi-modalities -- continuous 3D positions and discrete 2D topologies -- which jointly determine molecular geometries. By establishing the fact that noise schedules decide the Variational Lower Bound (VLB) for the twisted probability path, we propose VLB-Optimal Scheduling (VOS) strategy in this under-explored area, which optimizes VLB as a path integral for SBDD. Our model effectively enhances molecular geometries and interaction modeling, achieving state-of-the-art PoseBusters passing rate of 95.9% on CrossDock, more than 10% improvement upon strong baselines, while maintaining high affinities and robust intramolecular validity evaluated on held-out test set. Code is available at https://github.com/AlgoMole/MolCRAFT.", "date": "2025-05-12", "categories": []}
{"id": "2506.04303", "title": "Knowledge-guided Contextual Gene Set Analysis Using Large Language Models", "abstract": "Gene set analysis (GSA) is a foundational approach for interpreting genomic data of diseases by linking genes to biological processes. However, conventional GSA methods overlook clinical context of the analyses, often generating long lists of enriched pathways with redundant, nonspecific, or irrelevant results. Interpreting these requires extensive, ad-hoc manual effort, reducing both reliability and reproducibility. To address this limitation, we introduce cGSA, a novel AI-driven framework that enhances GSA by incorporating context-aware pathway prioritization. cGSA integrates gene cluster detection, enrichment analysis, and large language models to identify pathways that are not only statistically significant but also biologically meaningful. Benchmarking on 102 manually curated gene sets across 19 diseases and ten disease-related biological mechanisms shows that cGSA outperforms baseline methods by over 30%, with expert validation confirming its increased precision and interpretability. Two independent case studies in melanoma and breast cancer further demonstrate its potential to uncover context-specific insights and support targeted hypothesis generation.", "date": "2025-06-04", "categories": []}
{"id": "2502.09775", "title": "CellFlux: Simulating Cellular Morphology Changes via Flow Matching", "abstract": "Building a virtual cell capable of accurately simulating cellular behaviors in silico has long been a dream in computational biology. We introduce CellFlux, an image-generative model that simulates cellular morphology changes induced by chemical and genetic perturbations using flow matching. Unlike prior methods, CellFlux models distribution-wise transformations from unperturbed to perturbed cell states, effectively distinguishing actual perturbation effects from experimental artifacts such as batch effects -- a major challenge in biological data. Evaluated on chemical (BBBC021), genetic (RxRx1), and combined perturbation (JUMP) datasets, CellFlux generates biologically meaningful cell images that faithfully capture perturbation-specific morphological changes, achieving a 35% improvement in FID scores and a 12% increase in mode-of-action prediction accuracy over existing methods. Additionally, CellFlux enables continuous interpolation between cellular states, providing a potential tool for studying perturbation dynamics. These capabilities mark a significant step toward realizing virtual cell modeling for biomedical research. Project page: https://yuhui-zh15.github.io/CellFlux/.", "date": "2025-02-13", "categories": []}
{"id": "2503.09427", "title": "Language-Enhanced Representation Learning for Single-Cell Transcriptomics", "abstract": "Single-cell RNA sequencing (scRNA-seq) offers detailed insights into cellular heterogeneity. Recent advancements leverage single-cell large language models (scLLMs) for effective representation learning. These models focus exclusively on transcriptomic data, neglecting complementary biological knowledge from textual descriptions. To overcome this limitation, we propose scMMGPT, a novel multimodal framework designed for language-enhanced representation learning in single-cell transcriptomics. Unlike existing methods, scMMGPT employs robust cell representation extraction, preserving quantitative gene expression data, and introduces an innovative two-stage pre-training strategy combining discriminative precision with generative flexibility. Extensive experiments demonstrate that scMMGPT significantly outperforms unimodal and multimodal baselines across key downstream tasks, including cell annotation and clustering, and exhibits superior generalization in out-of-distribution scenarios.", "date": "2025-03-12", "categories": []}
{"id": "2505.14613", "title": "Virtual Cells: Predict, Explain, Discover", "abstract": "Drug discovery is fundamentally a process of inferring the effects of treatments on patients, and would therefore benefit immensely from computational models that can reliably simulate patient responses, enabling researchers to generate and test large numbers of therapeutic hypotheses safely and economically before initiating costly clinical trials. Even a more specific model that predicts the functional response of cells to a wide range of perturbations would be tremendously valuable for discovering safe and effective treatments that successfully translate to the clinic. Creating such virtual cells has long been a goal of the computational research community that unfortunately remains unachieved given the daunting complexity and scale of cellular biology. Nevertheless, recent advances in AI, computing power, lab automation, and high-throughput cellular profiling provide new opportunities for reaching this goal. In this perspective, we present a vision for developing and evaluating virtual cells that builds on our experience at Recursion. We argue that in order to be a useful tool to discover novel biology, virtual cells must accurately predict the functional response of a cell to perturbations and explain how the predicted response is a consequence of modifications to key biomolecular interactions. We then introduce key principles for designing therapeutically-relevant virtual cells, describe a lab-in-the-loop approach for generating novel insights with them, and advocate for biologically-grounded benchmarks to guide virtual cell development. Finally, we make the case that our approach to virtual cells provides a useful framework for building other models at higher levels of organization, including virtual patients. We hope that these directions prove useful to the research community in developing virtual models optimized for positive impact on drug discovery outcomes.", "date": "2025-05-20", "categories": []}
{"id": "2504.02628", "title": "Towards Computation- and Communication-efficient Computational Pathology", "abstract": "Despite the impressive performance across a wide range of applications, current computational pathology models face significant diagnostic efficiency challenges due to their reliance on high-magnification whole-slide image analysis. This limitation severely compromises their clinical utility, especially in time-sensitive diagnostic scenarios and situations requiring efficient data transfer. To address these issues, we present a novel computation- and communication-efficient framework called Magnification-Aligned Global-Local Transformer (MAG-GLTrans). Our approach significantly reduces computational time, file transfer requirements, and storage overhead by enabling effective analysis using low-magnification inputs rather than high-magnification ones. The key innovation lies in our proposed magnification alignment (MAG) mechanism, which employs self-supervised learning to bridge the information gap between low and high magnification levels by effectively aligning their feature representations. Through extensive evaluation across various fundamental CPath tasks, MAG-GLTrans demonstrates state-of-the-art classification performance while achieving remarkable efficiency gains: up to 10.7 times reduction in computational time and over 20 times reduction in file transfer and storage requirements. Furthermore, we highlight the versatility of our MAG framework through two significant extensions: (1) its applicability as a feature extractor to enhance the efficiency of any CPath architecture, and (2) its compatibility with existing foundation models and histopathology-specific encoders, enabling them to process low-magnification inputs with minimal information loss. These advancements position MAG-GLTrans as a particularly promising solution for time-sensitive applications, especially in the context of intraoperative frozen section diagnosis where both accuracy and efficiency are paramount.", "date": "2025-04-03", "categories": []}
{"id": "2506.02051", "title": "Phenotypic Profile-Informed Generation of Drug-Like Molecules via Dual-Channel Variational Autoencoders", "abstract": "The de novo generation of drug-like molecules capable of inducing desirable phenotypic changes is receiving increasing attention. However, previous methods predominantly rely on expression profiles to guide molecule generation, but overlook the perturbative effect of the molecules on cellular contexts. To overcome this limitation, we propose SmilesGEN, a novel generative model based on variational autoencoder (VAE) architecture to generate molecules with potential therapeutic effects. SmilesGEN integrates a pre-trained drug VAE (SmilesNet) with an expression profile VAE (ProfileNet), jointly modeling the interplay between drug perturbations and transcriptional responses in a common latent space. Specifically, ProfileNet is imposed to reconstruct pre-treatment expression profiles when eliminating drug-induced perturbations in the latent space, while SmilesNet is informed by desired expression profiles to generate drug-like molecules. Our empirical experiments demonstrate that SmilesGEN outperforms current state-of-the-art models in generating molecules with higher degree of validity, uniqueness, novelty, as well as higher Tanimoto similarity to known ligands targeting the relevant proteins. Moreover, we evaluate SmilesGEN for scaffold-based molecule optimization and generation of therapeutic agents, and confirmed its superior performance in generating molecules with higher similarity to approved drugs. SmilesGEN establishes a robust framework that leverages gene signatures to generate drug-like molecules that hold promising potential to induce desirable cellular phenotypic changes.", "date": "2025-06-01", "categories": []}
{"id": "2506.03028", "title": "Protein Inverse Folding From Structure Feedback", "abstract": "The inverse folding problem, aiming to design amino acid sequences that fold into desired three-dimensional structures, is pivotal for various biotechnological applications. Here, we introduce a novel approach leveraging Direct Preference Optimization (DPO) to fine-tune an inverse folding model using feedback from a protein folding model. Given a target protein structure, we begin by sampling candidate sequences from the inverse-folding model, then predict the three-dimensional structure of each sequence with the folding model to generate pairwise structural-preference labels. These labels are used to fine-tune the inverse-folding model under the DPO objective. Our results on the CATH 4.2 test set demonstrate that DPO fine-tuning not only improves sequence recovery of baseline models but also leads to a significant improvement in average TM-Score from 0.77 to 0.81, indicating enhanced structure similarity. Furthermore, iterative application of our DPO-based method on challenging protein structures yields substantial gains, with an average TM-Score increase of 79.5\\% with regard to the baseline model. This work establishes a promising direction for enhancing protein sequence design ability from structure feedback by effectively utilizing preference optimization.", "date": "2025-06-03", "categories": []}
{"id": "2402.01809", "title": "PhenoLinker: Phenotype-Gene Link Prediction and Explanation using Heterogeneous Graph Neural Networks", "abstract": "The association of a given human phenotype to a genetic variant remains a critical challenge for biology. We present a novel system called PhenoLinker capable of associating a score to a phenotype-gene relationship by using heterogeneous information networks and a convolutional neural network-based model for graphs, which can provide an explanation for the predictions. This system can aid in the discovery of new associations and in the understanding of the consequences of human genetic variation.", "date": "2024-02-02", "categories": []}
{"id": "2404.10242", "title": "Masked Autoencoders for Microscopy are Scalable Learners of Cellular Biology", "abstract": "Featurizing microscopy images for use in biological research remains a significant challenge, especially for large-scale experiments spanning millions of images. This work explores the scaling properties of weakly supervised classifiers and self-supervised masked autoencoders (MAEs) when training with increasingly larger model backbones and microscopy datasets. Our results show that ViT-based MAEs outperform weakly supervised classifiers on a variety of tasks, achieving as much as a 11.5% relative improvement when recalling known biological relationships curated from public databases. Additionally, we develop a new channel-agnostic MAE architecture (CA-MAE) that allows for inputting images of different numbers and orders of channels at inference time. We demonstrate that CA-MAEs effectively generalize by inferring and evaluating on a microscopy image dataset (JUMP-CP) generated under different experimental conditions with a different channel structure than our pretraining data (RPI-93M). Our findings motivate continued research into scaling self-supervised learning on microscopy data in order to create powerful foundation models of cellular biology that have the potential to catalyze advancements in drug discovery and beyond.", "date": "2024-04-16", "categories": []}
{"id": "2412.17780", "title": "PepTune: De Novo Generation of Therapeutic Peptides with Multi-Objective-Guided Discrete Diffusion", "abstract": "We present PepTune, a multi-objective discrete diffusion model for simultaneous generation and optimization of therapeutic peptide SMILES. Built on the Masked Discrete Language Model (MDLM) framework, PepTune ensures valid peptide structures with a novel bond-dependent masking schedule and invalid loss function. To guide the diffusion process, we introduce Monte Carlo Tree Guidance (MCTG), an inference-time multi-objective guidance algorithm that balances exploration and exploitation to iteratively refine Pareto-optimal sequences. MCTG integrates classifier-based rewards with search-tree expansion, overcoming gradient estimation challenges and data sparsity. Using PepTune, we generate diverse, chemically-modified peptides simultaneously optimized for multiple therapeutic properties, including target binding affinity, membrane permeability, solubility, hemolysis, and non-fouling for various disease-relevant targets. In total, our results demonstrate that MCTG for masked discrete diffusion is a powerful and modular approach for multi-objective sequence design in discrete state spaces.", "date": "2024-12-23", "categories": []}
{"id": "2502.04034", "title": "Fourier Asymmetric Attention on Domain Generalization for Pan-Cancer Drug Response Prediction", "abstract": "The accurate prediction of drug responses remains a formidable challenge, particularly at the single-cell level and in clinical treatment contexts. Some studies employ transfer learning techniques to predict drug responses in individual cells and patients, but they require access to target-domain data during training, which is often unavailable or only obtainable in future. In this study, we propose a novel domain generalization framework, termed FourierDrug, to address this challenge. Given the extracted feature from expression profile, we performed Fourier transforms and then introduced an asymmetric attention constraint that would cluster drug-sensitive samples into a compact group while drives resistant samples dispersed in the frequency domain. Our empirical experiments demonstrate that our model effectively learns task-relevant features from diverse source domains, and achieves accurate predictions of drug response for unseen cancer type. When evaluated on single-cell and patient-level drug response prediction tasks, FourierDrug--trained solely on in vitro cell line data without access to target-domain data--consistently outperforms or, at least, matched the performance of current state-of-the-art methods. These findings underscore the potential of our method for real-world clinical applications.", "date": "2025-02-06", "categories": []}
{"id": "2503.20158", "title": "RxRx3-core: Benchmarking drug-target interactions in High-Content Microscopy", "abstract": "High Content Screening (HCS) microscopy datasets have transformed the ability to profile cellular responses to genetic and chemical perturbations, enabling cell-based inference of drug-target interactions (DTI). However, the adoption of representation learning methods for HCS data has been hindered by the lack of accessible datasets and robust benchmarks. To address this gap, we present RxRx3-core, a curated and compressed subset of the RxRx3 dataset, and an associated DTI benchmarking task. At just 18GB, RxRx3-core significantly reduces the size barrier associated with large-scale HCS datasets while preserving critical data necessary for benchmarking representation learning models against a zero-shot DTI prediction task. RxRx3-core includes 222,601 microscopy images spanning 736 CRISPR knockouts and 1,674 compounds at 8 concentrations. RxRx3-core is available on HuggingFace and Polaris, along with pre-trained embeddings and benchmarking code, ensuring accessibility for the research community. By providing a compact dataset and robust benchmarks, we aim to accelerate innovation in representation learning methods for HCS data and support the discovery of novel biological insights.", "date": "2025-03-26", "categories": []}
{"id": "2505.19014", "title": "Tokenizing Electron Cloud in Protein-Ligand Interaction Learning", "abstract": "The affinity and specificity of protein-molecule binding directly impact functional outcomes, uncovering the mechanisms underlying biological regulation and signal transduction. Most deep-learning-based prediction approaches focus on structures of atoms or fragments. However, quantum chemical properties, such as electronic structures, are the key to unveiling interaction patterns but remain largely underexplored. To bridge this gap, we propose ECBind, a method for tokenizing electron cloud signals into quantized embeddings, enabling their integration into downstream tasks such as binding affinity prediction. By incorporating electron densities, ECBind helps uncover binding modes that cannot be fully represented by atom-level models. Specifically, to remove the redundancy inherent in electron cloud signals, a structure-aware transformer and hierarchical codebooks encode 3D binding sites enriched with electron structures into tokens. These tokenized codes are then used for specific tasks with labels. To extend its applicability to a wider range of scenarios, we utilize knowledge distillation to develop an electron-cloud-agnostic prediction model. Experimentally, ECBind demonstrates state-of-the-art performance across multiple tasks, achieving improvements of 6.42\\% and 15.58\\% in per-structure Pearson and Spearman correlation coefficients, respectively.", "date": "2025-05-25", "categories": []}
{"id": "2506.00053", "title": "Improving statistical learning methods via features selection without replacement sampling and random projection", "abstract": "Cancer is fundamentally a genetic disease characterized by genetic and epigenetic alterations that disrupt normal gene expression, leading to uncontrolled cell growth and metastasis. High-dimensional microarray datasets pose challenges for classification models due to the \"small n, large p\" problem, resulting in overfitting. This study makes three different key contributions: 1) we propose a machine learning-based approach integrating the Feature Selection Without Re-placement (FSWOR) technique and a projection method to improve classification accuracy. 2) We apply the Kendall statistical test to identify the most significant genes from the brain cancer mi-croarray dataset (GSE50161), reducing the feature space from 54,675 to 20,890 genes.3) we apply machine learning models using k-fold cross validation techniques in which our model incorpo-rates ensemble classifiers with LDA projection and Na\\\"ive Bayes, achieving a test score of 96%, outperforming existing methods by 9.09%. The results demonstrate the effectiveness of our ap-proach in high-dimensional gene expression analysis, improving classification accuracy while mitigating overfitting. This study contributes to cancer biomarker discovery, offering a robust computational method for analyzing microarray data.", "date": "2025-05-28", "categories": []}
{"id": "2506.00297", "title": "Improving Protein Sequence Design through Designability Preference Optimization", "abstract": "Protein sequence design methods have demonstrated strong performance in sequence generation for de novo protein design. However, as the training objective was sequence recovery, it does not guarantee designability--the likelihood that a designed sequence folds into the desired structure. To bridge this gap, we redefine the training objective by steering sequence generation toward high designability. To do this, we integrate Direct Preference Optimization (DPO), using AlphaFold pLDDT scores as the preference signal, which significantly improves the in silico design success rate. To further refine sequence generation at a finer, residue-level granularity, we introduce Residue-level Designability Preference Optimization (ResiDPO), which applies residue-level structural rewards and decouples optimization across residues. This enables direct improvement in designability while preserving regions that already perform well. Using a curated dataset with residue-level annotations, we fine-tune LigandMPNN with ResiDPO to obtain EnhancedMPNN, which achieves a nearly 3-fold increase in in silico design success rate (from 6.56% to 17.57%) on a challenging enzyme design benchmark.", "date": "2025-05-30", "categories": []}
{"id": "2506.01376", "title": "Modeling All-Atom Glycan Structures via Hierarchical Message Passing and Multi-Scale Pre-training", "abstract": "Understanding the various properties of glycans with machine learning has shown some preliminary promise. However, previous methods mainly focused on modeling the backbone structure of glycans as graphs of monosaccharides (i.e., sugar units), while they neglected the atomic structures underlying each monosaccharide, which are actually important indicators of glycan properties. We fill this blank by introducing the GlycanAA model for All-Atom-wise Glycan modeling. GlycanAA models a glycan as a heterogeneous graph with monosaccharide nodes representing its global backbone structure and atom nodes representing its local atomic-level structures. Based on such a graph, GlycanAA performs hierarchical message passing to capture from local atomic-level interactions to global monosaccharide-level interactions. To further enhance model capability, we pre-train GlycanAA on a high-quality unlabeled glycan dataset, deriving the PreGlycanAA model. We design a multi-scale mask prediction algorithm to endow the model about different levels of dependencies in a glycan. Extensive benchmark results show the superiority of GlycanAA over existing glycan encoders and verify the further improvements achieved by PreGlycanAA. We maintain all resources at https://github.com/kasawa1234/GlycanAA", "date": "2025-06-02", "categories": []}
{"id": "2506.01833", "title": "SPACE: Your Genomic Profile Predictor is a Powerful DNA Foundation Model", "abstract": "Inspired by the success of unsupervised pre-training paradigms, researchers have applied these approaches to DNA pre-training. However, we argue that these approaches alone yield suboptimal results because pure DNA sequences lack sufficient information, since their functions are regulated by genomic profiles like chromatin accessibility. Here, we demonstrate that supervised training for genomic profile prediction serves as a more effective alternative to pure sequence pre-training. Furthermore, considering the multi-species and multi-profile nature of genomic profile prediction, we introduce our $\\textbf{S}$pecies-$\\textbf{P}$rofile $\\textbf{A}$daptive $\\textbf{C}$ollaborative $\\textbf{E}$xperts (SPACE) that leverages Mixture of Experts (MoE) to better capture the relationships between DNA sequences across different species and genomic profiles, thereby learning more effective DNA representations. Through extensive experiments across various tasks, our model achieves state-of-the-art performance, establishing that DNA models trained with supervised genomic profiles serve as powerful DNA representation learners. The code is available at https://github.com/ZhuJiwei111/SPACE.", "date": "2025-06-02", "categories": []}
{"id": "2506.01918", "title": "Spatial Coordinates as a Cell Language: A Multi-Sentence Framework for Imaging Mass Cytometry Analysis", "abstract": "Image mass cytometry (IMC) enables high-dimensional spatial profiling by combining mass cytometry's analytical power with spatial distributions of cell phenotypes. Recent studies leverage large language models (LLMs) to extract cell states by translating gene or protein expression into biological context. However, existing single-cell LLMs face two major challenges: (1) Integration of spatial information: they struggle to generalize spatial coordinates and effectively encode spatial context as text, and (2) Treating each cell independently: they overlook cell-cell interactions, limiting their ability to capture biological relationships. To address these limitations, we propose Spatial2Sentence, a novel framework that integrates single-cell expression and spatial information into natural language using a multi-sentence approach. Spatial2Sentence constructs expression similarity and distance matrices, pairing spatially adjacent and expressionally similar cells as positive pairs while using distant and dissimilar cells as negatives. These multi-sentence representations enable LLMs to learn cellular interactions in both expression and spatial contexts. Equipped with multi-task learning, Spatial2Sentence outperforms existing single-cell LLMs on preprocessed IMC datasets, improving cell-type classification by 5.98% and clinical status prediction by 4.18% on the diabetes dataset while enhancing interpretability. The source code can be found here: https://github.com/UNITES-Lab/Spatial2Sentence.", "date": "2025-06-02", "categories": []}
{"id": "2504.21578", "title": "Glucagon and insulin production in pancreatic cells modeled using Petri nets and Boolean networks", "abstract": "Diabetes is a civilization chronic disease characterized by a constant elevated concentration of glucose in the blood. Many processes are involved in the glucose regulation, and their interactions are very complex. To better understand those processes we set ourselves a goal to create a Petri net model of the glucose regulation in the whole body. So far we have managed to create a model of glycolysis and synthesis of glucose in the liver, and the general overview models of the glucose regulation in a healthy and diabetic person. In this paper we introduce Petri nets models of insulin secretion in beta cell of the pancreas, and glucagon in the pancreas alpha cells. Those two hormones have mutually opposite effects: insulin preventing hyperglycemia, and glucagon preventing hypoglycemia. Understanding the mechanisms of insulin and glucagon secretion constitutes the basis for understanding diabetes. We also present a model in which both processes occur together, depending on the blood glucose level. The dynamics of each model is analysed. Additionally, we transform the overall insulin and glucagon secretion system to a Boolean network, following standard transformation rules.", "date": "2025-04-30", "categories": []}
{"id": "2505.17552", "title": "Universal Biological Sequence Reranking for Improved De Novo Peptide Sequencing", "abstract": "De novo peptide sequencing is a critical task in proteomics. However, the performance of current deep learning-based methods is limited by the inherent complexity of mass spectrometry data and the heterogeneous distribution of noise signals, leading to data-specific biases. We present RankNovo, the first deep reranking framework that enhances de novo peptide sequencing by leveraging the complementary strengths of multiple sequencing models. RankNovo employs a list-wise reranking approach, modeling candidate peptides as multiple sequence alignments and utilizing axial attention to extract informative features across candidates. Additionally, we introduce two new metrics, PMD (Peptide Mass Deviation) and RMD (residual Mass Deviation), which offer delicate supervision by quantifying mass differences between peptides at both the sequence and residue levels. Extensive experiments demonstrate that RankNovo not only surpasses its base models used to generate training candidates for reranking pre-training, but also sets a new state-of-the-art benchmark. Moreover, RankNovo exhibits strong zero-shot generalization to unseen models whose generations were not exposed during training, highlighting its robustness and potential as a universal reranking framework for peptide sequencing. Our work presents a novel reranking strategy that fundamentally challenges existing single-model paradigms and advances the frontier of accurate de novo sequencing. Our source code is provided on GitHub.", "date": "2025-05-23", "categories": []}
{"id": "2505.23861", "title": "BiBLDR: Bidirectional Behavior Learning for Drug Repositioning", "abstract": "Drug repositioning aims to identify potential new indications for existing drugs to reduce the time and financial costs associated with developing new drugs. Most existing deep learning-based drug repositioning methods predominantly utilize graph-based representations. However, graph-based drug repositioning methods struggle to perform effective inference in cold-start scenarios involving novel drugs because of the lack of association information with the diseases. Unlike traditional graph-based approaches, we propose a bidirectional behavior learning strategy for drug repositioning, known as BiBLDR. This innovative framework redefines drug repositioning as a behavior sequential learning task to capture drug-disease interaction patterns. First, we construct bidirectional behavioral sequences based on drug and disease sides. The consideration of bidirectional information ensures a more meticulous and rigorous characterization of the behavioral sequences. Subsequently, we propose a two-stage strategy for drug repositioning. In the first stage, we construct prototype spaces to characterize the representational attributes of drugs and diseases. In the second stage, these refined prototypes and bidirectional behavior sequence data are leveraged to predict potential drug-disease associations. Based on this learning approach, the model can more robustly and precisely capture the interactive relationships between drug and disease features from bidirectional behavioral sequences. Extensive experiments demonstrate that our method achieves state-of-the-art performance on benchmark datasets. Meanwhile, BiBLDR demonstrates significantly superior performance compared to previous methods in cold-start scenarios. Our code is published in https://github.com/Renyeeah/BiBLDR.", "date": "2025-05-29", "categories": []}
{"id": "2505.23862", "title": "A New Deep-learning-Based Approach For mRNA Optimization: High Fidelity, Computation Efficiency, and Multiple Optimization Factors", "abstract": "The mRNA optimization is critical for therapeutic and biotechnological applications, since sequence features directly govern protein expression levels and efficacy. However, current methods face significant challenges in simultaneously achieving three key objectives: (1) fidelity (preventing unintended amino acid changes), (2) computational efficiency (speed and scalability), and (3) the scope of optimization variables considered (multi-objective capability). Furthermore, existing methods often fall short of comprehensively incorporating the factors related to the mRNA lifecycle and translation process, including intrinsic mRNA sequence properties, secondary structure, translation elongation kinetics, and tRNA availability. To address these limitations, we introduce \\textbf{RNop}, a novel deep learning-based method for mRNA optimization. We collect a large-scale dataset containing over 3 million sequences and design four specialized loss functions, the GPLoss, CAILoss, tAILoss, and MFELoss, which simultaneously enable explicit control over sequence fidelity while optimizing species-specific codon adaptation, tRNA availability, and desirable mRNA secondary structure features. Then, we demonstrate RNop's effectiveness through extensive in silico and in vivo experiments. RNop ensures high sequence fidelity, achieves significant computational throughput up to 47.32 sequences/s, and yields optimized mRNA sequences resulting in a significant increase in protein expression for functional proteins compared to controls. RNop surpasses current methodologies in both quantitative metrics and experimental validation, enlightening a new dawn for efficient and effective mRNA design. Code and models will be available at https://github.com/HudenJear/RPLoss.", "date": "2025-05-29", "categories": []}
{"id": "2505.23879", "title": "CNN-LSTM Hybrid Model for AI-Driven Prediction of COVID-19 Severity from Spike Sequences and Clinical Data", "abstract": "The COVID-19 pandemic, caused by SARS-CoV-2, highlighted the critical need for accurate prediction of disease severity to optimize healthcare resource allocation and patient management. The spike protein, which facilitates viral entry into host cells, exhibits high mutation rates, particularly in the receptor-binding domain, influencing viral pathogenicity. Artificial intelligence approaches, such as deep learning, offer promising solutions for leveraging genomic and clinical data to predict disease outcomes. Objective: This study aimed to develop a hybrid CNN-LSTM deep learning model to predict COVID-19 severity using spike protein sequences and associated clinical metadata from South American patients. Methods: We retrieved 9,570 spike protein sequences from the GISAID database, of which 3,467 met inclusion criteria after standardization. The dataset included 2,313 severe and 1,154 mild cases. A feature engineering pipeline extracted features from sequences, while demographic and clinical variables were one-hot encoded. A hybrid CNN-LSTM architecture was trained, combining CNN layers for local pattern extraction and an LSTM layer for long-term dependency modeling. Results: The model achieved an F1 score of 82.92%, ROC-AUC of 0.9084, precision of 83.56%, and recall of 82.85%, demonstrating robust classification performance. Training stabilized at 85% accuracy with minimal overfitting. The most prevalent lineages (P.1, AY.99.2) and clades (GR, GK) aligned with regional epidemiological trends, suggesting potential associations between viral genetics and clinical outcomes. Conclusion: The CNN-LSTM hybrid model effectively predicted COVID-19 severity using spike protein sequences and clinical data, highlighting the utility of AI in genomic surveillance and precision public health. Despite limitations, this approach provides a framework for early severity prediction in future outbreaks.", "date": "2025-05-29", "categories": []}
{"id": "2505.24203", "title": "Aligning Protein Conformation Ensemble Generation with Physical Feedback", "abstract": "Protein dynamics play a crucial role in protein biological functions and properties, and their traditional study typically relies on time-consuming molecular dynamics (MD) simulations conducted in silico. Recent advances in generative modeling, particularly denoising diffusion models, have enabled efficient accurate protein structure prediction and conformation sampling by learning distributions over crystallographic structures. However, effectively integrating physical supervision into these data-driven approaches remains challenging, as standard energy-based objectives often lead to intractable optimization. In this paper, we introduce Energy-based Alignment (EBA), a method that aligns generative models with feedback from physical models, efficiently calibrating them to appropriately balance conformational states based on their energy differences. Experimental results on the MD ensemble benchmark demonstrate that EBA achieves state-of-the-art performance in generating high-quality protein ensembles. By improving the physical plausibility of generated structures, our approach enhances model predictions and holds promise for applications in structural biology and drug discovery.", "date": "2025-05-30", "categories": []}
{"id": "2410.03553", "title": "Structure-Enhanced Protein Instruction Tuning: Towards General-Purpose Protein Understanding with LLMs", "abstract": "Proteins, as essential biomolecules, play a central role in biological processes, including metabolic reactions and DNA replication. Accurate prediction of their properties and functions is crucial in biological applications. Recent development of protein language models (pLMs) with supervised fine tuning provides a promising solution to this problem. However, the fine-tuned model is tailored for particular downstream prediction task, and achieving general-purpose protein understanding remains a challenge. In this paper, we introduce Structure-Enhanced Protein Instruction Tuning (SEPIT) framework to bridge this gap. Our approach incorporates a novel structure-aware module into pLMs to enrich their structural knowledge, and subsequently integrates these enhanced pLMs with large language models (LLMs) to advance protein understanding. In this framework, we propose a novel instruction tuning pipeline. First, we warm up the enhanced pLMs using contrastive learning and structure denoising. Then, caption-based instructions are used to establish a basic understanding of proteins. Finally, we refine this understanding by employing a mixture of experts (MoEs) to capture more complex properties and functional information with the same number of activated parameters. Moreover, we construct the largest and most comprehensive protein instruction dataset to date, which allows us to train and evaluate the general-purpose protein understanding model. Extensive experiments on both open-ended generation and closed-set answer tasks demonstrate the superior performance of SEPIT over both closed-source general LLMs and open-source LLMs trained with protein knowledge.", "date": "2024-10-04", "categories": []}
{"id": "2505.22674", "title": "PSBench: a large-scale benchmark for estimating the accuracy of protein complex structural models", "abstract": "Predicting protein complex structures is essential for protein function analysis, protein design, and drug discovery. While AI methods like AlphaFold can predict accurate structural models for many protein complexes, reliably estimating the quality of these predicted models (estimation of model accuracy, or EMA) for model ranking and selection remains a major challenge. A key barrier to developing effective machine learning-based EMA methods is the lack of large, diverse, and well-annotated datasets for training and evaluation. To address this gap, we introduce PSBench, a benchmark suite comprising four large-scale, labeled datasets generated during the 15th and 16th community-wide Critical Assessment of Protein Structure Prediction (CASP15 and CASP16). PSBench includes over one million structural models covering a wide range of protein sequence lengths, complex stoichiometries, functional classes, and modeling difficulties. Each model is annotated with multiple complementary quality scores at the global, local, and interface levels. PSBench also provides multiple evaluation metrics and baseline EMA methods to facilitate rigorous comparisons. To demonstrate PSBench's utility, we trained and evaluated GATE, a graph transformer-based EMA method, on the CASP15 data. GATE was blindly tested in CASP16 (2024), where it ranked among the top-performing EMA methods. These results highlight PSBench as a valuable resource for advancing EMA research in protein complex modeling. PSBench is publicly available at: https://github.com/BioinfoMachineLearning/PSBench.", "date": "2025-05-13", "categories": []}
{"id": "2505.22746", "title": "StarBASE-GP: Biologically-Guided Automated Machine Learning for Genotype-to-Phenotype Association Analysis", "abstract": "We present the Star-Based Automated Single-locus and Epistasis analysis tool - Genetic Programming (StarBASE-GP), an automated framework for discovering meaningful genetic variants associated with phenotypic variation in large-scale genomic datasets. StarBASE-GP uses a genetic programming-based multi-objective optimization strategy to evolve machine learning pipelines that simultaneously maximize explanatory power (r2) and minimize pipeline complexity. Biological domain knowledge is integrated at multiple stages, including the use of nine inheritance encoding strategies to model deviations from additivity, a custom linkage disequilibrium pruning node that minimizes redundancy among features, and a dynamic variant recommendation system that prioritizes informative candidates for pipeline inclusion. We evaluate StarBASE-GP on a cohort of Rattus norvegicus (brown rat) to identify variants associated with body mass index, benchmarking its performance against a random baseline and a biologically naive version of the tool. StarBASE-GP consistently evolves Pareto fronts with superior performance, yielding higher accuracy in identifying both ground truth and novel quantitative trait loci, highlighting relevant targets for future validation. By incorporating evolutionary search and relevant biological theory into a flexible automated machine learning framework, StarBASE-GP demonstrates robust potential for advancing variant discovery in complex traits.", "date": "2025-05-28", "categories": []}
{"id": "2505.22869", "title": "CFP-Gen: Combinatorial Functional Protein Generation via Diffusion Language Models", "abstract": "Existing PLMs generate protein sequences based on a single-condition constraint from a specific modality, struggling to simultaneously satisfy multiple constraints across different modalities. In this work, we introduce CFP-Gen, a novel diffusion language model for Combinatorial Functional Protein GENeration. CFP-Gen facilitates the de novo protein design by integrating multimodal conditions with functional, sequence, and structural constraints. Specifically, an Annotation-Guided Feature Modulation (AGFM) module is introduced to dynamically adjust the protein feature distribution based on composable functional annotations, e.g., GO terms, IPR domains and EC numbers. Meanwhile, the Residue-Controlled Functional Encoding (RCFE) module captures residue-wise interaction to ensure more precise control. Additionally, off-the-shelf 3D structure encoders can be seamlessly integrated to impose geometric constraints. We demonstrate that CFP-Gen enables high-throughput generation of novel proteins with functionality comparable to natural proteins, while achieving a high success rate in designing multifunctional proteins. Code and data available at https://github.com/yinjunbo/cfpgen.", "date": "2025-05-28", "categories": []}
{"id": "2505.23334", "title": "X2Graph for Cancer Subtyping Prediction on Biological Tabular Data", "abstract": "Despite the transformative impact of deep learning on text, audio, and image datasets, its dominance in tabular data, especially in the medical domain where data are often scarce, remains less clear. In this paper, we propose X2Graph, a novel deep learning method that achieves strong performance on small biological tabular datasets. X2Graph leverages external knowledge about the relationships between table columns, such as gene interactions, to convert each sample into a graph structure. This transformation enables the application of standard message passing algorithms for graph modeling. Our X2Graph method demonstrates superior performance compared to existing tree-based and deep learning methods across three cancer subtyping datasets.", "date": "2025-05-29", "categories": []}
{"id": "2501.19200", "title": "A Variational Perspective on Generative Protein Fitness Optimization", "abstract": "The goal of protein fitness optimization is to discover new protein variants with enhanced fitness for a given use. The vast search space and the sparsely populated fitness landscape, along with the discrete nature of protein sequences, pose significant challenges when trying to determine the gradient towards configurations with higher fitness. We introduce Variational Latent Generative Protein Optimization (VLGPO), a variational perspective on fitness optimization. Our method embeds protein sequences in a continuous latent space to enable efficient sampling from the fitness distribution and combines a (learned) flow matching prior over sequence mutations with a fitness predictor to guide optimization towards sequences with high fitness. VLGPO achieves state-of-the-art results on two different protein benchmarks of varying complexity. Moreover, the variational design with explicit prior and likelihood functions offers a flexible plug-and-play framework that can be easily customized to suit various protein design tasks.", "date": "2025-01-31", "categories": []}
{"id": "2502.14637", "title": "ReQFlow: Rectified Quaternion Flow for Efficient and High-Quality Protein Backbone Generation", "abstract": "Protein backbone generation plays a central role in de novo protein design and is significant for many biological and medical applications. Although diffusion and flow-based generative models provide potential solutions to this challenging task, they often generate proteins with undesired designability and suffer computational inefficiency. In this study, we propose a novel rectified quaternion flow (ReQFlow) matching method for fast and high-quality protein backbone generation. In particular, our method generates a local translation and a 3D rotation from random noise for each residue in a protein chain, which represents each 3D rotation as a unit quaternion and constructs its flow by spherical linear interpolation (SLERP) in an exponential format. We train the model by quaternion flow (QFlow) matching with guaranteed numerical stability and rectify the QFlow model to accelerate its inference and improve the designability of generated protein backbones, leading to the proposed ReQFlow model. Experiments show that ReQFlow achieves on-par performance in protein backbone generation while requiring much fewer sampling steps and significantly less inference time (e.g., being 37x faster than RFDiffusion and 63x faster than Genie2 when generating a backbone of length 300), demonstrating its effectiveness and efficiency. The code is available at https://github.com/AngxiaoYue/ReQFlow.", "date": "2025-02-20", "categories": []}
{"id": "2505.20344", "title": "Genetic Influences on Brain Aging: Analyzing Sex Differences in the UK Biobank using Structural MRI", "abstract": "Brain aging trajectories differ between males and females, yet the genetic factors underlying these differences remain underexplored. Using structural MRI and genotyping data from 40,940 UK Biobank participants (aged 45-83), we computed Brain Age Gap Estimates (BrainAGE) for total brain, hippocampal, and ventricular volumes. We conducted sex-stratified genome-wide association studies (GWAS) and Post-GWAS analyses to identify genetic variants associated with accelerated brain aging. Distinct gene sets emerged by sex: in females, neurotransmitter transport and mitochondrial stress response genes were implicated; in males, immune and inflammation-related genes dominated. Shared genes, including GMNC and OSTN, were consistently linked to brain volumes across sexes, suggesting core roles in neurostructural maintenance. Tissue expression analyses revealed sex-specific enrichment in pathways tied to neurodegeneration. These findings highlight the importance of sex-stratified approaches in aging research and suggest genetic targets for personalized interventions against age-related cognitive decline.", "date": "2025-05-25", "categories": []}
{"id": "2505.21873", "title": "HelixDesign-Binder: A Scalable Production-Grade Platform for Binder Design Built on HelixFold3", "abstract": "Protein binder design is central to therapeutics, diagnostics, and synthetic biology, yet practical deployment remains challenging due to fragmented workflows, high computational costs, and complex tool integration. We present HelixDesign-Binder, a production-grade, high-throughput platform built on HelixFold3 that automates the full binder design pipeline, from backbone generation and sequence design to structural evaluation and multi-dimensional scoring. By unifying these stages into a scalable and user-friendly system, HelixDesign-Binder enables efficient exploration of binder candidates with favorable structural, energetic, and physicochemical properties. The platform leverages Baidu Cloud's high-performance infrastructure to support large-scale design and incorporates advanced scoring metrics, including ipTM, predicted binding free energy, and interface hydrophobicity. Benchmarking across six protein targets demonstrates that HelixDesign-Binder reliably produces diverse and high-quality binders, some of which match or exceed validated designs in predicted binding affinity. HelixDesign-Binder is accessible via an interactive web interface in PaddleHelix platform, supporting both academic research and industrial applications in antibody and protein binder development.", "date": "2025-05-28", "categories": []}
{"id": "2406.00984", "title": "Predicting drug-gene relations via analogy tasks with word embeddings", "abstract": "Natural language processing (NLP) is utilized in a wide range of fields, where words in text are typically transformed into feature vectors called embeddings. BioConceptVec is a specific example of embeddings tailored for biology, trained on approximately 30 million PubMed abstracts using models such as skip-gram. Generally, word embeddings are known to solve analogy tasks through simple vector arithmetic. For example, subtracting the vector for man from that of king and then adding the vector for woman yields a point that lies closer to queen in the embedding space. In this study, we demonstrate that BioConceptVec embeddings, along with our own embeddings trained on PubMed abstracts, contain information about drug-gene relations and can predict target genes from a given drug through analogy computations. We also show that categorizing drugs and genes using biological pathways improves performance. Furthermore, we illustrate that vectors derived from known relations in the past can predict unknown future relations in datasets divided by year. Despite the simplicity of implementing analogy tasks as vector additions, our approach demonstrated performance comparable to that of large language models such as GPT-4 in predicting drug-gene relations.", "date": "2024-06-03", "categories": []}
{"id": "2505.20098", "title": "Transformers in Protein: A Survey", "abstract": "As protein informatics advances rapidly, the demand for enhanced predictive accuracy, structural analysis, and functional understanding has intensified. Transformer models, as powerful deep learning architectures, have demonstrated unprecedented potential in addressing diverse challenges across protein research. However, a comprehensive review of Transformer applications in this field remains lacking. This paper bridges this gap by surveying over 100 studies, offering an in-depth analysis of practical implementations and research progress of Transformers in protein-related tasks. Our review systematically covers critical domains, including protein structure prediction, function prediction, protein-protein interaction analysis, functional annotation, and drug discovery/target identification. To contextualize these advancements across various protein domains, we adopt a domain-oriented classification system. We first introduce foundational concepts: the Transformer architecture and attention mechanisms, categorize Transformer variants tailored for protein science, and summarize essential protein knowledge. For each research domain, we outline its objectives and background, critically evaluate prior methods and their limitations, and highlight transformative contributions enabled by Transformer models. We also curate and summarize pivotal datasets and open-source code resources to facilitate reproducibility and benchmarking. Finally, we discuss persistent challenges in applying Transformers to protein informatics and propose future research directions. This review aims to provide a consolidated foundation for the synergistic integration of Transformer and protein informatics, fostering further innovation and expanded applications in the field.", "date": "2025-05-26", "categories": []}
{"id": "2505.20301", "title": "Sequence-Only Prediction of Binding Affinity Changes: A Robust and Interpretable Model for Antibody Engineering", "abstract": "A pivotal area of research in antibody engineering is to find effective modifications that enhance antibody-antigen binding affinity. Traditional wet-lab experiments assess mutants in a costly and time-consuming manner. Emerging deep learning solutions offer an alternative by modeling antibody structures to predict binding affinity changes. However, they heavily depend on high-quality complex structures, which are frequently unavailable in practice. Therefore, we propose ProtAttBA, a deep learning model that predicts binding affinity changes based solely on the sequence information of antibody-antigen complexes. ProtAttBA employs a pre-training phase to learn protein sequence patterns, following a supervised training phase using labeled antibody-antigen complex data to train a cross-attention-based regressor for predicting binding affinity changes. We evaluated ProtAttBA on three open benchmarks under different conditions. Compared to both sequence- and structure-based prediction methods, our approach achieves competitive performance, demonstrating notable robustness, especially with uncertain complex structures. Notably, our method possesses interpretability from the attention mechanism. We show that the learned attention scores can identify critical residues with impacts on binding affinity. This work introduces a rapid and cost-effective computational tool for antibody engineering, with the potential to accelerate the development of novel therapeutic antibodies.", "date": "2025-05-14", "categories": []}
{"id": "2505.20578", "title": "Ctrl-DNA: Controllable Cell-Type-Specific Regulatory DNA Design via Constrained RL", "abstract": "Designing regulatory DNA sequences that achieve precise cell-type-specific gene expression is crucial for advancements in synthetic biology, gene therapy and precision medicine. Although transformer-based language models (LMs) can effectively capture patterns in regulatory DNA, their generative approaches often struggle to produce novel sequences with reliable cell-specific activity. Here, we introduce Ctrl-DNA, a novel constrained reinforcement learning (RL) framework tailored for designing regulatory DNA sequences with controllable cell-type specificity. By formulating regulatory sequence design as a biologically informed constrained optimization problem, we apply RL to autoregressive genomic LMs, enabling the models to iteratively refine sequences that maximize regulatory activity in targeted cell types while constraining off-target effects. Our evaluation on human promoters and enhancers demonstrates that Ctrl-DNA consistently outperforms existing generative and RL-based approaches, generating high-fitness regulatory sequences and achieving state-of-the-art cell-type specificity. Moreover, Ctrl-DNA-generated sequences capture key cell-type-specific transcription factor binding sites (TFBS), short DNA motifs recognized by regulatory proteins that control gene expression, demonstrating the biological plausibility of the generated sequences.", "date": "2025-05-26", "categories": []}
{"id": "2505.21241", "title": "BindEnergyCraft: Casting Protein Structure Predictors as Energy-Based Models for Binder Design", "abstract": "Protein binder design has been transformed by hallucination-based methods that optimize structure prediction confidence metrics, such as the interface predicted TM-score (ipTM), via backpropagation. However, these metrics do not reflect the statistical likelihood of a binder-target complex under the learned distribution and yield sparse gradients for optimization. In this work, we propose a method to extract such likelihoods from structure predictors by reinterpreting their confidence outputs as an energy-based model (EBM). By leveraging the Joint Energy-based Modeling (JEM) framework, we introduce pTMEnergy, a statistical energy function derived from predicted inter-residue error distributions. We incorporate pTMEnergy into BindEnergyCraft (BECraft), a design pipeline that maintains the same optimization framework as BindCraft but replaces ipTM with our energy-based objective. BECraft outperforms BindCraft, RFDiffusion, and ESM3 across multiple challenging targets, achieving higher in silico binder success rates while reducing structural clashes. Furthermore, pTMEnergy establishes a new state-of-the-art in structure-based virtual screening tasks for miniprotein and RNA aptamer binders.", "date": "2025-05-27", "categories": []}
{"id": "2505.21452", "title": "Designing Cyclic Peptides via Harmonic SDE with Atom-Bond Modeling", "abstract": "Cyclic peptides offer inherent advantages in pharmaceuticals. For example, cyclic peptides are more resistant to enzymatic hydrolysis compared to linear peptides and usually exhibit excellent stability and affinity. Although deep generative models have achieved great success in linear peptide design, several challenges prevent the development of computational methods for designing diverse types of cyclic peptides. These challenges include the scarcity of 3D structural data on target proteins and associated cyclic peptide ligands, the geometric constraints that cyclization imposes, and the involvement of non-canonical amino acids in cyclization. To address the above challenges, we introduce CpSDE, which consists of two key components: AtomSDE, a generative structure prediction model based on harmonic SDE, and ResRouter, a residue type predictor. Utilizing a routed sampling algorithm that alternates between these two models to iteratively update sequences and structures, CpSDE facilitates the generation of cyclic peptides. By employing explicit all-atom and bond modeling, CpSDE overcomes existing data limitations and is proficient in designing a wide variety of cyclic peptides. Our experimental results demonstrate that the cyclic peptides designed by our method exhibit reliable stability and affinity.", "date": "2025-05-27", "categories": []}
{"id": "2302.07868", "title": "Target Specific De Novo Design of Drug Candidate Molecules with Graph Transformer-based Generative Adversarial Networks", "abstract": "Discovering novel drug candidate molecules is one of the most fundamental and critical steps in drug development. Generative deep learning models, which create synthetic data given a probability distribution, offer a high potential for designing de novo molecules. However, to be utilisable in real life drug development pipelines, these models should be able to design drug like and target centric molecules. In this study, we propose an end to end generative system, DrugGEN, for the de novo design of drug candidate molecules that interact with intended target proteins. The proposed method represents molecules as graphs and processes them via a generative adversarial network comprising graph transformer layers. The system is trained using a large dataset of drug like compounds and target specific bioactive molecules to design effective inhibitory molecules against the AKT1 protein, which is critically important in developing treatments for various types of cancer. We conducted molecular docking and dynamics to assess the target centric generation performance of the model, as well as attention score visualisation to examine model interpretability. In parallel, selected compounds were chemically synthesised and evaluated in the context of in vitro enzymatic assays, which identified two bioactive molecules that inhibited AKT1 at low micromolar concentrations. These results indicate that DrugGEN's de novo molecules have a high potential for interacting with the AKT1 protein at the level of its native ligands. Using the open access DrugGEN codebase, it is possible to easily train models for other druggable proteins, given a dataset of experimentally known bioactive molecules.", "date": "2023-02-15", "categories": []}
{"id": "2503.19300", "title": "UniMoMo: Unified Generative Modeling of 3D Molecules for De Novo Binder Design", "abstract": "The design of target-specific molecules such as small molecules, peptides, and antibodies is vital for biological research and drug discovery. Existing generative methods are restricted to single-domain molecules, failing to address versatile therapeutic needs or utilize cross-domain transferability to enhance model performance. In this paper, we introduce Unified generative Modeling of 3D Molecules (UniMoMo), the first framework capable of designing binders of multiple molecular domains using a single model. In particular, UniMoMo unifies the representations of different molecules as graphs of blocks, where each block corresponds to either a standard amino acid or a molecular fragment. Subsequently, UniMoMo utilizes a geometric latent diffusion model for 3D molecular generation, featuring an iterative full-atom autoencoder to compress blocks into latent space points, followed by an E(3)-equivariant diffusion process. Extensive benchmarks across peptides, antibodies, and small molecules demonstrate the superiority of our unified framework over existing domain-specific models, highlighting the benefits of multi-domain training.", "date": "2025-03-25", "categories": []}
{"id": "2505.19740", "title": "Machine Learning Algorithm for Noise Reduction and Disease-Causing Gene Feature Extraction in Gene Sequencing Data", "abstract": "In this study, we propose a machine learning-based method for noise reduction and disease-causing gene feature extraction in gene sequencing DeepSeqDenoise algorithm combines CNN and RNN to effectively remove the sequencing noise, and improves the signal-to-noise ratio by 9.4 dB. We screened 17 key features by feature engineering, and constructed an integrated learning model to predict disease-causing genes with 94.3% accuracy. We successfully identified 57 new candidate disease-causing genes in a cardiovascular disease cohort validation, and detected 3 missed variants in clinical applications. The method significantly outperforms existing tools and provides strong support for accurate diagnosis of genetic diseases.", "date": "2025-05-26", "categories": []}
{"id": "2505.18082", "title": "An Iterative Framework for Generative Backmapping of Coarse Grained Proteins", "abstract": "The techniques of data-driven backmapping from coarse-grained (CG) to fine-grained (FG) representation often struggle with accuracy, unstable training, and physical realism, especially when applied to complex systems such as proteins. In this work, we introduce a novel iterative framework by using conditional Variational Autoencoders and graph-based neural networks, specifically designed to tackle the challenges associated with such large-scale biomolecules. Our method enables stepwise refinement from CG beads to full atomistic details. We outline the theory of iterative generative backmapping and demonstrate via numerical experiments the advantages of multistep schemes by applying them to proteins of vastly different structures with very coarse representations. This multistep approach not only improves the accuracy of reconstructions but also makes the training process more computationally efficient for proteins with ultra-CG representations.", "date": "2025-05-23", "categories": []}
{"id": "2410.19236", "title": "SHAP zero Explains Biological Sequence Models with Near-zero Marginal Cost for Future Queries", "abstract": "The growing adoption of machine learning models for biological sequences has intensified the need for interpretable predictions, with Shapley values emerging as a theoretically grounded standard for model explanation. While effective for local explanations of individual input sequences, scaling Shapley-based interpretability to extract global biological insights requires evaluating thousands of sequences--incurring exponential computational cost per query. We introduce SHAP zero, a novel algorithm that amortizes the cost of Shapley value computation across large-scale biological datasets. After a one-time model sketching step, SHAP zero enables near-zero marginal cost for future queries by uncovering an underexplored connection between Shapley values, high-order feature interactions, and the sparse Fourier transform of the model. Applied to models of guide RNA efficacy, DNA repair outcomes, and protein fitness, SHAP zero explains predictions orders of magnitude faster than existing methods, recovering rich combinatorial interactions previously inaccessible at scale. This work opens the door to principled, efficient, and scalable interpretability for black-box sequence models in biology.", "date": "2024-10-25", "categories": []}
{"id": "2503.03360", "title": "Transformers for molecular property prediction: Domain adaptation efficiently improves performance", "abstract": "Over the past six years, molecular transformer models have become key tools in drug discovery. Most existing models are pre-trained on large, unlabeled datasets such as ZINC or ChEMBL. However, the extent to which large-scale pre-training improves molecular property prediction remains unclear. This study evaluates transformer models for this task while addressing their limitations. We explore how pre-training dataset size and chemically informed objectives impact performance. Our results show that increasing the dataset beyond approximately 400K to 800K molecules from large-scale unlabeled databases does not enhance performance across seven datasets covering five ADME endpoints: lipophilicity, permeability, solubility (two datasets), microsomal stability (two datasets), and plasma protein binding. In contrast, domain adaptation on a small, domain-specific dataset (less than or equal 4K molecules) using multi-task regression of physicochemical properties significantly boosts performance (P-value less than 0.001). A model pre-trained on 400K molecules and adapted with domain-specific data outperforms larger models such as MolFormer and performs comparably to MolBERT. Benchmarks against Random Forest (RF) baselines using descriptors and Morgan fingerprints show that chemically and physically informed features consistently yield better performance across model types. While RF remains a strong baseline, we identify concrete practices to enhance transformer performance. Aligning pre-training and adaptation with chemically meaningful tasks and domain-relevant data presents a promising direction for molecular property prediction. Our models are available on HuggingFace for easy use and adaptation.", "date": "2025-03-05", "categories": []}
{"id": "2505.15868", "title": "An Inclusive Foundation Model for Generalizable Cytogenetics in Precision Oncology", "abstract": "Chromosome analysis is vital for diagnosing genetic disorders and guiding cancer therapy decisions through the identification of somatic clonal aberrations. However, developing an AI model are hindered by the overwhelming complexity and diversity of chromosomal abnormalities, requiring extensive annotation efforts, while automated methods remain task-specific and lack generalizability due to the scarcity of comprehensive datasets spanning diverse resource conditions. Here, we introduce CHROMA, a foundation model for cytogenomics, designed to overcome these challenges by learning generalizable representations of chromosomal abnormalities. Pre-trained on over 84,000 specimens (~4 million chromosomal images) via self-supervised learning, CHROMA outperforms other methods across all types of abnormalities, even when trained on fewer labelled data and more imbalanced datasets. By facilitating comprehensive mapping of instability and clonal leisons across various aberration types, CHROMA offers a scalable and generalizable solution for reliable and automated clinical analysis, reducing the annotation workload for experts and advancing precision oncology through the early detection of rare genomic abnormalities, enabling broad clinical AI applications and making advanced genomic analysis more accessible.", "date": "2025-05-21", "categories": []}
{"id": "2505.16680", "title": "Learning Genomic Structure from $k$-mers", "abstract": "Sequencing a genome to determine an individual's DNA produces an enormous number of short nucleotide subsequences known as reads, which must be reassembled to reconstruct the full genome. We present a method for analyzing this type of data using contrastive learning, in which an encoder model is trained to produce embeddings that cluster together sequences from the same genomic region. The sequential nature of genomic regions is preserved in the form of trajectories through this embedding space. Trained solely to reflect the structure of the genome, the resulting model provides a general representation of $k$-mer sequences, suitable for a range of downstream tasks involving read data. We apply our framework to learn the structure of the $E.\\ coli$ genome, and demonstrate its use in simulated ancient DNA (aDNA) read mapping and identification of structural variations. Furthermore, we illustrate the potential of using this type of model for metagenomic species identification. We show how incorporating a domain-specific noise model can enhance embedding robustness, and how a supervised contrastive learning setting can be adopted when a linear reference genome is available, by introducing a distance thresholding parameter $\\Gamma$. The model can also be trained fully self-supervised on read data, enabling analysis without the need to construct a full genome assembly using specialized algorithms. Small prediction heads based on a pre-trained embedding are shown to perform on par with BWA-aln, the current gold standard approach for aDNA mapping, in terms of accuracy and runtime for short genomes. Given the method's favorable scaling properties with respect to total genome size, inference using our approach is highly promising for metagenomic applications and for mapping to genomes comparable in size to the human genome.", "date": "2025-05-22", "categories": []}
{"id": "2404.00050", "title": "Grappa -- A Machine Learned Molecular Mechanics Force Field", "abstract": "Simulating large molecular systems over long timescales requires force fields that are both accurate and efficient. In recent years, E(3) equivariant neural networks have lifted the tension between computational efficiency and accuracy of force fields, but they are still several orders of magnitude more expensive than established molecular mechanics (MM) force fields. Here, we propose Grappa, a machine learning framework to predict MM parameters from the molecular graph, employing a graph attentional neural network and a transformer with symmetry-preserving positional encoding. The resulting Grappa force field outperformstabulated and machine-learned MM force fields in terms of accuracy at the same computational efficiency and can be used in existing Molecular Dynamics (MD) engines like GROMACS and OpenMM. It predicts energies and forces of small molecules, peptides, RNA and - showcasing its extensibility to uncharted regions of chemical space - radicals at state-of-the-art MM accuracy. We demonstrate Grappa's transferability to macromolecules in MD simulations from a small fast folding protein up to a whole virus particle. Our force field sets the stage for biomolecular simulations closer to chemical accuracy, but with the same computational cost as established protein force fields.", "date": "2024-03-25", "categories": []}
{"id": "2505.14919", "title": "TxPert: Leveraging Biochemical Relationships for Out-of-Distribution Transcriptomic Perturbation Prediction", "abstract": "Accurately predicting cellular responses to genetic perturbations is essential for understanding disease mechanisms and designing effective therapies. Yet exhaustively exploring the space of possible perturbations (e.g., multi-gene perturbations or across tissues and cell types) is prohibitively expensive, motivating methods that can generalize to unseen conditions. In this work, we explore how knowledge graphs of gene-gene relationships can improve out-of-distribution (OOD) prediction across three challenging settings: unseen single perturbations; unseen double perturbations; and unseen cell lines. In particular, we present: (i) TxPert, a new state-of-the-art method that leverages multiple biological knowledge networks to predict transcriptional responses under OOD scenarios; (ii) an in-depth analysis demonstrating the impact of graphs, model architecture, and data on performance; and (iii) an expanded benchmarking framework that strengthens evaluation standards for perturbation modeling.", "date": "2025-05-20", "categories": []}
{"id": "2411.18463", "title": "Hotspot-Driven Peptide Design via Multi-Fragment Autoregressive Extension", "abstract": "Peptides, short chains of amino acids, interact with target proteins, making them a unique class of protein-based therapeutics for treating human diseases. Recently, deep generative models have shown great promise in peptide generation. However, several challenges remain in designing effective peptide binders. First, not all residues contribute equally to peptide-target interactions. Second, the generated peptides must adopt valid geometries due to the constraints of peptide bonds. Third, realistic tasks for peptide drug development are still lacking. To address these challenges, we introduce PepHAR, a hot-spot-driven autoregressive generative model for designing peptides targeting specific proteins. Building on the observation that certain hot spot residues have higher interaction potentials, we first use an energy-based density model to fit and sample these key residues. Next, to ensure proper peptide geometry, we autoregressively extend peptide fragments by estimating dihedral angles between residue frames. Finally, we apply an optimization process to iteratively refine fragment assembly, ensuring correct peptide structures. By combining hot spot sampling with fragment-based extension, our approach enables de novo peptide design tailored to a target protein and allows the incorporation of key hot spot residues into peptide scaffolds. Extensive experiments, including peptide design and peptide scaffold generation, demonstrate the strong potential of PepHAR in computational peptide binder design. Source code will be available at https://github.com/Ced3-han/PepHAR.", "date": "2024-11-26", "categories": []}
{"id": "2505.14402", "title": "OmniGenBench: A Modular Platform for Reproducible Genomic Foundation Models Benchmarking", "abstract": "The code of nature, embedded in DNA and RNA genomes since the origin of life, holds immense potential to impact both humans and ecosystems through genome modeling. Genomic Foundation Models (GFMs) have emerged as a transformative approach to decoding the genome. As GFMs scale up and reshape the landscape of AI-driven genomics, the field faces an urgent need for rigorous and reproducible evaluation. We present OmniGenBench, a modular benchmarking platform designed to unify the data, model, benchmarking, and interpretability layers across GFMs. OmniGenBench enables standardized, one-command evaluation of any GFM across five benchmark suites, with seamless integration of over 31 open-source models. Through automated pipelines and community-extensible features, the platform addresses critical reproducibility challenges, including data transparency, model interoperability, benchmark fragmentation, and black-box interpretability. OmniGenBench aims to serve as foundational infrastructure for reproducible genomic AI research, accelerating trustworthy discovery and collaborative innovation in the era of genome-scale modeling.", "date": "2025-05-20", "categories": []}
{"id": "2410.02647", "title": "Immunogenicity Prediction with Dual Attention Enables Vaccine Target Selection", "abstract": "Immunogenicity prediction is a central topic in reverse vaccinology for finding candidate vaccines that can trigger protective immune responses. Existing approaches typically rely on highly compressed features and simple model architectures, leading to limited prediction accuracy and poor generalizability. To address these challenges, we introduce VenusVaccine, a novel deep learning solution with a dual attention mechanism that integrates pre-trained latent vector representations of protein sequences and structures. We also compile the most comprehensive immunogenicity dataset to date, encompassing over 7000 antigen sequences, structures, and immunogenicity labels from bacteria, virus, and tumor. Extensive experiments demonstrate that VenusVaccine outperforms existing methods across a wide range of evaluation metrics. Furthermore, we establish a post-hoc validation protocol to assess the practical significance of deep learning models in tackling vaccine design challenges. Our work provides an effective tool for vaccine design and sets valuable benchmarks for future research. The implementation is at https://github.com/songleee/VenusVaccine.", "date": "2024-10-03", "categories": []}
{"id": "2505.11580", "title": "Flash Invariant Point Attention", "abstract": "Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models. However, its quadratic complexity limits the input sequence length. We introduce FlashIPA, a factorized reformulation of IPA that leverages hardware-efficient FlashAttention to achieve linear scaling in GPU memory and wall-clock time with sequence length. FlashIPA matches or exceeds standard IPA performance while substantially reducing computational costs. FlashIPA extends training to previously unattainable lengths, and we demonstrate this by re-training generative models without length restrictions and generating structures of thousands of residues. FlashIPA is available at https://github.com/flagshippioneering/flash_ipa.", "date": "2025-05-16", "categories": []}
{"id": "2505.11610", "title": "Foundation Models for AI-Enabled Biological Design", "abstract": "This paper surveys foundation models for AI-enabled biological design, focusing on recent developments in applying large-scale, self-supervised models to tasks such as protein engineering, small molecule design, and genomic sequence design. Though this domain is evolving rapidly, this survey presents and discusses a taxonomy of current models and methods. The focus is on challenges and solutions in adapting these models for biological applications, including biological sequence modeling architectures, controllability in generation, and multi-modal integration. The survey concludes with a discussion of open problems and future directions, offering concrete next-steps to improve the quality of biological sequence generation.", "date": "2025-05-16", "categories": []}
{"id": "2505.11812", "title": "VenusX: Unlocking Fine-Grained Functional Understanding of Proteins", "abstract": "Deep learning models have driven significant progress in predicting protein function and interactions at the protein level. While these advancements have been invaluable for many biological applications such as enzyme engineering and function annotation, a more detailed perspective is essential for understanding protein functional mechanisms and evaluating the biological knowledge captured by models. To address this demand, we introduce VenusX, the first large-scale benchmark for fine-grained functional annotation and function-based protein pairing at the residue, fragment, and domain levels. VenusX comprises three major task categories across six types of annotations, including residue-level binary classification, fragment-level multi-class classification, and pairwise functional similarity scoring for identifying critical active sites, binding sites, conserved sites, motifs, domains, and epitopes. The benchmark features over 878,000 samples curated from major open-source databases such as InterPro, BioLiP, and SAbDab. By providing mixed-family and cross-family splits at three sequence identity thresholds, our benchmark enables a comprehensive assessment of model performance on both in-distribution and out-of-distribution scenarios. For baseline evaluation, we assess a diverse set of popular and open-source models, including pre-trained protein language models, sequence-structure hybrids, structure-based methods, and alignment-based techniques. Their performance is reported across all benchmark datasets and evaluation settings using multiple metrics, offering a thorough comparison and a strong foundation for future research. Code and data are publicly available at https://github.com/ai4protein/VenusX.", "date": "2025-05-17", "categories": []}
{"id": "2505.12358", "title": "AbFlowNet: Optimizing Antibody-Antigen Binding Energy via Diffusion-GFlowNet Fusion", "abstract": "Complementarity Determining Regions (CDRs) are critical segments of an antibody that facilitate binding to specific antigens. Current computational methods for CDR design utilize reconstruction losses and do not jointly optimize binding energy, a crucial metric for antibody efficacy. Rather, binding energy optimization is done through computationally expensive Online Reinforcement Learning (RL) pipelines rely heavily on unreliable binding energy estimators. In this paper, we propose AbFlowNet, a novel generative framework that integrates GFlowNet with Diffusion models. By framing each diffusion step as a state in the GFlowNet framework, AbFlowNet jointly optimizes standard diffusion losses and binding energy by directly incorporating energy signals into the training process, thereby unifying diffusion and reward optimization in a single procedure. Experimental results show that AbFlowNet outperforms the base diffusion model by 3.06% in amino acid recovery, 20.40% in geometric reconstruction (RMSD), and 3.60% in binding energy improvement ratio. ABFlowNet also decreases Top-1 total energy and binding energy errors by 24.8% and 38.1% without pseudo-labeling the test dataset or using computationally expensive online RL regimes.", "date": "2025-05-18", "categories": []}
{"id": "2505.12511", "title": "DS-ProGen: A Dual-Structure Deep Language Model for Functional Protein Design", "abstract": "Inverse Protein Folding (IPF) is a critical subtask in the field of protein design, aiming to engineer amino acid sequences capable of folding correctly into a specified three-dimensional (3D) conformation. Although substantial progress has been achieved in recent years, existing methods generally rely on either backbone coordinates or molecular surface features alone, which restricts their ability to fully capture the complex chemical and geometric constraints necessary for precise sequence prediction. To address this limitation, we present DS-ProGen, a dual-structure deep language model for functional protein design, which integrates both backbone geometry and surface-level representations. By incorporating backbone coordinates as well as surface chemical and geometric descriptors into a next-amino-acid prediction paradigm, DS-ProGen is able to generate functionally relevant and structurally stable sequences while satisfying both global and local conformational constraints. On the PRIDE dataset, DS-ProGen attains the current state-of-the-art recovery rate of 61.47%, demonstrating the synergistic advantage of multi-modal structural encoding in protein design. Furthermore, DS-ProGen excels in predicting interactions with a variety of biological partners, including ligands, ions, and RNA, confirming its robust functional retention capabilities.", "date": "2025-05-18", "categories": []}
{"id": "2505.10711", "title": "GNN-Suite: a Graph Neural Network Benchmarking Framework for Biomedical Informatics", "abstract": "We present GNN-Suite, a robust modular framework for constructing and benchmarking Graph Neural Network (GNN) architectures in computational biology. GNN-Suite standardises experimentation and reproducibility using the Nextflow workflow to evaluate GNN performance. We demonstrate its utility in identifying cancer-driver genes by constructing molecular networks from protein-protein interaction (PPI) data from STRING and BioGRID and annotating nodes with features from the PCAWG, PID, and COSMIC-CGC repositories.\n  Our design enables fair comparisons among diverse GNN architectures including GAT, GAT3H, GCN, GCN2, GIN, GTN, HGCN, PHGCN, and GraphSAGE and a baseline Logistic Regression (LR) model. All GNNs were configured as standardised two-layer models and trained with uniform hyperparameters (dropout = 0.2; Adam optimiser with learning rate = 0.01; and an adjusted binary cross-entropy loss to address class imbalance) over an 80/20 train-test split for 300 epochs. Each model was evaluated over 10 independent runs with different random seeds to yield statistically robust performance metrics, with balanced accuracy (BACC) as the primary measure. Notably, GCN2 achieved the highest BACC (0.807 +/- 0.035) on a STRING-based network, although all GNN types outperformed the LR baseline, highlighting the advantage of network-based learning over feature-only approaches.\n  Our results show that a common framework for implementing and evaluating GNN architectures aids in identifying not only the best model but also the most effective means of incorporating complementary data. By making GNN-Suite publicly available, we aim to foster reproducible research and promote improved benchmarking standards in computational biology. Future work will explore additional omics datasets and further refine network architectures to enhance predictive accuracy and interpretability in biomedical applications.", "date": "2025-05-15", "categories": []}
{"id": "2505.11185", "title": "VitaGraph: Building a Knowledge Graph for Biologically Relevant Learning Tasks", "abstract": "The intrinsic complexity of human biology presents ongoing challenges to scientific understanding. Researchers collaborate across disciplines to expand our knowledge of the biological interactions that define human life. AI methodologies have emerged as powerful tools across scientific domains, particularly in computational biology, where graph data structures effectively model biological entities such as protein-protein interaction (PPI) networks and gene functional networks. Those networks are used as datasets for paramount network medicine tasks, such as gene-disease association prediction, drug repurposing, and polypharmacy side effect studies. Reliable predictions from machine learning models require high-quality foundational data. In this work, we present a comprehensive multi-purpose biological knowledge graph constructed by integrating and refining multiple publicly available datasets. Building upon the Drug Repurposing Knowledge Graph (DRKG), we define a pipeline tasked with a) cleaning inconsistencies and redundancies present in DRKG, b) coalescing information from the main available public data sources, and c) enriching the graph nodes with expressive feature vectors such as molecular fingerprints and gene ontologies. Biologically and chemically relevant features improve the capacity of machine learning models to generate accurate and well-structured embedding spaces. The resulting resource represents a coherent and reliable biological knowledge graph that serves as a state-of-the-art platform to advance research in computational biology and precision medicine. Moreover, it offers the opportunity to benchmark graph-based machine learning and network medicine models on relevant tasks. We demonstrate the effectiveness of the proposed dataset by benchmarking it against the task of drug repurposing, PPI prediction, and side-effect prediction, modeled as link prediction problems.", "date": "2025-05-16", "categories": []}
{"id": "2410.15229", "title": "Deep Learning-based Detection of Bacterial Swarm Motion Using a Single Image", "abstract": "Distinguishing between swarming and swimming, the two principal forms of bacterial movement, holds significant conceptual and clinical relevance. This is because bacteria that exhibit swarming capabilities often possess unique properties crucial to the pathogenesis of infectious diseases and may also have therapeutic potential. Here, we report a deep learning-based swarming classifier that rapidly and autonomously predicts swarming probability using a single blurry image. Compared with traditional video-based, manually-processed approaches, our method is particularly suited for high-throughput environments and provides objective, quantitative assessments of swarming probability. The swarming classifier demonstrated in our work was trained on Enterobacter sp. SM3 and showed good performance when blindly tested on new swarming (positive) and swimming (negative) test images of SM3, achieving a sensitivity of 97.44% and a specificity of 100%. Furthermore, this classifier demonstrated robust external generalization capabilities when applied to unseen bacterial species, such as Serratia marcescens DB10 and Citrobacter koseri H6. It blindly achieved a sensitivity of 97.92% and a specificity of 96.77% for DB10, and a sensitivity of 100% and a specificity of 97.22% for H6. This competitive performance indicates the potential to adapt our approach for diagnostic applications through portable devices or even smartphones. This adaptation would facilitate rapid, objective, on-site screening for bacterial swarming motility, potentially enhancing the early detection and treatment assessment of various diseases, including inflammatory bowel diseases (IBD) and urinary tract infections (UTI).", "date": "2024-10-19", "categories": []}
{"id": "2408.09106", "title": "Fragment-Masked Diffusion for Molecular Optimization", "abstract": "Molecular optimization is a crucial aspect of drug discovery, aimed at refining molecular structures to enhance drug efficacy and minimize side effects, ultimately accelerating the overall drug development process. Many molecular optimization methods have been proposed, significantly advancing drug discovery. These methods primarily on understanding the specific drug target structures or their hypothesized roles in combating diseases. However, challenges such as a limited number of available targets and a difficulty capturing clear structures hinder innovative drug development. In contrast, phenotypic drug discovery (PDD) does not depend on clear target structures and can identify hits with novel and unbiased polypharmacology signatures. As a result, PDD-based molecular optimization can reduce potential safety risks while optimizing phenotypic activity, thereby increasing the likelihood of clinical success. Therefore, we propose a fragment-masked molecular optimization method based on PDD (FMOP). FMOP employs a regression-free diffusion model to conditionally optimize the molecular masked regions, effectively generating new molecules with similar scaffolds. On the large-scale drug response dataset GDSCv2, we optimize the potential molecules across all 985 cell lines. The overall experiments demonstrate that the in-silico optimization success rate reaches 95.4\\%, with an average efficacy increase of 7.5\\%. Additionally, we conduct extensive ablation and visualization experiments, confirming that FMOP is an effective and robust molecular optimization method. The code is available at: https://anonymous.4open.science/r/FMOP-98C2.", "date": "2024-08-17", "categories": []}
{"id": "2505.07086", "title": "Multi-Objective-Guided Discrete Flow Matching for Controllable Biological Sequence Design", "abstract": "Designing biological sequences that satisfy multiple, often conflicting, functional and biophysical criteria remains a central challenge in biomolecule engineering. While discrete flow matching models have recently shown promise for efficient sampling in high-dimensional sequence spaces, existing approaches address only single objectives or require continuous embeddings that can distort discrete distributions. We present Multi-Objective-Guided Discrete Flow Matching (MOG-DFM), a general framework to steer any pretrained discrete flow matching generator toward Pareto-efficient trade-offs across multiple scalar objectives. At each sampling step, MOG-DFM computes a hybrid rank-directional score for candidate transitions and applies an adaptive hypercone filter to enforce consistent multi-objective progression. We also trained two unconditional discrete flow matching models, PepDFM for diverse peptide generation and EnhancerDFM for functional enhancer DNA generation, as base generation models for MOG-DFM. We demonstrate MOG-DFM's effectiveness in generating peptide binders optimized across five properties (hemolysis, non-fouling, solubility, half-life, and binding affinity), and in designing DNA sequences with specific enhancer classes and DNA shapes. In total, MOG-DFM proves to be a powerful tool for multi-property-guided biomolecule sequence design.", "date": "2025-05-11", "categories": []}
{"id": "2505.08918", "title": "When repeats drive the vocabulary: a Byte-Pair Encoding analysis of T2T primate genomes", "abstract": "The emergence of telomere-to-telomere (T2T) genome assemblies has opened new avenues for comparative genomics, yet effective tokenization strategies for genomic sequences remain underexplored. In this pilot study, we apply Byte Pair Encoding (BPE) to nine T2T primate genomes including three human assemblies by training independent BPE tokenizers with a fixed vocabulary of 512,000 tokens using our custom tool, dnaBPE. Our analysis reveals that only 11,569 tokens are shared across all assemblies, while nearly 991,854 tokens are unique to a single genome, indicating a rapid decline in shared vocabulary with increasing assembly comparisons. Moreover, phylogenetic trees derived from token overlap failed to recapitulate established primate relationships, a discrepancy attributed to the disproportionate influence of species-specific high-copy repetitive elements. These findings underscore the dual nature of BPE tokenization: while it effectively compresses repetitive sequences, its sensitivity to high-copy elements limits its utility as a universal tool for comparative genomics. We discuss potential hybrid strategies and repeat-masking approaches to refine genomic tokenization, emphasizing the need for domain-specific adaptations in the development of large-scale genomic language models. The dnaBPE tool used in this study is open-source and available at https://github.com/aglabx/dnaBPE.", "date": "2025-05-13", "categories": []}
{"id": "2505.09087", "title": "A Comparative Review of RNA Language Models", "abstract": "Given usefulness of protein language models (LMs) in structure and functional inference, RNA LMs have received increased attentions in the last few years. However, these RNA models are often not compared against the same standard. Here, we divided RNA LMs into three classes (pretrained on multiple RNA types (especially noncoding RNAs), specific-purpose RNAs, and LMs that unify RNA with DNA or proteins or both) and compared 13 RNA LMs along with 3 DNA and 1 protein LMs as controls in zero-shot prediction of RNA secondary structure and functional classification. Results shows that the models doing well on secondary structure prediction often perform worse in function classification or vice versa, suggesting that more balanced unsupervised training is needed.", "date": "2025-05-14", "categories": []}
{"id": "2504.14361", "title": "Integrating Single-Cell Foundation Models with Graph Neural Networks for Drug Response Prediction", "abstract": "AI-driven drug response prediction holds great promise for advancing personalized cancer treatment. However, the inherent heterogenity of cancer and high cost of data generation make accurate prediction challenging. In this study, we investigate whether incorporating the pretrained foundation model scGPT can enhance the performance of existing drug response prediction frameworks. Our approach builds on the DeepCDR framework, which encodes drug representations from graph structures and cell representations from multi-omics profiles. We adapt this framework by leveraging scGPT to generate enriched cell representations using its pretrained knowledge to compensate for limited amount of data. We evaluate our modified framework using IC$_{50}$ values on Pearson correlation coefficient (PCC) and a leave-one-drug out validation strategy, comparing it against the original DeepCDR framework and a prior scFoundation-based approach. scGPT not only outperforms previous approaches but also exhibits greater training stability, highlighting the value of leveraging scGPT-derived knowledge in this domain.", "date": "2025-04-19", "categories": []}
{"id": "2505.08608", "title": "Automated Model-Free Sorting of Single-Molecule Fluorescence Events Using a Deep Learning Based Hidden-State Model", "abstract": "Single-molecule fluorescence assays enable high-resolution analysis of biomolecular dynamics, but traditional analysis pipelines are labor-intensive and rely on users' experience, limiting scalability and reproducibility. Recent deep learning models have automated aspects of data processing, yet many still require manual thresholds, complex architectures, or extensive labeled data. Therefore, we present DASH, a fully streamlined architecture for trace classification, state assignment, and automatic sorting that requires no user input. DASH demonstrates robust performance across users and experimental conditions both in equilibrium and non-equilibrium systems such as Cas12a-mediated DNA cleavage. This paper proposes a novel strategy for the automatic and detailed sorting of single-molecule fluorescence events. The dynamic cleavage process of Cas12a is used as an example to provide a comprehensive analysis. This approach is crucial for studying biokinetic structural changes at the single-molecule level.", "date": "2025-05-13", "categories": []}
{"id": "2502.00694", "title": "Leveraging Large Language Models to Predict Antibody Biological Activity Against Influenza A Hemagglutinin", "abstract": "Monoclonal antibodies (mAbs) represent one of the most prevalent FDA-approved modalities for treating autoimmune diseases, infectious diseases, and cancers. However, discovery and development of therapeutic antibodies remains a time-consuming and expensive process. Recent advancements in machine learning (ML) and artificial intelligence (AI) have shown significant promise in revolutionizing antibody discovery and optimization. In particular, models that predict antibody biological activity enable in-silico evaluation of binding and functional properties; such models can prioritize antibodies with the highest likelihoods of success in costly and time-intensive laboratory testing procedures. We here explore an AI model for predicting the binding and receptor blocking activity of antibodies against influenza A hemagglutinin (HA) antigens. Our present model is developed with the MAMMAL framework for biologics discovery to predict antibody-antigen interactions using only sequence information. To evaluate the model's performance, we tested it under various data split conditions to mimic real-world scenarios.\n  Our models achieved an AUROC $\\geq$ 0.91 for predicting the activity of existing antibodies against seen HAs and an AUROC of 0.9 for unseen HAs. For novel antibody activity prediction, the AUROC was 0.73, which further declined to 0.63-0.66 under stringent constraints on similarity to existing antibodies. These results demonstrate the potential of AI foundation models to transform antibody design by reducing dependence on extensive laboratory testing and enabling more efficient prioritization of antibody candidates. Moreover, our findings emphasize the critical importance of diverse and comprehensive antibody datasets to improve the generalization of prediction models, particularly for novel antibody development.", "date": "2025-02-02", "categories": []}
{"id": "2503.09085", "title": "Differentiable Folding for Nearest Neighbor Model Optimization", "abstract": "The Nearest Neighbor model is the $\\textit{de facto}$ thermodynamic model of RNA secondary structure formation and is a cornerstone of RNA structure prediction and sequence design. The current functional form (Turner 2004) contains $\\approx13,000$ underlying thermodynamic parameters, and fitting these to both experimental and structural data is computationally challenging. Here, we leverage recent advances in $\\textit{differentiable folding}$, a method for directly computing gradients of the RNA folding algorithms, to devise an efficient, scalable, and flexible means of parameter optimization that uses known RNA structures and thermodynamic experiments. Our method yields a significantly improved parameter set that outperforms existing baselines on all metrics, including an increase in the average predicted probability of ground-truth sequence-structure pairs for a single RNA family by over 23 orders of magnitude. Our framework provides a path towards drastically improved RNA models, enabling the flexible incorporation of new experimental data, definition of novel loss terms, large training sets, and even treatment as a module in larger deep learning pipelines. We make available a new database, RNAometer, with experimentally-determined stabilities for small RNA model systems.", "date": "2025-03-12", "categories": []}
{"id": "2504.06304", "title": "Leveraging State Space Models in Long Range Genomics", "abstract": "Long-range dependencies are critical for understanding genomic structure and function, yet most conventional methods struggle with them. Widely adopted transformer-based models, while excelling at short-context tasks, are limited by the attention module's quadratic computational complexity and inability to extrapolate to sequences longer than those seen in training. In this work, we explore State Space Models (SSMs) as a promising alternative by benchmarking two SSM-inspired architectures, Caduceus and Hawk, on long-range genomics modeling tasks under conditions parallel to a 50M parameter transformer baseline. We discover that SSMs match transformer performance and exhibit impressive zero-shot extrapolation across multiple tasks, handling contexts 10 to 100 times longer than those seen during training, indicating more generalizable representations better suited for modeling the long and complex human genome. Moreover, we demonstrate that these models can efficiently process sequences of 1M tokens on a single GPU, allowing for modeling entire genomic regions at once, even in labs with limited compute. Our findings establish SSMs as efficient and scalable for long-context genomic analysis.", "date": "2025-04-07", "categories": []}
{"id": "2504.20127", "title": "Learning Hierarchical Interaction for Accurate Molecular Property Prediction", "abstract": "Discovering molecules with desirable molecular properties, including ADMET profiles, is of great importance in drug discovery. Existing approaches typically employ deep learning models, such as Graph Neural Networks (GNNs) and Transformers, to predict these molecular properties by learning from diverse chemical information. However, these models often fail to efficiently capture and utilize the hierarchical nature of molecular structures, and often lack mechanisms for effective interaction among multi-level features. To address these limitations, we propose a Hierarchical Interaction Message Passing Mechanism, which serves as the foundation of our novel model, the Hierarchical Interaction Message Net (HimNet). Our method enables interaction-aware representation learning across atomic, motif, and molecular levels via hierarchical attention-guided message passing. This design allows HimNet to effectively balance global and local information, ensuring rich and task-relevant feature extraction for downstream property prediction tasks, such as Blood-Brain Barrier Permeability (BBBP). We systematically evaluate HimNet on eleven datasets, including eight widely-used MoleculeNet benchmarks and three challenging, high-value datasets for metabolic stability, malaria activity, and liver microsomal clearance, covering a broad range of pharmacologically relevant properties. Extensive experiments demonstrate that HimNet achieves the best or near-best performance in most molecular property prediction tasks. Furthermore, our method exhibits promising hierarchical interpretability, aligning well with chemical intuition on representative molecules. We believe that HimNet offers an accurate and efficient solution for molecular activity and ADMET property prediction, contributing to advanced decision-making in the early stages of drug discovery.", "date": "2025-04-28", "categories": []}
{"id": "2505.06534", "title": "GBDTSVM: Combined Support Vector Machine and Gradient Boosting Decision Tree Framework for efficient snoRNA-disease association prediction", "abstract": "Small nucleolar RNAs (snoRNAs) are increasingly recognized for their critical role in the pathogenesis and characterization of various human diseases. Consequently, the precise identification of snoRNA-disease associations (SDAs) is essential for the progression of diseases and the advancement of treatment strategies. However, conventional biological experimental approaches are costly, time-consuming, and resource-intensive; therefore, machine learning-based computational methods offer a promising solution to mitigate these limitations. This paper proposes a model called 'GBDTSVM', representing a novel and efficient machine learning approach for predicting snoRNA-disease associations by leveraging a Gradient Boosting Decision Tree (GBDT) and Support Vector Machine (SVM). 'GBDTSVM' effectively extracts integrated snoRNA-disease feature representations utilizing GBDT and SVM is subsequently utilized to classify and identify potential associations. Furthermore, the method enhances the accuracy of these predictions by incorporating Gaussian kernel profile similarity for both snoRNAs and diseases. Experimental evaluation of the GBDTSVM model demonstrated superior performance compared to state-of-the-art methods in the field, achieving an area under the receiver operating characteristic (AUROC) of 0.96 and an area under the precision-recall curve (AUPRC) of 0.95 on MDRF dataset. Moreover, our model shows superior performance on two more datasets named LSGT and PsnoD. Additionally, a case study on the predicted snoRNA-disease associations verified the top 10 predicted snoRNAs across nine prevalent diseases, further validating the efficacy of the GBDTSVM approach. These results underscore the model's potential as a robust tool for advancing snoRNA-related disease research. Source codes and datasets our proposed framework can be obtained from: https://github.com/mariamuna04/gbdtsvm", "date": "2025-05-10", "categories": []}
{"id": "2409.09342", "title": "Persistent pseudopod splitting is an effective chemotaxis strategy in shallow gradients", "abstract": "Single-cell organisms and various cell types use a range of motility modes when following a chemical gradient, but it is unclear which mode is best suited for different gradients. Here, we model directional decision-making in chemotactic amoeboid cells as a stimulus-dependent actin recruitment contest. Pseudopods extending from the cell body compete for a finite actin pool to push the cell in their direction until one pseudopod wins and determines the direction of movement. Our minimal model provides a quantitative understanding of the strategies cells use to reach the physical limit of accurate chemotaxis, aligning with data without explicit gradient sensing or cellular memory for persistence. To generalize our model, we employ reinforcement learning optimization to study the effect of pseudopod suppression, a simple but effective cellular algorithm by which cells can suppress possible directions of movement. Different pseudopod-based chemotaxis strategies emerge naturally depending on the environment and its dynamics. For instance, in static gradients, cells can react faster at the cost of pseudopod accuracy, which is particularly useful in noisy, shallow gradients where it paradoxically increases chemotactic accuracy. In contrast, in dynamics gradients, cells form de novo pseudopods. Overall, our work demonstrates mechanical intelligence for high chemotaxis performance with minimal cellular regulation.", "date": "2024-09-14", "categories": []}
{"id": "2505.05612", "title": "scDrugMap: Benchmarking Large Foundation Models for Drug Response Prediction", "abstract": "Drug resistance presents a major challenge in cancer therapy. Single cell profiling offers insights into cellular heterogeneity, yet the application of large-scale foundation models for predicting drug response in single cell data remains underexplored. To address this, we developed scDrugMap, an integrated framework featuring both a Python command-line interface and a web server for drug response prediction. scDrugMap evaluates a wide range of foundation models, including eight single-cell models and two large language models, using a curated dataset of over 326,000 cells in the primary collection and 18,800 cells in the validation set, spanning 36 datasets and diverse tissue and cancer types. We benchmarked model performance under pooled-data and cross-data evaluation settings, employing both layer freezing and Low-Rank Adaptation (LoRA) fine-tuning strategies. In the pooled-data scenario, scFoundation achieved the best performance, with mean F1 scores of 0.971 (layer freezing) and 0.947 (fine-tuning), outperforming the lowest-performing model by over 50%. In the cross-data setting, UCE excelled post fine-tuning (mean F1: 0.774), while scGPT led in zero-shot learning (mean F1: 0.858). Overall, scDrugMap provides the first large-scale benchmark of foundation models for drug response prediction in single-cell data and serves as a user-friendly, flexible platform for advancing drug discovery and translational research.", "date": "2025-05-08", "categories": []}
{"id": "2505.05616", "title": "Leveraging Large Language Models for enzymatic reaction prediction and characterization", "abstract": "Predicting enzymatic reactions is crucial for applications in biocatalysis, metabolic engineering, and drug discovery, yet it remains a complex and resource-intensive task. Large Language Models (LLMs) have recently demonstrated remarkable success in various scientific domains, e.g., through their ability to generalize knowledge, reason over complex structures, and leverage in-context learning strategies. In this study, we systematically evaluate the capability of LLMs, particularly the Llama-3.1 family (8B and 70B), across three core biochemical tasks: Enzyme Commission number prediction, forward synthesis, and retrosynthesis. We compare single-task and multitask learning strategies, employing parameter-efficient fine-tuning via LoRA adapters. Additionally, we assess performance across different data regimes to explore their adaptability in low-data settings. Our results demonstrate that fine-tuned LLMs capture biochemical knowledge, with multitask learning enhancing forward- and retrosynthesis predictions by leveraging shared enzymatic information. We also identify key limitations, for example challenges in hierarchical EC classification schemes, highlighting areas for further improvement in LLM-driven biochemical modeling.", "date": "2025-05-08", "categories": []}
{"id": "2505.05874", "title": "A 3D pocket-aware and evolutionary conserved interaction guided diffusion model for molecular optimization", "abstract": "Generating molecules that bind to specific protein targets via diffusion models has shown good promise for structure-based drug design and molecule optimization. Especially, the diffusion models with binding interaction guidance enables molecule generation with high affinity through forming favorable interaction within protein pocket. However, the generated molecules may not form interactions with the highly conserved residues, which are important for protein functions and bioactivities of the ligands. Herein, we developed a new 3D target-aware diffusion model DiffDecip, which explicitly incorporates the protein-ligand binding interactions and evolutionary conservation information of protein residues into both diffusion and sampling process, for molecule optimization through scaffold decoration. The model performance revealed that DiffDecip outperforms baseline model DiffDec on molecule optimization towards higher affinity through forming more non-covalent interactions with highly conserved residues in the protein pocket.", "date": "2025-05-09", "categories": []}
{"id": "2505.03853", "title": "GRAPE: Heterogeneous Graph Representation Learning for Genetic\n  Perturbation with Coding and Non-Coding Biotype", "abstract": "  Predicting genetic perturbations enables the identification of potentially\ncrucial genes prior to wet-lab experiments, significantly improving overall\nexperimental efficiency. Since genes are the foundation of cellular life,\nbuilding gene regulatory networks (GRN) is essential to understand and predict\nthe effects of genetic perturbations. However, current methods fail to fully\nleverage gene-related information, and solely rely on simple evaluation metrics\nto construct coarse-grained GRN. More importantly, they ignore functional\ndifferences between biotypes, limiting the ability to capture potential gene\ninteractions. In this work, we leverage pre-trained large language model and\nDNA sequence model to extract features from gene descriptions and DNA sequence\ndata, respectively, which serve as the initialization for gene representations.\nAdditionally, we introduce gene biotype information for the first time in\ngenetic perturbation, simulating the distinct roles of genes with different\nbiotypes in regulating cellular processes, while capturing implicit gene\nrelationships through graph structure learning (GSL). We propose GRAPE, a\nheterogeneous graph neural network (HGNN) that leverages gene representations\ninitialized with features from descriptions and sequences, models the distinct\nroles of genes with different biotypes, and dynamically refines the GRN through\nGSL. The results on publicly available datasets show that our method achieves\nstate-of-the-art performance.\n", "date": "2025-05-06", "categories": []}
{"id": "2505.02887", "title": "CreoPep: A Universal Deep Learning Framework for Target-Specific Peptide\n  Design and Optimization", "abstract": "  Target-specific peptides, such as conotoxins, exhibit exceptional binding\naffinity and selectivity toward ion channels and receptors. However, their\ntherapeutic potential remains underutilized due to the limited diversity of\nnatural variants and the labor-intensive nature of traditional optimization\nstrategies. Here, we present CreoPep, a deep learning-based conditional\ngenerative framework that integrates masked language modeling with a\nprogressive masking scheme to design high-affinity peptide mutants while\nuncovering novel structural motifs. CreoPep employs an integrative augmentation\npipeline, combining FoldX-based energy screening with temperature-controlled\nmultinomial sampling, to generate structurally and functionally diverse\npeptides that retain key pharmacological properties. We validate this approach\nby designing conotoxin inhibitors targeting the $\\alpha$7 nicotinic\nacetylcholine receptor, achieving submicromolar potency in electrophysiological\nassays. Structural analysis reveals that CreoPep-generated variants engage in\nboth conserved and novel binding modes, including disulfide-deficient forms,\nthus expanding beyond conventional design paradigms. Overall, CreoPep offers a\nrobust and generalizable platform that bridges computational peptide design\nwith experimental validation, accelerating the discovery of next-generation\npeptide therapeutics.\n", "date": "2025-05-05", "categories": []}
{"id": "2505.03177", "title": "A Symbolic and Statistical Learning Framework to Discover Bioprocessing\n  Regulatory Mechanism: Cell Culture Example", "abstract": "  Bioprocess mechanistic modeling is essential for advancing intelligent\ndigital twin representation of biomanufacturing, yet challenges persist due to\ncomplex intracellular regulation, stochastic system behavior, and limited\nexperimental data. This paper introduces a symbolic and statistical learning\nframework to identify key regulatory mechanisms and quantify model uncertainty.\nBioprocess dynamics is formulated with stochastic differential equations\ncharacterizing intrinsic process variability, with a predefined set of\ncandidate regulatory mechanisms constructed from biological knowledge. A\nBayesian learning approach is developed, which is based on a joint learning of\nkinetic parameters and regulatory structure through a formulation of the\nmixture model. To enhance computational efficiency, a Metropolis-adjusted\nLangevin algorithm with adjoint sensitivity analysis is developed for posterior\nexploration. Compared to state-of-the-art Bayesian inference approaches, the\nproposed framework achieves improved sample efficiency and robust model\nselection. An empirical study demonstrates its ability to recover missing\nregulatory mechanisms and improve model fidelity under data-limited conditions.\n", "date": "2025-05-06", "categories": []}
{"id": "2505.02110", "title": "Eterna is Solved", "abstract": "  RNA design consists of discovering a nucleotide sequence that folds into a\ntarget secondary structure. It is useful for synthetic biology, medicine, and\nnanotechnology. We propose Montparnasse, a Multi Objective Generalized Nested\nRollout Policy Adaptation with Limited Repetition (MOGNRPALR) RNA design\nalgorithm. It solves the Eterna benchmark.\n", "date": "2025-05-04", "categories": []}
{"id": "2503.11347", "title": "Integrating Dynamical Systems Modeling with Spatiotemporal scRNA-seq\n  Data Analysis", "abstract": "  Understanding the dynamic nature of biological systems is fundamental to\ndeciphering cellular behavior, developmental processes, and disease\nprogression. Single-cell RNA sequencing (scRNA-seq) has provided static\nsnapshots of gene expression, offering valuable insights into cellular states\nat a single time point. Recent advancements in temporally resolved scRNA-seq,\nspatial transcriptomics (ST), and time-series spatial transcriptomics\n(temporal-ST) have further revolutionized our ability to study the\nspatiotemporal dynamics of individual cells. These technologies, when combined\nwith computational frameworks such as Markov chains, stochastic differential\nequations (SDEs), and generative models like optimal transport and\nSchr\\\"odinger bridges, enable the reconstruction of dynamic cellular\ntrajectories and cell fate decisions. This review discusses how these dynamical\nsystem approaches offer new opportunities to model and infer cellular dynamics\nfrom a systematic perspective.\n", "date": "2025-03-14", "categories": []}
{"id": "2505.00236", "title": "Node2Vec-DGI-EL: A Hierarchical Graph Representation Learning Model for\n  Ingredient-Disease Association Prediction", "abstract": "  Traditional Chinese medicine, as an essential component of traditional\nmedicine, contains active ingredients that serve as a crucial source for modern\ndrug development, holding immense therapeutic potential and development value.\nA multi-layered and complex network is formed from Chinese medicine to diseases\nand used to predict the potential associations between Chinese medicine\ningredients and diseases. This study proposes an ingredient-disease association\nprediction model (Node2Vec-DGI-EL) based on hierarchical graph representation\nlearning. First, the model uses the Node2Vec algorithm to extract node\nembedding vectors from the network as the initial features of the nodes. Next,\nthe network nodes are deeply represented and learned using the DGI algorithm to\nenhance the model's expressive power. To improve prediction accuracy and\nrobustness, an ensemble learning method is incorporated to achieve more\naccurate ingredient-disease association predictions. The effectiveness of the\nmodel is then evaluated through a series of theoretical verifications. The\nresults demonstrated that the proposed model significantly outperformed\nexisting methods, achieving an AUC of 0.9987 and an AUPR of 0.9545, thereby\nindicating superior predictive capability. Ablation experiments further\nrevealed the contribution and importance of each module. Additionally, case\nstudies explored potential associations, such as triptonide with hypertensive\nretinopathy and methyl ursolate with colorectal cancer. Molecular docking\nexperiments validated these findings, showing the triptonide-PGR interaction\nand the methyl ursolate-NFE2L2 interaction can bind stable. In conclusion, the\nNode2Vec-DGI-EL model focuses on TCM datasets and effectively predicts\ningredient-disease associations, overcoming the reliance on node semantic\ninformation.\n", "date": "2025-05-01", "categories": []}
{"id": "2505.00530", "title": "Leveraging Partial SMILES Validation Scheme for Enhanced Drug Design in\n  Reinforcement Learning Frameworks", "abstract": "  SMILES-based molecule generation has emerged as a powerful approach in drug\ndiscovery. Deep reinforcement learning (RL) using large language model (LLM)\nhas been incorporated into the molecule generation process to achieve high\nmatching score in term of likelihood of desired molecule candidates. However, a\ncritical challenge in this approach is catastrophic forgetting during the RL\nphase, where knowledge such as molecule validity, which often exceeds 99\\%\nduring pretraining, significantly deteriorates. Current RL algorithms applied\nin drug discovery, such as REINVENT, use prior models as anchors to retian\npretraining knowledge, but these methods lack robust exploration mechanisms. To\naddress these issues, we propose Partial SMILES Validation-PPO (PSV-PPO), a\nnovel RL algorithm that incorporates real-time partial SMILES validation to\nprevent catastrophic forgetting while encouraging exploration. Unlike\ntraditional RL approaches that validate molecule structures only after\ngenerating entire sequences, PSV-PPO performs stepwise validation at each\nauto-regressive step, evaluating not only the selected token candidate but also\nall potential branches stemming from the prior partial sequence. This enables\nearly detection of invalid partial SMILES across all potential paths. As a\nresult, PSV-PPO maintains high validity rates even during aggressive\nexploration of the vast chemical space. Our experiments on the PMO and GuacaMol\nbenchmark datasets demonstrate that PSV-PPO significantly reduces the number of\ninvalid generated structures while maintaining competitive exploration and\noptimization performance. While our work primarily focuses on maintaining\nvalidity, the framework of PSV-PPO can be extended in future research to\nincorporate additional forms of valuable domain knowledge, further enhancing\nreinforcement learning applications in drug discovery.\n", "date": "2025-05-01", "categories": []}
{"id": "2504.21065", "title": "A 3D pocket-aware and affinity-guided diffusion model for lead\n  optimization", "abstract": "  Molecular optimization, aimed at improving binding affinity or other\nmolecular properties, is a crucial task in drug discovery that often relies on\nthe expertise of medicinal chemists. Recently, deep learning-based 3D\ngenerative models showed promise in enhancing the efficiency of molecular\noptimization. However, these models often struggle to adequately consider\nbinding affinities with protein targets during lead optimization. Herein, we\npropose a 3D pocket-aware and affinity-guided diffusion model, named Diffleop,\nto optimize molecules with enhanced binding affinity. The model explicitly\nincorporates the knowledge of protein-ligand binding affinity to guide the\ndenoising sampling for molecule generation with high affinity. The\ncomprehensive evaluations indicated that Diffleop outperforms baseline models\nacross multiple metrics, especially in terms of binding affinity.\n", "date": "2025-04-29", "categories": []}
{"id": "2504.20102", "title": "HyboWaveNet: Hyperbolic Graph Neural Networks with Multi-Scale Wavelet\n  Transform for Protein-Protein Interaction Prediction", "abstract": "  Protein-protein interactions (PPIs) are fundamental for deciphering cellular\nfunctions,disease pathways,and drug discovery.Although existing neural networks\nand machine learning methods have achieved high accuracy in PPI\nprediction,their black-box nature leads to a lack of causal interpretation of\nthe prediction results and difficulty in capturing hierarchical geometries and\nmulti-scale dynamic interaction patterns among proteins.To address these\nchallenges, we propose HyboWaveNet,a novel deep learning framework that\ncollaborates with hyperbolic graphical neural networks (HGNNs) and multiscale\ngraphical wavelet transform for robust PPI prediction. Mapping protein features\nto Lorentz space simulates hierarchical topological relationships among\nbiomolecules via a hyperbolic distance metric,enabling node feature\nrepresentations that better fit biological a priori.HyboWaveNet inherently\nsimulates hierarchical and scale-free biological relationships, while the\nintegration of wavelet transforms enables adaptive extraction of local and\nglobal interaction features across different resolutions. Our framework\ngenerates node feature representations via a graph neural network under the\nLorenz model and generates pairs of positive samples under multiple different\nviews for comparative learning, followed by further feature extraction via\nmulti-scale graph wavelet transforms to predict potential PPIs. Experiments on\npublic datasets show that HyboWaveNet improves over both existing\nstate-of-the-art methods. We also demonstrate through ablation experimental\nstudies that the multi-scale graph wavelet transform module improves the\npredictive performance and generalization ability of HyboWaveNet. This work\nlinks geometric deep learning and signal processing to advance PPI prediction,\nproviding a principled approach for analyzing complex biological systems\n", "date": "2025-04-27", "categories": []}
{"id": "2504.02698", "title": "SCMPPI: Supervised Contrastive Multimodal Framework for Predicting\n  Protein-Protein Interactions", "abstract": "  Protein-protein interaction (PPI) prediction plays a pivotal role in\ndeciphering cellular functions and disease mechanisms. To address the\nlimitations of traditional experimental methods and existing computational\napproaches in cross-modal feature fusion and false-negative suppression, we\npropose SCMPPI-a novel supervised contrastive multimodal framework. By\neffectively integrating sequence-based features (AAC, DPC, ESMC-CKSAAP) with\nnetwork topology (Node2Vec embeddings) and incorporating an enhanced\ncontrastive learning strategy with negative sample filtering, SCMPPI achieves\nsuperior prediction performance. Extensive experiments on eight benchmark\ndatasets demonstrate its state-of-the-art accuracy(98.13%) and AUC(99.69%),\nalong with excellent cross-species generalization (AUC>99%). Successful\napplications in CD9 networks, Wnt pathway analysis, and cancer-specific\nnetworks further highlight its potential for disease target discovery,\nestablishing SCMPPI as a powerful tool for multimodal biological data analysis.\n", "date": "2025-04-03", "categories": []}
{"id": "2504.17162", "title": "A Comprehensive Review on RNA Subcellular Localization Prediction", "abstract": "  The subcellular localization of RNAs, including long non-coding RNAs\n(lncRNAs), messenger RNAs (mRNAs), microRNAs (miRNAs) and other smaller RNAs,\nplays a critical role in determining their biological functions. For instance,\nlncRNAs are predominantly associated with chromatin and act as regulators of\ngene transcription and chromatin structure, while mRNAs are distributed across\nthe nucleus and cytoplasm, facilitating the transport of genetic information\nfor protein synthesis. Understanding RNA localization sheds light on processes\nlike gene expression regulation with spatial and temporal precision. However,\ntraditional wet lab methods for determining RNA localization, such as in situ\nhybridization, are often time-consuming, resource-demanding, and costly. To\novercome these challenges, computational methods leveraging artificial\nintelligence (AI) and machine learning (ML) have emerged as powerful\nalternatives, enabling large-scale prediction of RNA subcellular localization.\nThis paper provides a comprehensive review of the latest advancements in\nAI-based approaches for RNA subcellular localization prediction, covering\nvarious RNA types and focusing on sequence-based, image-based, and hybrid\nmethodologies that combine both data types. We highlight the potential of these\nmethods to accelerate RNA research, uncover molecular pathways, and guide\ntargeted disease treatments. Furthermore, we critically discuss the challenges\nin AI/ML approaches for RNA subcellular localization, such as data scarcity and\nlack of benchmarks, and opportunities to address them. This review aims to\nserve as a valuable resource for researchers seeking to develop innovative\nsolutions in the field of RNA subcellular localization and beyond.\n", "date": "2025-04-24", "categories": []}
{"id": "2504.17624", "title": "Deciphering the unique dynamic activation pathway in a G protein-coupled\n  receptor enables unveiling biased signaling and identifying cryptic\n  allosteric sites in conformational intermediates", "abstract": "  Neurotensin receptor 1 (NTSR1), a member of the Class A G protein-coupled\nreceptor superfamily, plays an important role in modulating dopaminergic\nneuronal activity and eliciting opioid-independent analgesia. Recent studies\nsuggest that promoting \\{beta}-arrestin-biased signaling in NTSR1 may diminish\ndrugs of abuse, such as psychostimulants, thereby offering a potential avenue\nfor treating human addiction-related disorders. In this study, we utilized a\nnovel computational and experimental approach that combined nudged elastic\nband-based molecular dynamics simulations, Markov state models, temporal\ncommunication network analysis, site-directed mutagenesis, and conformational\nbiosensors, to explore the intricate mechanisms underlying NTSR1 activation and\nbiased signaling. Our study reveals a dynamic stepwise transition mechanism and\nactivated transmission network associated with NTSR1 activation. It also yields\nvaluable insights into the complex interplay between the unique polar network,\nnon-conserved ion locks, and aromatic clusters in NTSR1 signaling. Moreover, we\nidentified a cryptic allosteric site located in the intracellular region of the\nreceptor that exists in an intermediate state within the activation pathway.\nCollectively, these findings contribute to a more profound understanding of\nNTSR1 activation and biased signaling at the atomic level, thereby providing a\npotential strategy for the development of NTSR1 allosteric modulators in the\nrealm of G protein-coupled receptor biology, biophysics, and medicine.\n", "date": "2025-04-24", "categories": []}
{"id": "2312.13302", "title": "Longitudinal prediction of DNA methylation to forecast epigenetic\n  outcomes", "abstract": "  Interrogating the evolution of biological changes at early stages of life\nrequires longitudinal profiling of molecules, such as DNA methylation, which\ncan be challenging with children. We introduce a probabilistic and longitudinal\nmachine learning framework based on multi-mean Gaussian processes (GPs),\naccounting for individual and gene correlations across time. This method\nprovides future predictions of DNA methylation status at different individual\nages while accounting for uncertainty. Our model is trained on a birth cohort\nof children with methylation profiled at ages 0-4, and we demonstrated that the\nstatus of methylation sites for each child can be accurately predicted at ages\n5-7. We show that methylation profiles predicted by multi-mean GPs can be used\nto estimate other phenotypes, such as epigenetic age, and enable comparison to\nother health measures of interest. This approach encourages epigenetic studies\nto move towards longitudinal design for investigating epigenetic changes during\ndevelopment, ageing and disease progression.\n", "date": "2023-12-19", "categories": []}
{"id": "2406.04239", "title": "Solving Inverse Problems in Protein Space Using Diffusion-Based Priors", "abstract": "  The interaction of a protein with its environment can be understood and\ncontrolled via its 3D structure. Experimental methods for protein structure\ndetermination, such as X-ray crystallography or cryogenic electron microscopy,\nshed light on biological processes but introduce challenging inverse problems.\nLearning-based approaches have emerged as accurate and efficient methods to\nsolve these inverse problems for 3D structure determination, but are\nspecialized for a predefined type of measurement. Here, we introduce a\nversatile framework to turn biophysical measurements, such as cryo-EM density\nmaps, into 3D atomic models. Our method combines a physics-based forward model\nof the measurement process with a pretrained generative model providing a\ntask-agnostic, data-driven prior. Our method outperforms posterior sampling\nbaselines on linear and non-linear inverse problems. In particular, it is the\nfirst diffusion-based method for refining atomic models from cryo-EM maps and\nbuilding atomic models from sparse distance matrices.\n", "date": "2024-06-06", "categories": []}
{"id": "2410.21518", "title": "Predicting sub-population specific viral evolution", "abstract": "  Forecasting the change in the distribution of viral variants is crucial for\ntherapeutic design and disease surveillance. This task poses significant\nmodeling challenges due to the sharp differences in virus distributions across\nsub-populations (e.g., countries) and their dynamic interactions. Existing\nmachine learning approaches that model the variant distribution as a whole are\nincapable of making location-specific predictions and ignore transmissions that\nshape the viral landscape. In this paper, we propose a sub-population specific\nprotein evolution model, which predicts the time-resolved distributions of\nviral proteins in different locations. The algorithm explicitly models the\ntransmission rates between sub-populations and learns their interdependence\nfrom data. The change in protein distributions across all sub-populations is\ndefined through a linear ordinary differential equation (ODE) parametrized by\ntransmission rates. Solving this ODE yields the likelihood of a given protein\noccurring in particular sub-populations. Multi-year evaluation on both\nSARS-CoV-2 and influenza A/H3N2 demonstrates that our model outperforms\nbaselines in accurately predicting distributions of viral proteins across\ncontinents and countries. We also find that the transmission rates learned from\ndata are consistent with the transmission pathways discovered by retrospective\nphylogenetic analysis.\n", "date": "2024-10-28", "categories": []}
{"id": "2504.16479", "title": "The Dance of Atoms-De Novo Protein Design with Diffusion Model", "abstract": "  The de novo design of proteins refers to creating proteins with specific\nstructures and functions that do not naturally exist. In recent years, the\naccumulation of high-quality protein structure and sequence data and\ntechnological advancements have paved the way for the successful application of\ngenerative artificial intelligence (AI) models in protein design. These models\nhave surpassed traditional approaches that rely on fragments and\nbioinformatics. They have significantly enhanced the success rate of de novo\nprotein design, and reduced experimental costs, leading to breakthroughs in the\nfield. Among various generative AI models, diffusion models have yielded the\nmost promising results in protein design. In the past two to three years, more\nthan ten protein design models based on diffusion models have emerged. Among\nthem, the representative model, RFDiffusion, has demonstrated success rates in\n25 protein design tasks that far exceed those of traditional methods, and other\nAI-based approaches like RFjoint and hallucination. This review will\nsystematically examine the application of diffusion models in generating\nprotein backbones and sequences. We will explore the strengths and limitations\nof different models, summarize successful cases of protein design using\ndiffusion models, and discuss future development directions.\n", "date": "2025-04-23", "categories": []}
{"id": "2504.16886", "title": "Exploring zero-shot structure-based protein fitness prediction", "abstract": "  The ability to make zero-shot predictions about the fitness consequences of\nprotein sequence changes with pre-trained machine learning models enables many\npractical applications. Such models can be applied for downstream tasks like\ngenetic variant interpretation and protein engineering without additional\nlabeled data. The advent of capable protein structure prediction tools has led\nto the availability of orders of magnitude more precomputed predicted\nstructures, giving rise to powerful structure-based fitness prediction models.\nThrough our experiments, we assess several modeling choices for structure-based\nmodels and their effects on downstream fitness prediction. Zero-shot fitness\nprediction models can struggle to assess the fitness landscape within\ndisordered regions of proteins, those that lack a fixed 3D structure. We\nconfirm the importance of matching protein structures to fitness assays and\nfind that predicted structures for disordered regions can be misleading and\naffect predictive performance. Lastly, we evaluate an additional\nstructure-based model on the ProteinGym substitution benchmark and show that\nsimple multi-modal ensembles are strong baselines.\n", "date": "2025-04-23", "categories": []}
{"id": "2503.16816", "title": "PH2ST:ST-Prompt Guided Histological Hypergraph Learning for Spatial Gene\n  Expression Prediction", "abstract": "  Spatial Transcriptomics (ST) reveals the spatial distribution of gene\nexpression in tissues, offering critical insights into biological processes and\ndisease mechanisms. However, the high cost, limited coverage, and technical\ncomplexity of current ST technologies restrict their widespread use in clinical\nand research settings, making obtaining high-resolution transcriptomic profiles\nacross large tissue areas challenging. Predicting ST from H\\&E-stained\nhistology images has emerged as a promising alternative to address these\nlimitations but remains challenging due to the heterogeneous relationship\nbetween histomorphology and gene expression, which is affected by substantial\nvariability across patients and tissue sections. In response, we propose PH2ST,\na prompt-guided hypergraph learning framework, which leverages limited ST\nsignals to guide multi-scale histological representation learning for accurate\nand robust spatial gene expression prediction. Extensive evaluations on two\npublic ST datasets and multiple prompt sampling strategies simulating\nreal-world scenarios demonstrate that PH2ST not only outperforms existing\nstate-of-the-art methods, but also shows strong potential for practical\napplications such as imputing missing spots, ST super-resolution, and\nlocal-to-global prediction, highlighting its value for scalable and\ncost-effective spatial gene expression mapping in biomedical contexts.\n", "date": "2025-03-21", "categories": []}
{"id": "2504.13853", "title": "GenShin:geometry-enhanced structural graph embodies binding pose can\n  better predicting compound-protein interaction affinity", "abstract": "  AI-powered drug discovery typically relies on the successful prediction of\ncompound-protein interactions, which are pivotal for the evaluation of designed\ncompound molecules in structure-based drug design and represent a core\nchallenge in the field.\n  However, accurately predicting compound-protein affinity via regression\nmodels usually requires adequate-binding pose, which are derived from costly\nand complex experimental methods or time-consuming simulations with docking\nsoftware. In response, we have introduced the GenShin model, which constructs a\ngeometry-enhanced structural graph module that separately extracts additional\nfeatures from proteins and compounds. Consequently, it attains an accuracy on\npar with mainstream models in predicting compound-protein affinities, while\neliminating the need for adequate-binding pose as input. Our experimental\nfindings demonstrate that the GenShin model vastly outperforms other models\nthat rely on non-input docking conformations, achieving, or in some cases even\nexceeding, the performance of those requiring adequate-binding pose. Further\nexperiments indicate that our GenShin model is more robust to\ninadequate-binding pose, affirming its higher suitability for real-world drug\ndiscovery scenarios. We hope our work will inspire more endeavors to bridge the\ngap between AI models and practical drug discovery challenges.\n", "date": "2025-03-16", "categories": []}
{"id": "2504.14274", "title": "ProtPainter: Draw or Drag Protein via Topology-guided Diffusion", "abstract": "  Recent advances in protein backbone generation have achieved promising\nresults under structural, functional, or physical constraints. However,\nexisting methods lack the flexibility for precise topology control, limiting\nnavigation of the backbone space. We present ProtPainter, a diffusion-based\napproach for generating protein backbones conditioned on 3D curves. ProtPainter\nfollows a two-stage process: curve-based sketching and sketch-guided backbone\ngeneration. For the first stage, we propose CurveEncoder, which predicts\nsecondary structure annotations from a curve to parametrize sketch generation.\nFor the second stage, the sketch guides the generative process in Denoising\nDiffusion Probabilistic Modeling (DDPM) to generate backbones. During this\nprocess, we further introduce a fusion scheduling scheme, Helix-Gating, to\ncontrol the scaling factors. To evaluate, we propose the first benchmark for\ntopology-conditioned protein generation, introducing Protein Restoration Task\nand a new metric, self-consistency Topology Fitness (scTF). Experiments\ndemonstrate ProtPainter's ability to generate topology-fit (scTF > 0.8) and\ndesignable (scTM > 0.5) backbones, with drawing and dragging tasks showcasing\nits flexibility and versatility.\n", "date": "2025-04-19", "categories": []}
{"id": "2502.14944", "title": "Reward-Guided Iterative Refinement in Diffusion Models at Test-Time with\n  Applications to Protein and DNA Design", "abstract": "  To fully leverage the capabilities of diffusion models, we are often\ninterested in optimizing downstream reward functions during inference. While\nnumerous algorithms for reward-guided generation have been recently proposed\ndue to their significance, current approaches predominantly focus on\nsingle-shot generation, transitioning from fully noised to denoised states. We\npropose a novel framework for inference-time reward optimization with diffusion\nmodels inspired by evolutionary algorithms. Our approach employs an iterative\nrefinement process consisting of two steps in each iteration: noising and\nreward-guided denoising. This sequential refinement allows for the gradual\ncorrection of errors introduced during reward optimization. Besides, we provide\na theoretical guarantee for our framework. Finally, we demonstrate its superior\nempirical performance in protein and cell-type-specific regulatory DNA design.\nThe code is available at\n\\href{https://github.com/masa-ue/ProDifEvo-Refinement}{https://github.com/masa-ue/ProDifEvo-Refinement}.\n", "date": "2025-02-20", "categories": []}
{"id": "2504.12610", "title": "Machine Learning Methods for Gene Regulatory Network Inference", "abstract": "  Gene Regulatory Networks (GRNs) are intricate biological systems that control\ngene expression and regulation in response to environmental and developmental\ncues. Advances in computational biology, coupled with high throughput\nsequencing technologies, have significantly improved the accuracy of GRN\ninference and modeling. Modern approaches increasingly leverage artificial\nintelligence (AI), particularly machine learning techniques including\nsupervised, unsupervised, semi-supervised, and contrastive learning to analyze\nlarge scale omics data and uncover regulatory gene interactions. To support\nboth the application of GRN inference in studying gene regulation and the\ndevelopment of novel machine learning methods, we present a comprehensive\nreview of machine learning based GRN inference methodologies, along with the\ndatasets and evaluation metrics commonly used. Special emphasis is placed on\nthe emerging role of cutting edge deep learning techniques in enhancing\ninference performance. The potential future directions for improving GRN\ninference are also discussed.\n", "date": "2025-04-17", "categories": []}
{"id": "2504.13044", "title": "The Dissipation Theory of Aging: A Quantitative Analysis Using a\n  Cellular Aging Map", "abstract": "  We propose a new theory for aging based on dynamical systems and provide a\ndata-driven computational method to quantify the changes at the cellular level.\nWe use ergodic theory to decompose the dynamics of changes during aging and\nshow that aging is fundamentally a dissipative process within biological\nsystems, akin to dynamical systems where dissipation occurs due to\nnon-conservative forces. To quantify the dissipation dynamics, we employ a\ntransformer-based machine learning algorithm to analyze gene expression data,\nincorporating age as a token to assess how age-related dissipation is reflected\nin the embedding space. By evaluating the dynamics of gene and age embeddings,\nwe provide a cellular aging map (CAM) and identify patterns indicative of\ndivergence in gene embedding space, nonlinear transitions, and entropy\nvariations during aging for various tissues and cell types. Our results provide\na novel perspective on aging as a dissipative process and introduce a\ncomputational framework that enables measuring age-related changes with\nmolecular resolution.\n", "date": "2025-04-17", "categories": []}
{"id": "2410.01795", "title": "Knowledge-Driven Feature Selection and Engineering for Genotype Data\n  with Large Language Models", "abstract": "  Predicting phenotypes with complex genetic bases based on a small,\ninterpretable set of variant features remains a challenging task.\nConventionally, data-driven approaches are utilized for this task, yet the high\ndimensional nature of genotype data makes the analysis and prediction\ndifficult. Motivated by the extensive knowledge encoded in pre-trained LLMs and\ntheir success in processing complex biomedical concepts, we set to examine the\nability of LLMs in feature selection and engineering for tabular genotype data,\nwith a novel knowledge-driven framework. We develop FREEFORM, Free-flow\nReasoning and Ensembling for Enhanced Feature Output and Robust Modeling,\ndesigned with chain-of-thought and ensembling principles, to select and\nengineer features with the intrinsic knowledge of LLMs. Evaluated on two\ndistinct genotype-phenotype datasets, genetic ancestry and hereditary hearing\nloss, we find this framework outperforms several data-driven methods,\nparticularly on low-shot regimes. FREEFORM is available as open-source\nframework at GitHub: https://github.com/PennShenLab/FREEFORM.\n", "date": "2024-10-02", "categories": []}
{"id": "2407.08974", "title": "Topology-enhanced machine learning model (Top-ML) for anticancer peptide\n  prediction", "abstract": "  Recently, therapeutic peptides have demonstrated great promise for cancer\ntreatment. To explore powerful anticancer peptides, artificial intelligence\n(AI)-based approaches have been developed to systematically screen potential\ncandidates. However, the lack of efficient featurization of peptides has become\na bottleneck for these machine-learning models. In this paper, we propose a\ntopology-enhanced machine learning model (Top-ML) for anticancer peptides\nprediction. Our Top-ML employs peptide topological features derived from its\nsequence \"connection\" information characterized by vector and spectral\ndescriptors. Our Top-ML model, employing an Extra-Trees classifier, has been\nvalidated on the AntiCP 2.0 and mACPpred 2.0 benchmark datasets, achieving\nstate-of-the-art performance or results comparable to existing deep learning\nmodels, while providing greater interpretability. Our results highlight the\npotential of leveraging novel topology-based featurization to accelerate the\nidentification of anticancer peptides.\n", "date": "2024-07-12", "categories": []}
{"id": "2503.19821", "title": "IgCraft: A versatile sequence generation framework for antibody\n  discovery and engineering", "abstract": "  Designing antibody sequences to better resemble those observed in natural\nhuman repertoires is a key challenge in biologics development. We introduce\nIgCraft: a multi-purpose model for paired human antibody sequence generation,\nbuilt on Bayesian Flow Networks. IgCraft presents one of the first unified\ngenerative modeling frameworks capable of addressing multiple antibody sequence\ndesign tasks with a single model, including unconditional sampling, sequence\ninpainting, inverse folding, and CDR motif scaffolding. Our approach achieves\ncompetitive results across the full spectrum of these tasks while constraining\ngeneration to the space of human antibody sequences, exhibiting particular\nstrengths in CDR motif scaffolding (grafting) where we achieve state-of-the-art\nperformance in terms of humanness and preservation of structural properties. By\nintegrating previously separate tasks into a single scalable generative model,\nIgCraft provides a versatile platform for sampling human antibody sequences\nunder a variety of contexts relevant to antibody discovery and engineering.\nModel code and weights are publicly available at\nhttps://github.com/mgreenig/IgCraft.\n", "date": "2025-03-25", "categories": []}
{"id": "2504.03278", "title": "JanusDDG: A Thermodynamics-Compliant Model for Sequence-Based Protein\n  Stability via Two-Fronts Multi-Head Attention", "abstract": "  Understanding how residue variations affect protein stability is crucial for\ndesigning functional proteins and deciphering the molecular mechanisms\nunderlying disease-related mutations. Recent advances in protein language\nmodels (PLMs) have revolutionized computational protein analysis, enabling,\namong other things, more accurate predictions of mutational effects. In this\nwork, we introduce JanusDDG, a deep learning framework that leverages\nPLM-derived embeddings and a bidirectional cross-attention transformer\narchitecture to predict $\\Delta \\Delta G$ of single and multiple-residue\nmutations while simultaneously being constrained to respect fundamental\nthermodynamic properties, such as antisymmetry and transitivity. Unlike\nconventional self-attention, JanusDDG computes queries (Q) and values (V) as\nthe difference between wild-type and mutant embeddings, while keys (K)\nalternate between the two. This cross-interleaved attention mechanism enables\nthe model to capture mutation-induced perturbations while preserving essential\ncontextual information. Experimental results show that JanusDDG achieves\nstate-of-the-art performance in predicting $\\Delta \\Delta G$ from sequence\nalone, matching or exceeding the accuracy of structure-based methods for both\nsingle and multiple mutations. Code\nAvailability:https://github.com/compbiomed-unito/JanusDDG\n", "date": "2025-04-04", "categories": []}
{"id": "2504.10983", "title": "ProtFlow: Fast Protein Sequence Design via Flow Matching on Compressed\n  Protein Language Model Embeddings", "abstract": "  The design of protein sequences with desired functionalities is a fundamental\ntask in protein engineering. Deep generative methods, such as autoregressive\nmodels and diffusion models, have greatly accelerated the discovery of novel\nprotein sequences. However, these methods mainly focus on local or shallow\nresidual semantics and suffer from low inference efficiency, large modeling\nspace and high training cost. To address these challenges, we introduce\nProtFlow, a fast flow matching-based protein sequence design framework that\noperates on embeddings derived from semantically meaningful latent space of\nprotein language models. By compressing and smoothing the latent space,\nProtFlow enhances performance while training on limited computational\nresources. Leveraging reflow techniques, ProtFlow enables high-quality\nsingle-step sequence generation. Additionally, we develop a joint design\npipeline for the design scene of multichain proteins. We evaluate ProtFlow\nacross diverse protein design tasks, including general peptides and long-chain\nproteins, antimicrobial peptides, and antibodies. Experimental results\ndemonstrate that ProtFlow outperforms task-specific methods in these\napplications, underscoring its potential and broad applicability in\ncomputational protein sequence design and analysis.\n", "date": "2025-04-15", "categories": []}
{"id": "2504.09704", "title": "Transformer-Based Representation Learning for Robust Gene Expression\n  Modeling and Cancer Prognosis", "abstract": "  Transformer-based models have achieved remarkable success in natural language\nand vision tasks, but their application to gene expression analysis remains\nlimited due to data sparsity, high dimensionality, and missing values. We\npresent GexBERT, a transformer-based autoencoder framework for robust\nrepresentation learning of gene expression data. GexBERT learns context-aware\ngene embeddings by pretraining on large-scale transcriptomic profiles with a\nmasking and restoration objective that captures co-expression relationships\namong thousands of genes. We evaluate GexBERT across three critical tasks in\ncancer research: pan-cancer classification, cancer-specific survival\nprediction, and missing value imputation. GexBERT achieves state-of-the-art\nclassification accuracy from limited gene subsets, improves survival prediction\nby restoring expression of prognostic anchor genes, and outperforms\nconventional imputation methods under high missingness. Furthermore, its\nattention-based interpretability reveals biologically meaningful gene patterns\nacross cancer types. These findings demonstrate the utility of GexBERT as a\nscalable and effective tool for gene expression modeling, with translational\npotential in settings where gene coverage is limited or incomplete.\n", "date": "2025-04-13", "categories": []}
{"id": "2504.10343", "title": "Domain-Adversarial Neural Network and Explainable AI for Reducing\n  Tissue-of-Origin Signal in Pan-cancer Mortality Classification", "abstract": "  Tissue-of-origin signals dominate pan-cancer gene expression, often obscuring\nmolecular features linked to patient survival. This hampers the discovery of\ngeneralizable biomarkers, as models tend to overfit tissue-specific patterns\nrather than capture survival-relevant signals. To address this, we propose a\nDomain-Adversarial Neural Network (DANN) trained on TCGA RNA-seq data to learn\nrepresentations less biased by tissue and more focused on survival. Identifying\ntissue-independent genetic profiles is key to revealing core cancer programs.\nWe assess the DANN using: (1) Standard SHAP, based on the original input space\nand DANN's mortality classifier; (2) A layer-aware strategy applied to hidden\nactivations, including an unsupervised manifold from raw activations and a\nsupervised manifold from mortality-specific SHAP values. Standard SHAP remains\nconfounded by tissue signals due to biases inherent in its computation. The raw\nactivation manifold was dominated by high-magnitude activations, which masked\nsubtle tissue and mortality-related signals. In contrast, the layer-aware SHAP\nmanifold offers improved low-dimensional representations of both tissue and\nmortality signals, independent of activation strength, enabling subpopulation\nstratification and pan-cancer identification of survival-associated genes.\n", "date": "2025-04-14", "categories": []}
{"id": "2504.10388", "title": "Inferring genotype-phenotype maps using attention models", "abstract": "  Predicting phenotype from genotype is a central challenge in genetics.\nTraditional approaches in quantitative genetics typically analyze this problem\nusing methods based on linear regression. These methods generally assume that\nthe genetic architecture of complex traits can be parameterized in terms of an\nadditive model, where the effects of loci are independent, plus (in some cases)\npairwise epistatic interactions between loci. However, these models struggle to\nanalyze more complex patterns of epistasis or subtle gene-environment\ninteractions. Recent advances in machine learning, particularly attention-based\nmodels, offer a promising alternative. Initially developed for natural language\nprocessing, attention-based models excel at capturing context-dependent\ninteractions and have shown exceptional performance in predicting protein\nstructure and function. Here, we apply attention-based models to quantitative\ngenetics. We analyze the performance of this attention-based approach in\npredicting phenotype from genotype using simulated data across a range of\nmodels with increasing epistatic complexity, and using experimental data from a\nrecent quantitative trait locus mapping study in budding yeast. We find that\nour model demonstrates superior out-of-sample predictions in epistatic regimes\ncompared to standard methods. We also explore a more general multi-environment\nattention-based model to jointly analyze genotype-phenotype maps across\nmultiple environments and show that such architectures can be used for\n\"transfer learning\" - predicting phenotypes in novel environments with limited\ntraining data.\n", "date": "2025-04-14", "categories": []}
{"id": "2504.03976", "title": "OLAF: An Open Life Science Analysis Framework for Conversational\n  Bioinformatics Powered by Large Language Models", "abstract": "  OLAF (Open Life Science Analysis Framework) is an open-source platform that\nenables researchers to perform bioinformatics analyses using natural language.\nBy combining large language models (LLMs) with a modular agent-pipe-router\narchitecture, OLAF generates and executes bioinformatics code on real\nscientific data, including formats like .h5ad. The system includes an Angular\nfront end and a Python/Firebase backend, allowing users to run analyses such as\nsingle-cell RNA-seq workflows, gene annotation, and data visualization through\na simple web interface. Unlike general-purpose AI tools, OLAF integrates code\nexecution, data handling, and scientific libraries in a reproducible,\nuser-friendly environment. It is designed to lower the barrier to computational\nbiology for non-programmers and support transparent, AI-powered life science\nresearch.\n", "date": "2025-04-04", "categories": []}
{"id": "2504.08096", "title": "Cellular Development Follows the Path of Minimum Action", "abstract": "  Cellular development follows a stochastic yet rule-governed trajectory,\nthough the underlying principles remain elusive. Here, we propose that cellular\ndevelopment follows paths of least action, aligning with foundational physical\nlaws that govern dynamic systems across nature. We introduce a computational\nframework that takes advantage of the deep connection between the principle of\nleast action and maximum entropy to model developmental processes using\nTransformers architecture. This approach enables precise quantification of\nentropy production, information flow curvature, and local irreversibility for\ndevelopmental asymmetry in single-cell RNA sequence data. Within this unified\nframework, we provide interpretable metrics: entropy to capture\nexploration-exploitation trade-offs, curvature to assess plasticity-elasticity\ndynamics, and entropy production to characterize dedifferentiation and\ntransdifferentiation. We validate our method across both single-cell and\nembryonic development datasets, demonstrating its ability to reveal hidden\nthermodynamic and informational constraints shaping cellular fate decisions.\n", "date": "2025-04-10", "categories": []}
{"id": "2504.08328", "title": "Towards generalizable single-cell perturbation modeling via the\n  Conditional Monge Gap", "abstract": "  Learning the response of single-cells to various treatments offers great\npotential to enable targeted therapies. In this context, neural optimal\ntransport (OT) has emerged as a principled methodological framework because it\ninherently accommodates the challenges of unpaired data induced by cell\ndestruction during data acquisition. However, most existing OT approaches are\nincapable of conditioning on different treatment contexts (e.g., time, drug\ntreatment, drug dosage, or cell type) and we still lack methods that\nunanimously show promising generalization performance to unseen treatments.\nHere, we propose the Conditional Monge Gap which learns OT maps conditionally\non arbitrary covariates. We demonstrate its value in predicting single-cell\nperturbation responses conditional to one or multiple drugs, a drug dosage, or\ncombinations thereof. We find that our conditional models achieve results\ncomparable and sometimes even superior to the condition-specific\nstate-of-the-art on scRNA-seq as well as multiplexed protein imaging data.\nNotably, by aggregating data across conditions we perform cross-task learning\nwhich unlocks remarkable generalization abilities to unseen drugs or drug\ndosages, widely outperforming other conditional models in capturing\nheterogeneity (i.e., higher moments) in the perturbed population. Finally, by\nscaling to hundreds of conditions and testing on unseen drugs, we narrow the\ngap between structure-based and effect-based drug representations, suggesting a\npromising path to the successful prediction of perturbation effects for unseen\ntreatments.\n", "date": "2025-04-11", "categories": []}
{"id": "2410.19110", "title": "Bio2Token: All-atom tokenization of any biomolecular structure with\n  Mamba", "abstract": "  Efficient encoding and representation of large 3D molecular structures with\nhigh fidelity is critical for biomolecular design applications. Despite this,\nmany representation learning approaches restrict themselves to modeling smaller\nsystems or use coarse-grained approximations of the systems, for example\nmodeling proteins at the resolution of amino acid residues rather than at the\nlevel of individual atoms. To address this, we develop quantized auto-encoders\nthat learn atom-level tokenizations of complete proteins, RNA and small\nmolecule structures with reconstruction accuracies well below 1 Angstrom. We\ndemonstrate that a simple Mamba state space model architecture is efficient\ncompared to an SE(3)-invariant IPA architecture, reaches competitive accuracies\nand can scale to systems with almost 100,000 atoms. The learned structure\ntokens of bio2token may serve as the input for all-atom generative models in\nthe future.\n", "date": "2024-10-24", "categories": []}
{"id": "2504.07061", "title": "Teaching pathology foundation models to accurately predict gene\n  expression with parameter efficient knowledge transfer", "abstract": "  Gene expression profiling provides critical insights into cellular\nheterogeneity, biological processes and disease mechanisms. There has been an\nincreasing interest in computational approaches that can predict gene\nexpression directly from digitalized histopathology images. While image\nfoundation models have shown promise in a variety of pathology downstream\nanalysis, their performances on gene-expression prediction are still limited.\nExplicitly incorporating information from the transcriptomic models can help\nimage models to address domain shift, yet the fine-tuning and alignment of\nfoundation models can be expensive. In the work, we propose Parameter Efficient\nKnowledge trAnsfer (PEKA), a novel framework that leverages Block-Affine\nAdaptation and integrates knowledge distillation and structure alignment losses\nfor cross-modal knowledge transfer. We evaluated PEKA for gene expression\nprediction using multiple spatial transcriptomics datasets (comprising 206,123\nimage tiles with matched gene expression profiles) that encompassed various\ntypes of tissue. PEKA achieved at least 5\\% performance improvement over\nbaseline foundation models while also outperforming alternative\nparameter-efficient fine-tuning strategies. We will release the code, datasets\nand aligned models after peer-review to facilitate broader adoption and further\ndevelopment for parameter efficient model alignment.\n", "date": "2025-04-09", "categories": []}
{"id": "2406.15341", "title": "GenoTEX: An LLM Agent Benchmark for Automated Gene Expression Data\n  Analysis", "abstract": "  Recent advancements in machine learning have significantly improved the\nidentification of disease-associated genes from gene expression datasets.\nHowever, these processes often require extensive expertise and manual effort,\nlimiting their scalability. Large Language Model (LLM)-based agents have shown\npromise in automating these tasks due to their increasing problem-solving\nabilities. To support the evaluation and development of such methods, we\nintroduce GenoTEX, a benchmark dataset for the automated analysis of gene\nexpression data. GenoTEX provides analysis code and results for solving a wide\nrange of gene-trait association problems, encompassing dataset selection,\npreprocessing, and statistical analysis, in a pipeline that follows\ncomputational genomics standards. The benchmark includes expert-curated\nannotations from bioinformaticians to ensure accuracy and reliability. To\nprovide baselines for these tasks, we present GenoAgent, a team of LLM-based\nagents that adopt a multi-step programming workflow with flexible\nself-correction, to collaboratively analyze gene expression datasets. Our\nexperiments demonstrate the potential of LLM-based methods in analyzing genomic\ndata, while error analysis highlights the challenges and areas for future\nimprovement. We propose GenoTEX as a promising resource for benchmarking and\nenhancing automated methods for gene expression data analysis. The benchmark is\navailable at https://github.com/Liu-Hy/GenoTEX.\n", "date": "2024-06-21", "categories": []}
{"id": "2504.05844", "title": "Adaptive Substructure-Aware Expert Model for Molecular Property\n  Prediction", "abstract": "  Molecular property prediction is essential for applications such as drug\ndiscovery and toxicity assessment. While Graph Neural Networks (GNNs) have\nshown promising results by modeling molecules as molecular graphs, their\nreliance on data-driven learning limits their ability to generalize,\nparticularly in the presence of data imbalance and diverse molecular\nsubstructures. Existing methods often overlook the varying contributions of\ndifferent substructures to molecular properties, treating them uniformly. To\naddress these challenges, we propose ASE-Mol, a novel GNN-based framework that\nleverages a Mixture-of-Experts (MoE) approach for molecular property\nprediction. ASE-Mol incorporates BRICS decomposition and significant\nsubstructure awareness to dynamically identify positive and negative\nsubstructures. By integrating a MoE architecture, it reduces the adverse impact\nof negative motifs while improving adaptability to positive motifs.\nExperimental results on eight benchmark datasets demonstrate that ASE-Mol\nachieves state-of-the-art performance, with significant improvements in both\naccuracy and interpretability.\n", "date": "2025-04-08", "categories": []}
{"id": "2504.03733", "title": "Artificial Intelligence and Deep Learning Algorithms for Epigenetic\n  Sequence Analysis: A Review for Epigeneticists and AI Experts", "abstract": "  Epigenetics encompasses mechanisms that can alter the expression of genes\nwithout changing the underlying genetic sequence. The epigenetic regulation of\ngene expression is initiated and sustained by several mechanisms such as DNA\nmethylation, histone modifications, chromatin conformation, and non-coding RNA.\nThe changes in gene regulation and expression can manifest in the form of\nvarious diseases and disorders such as cancer and congenital deformities. Over\nthe last few decades, high throughput experimental approaches have been used to\nidentify and understand epigenetic changes, but these laboratory experimental\napproaches and biochemical processes are time-consuming and expensive. To\novercome these challenges, machine learning and artificial intelligence (AI)\napproaches have been extensively used for mapping epigenetic modifications to\ntheir phenotypic manifestations. In this paper we provide a narrative review of\npublished research on AI models trained on epigenomic data to address a variety\nof problems such as prediction of disease markers, gene expression, enhancer\npromoter interaction, and chromatin states. The purpose of this review is\ntwofold as it is addressed to both AI experts and epigeneticists. For AI\nresearchers, we provided a taxonomy of epigenetics research problems that can\nbenefit from an AI-based approach. For epigeneticists, given each of the above\nproblems we provide a list of candidate AI solutions in the literature. We have\nalso identified several gaps in the literature, research challenges, and\nrecommendations to address these challenges.\n", "date": "2025-04-01", "categories": []}
{"id": "2504.04698", "title": "scAgent: Universal Single-Cell Annotation via a LLM Agent", "abstract": "  Cell type annotation is critical for understanding cellular heterogeneity.\nBased on single-cell RNA-seq data and deep learning models, good progress has\nbeen made in annotating a fixed number of cell types within a specific tissue.\nHowever, universal cell annotation, which can generalize across tissues,\ndiscover novel cell types, and extend to novel cell types, remains less\nexplored. To fill this gap, this paper proposes scAgent, a universal cell\nannotation framework based on Large Language Models (LLMs). scAgent can\nidentify cell types and discover novel cell types in diverse tissues;\nfurthermore, it is data efficient to learn novel cell types. Experimental\nstudies in 160 cell types and 35 tissues demonstrate the superior performance\nof scAgent in general cell-type annotation, novel cell discovery, and\nextensibility to novel cell type.\n", "date": "2025-04-07", "categories": []}
{"id": "2306.00872", "title": "Is novelty predictable?", "abstract": "  Machine learning-based design has gained traction in the sciences, most\nnotably in the design of small molecules, materials, and proteins, with\nsocietal implications spanning drug development and manufacturing, plastic\ndegradation, and carbon sequestration. When designing objects to achieve novel\nproperty values with machine learning, one faces a fundamental challenge: how\nto push past the frontier of current knowledge, distilled from the training\ndata into the model, in a manner that rationally controls the risk of failure.\nIf one trusts learned models too much in extrapolation, one is likely to design\nrubbish. In contrast, if one does not extrapolate, one cannot find novelty.\nHerein, we ponder how one might strike a useful balance between these two\nextremes. We focus in particular on designing proteins with novel property\nvalues, although much of our discussion addresses machine learning-based design\nmore broadly.\n", "date": "2023-06-01", "categories": []}
{"id": "2504.01973", "title": "Universally applicable and tunable graph-based coarse-graining for\n  Machine learning force fields", "abstract": "  Coarse-grained (CG) force field methods for molecular systems are a crucial\ntool to simulate large biological macromolecules and are therefore essential\nfor characterisations of biomolecular systems. While state-of-the-art deep\nlearning (DL)-based models for all-atom force fields have improved immensely\nover recent years, we observe and analyse significant limitations of the\ncurrently available approaches for DL-based CG simulations. In this work, we\npresent the first transferable DL-based CG force field approach (i.e., not\nspecific to only one narrowly defined system type) applicable to a wide range\nof biosystems. To achieve this, our CG algorithm does not rely on hard-coded\nrules and is tuned to output coarse-grained systems optimised for minimal\nstatistical noise in the ground truth CG forces, which results in significant\nimprovement of model training. Our force field model is also the first CG\nvariant that is based on the MACE architecture and is trained on a custom\ndataset created by a new approach based on the fragmentation of large\nbiosystems covering protein, RNA and lipid chemistry. We demonstrate that our\nmodel can be applied in molecular dynamics simulations to obtain stable and\nqualitatively accurate trajectories for a variety of systems, while also\ndiscussing cases for which we observe limited reliability.\n", "date": "2025-03-24", "categories": []}
{"id": "2504.02014", "title": "HCAF-DTA: drug-target binding affinity prediction with cross-attention\n  fused hypergraph neural networks", "abstract": "  Accurate prediction of the binding affinity between drugs and target proteins\nis a core task in computer-aided drug design. Existing deep learning methods\ntend to ignore the information of internal sub-structural features of drug\nmolecules and drug-target interactions, resulting in limited prediction\nperformance. In this paper, we propose a drug-target association prediction\nmodel HCAF-DTA based on cross-attention fusion hypergraph neural network. The\nmodel innovatively introduces hypergraph representation in the feature\nextraction stage: drug molecule hypergraphs are constructed based on the tree\ndecomposition algorithm, and the sub-structural and global features extracted\nby fusing the hypergraph neural network with the graphical neural network\nthrough hopping connections, in which the hyper edges can efficiently\ncharacterise the functional functional groups and other key chemical features;\nfor the protein feature extraction, a weighted graph is constructed based on\nthe residues predicted by the ESM model contact maps to construct weighted\ngraphs, and multilayer graph neural networks were used to capture spatial\ndependencies. In the prediction stage, a bidirectional multi-head\ncross-attention mechanism is designed to model intermolecular interactions from\nthe dual viewpoints of atoms and amino acids, and cross-modal features with\ncorrelated information are fused by attention. Experiments on benchmark\ndatasets such as Davis and KIBA show that HCAF-DTA outperforms state of the\narts in all three performance evaluation metrics, with the MSE metrics reaching\n0.198 and 0.122, respectively, with an improvement of up to 4% from the optimal\nbaseline.\n", "date": "2025-04-02", "categories": []}
{"id": "2504.00146", "title": "Why risk matters for protein binder design", "abstract": "  Bayesian optimization (BO) has recently become more prevalent in protein\nengineering applications and hence has become a fruitful target of benchmarks.\nHowever, current BO comparisons often overlook real-world considerations like\nrisk and cost constraints. In this work, we compare 72 model combinations of\nencodings, surrogate models, and acquisition functions on 11 protein binder\nfitness landscapes, specifically from this perspective. Drawing from the\nportfolio optimization literature, we adopt metrics to quantify the cold-start\nperformance relative to a random baseline, to assess the risk of an\noptimization campaign, and to calculate the overall budget required to reach a\nfitness threshold. Our results suggest the existence of Pareto-optimal models\non the risk-performance axis, the shift of this preference depending on the\nlandscape explored, and the robust correlation between landscape properties\nsuch as epistasis with the average and worst-case model performance. They also\nhighlight that rigorous model selection requires substantial computational and\nstatistical efforts.\n", "date": "2025-03-31", "categories": []}
{"id": "2504.00306", "title": "LOCO-EPI: Leave-one-chromosome-out (LOCO) as a benchmarking paradigm for\n  deep learning based prediction of enhancer-promoter interactions", "abstract": "  In mammalian and vertebrate genomes, the promoter regions of the gene and\ntheir distal enhancers may be located millions of base-pairs from each other,\nwhile a promoter may not interact with the closest enhancer. Since base-pair\nproximity is not a good indicator of these interactions, there is considerable\nwork toward developing methods for predicting Enhancer-Promoter Interactions\n(EPI). Several machine learning methods have reported increasingly higher\naccuracies for predicting EPI. Typically, these approaches randomly split the\ndataset of Enhancer-Promoter (EP) pairs into training and testing subsets\nfollowed by model training. However, the aforementioned random splitting causes\ninformation leakage by assigning EP pairs from the same genomic region to both\ntesting and training sets, leading to performance overestimation. In this paper\nwe propose to use a more thorough training and testing paradigm i.e.,\nLeave-one-chromosome-out (LOCO) cross-validation for EPI-prediction. We\ndemonstrate that a deep learning algorithm, which gives higher accuracies when\ntrained and tested on random-splitting setting, drops drastically in\nperformance under LOCO setting, confirming overestimation of performance. We\nfurther propose a novel hybrid deep neural network for EPI-prediction that\nfuses k-mer features of the nucleotide sequence. We show that the hybrid\narchitecture performs significantly better in the LOCO setting, demonstrating\nit can learn more generalizable aspects of EP interactions. With this paper we\nare also releasing the LOCO splitting-based EPI dataset. Research data is\navailable in this public repository: https://github.com/malikmtahir/EPI\n", "date": "2025-04-01", "categories": []}
{"id": "2406.07025", "title": "Entropy-Reinforced Planning with Large Language Models for Drug\n  Discovery", "abstract": "  The objective of drug discovery is to identify chemical compounds that\npossess specific pharmaceutical properties toward a binding target. Existing\nlarge language models (LLMS) can achieve high token matching scores in terms of\nlikelihood for molecule generation. However, relying solely on LLM decoding\noften results in the generation of molecules that are either invalid due to a\nsingle misused token, or suboptimal due to unbalanced exploration and\nexploitation as a consequence of the LLMs prior experience. Here we propose\nERP, Entropy-Reinforced Planning for Transformer Decoding, which employs an\nentropy-reinforced planning algorithm to enhance the Transformer decoding\nprocess and strike a balance between exploitation and exploration. ERP aims to\nachieve improvements in multiple properties compared to direct sampling from\nthe Transformer. We evaluated ERP on the SARS-CoV-2 virus (3CLPro) and human\ncancer cell target protein (RTCB) benchmarks and demonstrated that, in both\nbenchmarks, ERP consistently outperforms the current state-of-the-art algorithm\nby 1-5 percent, and baselines by 5-10 percent, respectively. Moreover, such\nimprovement is robust across Transformer models trained with different\nobjectives. Finally, to further illustrate the capabilities of ERP, we tested\nour algorithm on three code generation benchmarks and outperformed the current\nstate-of-the-art approach as well. Our code is publicly available at:\nhttps://github.com/xuefeng-cs/ERP.\n", "date": "2024-06-11", "categories": []}
{"id": "2503.23014", "title": "MSNGO: multi-species protein function annotation based on 3D protein\n  structure and network propagation", "abstract": "  Motivation: In recent years, protein function prediction has broken through\nthe bottleneck of sequence features, significantly improving prediction\naccuracy using high-precision protein structures predicted by AlphaFold2. While\nsingle-species protein function prediction methods have achieved remarkable\nsuccess, multi-species protein function prediction methods are still in the\nstage of using PPI networks and sequence features. Providing effective\ncross-species label propagation for species with sparse protein annotations\nremains a challenging issue. To address this problem, we propose the MSNGO\nmodel, which integrates structural features and network propagation methods.\nOur validation shows that using structural features can significantly improve\nthe accuracy of multi-species protein function prediction. Results: We employ\ngraph representation learning techniques to extract amino acid representations\nfrom protein structure contact maps and train a structural model using a graph\nconvolution pooling module to derive protein-level structural features. After\nincorporating the sequence features from ESM-2, we apply a network propagation\nalgorithm to aggregate information and update node representations within a\nheterogeneous network. The results demonstrate that MSNGO outperforms previous\nmulti-species protein function prediction methods that rely on sequence\nfeatures and PPI networks. Availability: https://github.com/blingbell/MSNGO.\n", "date": "2025-03-29", "categories": []}
{"id": "2503.23535", "title": "In-silico biological discovery with large perturbation models", "abstract": "  Data generated in perturbation experiments link perturbations to the changes\nthey elicit and therefore contain information relevant to numerous biological\ndiscovery tasks -- from understanding the relationships between biological\nentities to developing therapeutics. However, these data encompass diverse\nperturbations and readouts, and the complex dependence of experimental outcomes\non their biological context makes it challenging to integrate insights across\nexperiments. Here, we present the Large Perturbation Model (LPM), a\ndeep-learning model that integrates multiple, heterogeneous perturbation\nexperiments by representing perturbation, readout, and context as disentangled\ndimensions. LPM outperforms existing methods across multiple biological\ndiscovery tasks, including in predicting post-perturbation transcriptomes of\nunseen experiments, identifying shared molecular mechanisms of action between\nchemical and genetic perturbations, and facilitating the inference of gene-gene\ninteraction networks.\n", "date": "2025-03-30", "categories": []}
{"id": "2503.23550", "title": "Addressing Model Overcomplexity in Drug-Drug Interaction Prediction With\n  Molecular Fingerprints", "abstract": "  Accurately predicting drug-drug interactions (DDIs) is crucial for\npharmaceutical research and clinical safety. Recent deep learning models often\nsuffer from high computational costs and limited generalization across\ndatasets. In this study, we investigate a simpler yet effective approach using\nmolecular representations such as Morgan fingerprints (MFPS), graph-based\nembeddings from graph convolutional networks (GCNs), and transformer-derived\nembeddings from MoLFormer integrated into a straightforward neural network. We\nbenchmark our implementation on DrugBank DDI splits and a drug-drug affinity\n(DDA) dataset from the Food and Drug Administration. MFPS along with MoLFormer\nand GCN representations achieve competitive performance across tasks, even in\nthe more challenging leak-proof split, highlighting the sufficiency of simple\nmolecular representations. Moreover, we are able to identify key molecular\nmotifs and structural patterns relevant to drug interactions via gradient-based\nanalyses using the representations under study. Despite these results, dataset\nlimitations such as insufficient chemical diversity, limited dataset size, and\ninconsistent labeling impact robust evaluation and challenge the need for more\ncomplex approaches. Our work provides a meaningful baseline and emphasizes the\nneed for better dataset curation and progressive complexity scaling.\n", "date": "2025-03-30", "categories": []}
{"id": "2408.12139", "title": "DRExplainer: Quantifiable Interpretability in Drug Response Prediction\n  with Directed Graph Convolutional Network", "abstract": "  Predicting the response of a cancer cell line to a therapeutic drug is\npivotal for personalized medicine. Despite numerous deep learning methods that\nhave been developed for drug response prediction, integrating diverse\ninformation about biological entities and predicting the directional response\nremain major challenges. Here, we propose a novel interpretable predictive\nmodel, DRExplainer, which leverages a directed graph convolutional network to\nenhance the prediction in a directed bipartite network framework. DRExplainer\nconstructs a directed bipartite network integrating multi-omics profiles of\ncell lines, the chemical structure of drugs and known drug response to achieve\ndirected prediction. Then, DRExplainer identifies the most relevant subgraph to\neach prediction in this directed bipartite network by learning a mask,\nfacilitating critical medical decision-making. Additionally, we introduce a\nquantifiable method for model interpretability that leverages a ground truth\nbenchmark dataset curated from biological features. In computational\nexperiments, DRExplainer outperforms state-of-the-art predictive methods and\nanother graph-based explanation method under the same experimental setting.\nFinally, the case studies further validate the interpretability and the\neffectiveness of DRExplainer in predictive novel drug response. Our code is\navailable at: https://github.com/vshy-dream/DRExplainer.\n", "date": "2024-08-22", "categories": []}
{"id": "2503.21914", "title": "PyUAT: Open-source Python framework for efficient and scalable cell\n  tracking", "abstract": "  Tracking individual cells in live-cell imaging provides fundamental insights,\ninevitable for studying causes and consequences of phenotypic heterogeneity,\nresponses to changing environmental conditions or stressors. Microbial cell\ntracking, characterized by stochastic cell movements and frequent cell\ndivisions, remains a challenging task when imaging frame rates must be limited\nto avoid counterfactual results. A promising way to overcome this limitation is\nuncertainty-aware tracking (UAT), which uses statistical models, calibrated to\nempirically observed cell behavior, to predict likely cell associations. We\npresent PyUAT, an efficient and modular Python implementation of UAT for\ntracking microbial cells in time-lapse imaging. We demonstrate its performance\non a large 2D+t data set and investigate the influence of modular biological\nmodels and imaging intervals on the tracking performance. The open-source PyUAT\nsoftware is available at https://github.com/JuBiotech/PyUAT, including example\nnotebooks for immediate use in Google Colab.\n", "date": "2025-03-27", "categories": []}
{"id": "2501.16391", "title": "Inductive-Associative Meta-learning Pipeline with Human Cognitive\n  Patterns for Unseen Drug-Target Interaction Prediction", "abstract": "  Significant differences in protein structures hinder the generalization of\nexisting drug-target interaction (DTI) models, which often rely heavily on\npre-learned binding principles or detailed annotations. In contrast, BioBridge\ndesigns an Inductive-Associative pipeline inspired by the workflow of\nscientists who base their accumulated expertise on drawing insights into novel\ndrug-target pairs from weakly related references. BioBridge predicts novel\ndrug-target interactions using limited sequence data, incorporating multi-level\nencoders with adversarial training to accumulate transferable binding\nprinciples. On these principles basis, BioBridge employs a dynamic prototype\nmeta-learning framework to associate insights from weakly related annotations,\nenabling robust predictions for previously unseen drug-target pairs. Extensive\nexperiments demonstrate that BioBridge surpasses existing models, especially\nfor unseen proteins. Notably, when only homologous protein binding data is\navailable, BioBridge proves effective for virtual screening of the epidermal\ngrowth factor receptor and adenosine receptor, underscoring its potential in\ndrug discovery.\n", "date": "2025-01-26", "categories": []}
{"id": "2503.20913", "title": "TransDiffSBDD: Causality-Aware Multi-Modal Structure-Based Drug Design", "abstract": "  Structure-based drug design (SBDD) is a critical task in drug discovery,\nrequiring the generation of molecular information across two distinct\nmodalities: discrete molecular graphs and continuous 3D coordinates. However,\nexisting SBDD methods often overlook two key challenges: (1) the multi-modal\nnature of this task and (2) the causal relationship between these modalities,\nlimiting their plausibility and performance. To address both challenges, we\npropose TransDiffSBDD, an integrated framework combining autoregressive\ntransformers and diffusion models for SBDD. Specifically, the autoregressive\ntransformer models discrete molecular information, while the diffusion model\nsamples continuous distributions, effectively resolving the first challenge. To\naddress the second challenge, we design a hybrid-modal sequence for\nprotein-ligand complexes that explicitly respects the causality between\nmodalities. Experiments on the CrossDocked2020 benchmark demonstrate that\nTransDiffSBDD outperforms existing baselines.\n", "date": "2025-03-26", "categories": []}
{"id": "2503.20027", "title": "A scalable gene network model of regulatory dynamics in single cells", "abstract": "  Single-cell data provide high-dimensional measurements of the transcriptional\nstates of cells, but extracting insights into the regulatory functions of\ngenes, particularly identifying transcriptional mechanisms affected by\nbiological perturbations, remains a challenge. Many perturbations induce\ncompensatory cellular responses, making it difficult to distinguish direct from\nindirect effects on gene regulation. Modeling how gene regulatory functions\nshape the temporal dynamics of these responses is key to improving our\nunderstanding of biological perturbations. Dynamical models based on\ndifferential equations offer a principled way to capture transcriptional\ndynamics, but their application to single-cell data has been hindered by\ncomputational constraints, stochasticity, sparsity, and noise. Existing methods\neither rely on low-dimensional representations or make strong simplifying\nassumptions, limiting their ability to model transcriptional dynamics at scale.\nWe introduce a Functional and Learnable model of Cell dynamicS, FLeCS, that\nincorporates gene network structure into coupled differential equations to\nmodel gene regulatory functions. Given (pseudo)time-series single-cell data,\nFLeCS accurately infers cell dynamics at scale, provides improved functional\ninsights into transcriptional mechanisms perturbed by gene knockouts, both in\nmyeloid differentiation and K562 Perturb-seq experiments, and simulates\nsingle-cell trajectories of A549 cells following small-molecule perturbations.\n", "date": "2025-03-25", "categories": []}
{"id": "2503.20049", "title": "Deep Learning Approaches for Blood Disease Diagnosis Across\n  Hematopoietic Lineages", "abstract": "  We present a foundation modeling framework that leverages deep learning to\nuncover latent genetic signatures across the hematopoietic hierarchy. Our\napproach trains a fully connected autoencoder on multipotent progenitor cells,\nreducing over 20,000 gene features to a 256-dimensional latent space that\ncaptures predictive information for both progenitor and downstream\ndifferentiated cells such as monocytes and lymphocytes. We validate the quality\nof these embeddings by training feed-forward, transformer, and graph\nconvolutional architectures for blood disease diagnosis tasks. We also explore\nzero-shot prediction using a progenitor disease state classification model to\nclassify downstream cell conditions. Our models achieve greater than 95%\naccuracy for multi-class classification, and in the zero-shot setting, we\nachieve greater than 0.7 F1-score on the binary classification task. Future\nwork should improve embeddings further to increase robustness on lymphocyte\nclassification specifically.\n", "date": "2025-03-25", "categories": []}
{"id": "2503.20730", "title": "Benchmarking and optimizing organism wide single-cell RNA alignment\n  methods", "abstract": "  Many methods have been proposed for removing batch effects and aligning\nsingle-cell RNA (scRNA) datasets. However, performance is typically evaluated\nbased on multiple parameters and few datasets, creating challenges in assessing\nwhich method is best for aligning data at scale. Here, we introduce the\nK-Neighbors Intersection (KNI) score, a single score that both penalizes batch\neffects and measures accuracy at cross-dataset cell-type label prediction\nalongside carefully curated small (scMARK) and large (scREF) benchmarks\ncomprising 11 and 46 human scRNA studies respectively, where we have\nstandardized author labels. Using the KNI score, we evaluate and optimize\napproaches for cross-dataset single-cell RNA integration. We introduce Batch\nAdversarial single-cell Variational Inference (BA-scVI), as a new variant of\nscVI that uses adversarial training to penalize batch-effects in the encoder\nand decoder, and show this approach outperforms other methods. In the resulting\naligned space, we find that the granularity of cell-type groupings is\nconserved, supporting the notion that whole-organism cell-type maps can be\ncreated by a single model without loss of information.\n", "date": "2025-03-26", "categories": []}
{"id": "2304.12693", "title": "Phylo2Vec: a vector representation for binary trees", "abstract": "  Binary phylogenetic trees inferred from biological data are central to\nunderstanding the shared history among evolutionary units. However, inferring\nthe placement of latent nodes in a tree is computationally expensive.\nState-of-the-art methods rely on carefully designed heuristics for tree search,\nusing different data structures for easy manipulation (e.g., classes in\nobject-oriented programming languages) and readable representation of trees\n(e.g., Newick-format strings). Here, we present Phylo2Vec, a parsimonious\nencoding for phylogenetic trees that serves as a unified approach for both\nmanipulating and representing phylogenetic trees. Phylo2Vec maps any binary\ntree with $n$ leaves to a unique integer vector of length $n-1$. The advantages\nof Phylo2Vec are fourfold: i) fast tree sampling, (ii) compressed tree\nrepresentation compared to a Newick string, iii) quick and unambiguous\nverification if two binary trees are identical topologically, and iv)\nsystematic ability to traverse tree space in very large or small jumps. As a\nproof of concept, we use Phylo2Vec for maximum likelihood inference on five\nreal-world datasets and show that a simple hill-climbing-based optimisation\nscheme can efficiently traverse the vastness of tree space from a random to an\noptimal tree.\n", "date": "2023-04-25", "categories": []}
{"id": "2503.16565", "title": "Gene42: Long-Range Genomic Foundation Model With Dense Attention", "abstract": "  We introduce Gene42, a novel family of Genomic Foundation Models (GFMs)\ndesigned to manage context lengths of up to 192,000 base pairs (bp) at a\nsingle-nucleotide resolution. Gene42 models utilize a decoder-only\n(LLaMA-style) architecture with a dense self-attention mechanism. Initially\ntrained on fixed-length sequences of 4,096 bp, our models underwent continuous\npretraining to extend the context length to 192,000 bp. This iterative\nextension allowed for the comprehensive processing of large-scale genomic data\nand the capture of intricate patterns and dependencies within the human genome.\nGene42 is the first dense attention model capable of handling such extensive\nlong context lengths in genomics, challenging state-space models that often\nrely on convolutional operators among other mechanisms. Our pretrained models\nexhibit notably low perplexity values and high reconstruction accuracy,\nhighlighting their strong ability to model genomic data. Extensive experiments\non various genomic benchmarks have demonstrated state-of-the-art performance\nacross multiple tasks, including biotype classification, regulatory region\nidentification, chromatin profiling prediction, variant pathogenicity\nprediction, and species classification. The models are publicly available at\nhuggingface.co/inceptionai.\n", "date": "2025-03-20", "categories": []}
{"id": "2503.19186", "title": "Protein Structure-Function Relationship: A Kernel-PCA Approach for\n  Reaction Coordinate Identification", "abstract": "  In this study, we propose a Kernel-PCA model designed to capture\nstructure-function relationships in a protein. This model also enables ranking\nof reaction coordinates according to their impact on protein properties. By\nleveraging machine learning techniques, including Kernel and principal\ncomponent analysis (PCA), our model uncovers meaningful patterns in\nhigh-dimensional protein data obtained from molecular dynamics (MD)\nsimulations. The effectiveness of our model in accurately identifying reaction\ncoordinates has been demonstrated through its application to a G\nprotein-coupled receptor. Furthermore, this model utilizes a network-based\napproach to uncover correlations in the dynamic behavior of residues associated\nwith a specific protein property. These findings underscore the potential of\nour model as a powerful tool for protein structure-function analysis and\nvisualization.\n", "date": "2025-03-24", "categories": []}
{"id": "2410.07511", "title": "CSGDN: Contrastive Signed Graph Diffusion Network for Predicting Crop\n  Gene-phenotype Associations", "abstract": "  Positive and negative association prediction between gene and phenotype helps\nto illustrate the underlying mechanism of complex traits in organisms. The\ntranscription and regulation activity of specific genes will be adjusted\naccordingly in different cell types, developmental stages, and physiological\nstates. There are the following two problems in obtaining the positive/negative\nassociations between gene and trait: 1) High-throughput DNA/RNA sequencing and\nphenotyping are expensive and time-consuming due to the need to process large\nsample sizes; 2) experiments introduce both random and systematic errors, and,\nmeanwhile, calculations or predictions using software or models may produce\nnoise. To address these two issues, we propose a Contrastive Signed Graph\nDiffusion Network, CSGDN, to learn robust node representations with fewer\ntraining samples to achieve higher link prediction accuracy. CSGDN employs a\nsigned graph diffusion method to uncover the underlying regulatory associations\nbetween genes and phenotypes. Then, stochastic perturbation strategies are used\nto create two views for both original and diffusive graphs. Lastly, a\nmulti-view contrastive learning paradigm loss is designed to unify the node\npresentations learned from the two views to resist interference and reduce\nnoise. We conduct experiments to validate the performance of CSGDN on three\ncrop datasets: Gossypium hirsutum, Brassica napus, and Triticum turgidum. The\nresults demonstrate that the proposed model outperforms state-of-the-art\nmethods by up to 9.28% AUC for link sign prediction in G. hirsutum dataset.\n", "date": "2024-10-10", "categories": []}
{"id": "2412.04526", "title": "Leveraging Multi-modal Representations to Predict Protein Melting\n  Temperatures", "abstract": "  Accurately predicting protein melting temperature changes (Delta Tm) is\nfundamental for assessing protein stability and guiding protein engineering.\nLeveraging multi-modal protein representations has shown great promise in\ncapturing the complex relationships among protein sequences, structures, and\nfunctions. In this study, we develop models based on powerful protein language\nmodels, including ESM-2, ESM-3 and AlphaFold, using various feature extraction\nmethods to enhance prediction accuracy. By utilizing the ESM-3 model, we\nachieve a new state-of-the-art performance on the s571 test dataset, obtaining\na Pearson correlation coefficient (PCC) of 0.50. Furthermore, we conduct a fair\nevaluation to compare the performance of different protein language models in\nthe Delta Tm prediction task. Our results demonstrate that integrating\nmulti-modal protein representations could advance the prediction of protein\nmelting temperatures.\n", "date": "2024-12-05", "categories": []}
{"id": "2412.10966", "title": "FlowDock: Geometric Flow Matching for Generative Protein-Ligand Docking\n  and Affinity Prediction", "abstract": "  Powerful generative AI models of protein-ligand structure have recently been\nproposed, but few of these methods support both flexible protein-ligand docking\nand affinity estimation. Of those that do, none can directly model multiple\nbinding ligands concurrently or have been rigorously benchmarked on\npharmacologically relevant drug targets, hindering their widespread adoption in\ndrug discovery efforts. In this work, we propose FlowDock, the first deep\ngeometric generative model based on conditional flow matching that learns to\ndirectly map unbound (apo) structures to their bound (holo) counterparts for an\narbitrary number of binding ligands. Furthermore, FlowDock provides predicted\nstructural confidence scores and binding affinity values with each of its\ngenerated protein-ligand complex structures, enabling fast virtual screening of\nnew (multi-ligand) drug targets. For the well-known PoseBusters Benchmark\ndataset, FlowDock outperforms single-sequence AlphaFold 3 with a 51% blind\ndocking success rate using unbound (apo) protein input structures and without\nany information derived from multiple sequence alignments, and for the\nchallenging new DockGen-E dataset, FlowDock outperforms single-sequence\nAlphaFold 3 and matches single-sequence Chai-1 for binding pocket\ngeneralization. Additionally, in the ligand category of the 16th community-wide\nCritical Assessment of Techniques for Structure Prediction (CASP16), FlowDock\nranked among the top-5 methods for pharmacological binding affinity estimation\nacross 140 protein-ligand complexes, demonstrating the efficacy of its learned\nrepresentations in virtual screening. Source code, data, and pre-trained models\nare available at https://github.com/BioinfoMachineLearning/FlowDock.\n", "date": "2024-12-14", "categories": []}
{"id": "2503.17368", "title": "Non-Canonical Crosslinks Confound Evolutionary Protein Structure Models", "abstract": "  Evolution-based protein structure prediction models have achieved\nbreakthrough success in recent years. However, they struggle to generalize\nbeyond evolutionary priors and on sequences lacking rich homologous data. Here\nwe present a novel, out-of-domain benchmark based on sactipeptides, a rare\nclass of ribosomally synthesized and post-translationally modified peptides\n(RiPPs) characterized by sulfur-to-$\\alpha$-carbon thioether bridges creating\ncross-links between cysteine residues and backbone. We evaluate recent models\non predicting conformations compatible with these cross-links bridges for the\n10 known sactipeptides with elucidated post-translational modifications.\nCrucially, the structures of 5 of them have not yet been experimentally\nresolved. This makes the task a challenging problem for evolution-based models,\nwhich we find exhibit limited performance (0.0% to 19.2% GDT-TS on\nsulfur-to-$\\alpha$-carbon distance). Our results point at the need for\nphysics-informed models to sustain progress in biomolecular structure\nprediction.\n", "date": "2025-03-09", "categories": []}
{"id": "2503.17666", "title": "Multi-Modality Representation Learning for Antibody-Antigen Interactions\n  Prediction", "abstract": "  While deep learning models play a crucial role in predicting antibody-antigen\ninteractions (AAI), the scarcity of publicly available sequence-structure\npairings constrains their generalization. Current AAI methods often focus on\nresidue-level static details, overlooking fine-grained structural\nrepresentations of antibodies and their inter-antibody similarities. To tackle\nthis challenge, we introduce a multi-modality representation approach that\nintegates 3D structural and 1D sequence data to unravel intricate\nintra-antibody hierarchical relationships. By harnessing these representations,\nwe present MuLAAIP, an AAI prediction framework that utilizes graph attention\nnetworks to illuminate graph-level structural features and normalized adaptive\ngraph convolution networks to capture inter-antibody sequence associations.\nFurthermore, we have curated an AAI benchmark dataset comprising both\nstructural and sequence information along with interaction labels. Through\nextensive experiments on this benchmark, our results demonstrate that MuLAAIP\noutperforms current state-of-the-art methods in terms of predictive\nperformance. The implementation code and dataset are publicly available at\nhttps://github.com/trashTian/MuLAAIP for reproducibility.\n", "date": "2025-03-22", "categories": []}
{"id": "2503.16582", "title": "Machine Learning-Based Genomic Linguistic Analysis (Gene Sequence\n  Feature Learning): A Case Study on Predicting Heavy Metal Response Genes in\n  Rice", "abstract": "  This study explores the application of machine learning-based genetic\nlinguistics for identifying heavy metal response genes in rice (Oryza sativa).\nBy integrating convolutional neural networks and random forest algorithms, we\ndeveloped a hybrid model capable of extracting and learning meaningful features\nfrom gene sequences, such as k-mer frequencies and physicochemical properties.\nThe model was trained and tested on datasets of genes, achieving high\npredictive performance (precision: 0.89, F1-score: 0.82). RNA-seq and qRT-PCR\nexperiments conducted on rice leaves which exposed to Hg0, revealed\ndifferential expression of genes associated with heavy metal responses, which\nvalidated the model's predictions. Co-expression network analysis identified\n103 related genes, and a literature review indicated that these genes are\nhighly likely to be involved in heavy metal-related biological processes. By\nintegrating and comparing the analysis results with those of differentially\nexpressed genes (DEGs), the validity of the new machine learning method was\nfurther demonstrated. This study highlights the efficacy of combining machine\nlearning with genetic linguistics for large-scale gene prediction. It\ndemonstrates a cost-effective and efficient approach for uncovering molecular\nmechanisms underlying heavy metal responses, with potential applications in\ndeveloping stress-tolerant crop varieties.\n", "date": "2025-03-20", "categories": []}
{"id": "2503.17361", "title": "Gumbel-Softmax Flow Matching with Straight-Through Guidance for\n  Controllable Biological Sequence Generation", "abstract": "  Flow matching in the continuous simplex has emerged as a promising strategy\nfor DNA sequence design, but struggles to scale to higher simplex dimensions\nrequired for peptide and protein generation. We introduce Gumbel-Softmax Flow\nand Score Matching, a generative framework on the simplex based on a novel\nGumbel-Softmax interpolant with a time-dependent temperature. Using this\ninterpolant, we introduce Gumbel-Softmax Flow Matching by deriving a\nparameterized velocity field that transports from smooth categorical\ndistributions to distributions concentrated at a single vertex of the simplex.\nWe alternatively present Gumbel-Softmax Score Matching which learns to regress\nthe gradient of the probability density. Our framework enables high-quality,\ndiverse generation and scales efficiently to higher-dimensional simplices. To\nenable training-free guidance, we propose Straight-Through Guided Flows\n(STGFlow), a classifier-based guidance method that leverages straight-through\nestimators to steer the unconditional velocity field toward optimal vertices of\nthe simplex. STGFlow enables efficient inference-time guidance using\nclassifiers pre-trained on clean sequences, and can be used with any discrete\nflow method. Together, these components form a robust framework for\ncontrollable de novo sequence generation. We demonstrate state-of-the-art\nperformance in conditional DNA promoter design, sequence-only protein\ngeneration, and target-binding peptide design for rare disease treatment.\n", "date": "2025-03-21", "categories": []}
{"id": "2412.02601", "title": "MERGE: Multi-faceted Hierarchical Graph-based GNN for Gene Expression\n  Prediction from Whole Slide Histopathology Images", "abstract": "  Recent advances in Spatial Transcriptomics (ST) pair histology images with\nspatially resolved gene expression profiles, enabling predictions of gene\nexpression across different tissue locations based on image patches. This opens\nup new possibilities for enhancing whole slide image (WSI) prediction tasks\nwith localized gene expression. However, existing methods fail to fully\nleverage the interactions between different tissue locations, which are crucial\nfor accurate joint prediction. To address this, we introduce MERGE\n(Multi-faceted hiErarchical gRaph for Gene Expressions), which combines a\nmulti-faceted hierarchical graph construction strategy with graph neural\nnetworks (GNN) to improve gene expression predictions from WSIs. By clustering\ntissue image patches based on both spatial and morphological features, and\nincorporating intra- and inter-cluster edges, our approach fosters interactions\nbetween distant tissue locations during GNN learning. As an additional\ncontribution, we evaluate different data smoothing techniques that are\nnecessary to mitigate artifacts in ST data, often caused by technical\nimperfections. We advocate for adopting gene-aware smoothing methods that are\nmore biologically justified. Experimental results on gene expression prediction\nshow that our GNN method outperforms state-of-the-art techniques across\nmultiple metrics.\n", "date": "2024-12-03", "categories": []}
{"id": "2501.15598", "title": "Diffusion Generative Modeling for Spatially Resolved Gene Expression\n  Inference from Histology Images", "abstract": "  Spatial Transcriptomics (ST) allows a high-resolution measurement of RNA\nsequence abundance by systematically connecting cell morphology depicted in\nHematoxylin and Eosin (H&E) stained histology images to spatially resolved gene\nexpressions. ST is a time-consuming, expensive yet powerful experimental\ntechnique that provides new opportunities to understand cancer mechanisms at a\nfine-grained molecular level, which is critical for uncovering new approaches\nfor disease diagnosis and treatments. Here, we present $\\textbf{Stem}$\n($\\textbf{S}$pa$\\textbf{T}$ially resolved gene $\\textbf{E}$xpression inference\nwith diffusion $\\textbf{M}$odel), a novel computational tool that leverages a\nconditional diffusion generative model to enable in silico gene expression\ninference from H&E stained images. Through better capturing the inherent\nstochasticity and heterogeneity in ST data, $\\textbf{Stem}$ achieves\nstate-of-the-art performance on spatial gene expression prediction and\ngenerates biologically meaningful gene profiles for new H&E stained images at\ntest time. We evaluate the proposed algorithm on datasets with various tissue\nsources and sequencing platforms, where it demonstrates clear improvement over\nexisting approaches. $\\textbf{Stem}$ generates high-fidelity gene expression\npredictions that share similar gene variation levels as ground truth data,\nsuggesting that our method preserves the underlying biological heterogeneity.\nOur proposed pipeline opens up the possibility of analyzing existing, easily\naccessible H&E stained histology images from a genomics point of view without\nphysically performing gene expression profiling and empowers potential\nbiological discovery from H&E stained histology images.\n", "date": "2025-01-26", "categories": []}
{"id": "2503.16351", "title": "Lyra: An Efficient and Expressive Subquadratic Architecture for Modeling\n  Biological Sequences", "abstract": "  Deep learning architectures such as convolutional neural networks and\nTransformers have revolutionized biological sequence modeling, with recent\nadvances driven by scaling up foundation and task-specific models. The\ncomputational resources and large datasets required, however, limit their\napplicability in biological contexts. We introduce Lyra, a subquadratic\narchitecture for sequence modeling, grounded in the biological framework of\nepistasis for understanding sequence-to-function relationships. Mathematically,\nwe demonstrate that state space models efficiently capture global epistatic\ninteractions and combine them with projected gated convolutions for modeling\nlocal relationships. We demonstrate that Lyra is performant across over 100\nwide-ranging biological tasks, achieving state-of-the-art (SOTA) performance in\nmany key areas, including protein fitness landscape prediction, biophysical\nproperty prediction (e.g. disordered protein region functions) peptide\nengineering applications (e.g. antibody binding, cell-penetrating peptide\nprediction), RNA structure analysis, RNA function prediction, and CRISPR guide\ndesign. It achieves this with orders-of-magnitude improvements in inference\nspeed and reduction in parameters (up to 120,000-fold in our tests) compared to\nrecent biology foundation models. Using Lyra, we were able to train and run\nevery task in this study on two or fewer GPUs in under two hours, democratizing\naccess to biological sequence modeling at SOTA performance, with potential\napplications to many fields.\n", "date": "2025-03-20", "categories": []}
{"id": "2503.13489", "title": "AI-driven control of bioelectric signalling for real-time topological\n  reorganization of cells", "abstract": "  Understanding and manipulating bioelectric signaling could present a new wave\nof progress in developmental biology, regenerative medicine, and synthetic\nbiology. Bioelectric signals, defined as voltage gradients across cell\nmembranes caused by ionic movements, play a role in regulating crucial\nprocesses including cellular differentiation, proliferation, apoptosis, and\ntissue morphogenesis. Recent studies demonstrate the ability to modulate these\nsignals to achieve controlled tissue regeneration and morphological outcomes in\norganisms such as planaria and frogs. However, significant knowledge gaps\nremain, particularly in predicting and controlling the spatial and temporal\ndynamics of membrane potentials (V_mem), understanding their regulatory roles\nin tissue and organ development, and exploring their therapeutic potential in\ndiseases. In this work we propose an experiment using Deep Reinforcement\nLearning (DRL) framework together with lab automation techniques for real-time\nmanipulation of bioelectric signals to guide tissue regeneration and\nmorphogenesis. The proposed framework should interact continuously with\nbiological systems, adapting strategies based on direct biological feedback.\nCombining DRL with real-time measurement techniques -- such as optogenetics,\nvoltage-sensitive dyes, fluorescent reporters, and advanced microscopy -- could\nprovide a comprehensive platform for precise bioelectric control, leading to\nimproved understanding of bioelectric mechanisms in morphogenesis, quantitative\nbioelectric models, identification of minimal experimental setups, and\nadvancements in bioelectric modulation techniques relevant to regenerative\nmedicine and cancer therapy. Ultimately, this research aims to utilize\nbioelectric signaling to develop new biomedical and bioengineering\napplications.\n", "date": "2025-03-10", "categories": []}
{"id": "2503.15438", "title": "VenusFactory: A Unified Platform for Protein Engineering Data Retrieval\n  and Language Model Fine-Tuning", "abstract": "  Natural language processing (NLP) has significantly influenced scientific\ndomains beyond human language, including protein engineering, where pre-trained\nprotein language models (PLMs) have demonstrated remarkable success. However,\ninterdisciplinary adoption remains limited due to challenges in data\ncollection, task benchmarking, and application. This work presents\nVenusFactory, a versatile engine that integrates biological data retrieval,\nstandardized task benchmarking, and modular fine-tuning of PLMs. VenusFactory\nsupports both computer science and biology communities with choices of both a\ncommand-line execution and a Gradio-based no-code interface, integrating $40+$\nprotein-related datasets and $40+$ popular PLMs. All implementations are\nopen-sourced on https://github.com/tyang816/VenusFactory.\n", "date": "2025-03-19", "categories": []}
{"id": "2406.05797", "title": "3D-MolT5: Leveraging Discrete Structural Information for Molecule-Text\n  Modeling", "abstract": "  The integration of molecular and natural language representations has emerged\nas a focal point in molecular science, with recent advancements in Language\nModels (LMs) demonstrating significant potential for comprehensive modeling of\nboth domains. However, existing approaches face notable limitations,\nparticularly in their neglect of three-dimensional (3D) information, which is\ncrucial for understanding molecular structures and functions. While some\nefforts have been made to incorporate 3D molecular information into LMs using\nexternal structure encoding modules, significant difficulties remain, such as\ninsufficient interaction across modalities in pre-training and challenges in\nmodality alignment. To address the limitations, we propose \\textbf{3D-MolT5}, a\nunified framework designed to model molecule in both sequence and 3D structure\nspaces. The key innovation of our approach lies in mapping fine-grained 3D\nsubstructure representations into a specialized 3D token vocabulary. This\nmethodology facilitates the seamless integration of sequence and structure\nrepresentations in a tokenized format, enabling 3D-MolT5 to encode molecular\nsequences, molecular structures, and text sequences within a unified\narchitecture. Leveraging this tokenized input strategy, we build a foundation\nmodel that unifies the sequence and structure data formats. We then conduct\njoint pre-training with multi-task objectives to enhance the model's\ncomprehension of these diverse modalities within a shared representation space.\nThus, our approach significantly improves cross-modal interaction and\nalignment, addressing key challenges in previous work. Further instruction\ntuning demonstrated that our 3D-MolT5 has strong generalization ability and\nsurpasses existing methods with superior performance in multiple downstream\ntasks. Our code is available at https://github.com/QizhiPei/3D-MolT5.\n", "date": "2024-06-09", "categories": []}
{"id": "2410.13643", "title": "Fine-Tuning Discrete Diffusion Models via Reward Optimization with\n  Applications to DNA and Protein Design", "abstract": "  Recent studies have demonstrated the strong empirical performance of\ndiffusion models on discrete sequences across domains from natural language to\nbiological sequence generation. For example, in the protein inverse folding\ntask, conditional diffusion models have achieved impressive results in\ngenerating natural-like sequences that fold back into the original structure.\nHowever, practical design tasks often require not only modeling a conditional\ndistribution but also optimizing specific task objectives. For instance, we may\nprefer protein sequences with high stability. To address this, we consider the\nscenario where we have pre-trained discrete diffusion models that can generate\nnatural-like sequences, as well as reward models that map sequences to task\nobjectives. We then formulate the reward maximization problem within discrete\ndiffusion models, analogous to reinforcement learning (RL), while minimizing\nthe KL divergence against pretrained diffusion models to preserve naturalness.\nTo solve this RL problem, we propose a novel algorithm, DRAKES, that enables\ndirect backpropagation of rewards through entire trajectories generated by\ndiffusion models, by making the originally non-differentiable trajectories\ndifferentiable using the Gumbel-Softmax trick. Our theoretical analysis\nindicates that our approach can generate sequences that are both natural-like\nand yield high rewards. While similar tasks have been recently explored in\ndiffusion models for continuous domains, our work addresses unique algorithmic\nand theoretical challenges specific to discrete diffusion models, which arise\nfrom their foundation in continuous-time Markov chains rather than Brownian\nmotion. Finally, we demonstrate the effectiveness of DRAKES in generating DNA\nand protein sequences that optimize enhancer activity and protein stability,\nrespectively, important tasks for gene therapies and protein-based\ntherapeutics.\n", "date": "2024-10-17", "categories": []}
{"id": "2503.12377", "title": "GCBLANE: A graph-enhanced convolutional BiLSTM attention network for\n  improved transcription factor binding site prediction", "abstract": "  Identifying transcription factor binding sites (TFBS) is crucial for\nunderstanding gene regulation, as these sites enable transcription factors\n(TFs) to bind to DNA and modulate gene expression. Despite advances in\nhigh-throughput sequencing, accurately identifying TFBS remains challenging due\nto the vast genomic data and complex binding patterns. GCBLANE, a\ngraph-enhanced convolutional bidirectional Long Short-Term Memory (LSTM)\nattention network, is introduced to address this issue. It integrates\nconvolutional, multi-head attention, and recurrent layers with a graph neural\nnetwork to detect key features for TFBS prediction. On 690 ENCODE ChIP-Seq\ndatasets, GCBLANE achieved an average AUC of 0.943, and on 165 ENCODE datasets,\nit reached an AUC of 0.9495, outperforming advanced models that utilize\nmultimodal approaches, including DNA shape information. This result underscores\nGCBLANE's effectiveness compared to other methods. By combining graph-based\nlearning with sequence analysis, GCBLANE significantly advances TFBS\nprediction.\n", "date": "2025-03-16", "categories": []}
{"id": "2306.12360", "title": "Protein Discovery with Discrete Walk-Jump Sampling", "abstract": "  We resolve difficulties in training and sampling from a discrete generative\nmodel by learning a smoothed energy function, sampling from the smoothed data\nmanifold with Langevin Markov chain Monte Carlo (MCMC), and projecting back to\nthe true data manifold with one-step denoising. Our Discrete Walk-Jump Sampling\nformalism combines the contrastive divergence training of an energy-based model\nand improved sample quality of a score-based model, while simplifying training\nand sampling by requiring only a single noise level. We evaluate the robustness\nof our approach on generative modeling of antibody proteins and introduce the\ndistributional conformity score to benchmark protein generative models. By\noptimizing and sampling from our models for the proposed distributional\nconformity score, 97-100% of generated samples are successfully expressed and\npurified and 70% of functional designs show equal or improved binding affinity\ncompared to known functional antibodies on the first attempt in a single round\nof laboratory experiments. We also report the first demonstration of long-run\nfast-mixing MCMC chains where diverse antibody protein classes are visited in a\nsingle MCMC chain.\n", "date": "2023-06-08", "categories": []}
{"id": "2503.10655", "title": "Language modelling techniques for analysing the impact of human genetic\n  variation", "abstract": "  Interpreting the effects of variants within the human genome and proteome is\nessential for analysing disease risk, predicting medication response, and\ndeveloping personalised health interventions. Due to the intrinsic similarities\nbetween the structure of natural languages and genetic sequences, natural\nlanguage processing techniques have demonstrated great applicability in\ncomputational variant effect prediction. In particular, the advent of the\nTransformer has led to significant advancements in the field. However,\nTransformer-based models are not without their limitations, and a number of\nextensions and alternatives have been developed to improve results and enhance\ncomputational efficiency. This review explores the use of language models for\ncomputational variant effect prediction over the past decade, analysing the\nmain architectures, and identifying key trends and future directions.\n", "date": "2025-03-07", "categories": []}
{"id": "2503.10713", "title": "HiCMamba: Enhancing Hi-C Resolution and Identifying 3D Genome Structures\n  with State Space Modeling", "abstract": "  Hi-C technology measures genome-wide interaction frequencies, providing a\npowerful tool for studying the 3D genomic structure within the nucleus.\nHowever, high sequencing costs and technical challenges often result in Hi-C\ndata with limited coverage, leading to imprecise estimates of chromatin\ninteraction frequencies. To address this issue, we present a novel deep\nlearning-based method HiCMamba to enhance the resolution of Hi-C contact maps\nusing a state space model. We adopt the UNet-based auto-encoder architecture to\nstack the proposed holistic scan block, enabling the perception of both global\nand local receptive fields at multiple scales. Experimental results demonstrate\nthat HiCMamba outperforms state-of-the-art methods while significantly reducing\ncomputational resources. Furthermore, the 3D genome structures, including\ntopologically associating domains (TADs) and loops, identified in the contact\nmaps recovered by HiCMamba are validated through associated epigenomic\nfeatures. Our work demonstrates the potential of a state space model as\nfoundational frameworks in the field of Hi-C resolution enhancement.\n", "date": "2025-03-13", "categories": []}
{"id": "2410.12459", "title": "HELM: Hierarchical Encoding for mRNA Language Modeling", "abstract": "  Messenger RNA (mRNA) plays a crucial role in protein synthesis, with its\ncodon structure directly impacting biological properties. While Language Models\n(LMs) have shown promise in analyzing biological sequences, existing approaches\nfail to account for the hierarchical nature of mRNA's codon structure. We\nintroduce Hierarchical Encoding for mRNA Language Modeling (HELM), a novel\npre-training strategy that incorporates codon-level hierarchical structure into\nlanguage model training. HELM modulates the loss function based on codon\nsynonymity, aligning the model's learning process with the biological reality\nof mRNA sequences. We evaluate HELM on diverse mRNA datasets and tasks,\ndemonstrating that HELM outperforms standard language model pre-training as\nwell as existing foundation model baselines on seven diverse downstream\nproperty prediction tasks and an antibody region annotation tasks on average by\naround 8%. Additionally, HELM enhances the generative capabilities of language\nmodel, producing diverse mRNA sequences that better align with the underlying\ntrue data distribution compared to non-hierarchical baselines.\n", "date": "2024-10-16", "categories": []}
{"id": "2503.08732", "title": "Quantifying Circadian Desynchrony in ICU Patients and Its Association\n  with Delirium", "abstract": "  Background: Circadian desynchrony characterized by the misalignment between\nan individual's internal biological rhythms and external environmental cues,\nsignificantly affects various physiological processes and health outcomes.\nQuantifying circadian desynchrony often requires prolonged and frequent\nmonitoring, and currently, an easy tool for this purpose is missing.\nAdditionally, its association with the incidence of delirium has not been\nclearly explored. Methods: A prospective observational study was carried out in\nintensive care units (ICU) of a tertiary hospital. Circadian transcriptomics of\nblood monocytes from 86 individuals were collected on two consecutive days,\nalthough a second sample could not be obtained from all participants. Using two\npublic datasets comprised of healthy volunteers, we replicated a model for\ndetermining internal circadian time. We developed an approach to quantify\ncircadian desynchrony by comparing internal circadian time and external blood\ncollection time. We applied the model and quantified circadian desynchrony\nindex among ICU patients, and investigated its association with the incidence\nof delirium. Results: The replicated model for determining internal circadian\ntime achieved comparable high accuracy. The quantified circadian desynchrony\nindex was significantly higher among critically ill ICU patients compared to\nhealthy subjects, with values of 10.03 hours vs 2.50-2.95 hours (p < 0.001).\nMost ICU patients had a circadian desynchrony index greater than 9 hours.\nAdditionally, the index was lower in patients whose blood samples were drawn\nafter 3pm, with values of 5.00 hours compared to 10.01-10.90 hours in other\ngroups (p < 0.001)...\n", "date": "2025-03-11", "categories": []}
{"id": "2503.08764", "title": "Towards Interpretable Protein Structure Prediction with Sparse\n  Autoencoders", "abstract": "  Protein language models have revolutionized structure prediction, but their\nnonlinear nature obscures how sequence representations inform structure\nprediction. While sparse autoencoders (SAEs) offer a path to interpretability\nhere by learning linear representations in high-dimensional space, their\napplication has been limited to smaller protein language models unable to\nperform structure prediction. In this work, we make two key advances: (1) we\nscale SAEs to ESM2-3B, the base model for ESMFold, enabling mechanistic\ninterpretability of protein structure prediction for the first time, and (2) we\nadapt Matryoshka SAEs for protein language models, which learn hierarchically\norganized features by forcing nested groups of latents to reconstruct inputs\nindependently. We demonstrate that our Matryoshka SAEs achieve comparable or\nbetter performance than standard architectures. Through comprehensive\nevaluations, we show that SAEs trained on ESM2-3B significantly outperform\nthose trained on smaller models for both biological concept discovery and\ncontact map prediction. Finally, we present an initial case study demonstrating\nhow our approach enables targeted steering of ESMFold predictions, increasing\nstructure solvent accessibility while fixing the input sequence. To facilitate\nfurther investigation by the broader community, we open-source our code,\ndataset, pretrained models https://github.com/johnyang101/reticular-sae , and\nvisualizer https://sae.reticular.ai .\n", "date": "2025-03-11", "categories": []}
{"id": "2503.09244", "title": "How To Make Your Cell Tracker Say \"I dunno!\"", "abstract": "  Cell tracking is a key computational task in live-cell microscopy, but fully\nautomated analysis of high-throughput imaging requires reliable and, thus,\nuncertainty-aware data analysis tools, as the amount of data recorded within a\nsingle experiment exceeds what humans are able to overlook. We here propose and\nbenchmark various methods to reason about and quantify uncertainty in linear\nassignment-based cell tracking algorithms. Our methods take inspiration from\nstatistics and machine learning, leveraging two perspectives on the cell\ntracking problem explored throughout this work: Considering it as a Bayesian\ninference problem and as a classification problem. Our methods admit a\nframework-like character in that they equip any frame-to-frame tracking method\nwith uncertainty quantification. We demonstrate this by applying it to various\nexisting tracking algorithms including the recently presented Transformer-based\ntrackers. We demonstrate empirically that our methods yield useful and\nwell-calibrated tracking uncertainties.\n", "date": "2025-03-12", "categories": []}
{"id": "2502.13785", "title": "Helix-mRNA: A Hybrid Foundation Model For Full Sequence mRNA\n  Therapeutics", "abstract": "  mRNA-based vaccines have become a major focus in the pharmaceutical industry.\nThe coding sequence as well as the Untranslated Regions (UTRs) of an mRNA can\nstrongly influence translation efficiency, stability, degradation, and other\nfactors that collectively determine a vaccine's effectiveness. However,\noptimizing mRNA sequences for those properties remains a complex challenge.\nExisting deep learning models often focus solely on coding region optimization,\noverlooking the UTRs. We present Helix-mRNA, a structured state-space-based and\nattention hybrid model to address these challenges. In addition to a first\npre-training, a second pre-training stage allows us to specialise the model\nwith high-quality data. We employ single nucleotide tokenization of mRNA\nsequences with codon separation, ensuring prior biological and structural\ninformation from the original mRNA sequence is not lost. Our model, Helix-mRNA,\noutperforms existing methods in analysing both UTRs and coding region\nproperties. It can process sequences 6x longer than current approaches while\nusing only 10% of the parameters of existing foundation models. Its predictive\ncapabilities extend to all mRNA regions. We open-source the model\n(https://github.com/helicalAI/helical) and model weights\n(https://huggingface.co/helical-ai/helix-mRNA).\n", "date": "2025-02-19", "categories": []}
{"id": "2503.07981", "title": "Regulatory DNA sequence Design with Reinforcement Learning", "abstract": "  Cis-regulatory elements (CREs), such as promoters and enhancers, are\nrelatively short DNA sequences that directly regulate gene expression. The\nfitness of CREs, measured by their ability to modulate gene expression, highly\ndepends on the nucleotide sequences, especially specific motifs known as\ntranscription factor binding sites (TFBSs). Designing high-fitness CREs is\ncrucial for therapeutic and bioengineering applications. Current CRE design\nmethods are limited by two major drawbacks: (1) they typically rely on\niterative optimization strategies that modify existing sequences and are prone\nto local optima, and (2) they lack the guidance of biological prior knowledge\nin sequence optimization. In this paper, we address these limitations by\nproposing a generative approach that leverages reinforcement learning (RL) to\nfine-tune a pre-trained autoregressive (AR) model. Our method incorporates\ndata-driven biological priors by deriving computational inference-based rewards\nthat simulate the addition of activator TFBSs and removal of repressor TFBSs,\nwhich are then integrated into the RL process. We evaluate our method on\npromoter design tasks in two yeast media conditions and enhancer design tasks\nfor three human cell types, demonstrating its ability to generate high-fitness\nCREs while maintaining sequence diversity. The code is available at\nhttps://github.com/yangzhao1230/TACO.\n", "date": "2025-03-11", "categories": []}
{"id": "2503.08160", "title": "Concept-Driven Deep Learning for Enhanced Protein-Specific Molecular\n  Generation", "abstract": "  In recent years, deep learning techniques have made significant strides in\nmolecular generation for specific targets, driving advancements in drug\ndiscovery. However, existing molecular generation methods present significant\nlimitations: those operating at the atomic level often lack synthetic\nfeasibility, drug-likeness, and interpretability, while fragment-based\napproaches frequently overlook comprehensive factors that influence\nprotein-molecule interactions. To address these challenges, we propose a novel\nfragment-based molecular generation framework tailored for specific proteins.\nOur method begins by constructing a protein subpocket and molecular arm\nconcept-based neural network, which systematically integrates interaction force\ninformation and geometric complementarity to sample molecular arms for specific\nprotein subpockets. Subsequently, we introduce a diffusion model to generate\nmolecular backbones that connect these arms, ensuring structural integrity and\nchemical diversity. Our approach significantly improves synthetic feasibility\nand binding affinity, with a 4% increase in drug-likeness and a 6% improvement\nin synthetic feasibility. Furthermore, by integrating explicit interaction data\nthrough a concept-based model, our framework enhances interpretability,\noffering valuable insights into the molecular design process.\n", "date": "2025-03-11", "categories": []}
{"id": "2405.17631", "title": "BioDiscoveryAgent: An AI Agent for Designing Genetic Perturbation\n  Experiments", "abstract": "  Agents based on large language models have shown great potential in\naccelerating scientific discovery by leveraging their rich background knowledge\nand reasoning capabilities. In this paper, we introduce BioDiscoveryAgent, an\nagent that designs new experiments, reasons about their outcomes, and\nefficiently navigates the hypothesis space to reach desired solutions. We\ndemonstrate our agent on the problem of designing genetic perturbation\nexperiments, where the aim is to find a small subset out of many possible genes\nthat, when perturbed, result in a specific phenotype (e.g., cell growth).\nUtilizing its biological knowledge, BioDiscoveryAgent can uniquely design new\nexperiments without the need to train a machine learning model or explicitly\ndesign an acquisition function as in Bayesian optimization. Moreover,\nBioDiscoveryAgent, using Claude 3.5 Sonnet, achieves an average of 21%\nimprovement in predicting relevant genetic perturbations across six datasets,\nand a 46% improvement in the harder task of non-essential gene perturbation,\ncompared to existing Bayesian optimization baselines specifically trained for\nthis task. Our evaluation includes one dataset that is unpublished, ensuring it\nis not part of the language model's training data. Additionally,\nBioDiscoveryAgent predicts gene combinations to perturb more than twice as\naccurately as a random baseline, a task so far not explored in the context of\nclosed-loop experiment design. The agent also has access to tools for searching\nthe biomedical literature, executing code to analyze biological datasets, and\nprompting another agent to critically evaluate its predictions. Overall,\nBioDiscoveryAgent is interpretable at every stage, representing an accessible\nnew paradigm in the computational design of biological experiments with the\npotential to augment scientists' efficacy.\n", "date": "2024-05-27", "categories": []}
{"id": "2407.19838", "title": "RNACG: A Universal RNA Sequence Conditional Generation model based on\n  Flow-Matching", "abstract": "  RNA plays a pivotal role in diverse biological processes, ranging from gene\nregulation to catalysis. Recent advances in RNA design, such as RfamGen,\nRibodiffusion and RDesign, have demonstrated promising results, with successful\ndesigns of functional sequences. However, RNA design remains challenging due to\nthe inherent flexibility of RNA molecules and the scarcity of experimental data\non tertiary and secondary structures compared to proteins. These limitations\nhighlight the need for a more universal and comprehensive approach to RNA\ndesign that integrates diverse annotation information at the sequence level. To\naddress these challenges, we propose RNACG (RNA Conditional Generator), a\nuniversal framework for RNA sequence design based on flow matching. RNACG\nsupports diverse conditional inputs, including structural, functional, and\nfamily-specific annotations, and offers a modular design that allows users to\ncustomize the encoding network for specific tasks. By unifying sequence\ngeneration under a single framework, RNACG enables the integration of multiple\nRNA design paradigms, from family-specific generation to tertiary structure\ninverse folding.\n", "date": "2024-07-29", "categories": []}
{"id": "2502.04684", "title": "G2PDiffusion: Cross-Species Genotype-to-Phenotype Prediction via\n  Evolutionary Diffusion", "abstract": "  Understanding how genes influence phenotype across species is a fundamental\nchallenge in genetic engineering, which will facilitate advances in various\nfields such as crop breeding, conservation biology, and personalized medicine.\nHowever, current phenotype prediction models are limited to individual species\nand expensive phenotype labeling process, making the genotype-to-phenotype\nprediction a highly domain-dependent and data-scarce problem. To this end, we\nsuggest taking images as morphological proxies, facilitating cross-species\ngeneralization through large-scale multimodal pretraining. We propose the first\ngenotype-to-phenotype diffusion model (G2PDiffusion) that generates\nmorphological images from DNA considering two critical evolutionary signals,\ni.e., multiple sequence alignments (MSA) and environmental contexts. The model\ncontains three novel components: 1) a MSA retrieval engine that identifies\nconserved and co-evolutionary patterns; 2) an environment-aware MSA conditional\nencoder that effectively models complex genotype-environment interactions; and\n3) an adaptive phenomic alignment module to improve genotype-phenotype\nconsistency. Extensive experiments show that integrating evolutionary signals\nwith environmental context enriches the model's understanding of phenotype\nvariability across species, thereby offering a valuable and promising\nexploration into advanced AI-assisted genomic analysis.\n", "date": "2025-02-07", "categories": []}
{"id": "2406.09898", "title": "Positive-Unlabelled Learning for identifying new candidate Dietary\n  Restriction-related genes among Ageing-related genes", "abstract": "  Dietary Restriction (DR) is one of the most popular anti-ageing\ninterventions; recently, Machine Learning (ML) has been explored to identify\npotential DR-related genes among ageing-related genes, aiming to minimize\ncostly wet lab experiments needed to expand our knowledge on DR. However, to\ntrain a model from positive (DR-related) and negative (non-DR-related)\nexamples, the existing ML approach naively labels genes without known DR\nrelation as negative examples, assuming that lack of DR-related annotation for\na gene represents evidence of absence of DR-relatedness, rather than absence of\nevidence. This hinders the reliability of the negative examples (non-DR-related\ngenes) and the method's ability to identify novel DR-related genes. This work\nintroduces a novel gene prioritisation method based on the two-step\nPositive-Unlabelled (PU) Learning paradigm: using a similarity-based,\nKNN-inspired approach, our method first selects reliable negative examples\namong the genes without known DR associations. Then, these reliable negatives\nand all known positives are used to train a classifier that effectively\ndifferentiates DR-related and non-DR-related genes, which is finally employed\nto generate a more reliable ranking of promising genes for novel\nDR-relatedness. Our method significantly outperforms (p<0.05) the existing\nstate-of-the-art approach in three predictive accuracy metrics with up to 40%\nlower computational cost in the best case, and we identify 4 new promising\nDR-related genes (PRKAB1, PRKAB2, IRS2, PRKAG1), all with evidence from the\nexisting literature supporting their potential DR-related role.\n", "date": "2024-06-14", "categories": []}
{"id": "2406.12723", "title": "BIOSCAN-5M: A Multimodal Dataset for Insect Biodiversity", "abstract": "  As part of an ongoing worldwide effort to comprehend and monitor insect\nbiodiversity, this paper presents the BIOSCAN-5M Insect dataset to the machine\nlearning community and establish several benchmark tasks. BIOSCAN-5M is a\ncomprehensive dataset containing multi-modal information for over 5 million\ninsect specimens, and it significantly expands existing image-based biological\ndatasets by including taxonomic labels, raw nucleotide barcode sequences,\nassigned barcode index numbers, geographical, and size information. We propose\nthree benchmark experiments to demonstrate the impact of the multi-modal data\ntypes on the classification and clustering accuracy. First, we pretrain a\nmasked language model on the DNA barcode sequences of the BIOSCAN-5M dataset,\nand demonstrate the impact of using this large reference library on species-\nand genus-level classification performance. Second, we propose a zero-shot\ntransfer learning task applied to images and DNA barcodes to cluster feature\nembeddings obtained from self-supervised learning, to investigate whether\nmeaningful clusters can be derived from these representation embeddings. Third,\nwe benchmark multi-modality by performing contrastive learning on DNA barcodes,\nimage data, and taxonomic information. This yields a general shared embedding\nspace enabling taxonomic classification using multiple types of information and\nmodalities. The code repository of the BIOSCAN-5M Insect dataset is available\nat https://github.com/bioscan-ml/BIOSCAN-5M.\n", "date": "2024-06-18", "categories": []}
{"id": "2410.05670", "title": "Improving Disease Comorbidity Prediction Based on Human Interactome with\n  Biologically Supervised Graph Embedding", "abstract": "  Comorbidity carries significant implications for disease understanding and\nmanagement. The genetic causes for comorbidity often trace back to mutations\noccurred either in the same gene associated with two diseases or in different\ngenes associated with different diseases respectively but coming into\nconnection via protein-protein interactions. Therefore, human interactome has\nbeen used in more sophisticated study of disease comorbidity. Human\ninteractome, as a large incomplete graph, presents its own challenges to\nextracting useful features for comorbidity prediction. In this work, we\nintroduce a novel approach named Biologically Supervised Graph Embedding (BSE)\nto allow for selecting most relevant features to enhance the prediction\naccuracy of comorbid disease pairs. Our investigation into BSE's impact on both\ncentered and uncentered embedding methods showcases its consistent superiority\nover the state-of-the-art techniques and its adeptness in selecting dimensions\nenriched with vital biological insights, thereby improving prediction\nperformance significantly, up to 50% when measured by ROC for some variations.\nFurther analysis indicates that BSE consistently and substantially improves the\nratio of disease associations to gene connectivity, affirming its potential in\nuncovering latent biological factors affecting comorbidity. The statistically\nsignificant enhancements across diverse metrics underscore BSE's potential to\nintroduce novel avenues for precise disease comorbidity predictions and other\npotential applications. The GitHub repository containing the source code can be\naccessed at the following link:\nhttps://github.com/xihan-qin/Biologically-Supervised-Graph-Embedding.\n", "date": "2024-10-08", "categories": []}
{"id": "2503.03904", "title": "The Signed Two-Space Proximity Model for Learning Representations in\n  Protein-Protein Interaction Networks", "abstract": "  Accurately predicting complex protein-protein interactions (PPIs) is crucial\nfor decoding biological processes, from cellular functioning to disease\nmechanisms. However, experimental methods for determining PPIs are\ncomputationally expensive. Thus, attention has been recently drawn to machine\nlearning approaches. Furthermore, insufficient effort has been made toward\nanalyzing signed PPI networks, which capture both activating (positive) and\ninhibitory (negative) interactions. To accurately represent biological\nrelationships, we present the Signed Two-Space Proximity Model (S2-SPM) for\nsigned PPI networks, which explicitly incorporates both types of interactions,\nreflecting the complex regulatory mechanisms within biological systems. This is\nachieved by leveraging two independent latent spaces to differentiate between\npositive and negative interactions while representing protein similarity\nthrough proximity in these spaces. Our approach also enables the identification\nof archetypes representing extreme protein profiles. S2-SPM's superior\nperformance in predicting the presence and sign of interactions in SPPI\nnetworks is demonstrated in link prediction tasks against relevant baseline\nmethods. Additionally, the biological prevalence of the identified archetypes\nis confirmed by an enrichment analysis of Gene Ontology (GO) terms, which\nreveals that distinct biological tasks are associated with archetypal groups\nformed by both interactions. This study is also validated regarding statistical\nsignificance and sensitivity analysis, providing insights into the functional\nroles of different interaction types. Finally, the robustness and consistency\nof the extracted archetype structures are confirmed using the Bayesian\nNormalized Mutual Information (BNMI) metric, proving the model's reliability in\ncapturing meaningful SPPI patterns.\n", "date": "2025-03-05", "categories": []}
{"id": "2503.03989", "title": "Integrating Protein Dynamics into Structure-Based Drug Design via\n  Full-Atom Stochastic Flows", "abstract": "  The dynamic nature of proteins, influenced by ligand interactions, is\nessential for comprehending protein function and progressing drug discovery.\nTraditional structure-based drug design (SBDD) approaches typically target\nbinding sites with rigid structures, limiting their practical application in\ndrug development. While molecular dynamics simulation can theoretically capture\nall the biologically relevant conformations, the transition rate is dictated by\nthe intrinsic energy barrier between them, making the sampling process\ncomputationally expensive. To overcome the aforementioned challenges, we\npropose to use generative modeling for SBDD considering conformational changes\nof protein pockets. We curate a dataset of apo and multiple holo states of\nprotein-ligand complexes, simulated by molecular dynamics, and propose a\nfull-atom flow model (and a stochastic version), named DynamicFlow, that learns\nto transform apo pockets and noisy ligands into holo pockets and corresponding\n3D ligand molecules. Our method uncovers promising ligand molecules and\ncorresponding holo conformations of pockets. Additionally, the resultant\nholo-like states provide superior inputs for traditional SBDD approaches,\nplaying a significant role in practical drug discovery.\n", "date": "2025-03-06", "categories": []}
{"id": "2503.04135", "title": "Biological Sequence with Language Model Prompting: A Survey", "abstract": "  Large Language models (LLMs) have emerged as powerful tools for addressing\nchallenges across diverse domains. Notably, recent studies have demonstrated\nthat large language models significantly enhance the efficiency of biomolecular\nanalysis and synthesis, attracting widespread attention from academics and\nmedicine. In this paper, we systematically investigate the application of\nprompt-based methods with LLMs to biological sequences, including DNA, RNA,\nproteins, and drug discovery tasks. Specifically, we focus on how prompt\nengineering enables LLMs to tackle domain-specific problems, such as promoter\nsequence prediction, protein structure modeling, and drug-target binding\naffinity prediction, often with limited labeled data. Furthermore, our\ndiscussion highlights the transformative potential of prompting in\nbioinformatics while addressing key challenges such as data scarcity,\nmultimodal fusion, and computational resource limitations. Our aim is for this\npaper to function both as a foundational primer for newcomers and a catalyst\nfor continued innovation within this dynamic field of study.\n", "date": "2025-03-06", "categories": []}
{"id": "2503.04362", "title": "A Generalist Cross-Domain Molecular Learning Framework for\n  Structure-Based Drug Discovery", "abstract": "  Structure-based drug discovery (SBDD) is a systematic scientific process that\ndevelops new drugs by leveraging the detailed physical structure of the target\nprotein. Recent advancements in pre-trained models for biomolecules have\ndemonstrated remarkable success across various biochemical applications,\nincluding drug discovery and protein engineering. However, in most approaches,\nthe pre-trained models primarily focus on the characteristics of either small\nmolecules or proteins, without delving into their binding interactions which\nare essential cross-domain relationships pivotal to SBDD. To fill this gap, we\npropose a general-purpose foundation model named BIT (an abbreviation for\nBiomolecular Interaction Transformer), which is capable of encoding a range of\nbiochemical entities, including small molecules, proteins, and protein-ligand\ncomplexes, as well as various data formats, encompassing both 2D and 3D\nstructures. Specifically, we introduce Mixture-of-Domain-Experts (MoDE) to\nhandle the biomolecules from diverse biochemical domains and\nMixture-of-Structure-Experts (MoSE) to capture positional dependencies in the\nmolecular structures. The proposed mixture-of-experts approach enables BIT to\nachieve both deep fusion and domain-specific encoding, effectively capturing\nfine-grained molecular interactions within protein-ligand complexes. Then, we\nperform cross-domain pre-training on the shared Transformer backbone via\nseveral unified self-supervised denoising tasks. Experimental results on\nvarious benchmarks demonstrate that BIT achieves exceptional performance in\ndownstream tasks, including binding affinity prediction, structure-based\nvirtual screening, and molecular property prediction.\n", "date": "2025-03-06", "categories": []}
{"id": "2503.04643", "title": "Adaptive Prototype Learning for Multimodal Cancer Survival Analysis", "abstract": "  Leveraging multimodal data, particularly the integration of whole-slide\nhistology images (WSIs) and transcriptomic profiles, holds great promise for\nimproving cancer survival prediction. However, excessive redundancy in\nmultimodal data can degrade model performance. In this paper, we propose\nAdaptive Prototype Learning (APL), a novel and effective approach for\nmultimodal cancer survival analysis. APL adaptively learns representative\nprototypes in a data-driven manner, reducing redundancy while preserving\ncritical information. Our method employs two sets of learnable query vectors\nthat serve as a bridge between high-dimensional representations and survival\nprediction, capturing task-relevant features. Additionally, we introduce a\nmultimodal mixed self-attention mechanism to enable cross-modal interactions,\nfurther enhancing information fusion. Extensive experiments on five benchmark\ncancer datasets demonstrate the superiority of our approach over existing\nmethods. The code is available at https://github.com/HongLiuuuuu/APL.\n", "date": "2025-03-06", "categories": []}
{"id": "2503.04650", "title": "Joint Masked Reconstruction and Contrastive Learning for Mining\n  Interactions Between Proteins", "abstract": "  Protein-protein interaction (PPI) prediction is an instrumental means in\nelucidating the mechanisms underlying cellular operations, holding significant\npractical implications for the realms of pharmaceutical development and\nclinical treatment. Presently, the majority of research methods primarily\nconcentrate on the analysis of amino acid sequences, while investigations\npredicated on protein structures remain in the nascent stages of exploration.\nDespite the emergence of several structure-based algorithms in recent years,\nthese are still confronted with inherent challenges: (1) the extraction of\nintrinsic structural information of proteins typically necessitates the\nexpenditure of substantial computational resources; (2) these models are overly\nreliant on seen protein data, struggling to effectively unearth interaction\ncues between unknown proteins. To further propel advancements in this domain,\nthis paper introduces a novel PPI prediction method jointing masked\nreconstruction and contrastive learning, termed JmcPPI. This methodology\ndissects the PPI prediction task into two distinct phases: during the residue\nstructure encoding phase, JmcPPI devises two feature reconstruction tasks and\nemploys graph attention mechanism to capture structural information between\nresidues; during the protein interaction inference phase, JmcPPI perturbs the\noriginal PPI graph and employs a multi-graph contrastive learning strategy to\nthoroughly mine extrinsic interaction information of novel proteins. Extensive\nexperiments conducted on three widely utilized PPI datasets demonstrate that\nJmcPPI surpasses existing optimal baseline models across various data partition\nschemes. The associated code can be accessed via\nhttps://github.com/lijfrank-open/JmcPPI.\n", "date": "2025-03-06", "categories": []}
{"id": "2405.05998", "title": "Whole Genome Transformer for Gene Interaction Effects in Microbiome\n  Habitat Specificity", "abstract": "  Leveraging the vast genetic diversity within microbiomes offers unparalleled\ninsights into complex phenotypes, yet the task of accurately predicting and\nunderstanding such traits from genomic data remains challenging. We propose a\nframework taking advantage of existing large models for gene vectorization to\npredict habitat specificity from entire microbial genome sequences. Based on\nour model, we develop attribution techniques to elucidate gene interaction\neffects that drive microbial adaptation to diverse environments. We train and\nvalidate our approach on a large dataset of high quality microbiome genomes\nfrom different habitats. We not only demonstrate solid predictive performance,\nbut also how sequence-level information of entire genomes allows us to identify\ngene associations underlying complex phenotypes. Our attribution recovers known\nimportant interaction networks and proposes new candidates for experimental\nfollow up.\n", "date": "2024-05-09", "categories": []}
{"id": "2503.01910", "title": "dyAb: Flow Matching for Flexible Antibody Design with AlphaFold-driven\n  Pre-binding Antigen", "abstract": "  The development of therapeutic antibodies heavily relies on accurate\npredictions of how antigens will interact with antibodies. Existing\ncomputational methods in antibody design often overlook crucial conformational\nchanges that antigens undergo during the binding process, significantly\nimpacting the reliability of the resulting antibodies. To bridge this gap, we\nintroduce dyAb, a flexible framework that incorporates AlphaFold2-driven\npredictions to model pre-binding antigen structures and specifically addresses\nthe dynamic nature of antigen conformation changes. Our dyAb model leverages a\nunique combination of coarse-grained interface alignment and fine-grained flow\nmatching techniques to simulate the interaction dynamics and structural\nevolution of the antigen-antibody complex, providing a realistic representation\nof the binding process. Extensive experiments show that dyAb significantly\noutperforms existing models in antibody design involving changing antigen\nconformations. These results highlight dyAb's potential to streamline the\ndesign process for therapeutic antibodies, promising more efficient development\ncycles and improved outcomes in clinical applications.\n", "date": "2025-03-01", "categories": []}
{"id": "2402.08491", "title": "Deep Reinforcement Learning for Controlled Traversing of the Attractor\n  Landscape of Boolean Models in the Context of Cellular Reprogramming", "abstract": "  Cellular reprogramming can be used for both the prevention and cure of\ndifferent diseases. However, the efficiency of discovering reprogramming\nstrategies with classical wet-lab experiments is hindered by lengthy time\ncommitments and high costs. In this study, we develop a novel computational\nframework based on deep reinforcement learning that facilitates the\nidentification of reprogramming strategies. For this aim, we formulate a\ncontrol problem in the context of cellular reprogramming for the frameworks of\nBNs and PBNs under the asynchronous update mode. Furthermore, we introduce the\nnotion of a pseudo-attractor and a procedure for identification of\npseudo-attractor state during training. Finally, we devise a computational\nframework for solving the control problem, which we test on a number of\ndifferent models.\n", "date": "2024-02-13", "categories": []}
{"id": "2407.11734", "title": "Multi-Modal and Multi-Attribute Generation of Single Cells with CFGen", "abstract": "  Generative modeling of single-cell RNA-seq data is crucial for tasks like\ntrajectory inference, batch effect removal, and simulation of realistic\ncellular data. However, recent deep generative models simulating synthetic\nsingle cells from noise operate on pre-processed continuous gene expression\napproximations, overlooking the discrete nature of single-cell data, which\nlimits their effectiveness and hinders the incorporation of robust noise\nmodels. Additionally, aspects like controllable multi-modal and multi-label\ngeneration of cellular data remain underexplored. This work introduces CellFlow\nfor Generation (CFGen), a flow-based conditional generative model that\npreserves the inherent discreteness of single-cell data. CFGen generates\nwhole-genome multi-modal single-cell data reliably, improving the recovery of\ncrucial biological data characteristics while tackling relevant generative\ntasks such as rare cell type augmentation and batch correction. We also\nintroduce a novel framework for compositional data generation using Flow\nMatching. By showcasing CFGen on a diverse set of biological datasets and\nsettings, we provide evidence of its value to the fields of computational\nbiology and deep generative models.\n", "date": "2024-07-16", "categories": []}
{"id": "2409.13664", "title": "Analysis of Gene Regulatory Networks from Gene Expression Using Graph\n  Neural Networks", "abstract": "  Unraveling the complexities of Gene Regulatory Networks (GRNs) is crucial for\nunderstanding cellular processes and disease mechanisms. Traditional\ncomputational methods often struggle with the dynamic nature of these networks.\nThis study explores the use of Graph Neural Networks (GNNs), a powerful\napproach for modeling graph-structured data like GRNs. Utilizing a Graph\nAttention Network v2 (GATv2), our study presents a novel approach to the\nconstruction and interrogation of GRNs, informed by gene expression data and\nBoolean models derived from literature. The model's adeptness in accurately\npredicting regulatory interactions and pinpointing key regulators is attributed\nto advanced attention mechanisms, a hallmark of the GNN framework. These\ninsights suggest that GNNs are primed to revolutionize GRN analysis, addressing\ntraditional limitations and offering richer biological insights. The success of\nGNNs, as highlighted by our model's reliance on high-quality data, calls for\nenhanced data collection methods to sustain progress. The integration of GNNs\nin GRN research is set to pioneer developments in personalized medicine, drug\ndiscovery, and our grasp of biological systems, bolstered by the structural\nanalysis of networks for improved node and edge prediction.\n", "date": "2024-09-20", "categories": []}
{"id": "2409.19788", "title": "Exploring Adversarial Robustness in Classification tasks using DNA\n  Language Models", "abstract": "  DNA Language Models, such as GROVER, DNABERT2 and the Nucleotide Transformer,\noperate on DNA sequences that inherently contain sequencing errors, mutations,\nand laboratory-induced noise, which may significantly impact model performance.\nDespite the importance of this issue, the robustness of DNA language models\nremains largely underexplored. In this paper, we comprehensivly investigate\ntheir robustness in DNA classification by applying various adversarial attack\nstrategies: the character (nucleotide substitutions), word (codon\nmodifications), and sentence levels (back-translation-based transformations) to\nsystematically analyze model vulnerabilities. Our results demonstrate that DNA\nlanguage models are highly susceptible to adversarial attacks, leading to\nsignificant performance degradation. Furthermore, we explore adversarial\ntraining method as a defense mechanism, which enhances both robustness and\nclassification accuracy. This study highlights the limitations of DNA language\nmodels and underscores the necessity of robustness in bioinformatics.\n", "date": "2024-09-29", "categories": []}
{"id": "2503.00710", "title": "Proteina: Scaling Flow-based Protein Structure Generative Models", "abstract": "  Recently, diffusion- and flow-based generative models of protein structures\nhave emerged as a powerful tool for de novo protein design. Here, we develop\nProteina, a new large-scale flow-based protein backbone generator that utilizes\nhierarchical fold class labels for conditioning and relies on a tailored\nscalable transformer architecture with up to 5x as many parameters as previous\nmodels. To meaningfully quantify performance, we introduce a new set of metrics\nthat directly measure the distributional similarity of generated proteins with\nreference sets, complementing existing metrics. We further explore scaling\ntraining data to millions of synthetic protein structures and explore improved\ntraining and sampling recipes adapted to protein backbone generation. This\nincludes fine-tuning strategies like LoRA for protein backbones, new guidance\nmethods like classifier-free guidance and autoguidance for protein backbones,\nand new adjusted training objectives. Proteina achieves state-of-the-art\nperformance on de novo protein backbone design and produces diverse and\ndesignable proteins at unprecedented length, up to 800 residues. The\nhierarchical conditioning offers novel control, enabling high-level\nsecondary-structure guidance as well as low-level fold-specific generation.\n", "date": "2025-03-02", "categories": []}
{"id": "2503.00975", "title": "Molecule Generation for Target Protein Binding with Hierarchical\n  Consistency Diffusion Model", "abstract": "  Effective generation of molecular structures, or new chemical entities, that\nbind to target proteins is crucial for lead identification and optimization in\ndrug discovery. Despite advancements in atom- and motif-wise deep learning\nmodels for 3D molecular generation, current methods often struggle with\nvalidity and reliability. To address these issues, we develop the Atom-Motif\nConsistency Diffusion Model (AMDiff), utilizing a joint-training paradigm for\nmulti-view learning. This model features a hierarchical diffusion architecture\nthat integrates both atom- and motif-level views of molecules, allowing for\ncomprehensive exploration of complementary information. By leveraging\nclassifier-free guidance and incorporating binding site features as conditional\ninputs, AMDiff ensures robust molecule generation across diverse targets.\nCompared to existing approaches, AMDiff exhibits superior validity and novelty\nin generating molecules tailored to fit various protein pockets. Case studies\ntargeting protein kinases, including Anaplastic Lymphoma Kinase (ALK) and\nCyclin-dependent kinase 4 (CDK4), demonstrate the model's capability in\nstructure-based de novo drug design. Overall, AMDiff bridges the gap between\natom-view and motif-view drug discovery and speeds up the process of\ntarget-aware molecular generation.\n", "date": "2025-03-02", "categories": []}
{"id": "2503.01488", "title": "InversionGNN: A Dual Path Network for Multi-Property Molecular\n  Optimization", "abstract": "  Exploring chemical space to find novel molecules that simultaneously satisfy\nmultiple properties is crucial in drug discovery. However, existing methods\noften struggle with trading off multiple properties due to the conflicting or\ncorrelated nature of chemical properties. To tackle this issue, we introduce\nInversionGNN framework, an effective yet sample-efficient dual-path graph\nneural network (GNN) for multi-objective drug discovery. In the direct\nprediction path of InversionGNN, we train the model for multi-property\nprediction to acquire knowledge of the optimal combination of functional\ngroups. Then the learned chemical knowledge helps the inversion generation path\nto generate molecules with required properties. In order to decode the complex\nknowledge of multiple properties in the inversion path, we propose a\ngradient-based Pareto search method to balance conflicting properties and\ngenerate Pareto optimal molecules. Additionally, InversionGNN is able to search\nthe full Pareto front approximately in discrete chemical space. Comprehensive\nexperimental evaluations show that InversionGNN is both effective and\nsample-efficient in various discrete multi-objective settings including drug\ndiscovery.\n", "date": "2025-03-03", "categories": []}
{"id": "2503.01682", "title": "GRNFormer: A Biologically-Guided Framework for Integrating Gene\n  Regulatory Networks into RNA Foundation Models", "abstract": "  Foundation models for single-cell RNA sequencing (scRNA-seq) have shown\npromising capabilities in capturing gene expression patterns. However, current\napproaches face critical limitations: they ignore biological prior knowledge\nencoded in gene regulatory relationships and fail to leverage multi-omics\nsignals that could provide complementary regulatory insights. In this paper, we\npropose GRNFormer, a new framework that systematically integrates multi-scale\nGene Regulatory Networks (GRNs) inferred from multi-omics data into RNA\nfoundation model training. Our framework introduces two key innovations. First,\nwe introduce a pipeline for constructing hierarchical GRNs that capture\nregulatory relationships at both cell-type-specific and cell-specific\nresolutions. Second, we design a structure-aware integration framework that\naddresses the information asymmetry in GRNs through two technical advances: (1)\nA graph topological adapter using multi-head cross-attention to weight\nregulatory relationships dynamically, and (2) a novel edge perturbation\nstrategy that perturb GRNs with biologically-informed co-expression links to\naugment graph neural network training. Comprehensive experiments have been\nconducted on three representative downstream tasks across multiple model\narchitectures to demonstrate the effectiveness of GRNFormer. It achieves\nconsistent improvements over state-of-the-art (SoTA) baselines: $3.6\\%$\nincrease in drug response prediction correlation, $9.6\\%$ improvement in\nsingle-cell drug classification AUC, and $1.1\\%$ average gain in gene\nperturbation prediction accuracy.\n", "date": "2025-03-03", "categories": []}
{"id": "2408.12373", "title": "Cell-ontology guided transcriptome foundation model", "abstract": "  Transcriptome foundation models TFMs hold great promises of deciphering the\ntranscriptomic language that dictate diverse cell functions by self-supervised\nlearning on large-scale single-cell gene expression data, and ultimately\nunraveling the complex mechanisms of human diseases. However, current TFMs\ntreat cells as independent samples and ignore the taxonomic relationships\nbetween cell types, which are available in cell ontology graphs. We argue that\neffectively leveraging this ontology information during the TFM pre-training\ncan improve learning biologically meaningful gene co-expression patterns while\npreserving TFM as a general purpose foundation model for downstream zero-shot\nand fine-tuning tasks. To this end, we present single cell, Cell-ontology\nguided TFM scCello. We introduce cell-type coherence loss and ontology\nalignment loss, which are minimized along with the masked gene expression\nprediction loss during the pre-training. The novel loss component guide scCello\nto learn the cell-type-specific representation and the structural relation\nbetween cell types from the cell ontology graph, respectively. We pre-trained\nscCello on 22 million cells from CellxGene database leveraging their cell-type\nlabels mapped to the cell ontology graph from Open Biological and Biomedical\nOntology Foundry. Our TFM demonstrates competitive generalization and\ntransferability performance over the existing TFMs on biologically important\ntasks including identifying novel cell types of unseen cells, prediction of\ncell-type-specific marker genes, and cancer drug responses.\n", "date": "2024-08-22", "categories": []}
{"id": "2403.12117", "title": "Transfer Learning for T-Cell Response Prediction", "abstract": "  We study the prediction of T-cell response for specific given peptides, which\ncould, among other applications, be a crucial step towards the development of\npersonalized cancer vaccines. It is a challenging task due to limited,\nheterogeneous training data featuring a multi-domain structure; such data\nentail the danger of shortcut learning, where models learn general\ncharacteristics of peptide sources, such as the source organism, rather than\nspecific peptide characteristics associated with T-cell response.\n  Using a transformer model for T-cell response prediction, we show that the\ndanger of inflated predictive performance is not merely theoretical but occurs\nin practice. Consequently, we propose a domain-aware evaluation scheme. We then\nstudy different transfer learning techniques to deal with the multi-domain\nstructure and shortcut learning. We demonstrate a per-source fine tuning\napproach to be effective across a wide range of peptide sources and further\nshow that our final model is competitive with existing state-of-the-art\napproaches for predicting T-cell responses for human peptides.\n", "date": "2024-03-18", "categories": []}
{"id": "2411.16084", "title": "Deciphering genomic codes using advanced NLP techniques: a scoping\n  review", "abstract": "  Objectives: The vast and complex nature of human genomic sequencing data\npresents challenges for effective analysis. This review aims to investigate the\napplication of Natural Language Processing (NLP) techniques, particularly Large\nLanguage Models (LLMs) and transformer architectures, in deciphering genomic\ncodes, focusing on tokenization, transformer models, and regulatory annotation\nprediction. The goal of this review is to assess data and model accessibility\nin the most recent literature, gaining a better understanding of the existing\ncapabilities and constraints of these tools in processing genomic sequencing\ndata.\n  Methods: Following Preferred Reporting Items for Systematic Reviews and\nMeta-Analyses (PRISMA) guidelines, our scoping review was conducted across\nPubMed, Medline, Scopus, Web of Science, Embase, and ACM Digital Library.\nStudies were included if they focused on NLP methodologies applied to genomic\nsequencing data analysis, without restrictions on publication date or article\ntype.\n  Results: A total of 26 studies published between 2021 and April 2024 were\nselected for review. The review highlights that tokenization and transformer\nmodels enhance the processing and understanding of genomic data, with\napplications in predicting regulatory annotations like transcription-factor\nbinding sites and chromatin accessibility.\n  Discussion: The application of NLP and LLMs to genomic sequencing data\ninterpretation is a promising field that can help streamline the processing of\nlarge-scale genomic data while also providing a better understanding of its\ncomplex structures. It has the potential to drive advancements in personalized\nmedicine by offering more efficient and scalable solutions for genomic\nanalysis. Further research is also needed to discuss and overcome current\nlimitations, enhancing model transparency and applicability.\n", "date": "2024-11-25", "categories": []}
{"id": "2502.19429", "title": "scMamba: A Pre-Trained Model for Single-Nucleus RNA Sequencing Analysis\n  in Neurodegenerative Disorders", "abstract": "  Single-nucleus RNA sequencing (snRNA-seq) has significantly advanced our\nunderstanding of the disease etiology of neurodegenerative disorders. However,\nthe low quality of specimens derived from postmortem brain tissues, combined\nwith the high variability caused by disease heterogeneity, makes it challenging\nto integrate snRNA-seq data from multiple sources for precise analyses. To\naddress these challenges, we present scMamba, a pre-trained model designed to\nimprove the quality and utility of snRNA-seq analysis, with a particular focus\non neurodegenerative diseases. Inspired by the recent Mamba model, scMamba\nintroduces a novel architecture that incorporates a linear adapter layer, gene\nembeddings, and bidirectional Mamba blocks, enabling efficient processing of\nsnRNA-seq data while preserving information from the raw input. Notably,\nscMamba learns generalizable features of cells and genes through pre-training\non snRNA-seq data, without relying on dimension reduction or selection of\nhighly variable genes. We demonstrate that scMamba outperforms benchmark\nmethods in various downstream tasks, including cell type annotation, doublet\ndetection, imputation, and the identification of differentially expressed\ngenes.\n", "date": "2025-02-12", "categories": []}
{"id": "2502.19568", "title": "PhenoProfiler: Advancing Phenotypic Learning for Image-based Drug\n  Discovery", "abstract": "  In the field of image-based drug discovery, capturing the phenotypic response\nof cells to various drug treatments and perturbations is a crucial step.\nHowever, existing methods require computationally extensive and complex\nmulti-step procedures, which can introduce inefficiencies, limit\ngeneralizability, and increase potential errors. To address these challenges,\nwe present PhenoProfiler, an innovative model designed to efficiently and\neffectively extract morphological representations, enabling the elucidation of\nphenotypic changes induced by treatments. PhenoProfiler is designed as an\nend-to-end tool that processes whole-slide multi-channel images directly into\nlow-dimensional quantitative representations, eliminating the extensive\ncomputational steps required by existing methods. It also includes a\nmulti-objective learning module to enhance robustness, accuracy, and\ngeneralization in morphological representation learning. PhenoProfiler is\nrigorously evaluated on large-scale publicly available datasets, including over\n230,000 whole-slide multi-channel images in end-to-end scenarios and more than\n8.42 million single-cell images in non-end-to-end settings. Across these\nbenchmarks, PhenoProfiler consistently outperforms state-of-the-art methods by\nup to 20%, demonstrating substantial improvements in both accuracy and\nrobustness. Furthermore, PhenoProfiler uses a tailored phenotype correction\nstrategy to emphasize relative phenotypic changes under treatments,\nfacilitating the detection of biologically meaningful signals. UMAP\nvisualizations of treatment profiles demonstrate PhenoProfiler ability to\neffectively cluster treatments with similar biological annotations, thereby\nenhancing interpretability. These findings establish PhenoProfiler as a\nscalable, generalizable, and robust tool for phenotypic learning.\n", "date": "2025-02-26", "categories": []}
{"id": "2502.13080", "title": "BOLIMES: Boruta and LIME optiMized fEature Selection for Gene Expression\n  Classification", "abstract": "  Gene expression classification is a pivotal yet challenging task in\nbioinformatics, primarily due to the high dimensionality of genomic data and\nthe risk of overfitting. To bridge this gap, we propose BOLIMES, a novel\nfeature selection algorithm designed to enhance gene expression classification\nby systematically refining the feature subset. Unlike conventional methods that\nrely solely on statistical ranking or classifier-specific selection, we\nintegrate the robustness of Boruta with the interpretability of LIME, ensuring\nthat only the most relevant and influential genes are retained. BOLIMES first\nemploys Boruta to filter out non-informative genes by comparing each feature\nagainst its randomized counterpart, thus preserving valuable information. It\nthen uses LIME to rank the remaining genes based on their local importance to\nthe classifier. Finally, an iterative classification evaluation determines the\noptimal feature subset by selecting the number of genes that maximizes\npredictive accuracy. By combining exhaustive feature selection with\ninterpretability-driven refinement, our solution effectively balances\ndimensionality reduction with high classification performance, offering a\npowerful solution for high-dimensional gene expression analysis.\n", "date": "2025-02-18", "categories": []}
{"id": "2502.18875", "title": "SE(3)-Equivariant Ternary Complex Prediction Towards Target Protein\n  Degradation", "abstract": "  Targeted protein degradation (TPD) induced by small molecules has emerged as\na rapidly evolving modality in drug discovery, targeting proteins traditionally\nconsidered \"undruggable\". Proteolysis-targeting chimeras (PROTACs) and\nmolecular glue degraders (MGDs) are the primary small molecules that induce\nTPD. Both types of molecules form a ternary complex linking an E3 ligase with a\ntarget protein, a crucial step for drug discovery. While significant advances\nhave been made in binary structure prediction for proteins and small molecules,\nternary structure prediction remains challenging due to obscure interaction\nmechanisms and insufficient training data. Traditional methods relying on\nmanually assigned rules perform poorly and are computationally demanding due to\nextensive random sampling. In this work, we introduce DeepTernary, a novel deep\nlearning-based approach that directly predicts ternary structures in an\nend-to-end manner using an encoder-decoder architecture. DeepTernary leverages\nan SE(3)-equivariant graph neural network (GNN) with both intra-graph and\nternary inter-graph attention mechanisms to capture intricate ternary\ninteractions from our collected high-quality training dataset, TernaryDB. The\nproposed query-based Pocket Points Decoder extracts the 3D structure of the\nfinal binding ternary complex from learned ternary embeddings, demonstrating\nstate-of-the-art accuracy and speed in existing PROTAC benchmarks without prior\nknowledge from known PROTACs. It also achieves notable accuracy on the more\nchallenging MGD benchmark under the blind docking protocol. Remarkably, our\nexperiments reveal that the buried surface area calculated from predicted\nstructures correlates with experimentally obtained degradation potency-related\nmetrics. Consequently, DeepTernary shows potential in effectively assisting and\naccelerating the development of TPDs for previously undruggable targets.\n", "date": "2025-02-26", "categories": []}
{"id": "2502.19391", "title": "Towards More Accurate Full-Atom Antibody Co-Design", "abstract": "  Antibody co-design represents a critical frontier in drug development, where\naccurate prediction of both 1D sequence and 3D structure of\ncomplementarity-determining regions (CDRs) is essential for targeting specific\nepitopes. Despite recent advances in equivariant graph neural networks for\nantibody design, current approaches often fall short in capturing the intricate\ninteractions that govern antibody-antigen recognition and binding specificity.\nIn this work, we present Igformer, a novel end-to-end framework that addresses\nthese limitations through innovative modeling of antibody-antigen binding\ninterfaces. Our approach refines the inter-graph representation by integrating\npersonalized propagation with global attention mechanisms, enabling\ncomprehensive capture of the intricate interplay between local chemical\ninteractions and global conformational dependencies that characterize effective\nantibody-antigen binding. Through extensive validation on epitope-binding CDR\ndesign and structure prediction tasks, Igformer demonstrates significant\nimprovements over existing methods, suggesting that explicit modeling of\nmulti-scale residue interactions can substantially advance computational\nantibody design for therapeutic applications.\n", "date": "2025-02-11", "categories": []}
{"id": "2502.19395", "title": "Fast and Accurate Antibody Sequence Design via Structure Retrieval", "abstract": "  Recent advancements in protein design have leveraged diffusion models to\ngenerate structural scaffolds, followed by a process known as protein inverse\nfolding, which involves sequence inference on these scaffolds. However, these\nmethodologies face significant challenges when applied to hyper-variable\nstructures such as antibody Complementarity-Determining Regions (CDRs), where\nsequence inference frequently results in non-functional sequences due to\nhallucinations. Distinguished from prevailing protein inverse folding\napproaches, this paper introduces Igseek, a novel structure-retrieval framework\nthat infers CDR sequences by retrieving similar structures from a natural\nantibody database. Specifically, Igseek employs a simple yet effective\nmulti-channel equivariant graph neural network to generate high-quality\ngeometric representations of CDR backbone structures. Subsequently, it aligns\nsequences of structurally similar CDRs and utilizes structurally conserved\nsequence motifs to enhance inference accuracy. Our experiments demonstrate that\nIgseek not only proves to be highly efficient in structural retrieval but also\noutperforms state-of-the-art approaches in sequence recovery for both\nantibodies and T-Cell Receptors, offering a new retrieval-based perspective for\ntherapeutic protein design.\n", "date": "2025-02-11", "categories": []}
{"id": "2305.14749", "title": "gRNAde: Geometric Deep Learning for 3D RNA inverse design", "abstract": "  Computational RNA design tasks are often posed as inverse problems, where\nsequences are designed based on adopting a single desired secondary structure\nwithout considering 3D conformational diversity. We introduce gRNAde, a\ngeometric RNA design pipeline operating on 3D RNA backbones to design sequences\nthat explicitly account for structure and dynamics. gRNAde uses a multi-state\nGraph Neural Network and autoregressive decoding to generates candidate RNA\nsequences conditioned on one or more 3D backbone structures where the\nidentities of the bases are unknown. On a single-state fixed backbone re-design\nbenchmark of 14 RNA structures from the PDB identified by Das et al. (2010),\ngRNAde obtains higher native sequence recovery rates (56% on average) compared\nto Rosetta (45% on average), taking under a second to produce designs compared\nto the reported hours for Rosetta. We further demonstrate the utility of gRNAde\non a new benchmark of multi-state design for structurally flexible RNAs, as\nwell as zero-shot ranking of mutational fitness landscapes in a retrospective\nanalysis of a recent ribozyme. Experimental wet lab validation on 10 different\nstructured RNA backbones finds that gRNAde has a success rate of 50% at\ndesigning pseudoknotted RNA structures, a significant advance over 35% for\nRosetta. Open source code and tutorials are available at:\nhttps://github.com/chaitjo/geometric-rna-design\n", "date": "2023-05-24", "categories": []}
{"id": "2402.12714", "title": "An Equivariant Pretrained Transformer for Unified 3D Molecular\n  Representation Learning", "abstract": "  Pretraining on a large number of unlabeled 3D molecules has showcased\nsuperiority in various scientific applications. However, prior efforts\ntypically focus on pretraining models in a specific domain, either proteins or\nsmall molecules, missing the opportunity to leverage cross-domain knowledge. To\nmitigate this gap, we introduce Equivariant Pretrained Transformer (EPT), an\nall-atom foundation model that can be pretrained from multiple domain 3D\nmolecules. Built upon an E(3)-equivariant transformer, EPT is able to not only\nprocess atom-level information but also incorporate block-level features (e.g.\nresiduals in proteins). Additionally, we employ a block-level denoising task,\nrather than the conventional atom-level denoising, as the pretraining\nobjective. To pretrain EPT, we construct a large-scale dataset of 5.89M\nentries, comprising small molecules, proteins, protein-protein complexes, and\nprotein-molecule complexes. Experimental evaluations on downstream tasks\nincluding ligand binding affinity prediction, protein property prediction, and\nmolecular property prediction, show that EPT significantly outperforms previous\nstate-of-the-art methods in the first task and achieves competitively superior\nperformance for the remaining two tasks. Furthermore, we demonstrate the\npotential of EPT in identifying small molecule drug candidates targeting 3CL\nprotease, a critical target in the replication of SARS-CoV-2. Among 1,978\nFDA-approved drugs, EPT ranks 7 out of 8 known anti-COVID-19 drugs in the top\n200, indicating the high recall of EPT. By using Molecular Dynamics (MD)\nsimulations, EPT further discoveries 7 novel compounds whose binding affinities\nare higher than that of the top-ranked known anti-COVID-19 drug, showcasing its\npowerful capabilities in drug discovery.\n", "date": "2024-02-20", "categories": []}
{"id": "2403.20261", "title": "FABind+: Enhancing Molecular Docking through Improved Pocket Prediction\n  and Pose Generation", "abstract": "  Molecular docking is a pivotal process in drug discovery. While traditional\ntechniques rely on extensive sampling and simulation governed by physical\nprinciples, these methods are often slow and costly. The advent of deep\nlearning-based approaches has shown significant promise, offering increases in\nboth accuracy and efficiency. Building upon the foundational work of FABind, a\nmodel designed with a focus on speed and accuracy, we present FABind+, an\nenhanced iteration that largely boosts the performance of its predecessor. We\nidentify pocket prediction as a critical bottleneck in molecular docking and\npropose a novel methodology that significantly refines pocket prediction,\nthereby streamlining the docking process. Furthermore, we introduce\nmodifications to the docking module to enhance its pose generation\ncapabilities. In an effort to bridge the gap with conventional\nsampling/generative methods, we incorporate a simple yet effective sampling\ntechnique coupled with a confidence model, requiring only minor adjustments to\nthe regression framework of FABind. Experimental results and analysis reveal\nthat FABind+ remarkably outperforms the original FABind, achieves competitive\nstate-of-the-art performance, and delivers insightful modeling strategies. This\ndemonstrates FABind+ represents a substantial step forward in molecular docking\nand drug discovery. Our code is in https://github.com/QizhiPei/FABind.\n", "date": "2024-03-29", "categories": []}
{"id": "2410.13178", "title": "GeSubNet: Gene Interaction Inference for Disease Subtype Network\n  Generation", "abstract": "  Retrieving gene functional networks from knowledge databases presents a\nchallenge due to the mismatch between disease networks and subtype-specific\nvariations. Current solutions, including statistical and deep learning methods,\noften fail to effectively integrate gene interaction knowledge from databases\nor explicitly learn subtype-specific interactions. To address this mismatch, we\npropose GeSubNet, which learns a unified representation capable of predicting\ngene interactions while distinguishing between different disease subtypes.\nGraphs generated by such representations can be considered subtype-specific\nnetworks. GeSubNet is a multi-step representation learning framework with three\nmodules: First, a deep generative model learns distinct disease subtypes from\npatient gene expression profiles. Second, a graph neural network captures\nrepresentations of prior gene networks from knowledge databases, ensuring\naccurate physical gene interactions. Finally, we integrate these two\nrepresentations using an inference loss that leverages graph generation\ncapabilities, conditioned on the patient separation loss, to refine\nsubtype-specific information in the learned representation. GeSubNet\nconsistently outperforms traditional methods, with average improvements of\n30.6%, 21.0%, 20.1%, and 56.6% across four graph evaluation metrics, averaged\nover four cancer datasets. Particularly, we conduct a biological simulation\nexperiment to assess how the behavior of selected genes from over 11,000\ncandidates affects subtypes or patient distributions. The results show that the\ngenerated network has the potential to identify subtype-specific genes with an\n83% likelihood of impacting patient distribution shifts.\n", "date": "2024-10-17", "categories": []}
{"id": "2502.15867", "title": "Strategic priorities for transformative progress in advancing biology\n  with proteomics and artificial intelligence", "abstract": "  Artificial intelligence (AI) is transforming scientific research, including\nproteomics. Advances in mass spectrometry (MS)-based proteomics data quality,\ndiversity, and scale, combined with groundbreaking AI techniques, are unlocking\nnew challenges and opportunities in biological discovery. Here, we highlight\nkey areas where AI is driving innovation, from data analysis to new biological\ninsights. These include developing an AI-friendly ecosystem for proteomics data\ngeneration, sharing, and analysis; improving peptide and protein identification\nand quantification; characterizing protein-protein interactions and protein\ncomplexes; advancing spatial and perturbation proteomics; integrating\nmulti-omics data; and ultimately enabling AI-empowered virtual cells.\n", "date": "2025-02-21", "categories": []}
{"id": "2502.17246", "title": "Overconfident Oracles: Limitations of In Silico Sequence Design\n  Benchmarking", "abstract": "  Machine learning methods can automate the in silico design of biological\nsequences, aiming to reduce costs and accelerate medical research. Given the\nlimited access to wet labs, in silico design methods commonly use an oracle\nmodel to evaluate de novo generated sequences. However, the use of different\noracle models across methods makes it challenging to compare them reliably,\nmotivating the question: are in silico sequence design benchmarks reliable? In\nthis work, we examine 12 sequence design methods that utilise ML oracles common\nin the literature and find that there are significant challenges with their\ncross-consistency and reproducibility. Indeed, oracles differing by\narchitecture, or even just training seed, are shown to yield conflicting\nrelative performance with our analysis suggesting poor out-of-distribution\ngeneralisation as a key issue. To address these challenges, we propose\nsupplementing the evaluation with a suite of biophysical measures to assess the\nviability of generated sequences and limit out-of-distribution sequences the\noracle is required to score, thereby improving the robustness of the design\nprocedure. Our work aims to highlight potential pitfalls in the current\nevaluation process and contribute to the development of robust benchmarks,\nultimately driving the improvement of in silico design methods.\n", "date": "2025-02-24", "categories": []}
{"id": "2502.14919", "title": "Optimizing Gene-Based Testing for Antibiotic Resistance Prediction", "abstract": "  Antibiotic Resistance (AR) is a critical global health challenge that\nnecessitates the development of cost-effective, efficient, and accurate\ndiagnostic tools. Given the genetic basis of AR, techniques such as Polymerase\nChain Reaction (PCR) that target specific resistance genes offer a promising\napproach for predictive diagnostics using a limited set of key genes. This\nstudy introduces GenoARM, a novel framework that integrates reinforcement\nlearning (RL) with transformer-based models to optimize the selection of PCR\ngene tests and improve AR predictions, leveraging observed metadata for\nimproved accuracy. In our evaluation, we developed several high-performing\nbaselines and compared them using publicly available datasets derived from\nreal-world bacterial samples representing multiple clinically relevant\npathogens. The results show that all evaluated methods achieve strong and\nreliable performance when metadata is not utilized. When metadata is introduced\nand the number of selected genes increases, GenoARM demonstrates superior\nperformance due to its capacity to approximate rewards for unseen and sparse\ncombinations. Overall, our framework represents a major advancement in\noptimizing diagnostic tools for AR in clinical settings.\n", "date": "2025-02-19", "categories": []}
{"id": "2502.14934", "title": "Fast and Accurate Blind Flexible Docking", "abstract": "  Molecular docking that predicts the bound structures of small molecules\n(ligands) to their protein targets, plays a vital role in drug discovery.\nHowever, existing docking methods often face limitations: they either overlook\ncrucial structural changes by assuming protein rigidity or suffer from low\ncomputational efficiency due to their reliance on generative models for\nstructure sampling. To address these challenges, we propose FABFlex, a fast and\naccurate regression-based multi-task learning model designed for realistic\nblind flexible docking scenarios, where proteins exhibit flexibility and\nbinding pocket sites are unknown (blind). Specifically, FABFlex's architecture\ncomprises three specialized modules working in concert: (1) A pocket prediction\nmodule that identifies potential binding sites, addressing the challenges\ninherent in blind docking scenarios. (2) A ligand docking module that predicts\nthe bound (holo) structures of ligands from their unbound (apo) states. (3) A\npocket docking module that forecasts the holo structures of protein pockets\nfrom their apo conformations. Notably, FABFlex incorporates an iterative update\nmechanism that serves as a conduit between the ligand and pocket docking\nmodules, enabling continuous structural refinements. This approach effectively\nintegrates the three subtasks of blind flexible docking-pocket identification,\nligand conformation prediction, and protein flexibility modeling-into a\nunified, coherent framework. Extensive experiments on public benchmark datasets\ndemonstrate that FABFlex not only achieves superior effectiveness in predicting\naccurate binding modes but also exhibits a significant speed advantage (208\n$\\times$) compared to existing state-of-the-art methods. Our code is released\nat https://github.com/tmlr-group/FABFlex.\n", "date": "2025-02-20", "categories": []}
{"id": "2502.13991", "title": "Learning to Discover Regulatory Elements for Gene Expression Prediction", "abstract": "  We consider the problem of predicting gene expressions from DNA sequences. A\nkey challenge of this task is to find the regulatory elements that control gene\nexpressions. Here, we introduce Seq2Exp, a Sequence to Expression network\nexplicitly designed to discover and extract regulatory elements that drive\ntarget gene expression, enhancing the accuracy of the gene expression\nprediction. Our approach captures the causal relationship between epigenomic\nsignals, DNA sequences and their associated regulatory elements. Specifically,\nwe propose to decompose the epigenomic signals and the DNA sequence conditioned\non the causal active regulatory elements, and apply an information bottleneck\nwith the Beta distribution to combine their effects while filtering out\nnon-causal components. Our experiments demonstrate that Seq2Exp outperforms\nexisting baselines in gene expression prediction tasks and discovers\ninfluential regions compared to commonly used statistical methods for peak\ndetection such as MACS3. The source code is released as part of the AIRS\nlibrary (https://github.com/divelab/AIRS/).\n", "date": "2025-02-19", "categories": []}
{"id": "2502.12479", "title": "MotifBench: A standardized protein design benchmark for\n  motif-scaffolding problems", "abstract": "  The motif-scaffolding problem is a central task in computational protein\ndesign: Given the coordinates of atoms in a geometry chosen to confer a desired\nbiochemical function (a motif), the task is to identify diverse protein\nstructures (scaffolds) that include the motif and maintain its geometry.\nSignificant recent progress on motif-scaffolding has been made due to\ncomputational evaluation with reliable protein structure prediction and\nfixed-backbone sequence design methods. However, significant variability in\nevaluation strategies across publications has hindered comparability of\nresults, challenged reproducibility, and impeded robust progress. In response\nwe introduce MotifBench, comprising (1) a precisely specified pipeline and\nevaluation metrics, (2) a collection of 30 benchmark problems, and (3) an\nimplementation of this benchmark and leaderboard at\ngithub.com/blt2114/MotifBench. The MotifBench test cases are more difficult\ncompared to earlier benchmarks, and include protein design problems for which\nsolutions are known but on which, to the best of our knowledge,\nstate-of-the-art methods fail to identify any solution.\n", "date": "2025-02-18", "categories": []}
{"id": "2502.10807", "title": "HybriDNA: A Hybrid Transformer-Mamba2 Long-Range DNA Language Model", "abstract": "  Advances in natural language processing and large language models have\nsparked growing interest in modeling DNA, often referred to as the \"language of\nlife\". However, DNA modeling poses unique challenges. First, it requires the\nability to process ultra-long DNA sequences while preserving single-nucleotide\nresolution, as individual nucleotides play a critical role in DNA function.\nSecond, success in this domain requires excelling at both generative and\nunderstanding tasks: generative tasks hold potential for therapeutic and\nindustrial applications, while understanding tasks provide crucial insights\ninto biological mechanisms and diseases. To address these challenges, we\npropose HybriDNA, a decoder-only DNA language model that incorporates a hybrid\nTransformer-Mamba2 architecture, seamlessly integrating the strengths of\nattention mechanisms with selective state-space models. This hybrid design\nenables HybriDNA to efficiently process DNA sequences up to 131kb in length\nwith single-nucleotide resolution. HybriDNA achieves state-of-the-art\nperformance across 33 DNA understanding datasets curated from the BEND, GUE,\nand LRB benchmarks, and demonstrates exceptional capability in generating\nsynthetic cis-regulatory elements (CREs) with desired properties. Furthermore,\nwe show that HybriDNA adheres to expected scaling laws, with performance\nimproving consistently as the model scales from 300M to 3B and 7B parameters.\nThese findings underscore HybriDNA's versatility and its potential to advance\nDNA research and applications, paving the way for innovations in understanding\nand engineering the \"language of life\".\n", "date": "2025-02-15", "categories": []}
{"id": "2502.12167", "title": "TastepepAI, An artificial intelligence platform for taste peptide de\n  novo design", "abstract": "  Taste peptides have emerged as promising natural flavoring agents attributed\nto their unique organoleptic properties, high safety profile, and potential\nhealth benefits. However, the de novo identification of taste peptides derived\nfrom animal, plant, or microbial sources remains a time-consuming and\nresource-intensive process, significantly impeding their widespread application\nin the food industry. Here, we present TastePepAI, a comprehensive artificial\nintelligence framework for customized taste peptide design and safety\nassessment. As the key element of this framework, a loss-supervised adaptive\nvariational autoencoder (LA-VAE) is implemented to efficiently optimizes the\nlatent representation of sequences during training and facilitates the\ngeneration of target peptides with desired taste profiles. Notably, our model\nincorporates a novel taste-avoidance mechanism, allowing for selective flavor\nexclusion. Subsequently, our in-house developed toxicity prediction algorithm\n(SpepToxPred) is integrated in the framework to undergo rigorous safety\nevaluation of generated peptides. Using this integrated platform, we\nsuccessfully identified 73 peptides exhibiting sweet, salty, and umami,\nsignificantly expanding the current repertoire of taste peptides. This work\ndemonstrates the potential of TastePepAI in accelerating taste peptide\ndiscovery for food applications and provides a versatile framework adaptable to\nbroader peptide engineering challenges.\n", "date": "2025-02-13", "categories": []}
{"id": "2410.04815", "title": "A Review of BioTree Construction in the Context of Information Fusion:\n  Priors, Methods, Applications and Trends", "abstract": "  Biological tree (BioTree) analysis is a foundational tool in biology,\nenabling the exploration of evolutionary and differentiation relationships\namong organisms, genes, and cells. Traditional tree construction methods, while\ninstrumental in early research, face significant challenges in handling the\ngrowing complexity and scale of modern biological data, particularly in\nintegrating multimodal datasets. Advances in deep learning (DL) offer\ntransformative opportunities by enabling the fusion of biological prior\nknowledge with data-driven models. These approaches address key limitations of\ntraditional methods, facilitating the construction of more accurate and\ninterpretable BioTrees. This review highlights critical biological priors\nessential for phylogenetic and differentiation tree analyses and explores\nstrategies for integrating these priors into DL models to enhance accuracy and\ninterpretability. Additionally, the review systematically examines commonly\nused data modalities and databases, offering a valuable resource for developing\nand evaluating multimodal fusion models. Traditional tree construction methods\nare critically assessed, focusing on their biological assumptions, technical\nlimitations, and scalability issues. Recent advancements in DL-based tree\ngeneration methods are reviewed, emphasizing their innovative approaches to\nmultimodal integration and prior knowledge incorporation. Finally, the review\ndiscusses diverse applications of BioTrees in various biological disciplines,\nfrom phylogenetics to developmental biology, and outlines future trends in\nleveraging DL to advance BioTree research. By addressing the challenges of data\ncomplexity and prior knowledge integration, this review aims to inspire\ninterdisciplinary innovation at the intersection of biology and DL.\n", "date": "2024-10-07", "categories": []}
{"id": "2410.12522", "title": "MING: A Functional Approach to Learning Molecular Generative Models", "abstract": "  Traditional molecule generation methods often rely on sequence- or\ngraph-based representations, which can limit their expressive power or require\ncomplex permutation-equivariant architectures. This paper introduces a novel\nparadigm for learning molecule generative models based on functional\nrepresentations. Specifically, we propose Molecular Implicit Neural Generation\n(MING), a diffusion-based model that learns molecular distributions in the\nfunction space. Unlike standard diffusion processes in the data space, MING\nemploys a novel functional denoising probabilistic process, which jointly\ndenoises information in both the function's input and output spaces by\nleveraging an expectation-maximization procedure for latent implicit neural\nrepresentations of data. This approach enables a simple yet effective model\ndesign that accurately captures underlying function distributions. Experimental\nresults on molecule-related datasets demonstrate MING's superior performance\nand ability to generate plausible molecular samples, surpassing\nstate-of-the-art data-space methods while offering a more streamlined\narchitecture and significantly faster generation times. The code is available\nat https://github.com/v18nguye/MING.\n", "date": "2024-10-16", "categories": []}
{"id": "2502.10365", "title": "AffinityFlow: Guided Flows for Antibody Affinity Maturation", "abstract": "  Antibodies are widely used as therapeutics, but their development requires\ncostly affinity maturation, involving iterative mutations to enhance binding\naffinity.This paper explores a sequence-only scenario for affinity maturation,\nusing solely antibody and antigen sequences. Recently AlphaFlow wraps AlphaFold\nwithin flow matching to generate diverse protein structures, enabling a\nsequence-conditioned generative model of structure. Building on this, we\npropose an alternating optimization framework that (1) fixes the sequence to\nguide structure generation toward high binding affinity using a structure-based\naffinity predictor, then (2) applies inverse folding to create sequence\nmutations, refined by a sequence-based affinity predictor for post selection. A\nkey challenge is the lack of labeled data for training both predictors. To\naddress this, we develop a co-teaching module that incorporates valuable\ninformation from noisy biophysical energies into predictor refinement. The\nsequence-based predictor selects consensus samples to teach the structure-based\npredictor, and vice versa. Our method, AffinityFlow, achieves state-of-the-art\nperformance in affinity maturation experiments. We plan to open-source our code\nafter acceptance.\n", "date": "2025-02-14", "categories": []}
{"id": "2502.12049", "title": "Classifying the Stoichiometry of Virus-like Particles with Interpretable\n  Machine Learning", "abstract": "  Virus-like particles (VLPs) are valuable for vaccine development due to their\nimmune-triggering properties. Understanding their stoichiometry, the number of\nprotein subunits to form a VLP, is critical for vaccine optimisation. However,\ncurrent experimental methods to determine stoichiometry are time-consuming and\nrequire highly purified proteins. To efficiently classify stoichiometry classes\nin proteins, we curate a new dataset and propose an interpretable, data-driven\npipeline leveraging linear machine learning models. We also explore the impact\nof feature encoding on model performance and interpretability, as well as\nmethods to identify key protein sequence features influencing classification.\nThe evaluation of our pipeline demonstrates that it can classify stoichiometry\nwhile revealing protein features that possibly influence VLP assembly. The data\nand code used in this work are publicly available at\nhttps://github.com/Shef-AIRE/StoicIML.\n", "date": "2025-02-17", "categories": []}
{"id": "2502.10173", "title": "Agentic End-to-End De Novo Protein Design for Tailored Dynamics Using a\n  Language Diffusion Model", "abstract": "  Proteins are dynamic molecular machines whose biological functions, spanning\nenzymatic catalysis, signal transduction, and structural adaptation, are\nintrinsically linked to their motions. Designing proteins with targeted dynamic\nproperties, however, remains a challenge due to the complex, degenerate\nrelationships between sequence, structure, and molecular motion. Here, we\nintroduce VibeGen, a generative AI framework that enables end-to-end de novo\nprotein design conditioned on normal mode vibrations. VibeGen employs an\nagentic dual-model architecture, comprising a protein designer that generates\nsequence candidates based on specified vibrational modes and a protein\npredictor that evaluates their dynamic accuracy. This approach synergizes\ndiversity, accuracy, and novelty during the design process. Via full-atom\nmolecular simulations as direct validation, we demonstrate that the designed\nproteins accurately reproduce the prescribed normal mode amplitudes across the\nbackbone while adopting various stable, functionally relevant structures.\nNotably, generated sequences are de novo, exhibiting no significant similarity\nto natural proteins, thereby expanding the accessible protein space beyond\nevolutionary constraints. Our work integrates protein dynamics into generative\nprotein design, and establishes a direct, bidirectional link between sequence\nand vibrational behavior, unlocking new pathways for engineering biomolecules\nwith tailored dynamical and functional properties. This framework holds broad\nimplications for the rational design of flexible enzymes, dynamic scaffolds,\nand biomaterials, paving the way toward dynamics-informed AI-driven protein\nengineering.\n", "date": "2025-02-14", "categories": []}
{"id": "2502.06913", "title": "A Simple yet Effective DDG Predictor is An Unsupervised Antibody\n  Optimizer and Explainer", "abstract": "  The proteins that exist today have been optimized over billions of years of\nnatural evolution, during which nature creates random mutations and selects\nthem. The discovery of functionally promising mutations is challenged by the\nlimited evolutionary accessible regions, i.e., only a small region on the\nfitness landscape is beneficial. There have been numerous priors used to\nconstrain protein evolution to regions of landscapes with high-fitness\nvariants, among which the change in binding free energy (DDG) of protein\ncomplexes upon mutations is one of the most commonly used priors. However, the\nhuge mutation space poses two challenges: (1) how to improve the efficiency of\nDDG prediction for fast mutation screening; and (2) how to explain mutation\npreferences and efficiently explore accessible evolutionary regions. To address\nthese challenges, we propose a lightweight DDG predictor (Light-DDG), which\nadopts a structure-aware Transformer as the backbone and enhances it by\nknowledge distilled from existing powerful but computationally heavy DDG\npredictors. Additionally, we augmented, annotated, and released a large-scale\ndataset containing millions of mutation data for pre-training Light-DDG. We\nfind that such a simple yet effective Light-DDG can serve as a good\nunsupervised antibody optimizer and explainer. For the target antibody, we\npropose a novel Mutation Explainer to learn mutation preferences, which\naccounts for the marginal benefit of each mutation per residue. To further\nexplore accessible evolutionary regions, we conduct preference-guided antibody\noptimization and evaluate antibody candidates quickly using Light-DDG to\nidentify desirable mutations.\n", "date": "2025-02-10", "categories": []}
{"id": "2502.09135", "title": "Interpreting and Steering Protein Language Models through Sparse\n  Autoencoders", "abstract": "  The rapid advancements in transformer-based language models have\nrevolutionized natural language processing, yet understanding the internal\nmechanisms of these models remains a significant challenge. This paper explores\nthe application of sparse autoencoders (SAE) to interpret the internal\nrepresentations of protein language models, specifically focusing on the ESM-2\n8M parameter model. By performing a statistical analysis on each latent\ncomponent's relevance to distinct protein annotations, we identify potential\ninterpretations linked to various protein characteristics, including\ntransmembrane regions, binding sites, and specialized motifs.\n  We then leverage these insights to guide sequence generation, shortlisting\nthe relevant latent components that can steer the model towards desired targets\nsuch as zinc finger domains. This work contributes to the emerging field of\nmechanistic interpretability in biological sequence models, offering new\nperspectives on model steering for sequence design.\n", "date": "2025-02-13", "categories": []}
{"id": "2502.09335", "title": "Graph Diffusion Network for Drug-Gene Prediction", "abstract": "  Predicting drug-gene associations is crucial for drug development and disease\ntreatment. While graph neural networks (GNN) have shown effectiveness in this\ntask, they face challenges with data sparsity and efficient contrastive\nlearning implementation. We introduce a graph diffusion network for drug-gene\nprediction (GDNDGP), a framework that addresses these limitations through two\nkey innovations. First, it employs meta-path-based homogeneous graph learning\nto capture drug-drug and gene-gene relationships, ensuring similar entities\nshare embedding spaces. Second, it incorporates a parallel diffusion network\nthat generates hard negative samples during training, eliminating the need for\nexhaustive negative sample retrieval. Our model achieves superior performance\non the DGIdb 4.0 dataset and demonstrates strong generalization capability on\ntripartite drug-gene-disease networks. Results show significant improvements\nover existing methods in drug-gene prediction tasks, particularly in handling\ncomplex heterogeneous relationships. The source code is publicly available at\nhttps://github.com/csjywu1/GDNDGP.\n", "date": "2025-02-13", "categories": []}
{"id": "2210.00116", "title": "Predicting Cellular Responses with Variational Causal Inference and\n  Refined Relational Information", "abstract": "  Predicting the responses of a cell under perturbations may bring important\nbenefits to drug discovery and personalized therapeutics. In this work, we\npropose a novel graph variational Bayesian causal inference framework to\npredict a cell's gene expressions under counterfactual perturbations\n(perturbations that this cell did not factually receive), leveraging\ninformation representing biological knowledge in the form of gene regulatory\nnetworks (GRNs) to aid individualized cellular response predictions. Aiming at\na data-adaptive GRN, we also developed an adjacency matrix updating technique\nfor graph convolutional networks and used it to refine GRNs during\npre-training, which generated more insights on gene relations and enhanced\nmodel performance. Additionally, we propose a robust estimator within our\nframework for the asymptotically efficient estimation of marginal perturbation\neffect, which is yet to be carried out in previous works. With extensive\nexperiments, we exhibited the advantage of our approach over state-of-the-art\ndeep learning models for individual response prediction.\n", "date": "2022-09-30", "categories": []}
{"id": "2407.18811", "title": "Interpreting artificial neural networks to detect genome-wide\n  association signals for complex traits", "abstract": "  Investigating the genetic architecture of complex diseases is challenging due\nto the multifactorial and interactive landscape of genomic and environmental\ninfluences. Although genome-wide association studies (GWAS) have identified\nthousands of variants for multiple complex traits, conventional statistical\napproaches can be limited by simplified assumptions such as linearity and lack\nof epistasis in models. In this work, we trained artificial neural networks to\npredict complex traits using both simulated and real genotype-phenotype\ndatasets. We extracted feature importance scores via different post hoc\ninterpretability methods to identify potentially associated loci (PAL) for the\ntarget phenotype and devised an approach for obtaining p-values for the\ndetected PAL. Simulations with various parameters demonstrated that associated\nloci can be detected with good precision using strict selection criteria. By\napplying our approach to the schizophrenia cohort in the Estonian Biobank, we\ndetected multiple loci associated with this highly polygenic and heritable\ndisorder. There was significant concordance between PAL and loci previously\nassociated with schizophrenia and bipolar disorder, with enrichment analyses of\ngenes within the identified PAL predominantly highlighting terms related to\nbrain morphology and function. With advancements in model optimization and\nuncertainty quantification, artificial neural networks have the potential to\nenhance the identification of genomic loci associated with complex diseases,\noffering a more comprehensive approach for GWAS and serving as initial\nscreening tools for subsequent functional studies.\n", "date": "2024-07-26", "categories": []}
{"id": "2502.07749", "title": "Whole-Genome Phenotype Prediction with Machine Learning: Open Problems\n  in Bacterial Genomics", "abstract": "  How can we identify causal genetic mechanisms that govern bacterial traits?\nInitial efforts entrusting machine learning models to handle the task of\npredicting phenotype from genotype return high accuracy scores. However,\nattempts to extract any meaning from the predictive models are found to be\ncorrupted by falsely identified \"causal\" features. Relying solely on pattern\nrecognition and correlations is unreliable, significantly so in bacterial\ngenomics settings where high-dimensionality and spurious associations are the\nnorm. Though it is not yet clear whether we can overcome this hurdle,\nsignificant efforts are being made towards discovering potential high-risk\nbacterial genetic variants. In view of this, we set up open problems\nsurrounding phenotype prediction from bacterial whole-genome datasets and\nextending those to learning causal effects, and discuss challenges that impact\nthe reliability of a machine's decision-making when faced with datasets of this\nnature.\n", "date": "2025-02-11", "categories": []}
{"id": "2502.06126", "title": "Graph Pseudotime Analysis and Neural Stochastic Differential Equations\n  for Analyzing Retinal Degeneration Dynamics and Beyond", "abstract": "  Understanding disease progression at the molecular pathway level usually\nrequires capturing both structural dependencies between pathways and the\ntemporal dynamics of disease evolution. In this work, we solve the former\nchallenge by developing a biologically informed graph-forming method to\nefficiently construct pathway graphs for subjects from our newly curated JR5558\nmouse transcriptomics dataset. We then develop Graph-level Pseudotime Analysis\n(GPA) to infer graph-level trajectories that reveal how disease progresses at\nthe population level, rather than in individual subjects. Based on the\ntrajectories estimated by GPA, we identify the most sensitive pathways that\ndrive disease stage transitions. In addition, we measure changes in pathway\nfeatures using neural stochastic differential equations (SDEs), which enables\nus to formally define and compute pathway stability and disease bifurcation\npoints (points of no return), two fundamental problems in disease progression\nresearch. We further extend our theory to the case when pathways can interact\nwith each other, enabling a more comprehensive and multi-faceted\ncharacterization of disease phenotypes. The comprehensive experimental results\ndemonstrate the effectiveness of our framework in reconstructing the dynamics\nof the pathway, identifying critical transitions, and providing novel insights\ninto the mechanistic understanding of disease evolution.\n", "date": "2025-02-10", "categories": []}
{"id": "2502.04730", "title": "PhyloVAE: Unsupervised Learning of Phylogenetic Trees via Variational\n  Autoencoders", "abstract": "  Learning informative representations of phylogenetic tree structures is\nessential for analyzing evolutionary relationships. Classical distance-based\nmethods have been widely used to project phylogenetic trees into Euclidean\nspace, but they are often sensitive to the choice of distance metric and may\nlack sufficient resolution. In this paper, we introduce phylogenetic\nvariational autoencoders (PhyloVAEs), an unsupervised learning framework\ndesigned for representation learning and generative modeling of tree\ntopologies. Leveraging an efficient encoding mechanism inspired by\nautoregressive tree topology generation, we develop a deep latent-variable\ngenerative model that facilitates fast, parallelized topology generation.\nPhyloVAE combines this generative model with a collaborative inference model\nbased on learnable topological features, allowing for high-resolution\nrepresentations of phylogenetic tree samples. Extensive experiments demonstrate\nPhyloVAE's robust representation learning capabilities and fast generation of\nphylogenetic tree topologies.\n", "date": "2025-02-07", "categories": []}
{"id": "2412.16483", "title": "MOL-Mamba: Enhancing Molecular Representation with Structural &\n  Electronic Insights", "abstract": "  Molecular representation learning plays a crucial role in various downstream\ntasks, such as molecular property prediction and drug design. To accurately\nrepresent molecules, Graph Neural Networks (GNNs) and Graph Transformers (GTs)\nhave shown potential in the realm of self-supervised pretraining. However,\nexisting approaches often overlook the relationship between molecular structure\nand electronic information, as well as the internal semantic reasoning within\nmolecules. This omission of fundamental chemical knowledge in graph semantics\nleads to incomplete molecular representations, missing the integration of\nstructural and electronic data. To address these issues, we introduce\nMOL-Mamba, a framework that enhances molecular representation by combining\nstructural and electronic insights. MOL-Mamba consists of an Atom & Fragment\nMamba-Graph (MG) for hierarchical structural reasoning and a Mamba-Transformer\n(MT) fuser for integrating molecular structure and electronic correlation\nlearning. Additionally, we propose a Structural Distribution Collaborative\nTraining and E-semantic Fusion Training framework to further enhance molecular\nrepresentation learning. Extensive experiments demonstrate that MOL-Mamba\noutperforms state-of-the-art baselines across eleven chemical-biological\nmolecular datasets.\n", "date": "2024-12-21", "categories": []}
{"id": "2502.03499", "title": "Omni-DNA: A Unified Genomic Foundation Model for Cross-Modal and\n  Multi-Task Learning", "abstract": "  Large Language Models (LLMs) demonstrate remarkable generalizability across\ndiverse tasks, yet genomic foundation models (GFMs) still require separate\nfinetuning for each downstream application, creating significant overhead as\nmodel sizes grow. Moreover, existing GFMs are constrained by rigid output\nformats, limiting their applicability to various genomic tasks. In this work,\nwe revisit the transformer-based auto-regressive models and introduce Omni-DNA,\na family of cross-modal multi-task models ranging from 20 million to 1 billion\nparameters. Our approach consists of two stages: (i) pretraining on DNA\nsequences with next token prediction objective, and (ii) expanding the\nmulti-modal task-specific tokens and finetuning for multiple downstream tasks\nsimultaneously. When evaluated on the Nucleotide Transformer and GB benchmarks,\nOmni-DNA achieves state-of-the-art performance on 18 out of 26 tasks. Through\nmulti-task finetuning, Omni-DNA addresses 10 acetylation and methylation tasks\nat once, surpassing models trained on each task individually. Finally, we\ndesign two complex genomic tasks, DNA2Function and Needle-in-DNA, which map DNA\nsequences to textual functional descriptions and images, respectively,\nindicating Omni-DNA's cross-modal capabilities to broaden the scope of genomic\napplications. All the models are available through\nhttps://huggingface.co/collections/zehui127\n", "date": "2025-02-05", "categories": []}
{"id": "2502.03938", "title": "Unravelling Causal Genetic Biomarkers of Alzheimer's Disease via Neuron\n  to Gene-token Backtracking in Neural Architecture: A Groundbreaking\n  Reverse-Gene-Finder Approach", "abstract": "  Alzheimer's Disease (AD) affects over 55 million people globally, yet the key\ngenetic contributors remain poorly understood. Leveraging recent advancements\nin genomic foundation models, we present the innovative Reverse-Gene-Finder\ntechnology, a ground-breaking neuron-to-gene-token backtracking approach in a\nneural network architecture to elucidate the novel causal genetic biomarkers\ndriving AD onset. Reverse-Gene-Finder comprises three key innovations. Firstly,\nwe exploit the observation that genes with the highest probability of causing\nAD, defined as the most causal genes (MCGs), must have the highest probability\nof activating those neurons with the highest probability of causing AD, defined\nas the most causal neurons (MCNs). Secondly, we utilize a gene token\nrepresentation at the input layer to allow each gene (known or novel to AD) to\nbe represented as a discrete and unique entity in the input space. Lastly, in\ncontrast to the existing neural network architectures, which track neuron\nactivations from the input layer to the output layer in a feed-forward manner,\nwe develop an innovative backtracking method to track backwards from the MCNs\nto the input layer, identifying the Most Causal Tokens (MCTs) and the\ncorresponding MCGs. Reverse-Gene-Finder is highly interpretable, generalizable,\nand adaptable, providing a promising avenue for application in other disease\nscenarios.\n", "date": "2025-02-06", "categories": []}
{"id": "2401.02564", "title": "Predicting Future States with Spatial Point Processes in Single Molecule\n  Resolution Spatial Transcriptomics", "abstract": "  In this paper, we introduce a pipeline based on XGboost to predict the future\ndistribution of cells that are expressed by the Sog-D gene (active cells) in\nboth the Anterior to posterior (AP) and the Dorsal to Ventral (DV) axis of the\nDrosophila in embryogenesis process. This method provides insights about how\ncells and living organisms control gene expression in super resolution whole\nembryo spatial transcriptomics imaging at sub cellular, single molecule\nresolution. An XGboost model was used to predict the next stage active\ndistribution based on the previous one. To achieve this goal, we leveraged\ntemporally resolved, spatial point processes by including Ripley's K-function\nin conjunction with the cell's state in each stage of embryogenesis, and found\naverage predictive accuracy of active cell distribution. This tool is analogous\nto RNA Velocity for spatially resolved developmental biology, from one data\npoint we can predict future spatially resolved gene expression using features\nfrom the spatial point processes.\n", "date": "2024-01-04", "categories": []}
{"id": "2502.02629", "title": "Graph Structure Learning for Tumor Microenvironment with Cell Type\n  Annotation from non-spatial scRNA-seq data", "abstract": "  The exploration of cellular heterogeneity within the tumor microenvironment\n(TME) via single-cell RNA sequencing (scRNA-seq) is essential for understanding\ncancer progression and response to therapy. Current scRNA-seq approaches,\nhowever, lack spatial context and rely on incomplete datasets of\nligand-receptor interactions (LRIs), limiting accurate cell type annotation and\ncell-cell communication (CCC) inference. This study addresses these challenges\nusing a novel graph neural network (GNN) model that enhances cell type\nprediction and cell interaction analysis. Our study utilized a dataset\nconsisting of 49,020 cells from 19 patients across three cancer types:\nLeukemia, Breast Invasive Carcinoma, and Colorectal Cancer. The proposed scGSL\nmodel demonstrated robust performance, achieving an average accuracy of 84.83%,\nprecision of 86.23%, recall of 81.51%, and an F1 score of 80.92% across all\ndatasets. These metrics represent a significant enhancement over existing\nmethods, which typically exhibit lower performance metrics. Additionally, by\nreviewing existing literature on gene interactions within the TME, the scGSL\nmodel proves to robustly identify biologically meaningful gene interactions in\nan unsupervised manner, validated by significant expression differences in key\ngene pairs across various cancers. The source code and data used in this paper\ncan be found in https://github.com/LiYuechao1998/scGSL.\n", "date": "2025-02-04", "categories": []}
{"id": "2502.01689", "title": "scGSDR: Harnessing Gene Semantics for Single-Cell Pharmacological\n  Profiling", "abstract": "  The rise of single-cell sequencing technologies has revolutionized the\nexploration of drug resistance, revealing the crucial role of cellular\nheterogeneity in advancing precision medicine. By building computational models\nfrom existing single-cell drug response data, we can rapidly annotate cellular\nresponses to drugs in subsequent trials. To this end, we developed scGSDR, a\nmodel that integrates two computational pipelines grounded in the knowledge of\ncellular states and gene signaling pathways, both essential for understanding\nbiological gene semantics. scGSDR enhances predictive performance by\nincorporating gene semantics and employs an interpretability module to identify\nkey pathways contributing to drug resistance phenotypes. Our extensive\nvalidation, which included 16 experiments covering 11 drugs, demonstrates\nscGSDR's superior predictive accuracy, when trained with either bulk-seq or\nscRNA-seq data, achieving high AUROC, AUPR, and F1 Scores. The model's\napplication has extended from single-drug predictions to scenarios involving\ndrug combinations. Leveraging pathways of known drug target genes, we found\nthat scGSDR's cell-pathway attention scores are biologically interpretable,\nwhich helped us identify other potential drug-related genes. Literature review\nof top-ranking genes in our predictions such as BCL2, CCND1, the AKT family,\nand PIK3CA for PLX4720; and ICAM1, VCAM1, NFKB1, NFKBIA, and RAC1 for\nPaclitaxel confirmed their relevance. In conclusion, scGSDR, by incorporating\ngene semantics, enhances predictive modeling of cellular responses to diverse\ndrugs, proving invaluable for scenarios involving both single drug and\ncombination therapies and effectively identifying key resistance-related\npathways, thus advancing precision medicine and targeted therapy development.\n", "date": "2025-02-02", "categories": []}
{"id": "2502.02129", "title": "Deep Neural Cellular Potts Models", "abstract": "  The cellular Potts model (CPM) is a powerful computational method for\nsimulating collective spatiotemporal dynamics of biological cells. To drive the\ndynamics, CPMs rely on physics-inspired Hamiltonians. However, as first\nprinciples remain elusive in biology, these Hamiltonians only approximate the\nfull complexity of real multicellular systems. To address this limitation, we\npropose NeuralCPM, a more expressive cellular Potts model that can be trained\ndirectly on observational data. At the core of NeuralCPM lies the Neural\nHamiltonian, a neural network architecture that respects universal symmetries\nin collective cellular dynamics. Moreover, this approach enables seamless\nintegration of domain knowledge by combining known biological mechanisms and\nthe expressive Neural Hamiltonian into a hybrid model. Our evaluation with\nsynthetic and real-world multicellular systems demonstrates that NeuralCPM is\nable to model cellular dynamics that cannot be accounted for by traditional\nanalytical Hamiltonians.\n", "date": "2025-02-04", "categories": []}
{"id": "2401.04155", "title": "Advancing bioinformatics with large language models: components,\n  applications and perspectives", "abstract": "  Large language models (LLMs) are a class of artificial intelligence models\nbased on deep learning, which have great performance in various tasks,\nespecially in natural language processing (NLP). Large language models\ntypically consist of artificial neural networks with numerous parameters,\ntrained on large amounts of unlabeled input using self-supervised or\nsemi-supervised learning. However, their potential for solving bioinformatics\nproblems may even exceed their proficiency in modeling human language. In this\nreview, we will provide a comprehensive overview of the essential components of\nlarge language models (LLMs) in bioinformatics, spanning genomics,\ntranscriptomics, proteomics, drug discovery, and single-cell analysis. Key\naspects covered include tokenization methods for diverse data types, the\narchitecture of transformer models, the core attention mechanism, and the\npre-training processes underlying these models. Additionally, we will introduce\ncurrently available foundation models and highlight their downstream\napplications across various bioinformatics domains. Finally, drawing from our\nexperience, we will offer practical guidance for both LLM users and developers,\nemphasizing strategies to optimize their use and foster further innovation in\nthe field.\n", "date": "2024-01-08", "categories": []}
{"id": "2410.20660", "title": "TurboHopp: Accelerated Molecule Scaffold Hopping with Consistency Models", "abstract": "  Navigating the vast chemical space of druggable compounds is a formidable\nchallenge in drug discovery, where generative models are increasingly employed\nto identify viable candidates. Conditional 3D structure-based drug design\n(3D-SBDD) models, which take into account complex three-dimensional\ninteractions and molecular geometries, are particularly promising. Scaffold\nhopping is an efficient strategy that facilitates the identification of similar\nactive compounds by strategically modifying the core structure of molecules,\neffectively narrowing the wide chemical space and enhancing the discovery of\ndrug-like products. However, the practical application of 3D-SBDD generative\nmodels is hampered by their slow processing speeds. To address this bottleneck,\nwe introduce TurboHopp, an accelerated pocket-conditioned 3D scaffold hopping\nmodel that merges the strategic effectiveness of traditional scaffold hopping\nwith rapid generation capabilities of consistency models. This synergy not only\nenhances efficiency but also significantly boosts generation speeds, achieving\nup to 30 times faster inference speed as well as superior generation quality\ncompared to existing diffusion-based models, establishing TurboHopp as a\npowerful tool in drug discovery. Supported by faster inference speed, we\nfurther optimize our model, using Reinforcement Learning for Consistency Models\n(RLCM), to output desirable molecules. We demonstrate the broad applicability\nof TurboHopp across multiple drug discovery scenarios, underscoring its\npotential in diverse molecular settings.\n", "date": "2024-10-28", "categories": []}
{"id": "2502.01012", "title": "Deep Active Learning based Experimental Design to Uncover Synergistic\n  Genetic Interactions for Host Targeted Therapeutics", "abstract": "  Recent technological advances have introduced new high-throughput methods for\nstudying host-virus interactions, but testing synergistic interactions between\nhost gene pairs during infection remains relatively slow and labor intensive.\nIdentification of multiple gene knockdowns that effectively inhibit viral\nreplication requires a search over the combinatorial space of all possible\ntarget gene pairs and is infeasible via brute-force experiments. Although\nactive learning methods for sequential experimental design have shown promise,\nexisting approaches have generally been restricted to single-gene knockdowns or\nsmall-scale double knockdown datasets. In this study, we present an integrated\nDeep Active Learning (DeepAL) framework that incorporates information from a\nbiological knowledge graph (SPOKE, the Scalable Precision Medicine Open\nKnowledge Engine) to efficiently search the configuration space of a large\ndataset of all pairwise knockdowns of 356 human genes in HIV infection. Through\ngraph representation learning, the framework is able to generate task-specific\nrepresentations of genes while also balancing the exploration-exploitation\ntrade-off to pinpoint highly effective double-knockdown pairs. We additionally\npresent an ensemble method for uncertainty quantification and an interpretation\nof the gene pairs selected by our algorithm via pathway analysis. To our\nknowledge, this is the first work to show promising results on double-gene\nknockdown experimental data of appreciable scale (356 by 356 matrix).\n", "date": "2025-02-03", "categories": []}
{"id": "2502.01311", "title": "TFBS-Finder: Deep Learning-based Model with DNABERT and Convolutional\n  Networks to Predict Transcription Factor Binding Sites", "abstract": "  Transcription factors are proteins that regulate the expression of genes by\nbinding to specific genomic regions known as Transcription Factor Binding Sites\n(TFBSs), typically located in the promoter regions of those genes. Accurate\nprediction of these binding sites is essential for understanding the complex\ngene regulatory networks underlying various cellular functions. In this regard,\nmany deep learning models have been developed for such prediction, but there is\nstill scope of improvement. In this work, we have developed a deep learning\nmodel which uses pre-trained DNABERT, a Convolutional Neural Network (CNN)\nmodule, a Modified Convolutional Block Attention Module (MCBAM), a Multi-Scale\nConvolutions with Attention (MSCA) module and an output module. The pre-trained\nDNABERT is used for sequence embedding, thereby capturing the long-term\ndependencies in the DNA sequences while the CNN, MCBAM and MSCA modules are\nuseful in extracting higher-order local features. TFBS-Finder is trained and\ntested on 165 ENCODE ChIP-seq datasets. We have also performed ablation studies\nas well as cross-cell line validations and comparisons with other models. The\nexperimental results show the superiority of the proposed method in predicting\nTFBSs compared to the existing methodologies. The codes and the relevant\ndatasets are publicly available at\nhttps://github.com/NimishaGhosh/TFBS-Finder/.\n", "date": "2025-02-03", "categories": []}
{"id": "2502.01461", "title": "Docking-Aware Attention: Dynamic Protein Representations through\n  Molecular Context Integration", "abstract": "  Computational prediction of enzymatic reactions represents a crucial\nchallenge in sustainable chemical synthesis across various scientific domains,\nranging from drug discovery to materials science and green chemistry. These\nsyntheses rely on proteins that selectively catalyze complex molecular\ntransformations. These protein catalysts exhibit remarkable substrate\nadaptability, with the same protein often catalyzing different chemical\ntransformations depending on its molecular partners. Current approaches to\nprotein representation in reaction prediction either ignore protein structure\nentirely or rely on static embeddings, failing to capture how proteins\ndynamically adapt their behavior to different substrates. We present\nDocking-Aware Attention (DAA), a novel architecture that generates dynamic,\ncontext-dependent protein representations by incorporating molecular docking\ninformation into the attention mechanism. DAA combines physical interaction\nscores from docking predictions with learned attention patterns to focus on\nprotein regions most relevant to specific molecular interactions. We evaluate\nour method on enzymatic reaction prediction, where it outperforms previous\nstate-of-the-art methods, achieving 62.2\\% accuracy versus 56.79\\% on complex\nmolecules and 55.54\\% versus 49.45\\% on innovative reactions. Through detailed\nablation studies and visualizations, we demonstrate how DAA generates\ninterpretable attention patterns that adapt to different molecular contexts.\nOur approach represents a general framework for context-aware protein\nrepresentation in biocatalysis prediction, with potential applications across\nenzymatic synthesis planning. We open-source our implementation and pre-trained\nmodels to facilitate further research.\n", "date": "2025-02-03", "categories": []}
{"id": "2501.18650", "title": "Constructing Cell-type Taxonomy by Optimal Transport with Relaxed\n  Marginal Constraints", "abstract": "  The rapid emergence of single-cell data has facilitated the study of many\ndifferent biological conditions at the cellular level. Cluster analysis has\nbeen widely applied to identify cell types, capturing the essential patterns of\nthe original data in a much more concise form. One challenge in the cluster\nanalysis of cells is matching clusters extracted from datasets of different\norigins or conditions. Many existing algorithms cannot recognize new cell types\npresent in only one of the two samples when establishing a correspondence\nbetween clusters obtained from two samples. Additionally, when there are more\nthan two samples, it is advantageous to align clusters across all samples\nsimultaneously rather than performing pairwise alignment. Our approach aims to\nconstruct a taxonomy for cell clusters across all samples to better annotate\nthese clusters and effectively extract features for downstream analysis. A new\nsystem for constructing cell-type taxonomy has been developed by combining the\ntechnique of Optimal Transport with Relaxed Marginal Constraints (OT-RMC) and\nthe simultaneous alignment of clusters across multiple samples. OT-RMC allows\nus to address challenges that arise when the proportions of clusters vary\nsubstantially between samples or when some clusters do not appear in all the\nsamples. Experiments on more than twenty datasets demonstrate that the taxonomy\nconstructed by this new system can yield highly accurate annotation of cell\ntypes. Additionally, sample-level features extracted based on the taxonomy\nresult in accurate classification of samples.\n", "date": "2025-01-29", "categories": []}
{"id": "2501.18794", "title": "Survey and Improvement Strategies for Gene Prioritization with Large\n  Language Models", "abstract": "  Rare diseases are challenging to diagnose due to limited patient data and\ngenetic diversity. Despite advances in variant prioritization, many cases\nremain undiagnosed. While large language models (LLMs) have performed well in\nmedical exams, their effectiveness in diagnosing rare genetic diseases has not\nbeen assessed. To identify causal genes, we benchmarked various LLMs for gene\nprioritization. Using multi-agent and Human Phenotype Ontology (HPO)\nclassification, we categorized patients based on phenotypes and solvability\nlevels. As gene set size increased, LLM performance deteriorated, so we used a\ndivide-and-conquer strategy to break the task into smaller subsets. At\nbaseline, GPT-4 outperformed other LLMs, achieving near 30% accuracy in ranking\ncausal genes correctly. The multi-agent and HPO approaches helped distinguish\nconfidently solved cases from challenging ones, highlighting the importance of\nknown gene-phenotype associations and phenotype specificity. We found that\ncases with specific phenotypes or clear associations were more accurately\nsolved. However, we observed biases toward well-studied genes and input order\nsensitivity, which hindered gene prioritization. Our divide-and-conquer\nstrategy improved accuracy by overcoming these biases. By utilizing HPO\nclassification, novel multi-agent techniques, and our LLM strategy, we improved\ncausal gene identification accuracy compared to our baseline evaluation. This\napproach streamlines rare disease diagnosis, facilitates reanalysis of unsolved\ncases, and accelerates gene discovery, supporting the development of targeted\ndiagnostics and therapies.\n", "date": "2025-01-30", "categories": []}
{"id": "2501.18973", "title": "GPO-VAE: Modeling Explainable Gene Perturbation Responses utilizing\n  GRN-Aligned Parameter Optimization", "abstract": "  Motivation: Predicting cellular responses to genetic perturbations is\nessential for understanding biological systems and developing targeted\ntherapeutic strategies. While variational autoencoders (VAEs) have shown\npromise in modeling perturbation responses, their limited explainability poses\na significant challenge, as the learned features often lack clear biological\nmeaning. Nevertheless, model explainability is one of the most important\naspects in the realm of biological AI. One of the most effective ways to\nachieve explainability is incorporating the concept of gene regulatory networks\n(GRNs) in designing deep learning models such as VAEs. GRNs elicit the\nunderlying causal relationships between genes and are capable of explaining the\ntranscriptional responses caused by genetic perturbation treatments. Results:\nWe propose GPO-VAE, an explainable VAE enhanced by GRN-aligned Parameter\nOptimization that explicitly models gene regulatory networks in the latent\nspace. Our key approach is to optimize the learnable parameters related to\nlatent perturbation effects towards GRN-aligned explainability. Experimental\nresults on perturbation prediction show our model achieves state-of-the-art\nperformance in predicting transcriptional responses across multiple benchmark\ndatasets. Furthermore, additional results on evaluating the GRN inference task\nreveal our model's ability to generate meaningful GRNs compared to other\nmethods. According to qualitative analysis, GPO-VAE posseses the ability to\nconstruct biologically explainable GRNs that align with experimentally\nvalidated regulatory pathways. GPO-VAE is available at\nhttps://github.com/dmis-lab/GPO-VAE\n", "date": "2025-01-31", "categories": []}
{"id": "2501.19178", "title": "No Foundations without Foundations -- Why semi-mechanistic models are\n  essential for regulatory biology", "abstract": "  Despite substantial efforts, deep learning has not yet delivered a\ntransformative impact on elucidating regulatory biology, particularly in the\nrealm of predicting gene expression profiles. Here, we argue that genuine\n\"foundation models\" of regulatory biology will remain out of reach unless\nguided by frameworks that integrate mechanistic insight with principled\nexperimental design. We present one such ground-up, semi-mechanistic framework\nthat unifies perturbation-based experimental designs across both in vitro and\nin vivo CRISPR screens, accounting for differentiating and non-differentiating\ncellular systems. By revealing previously unrecognised assumptions in published\nmachine learning methods, our approach clarifies links with popular techniques\nsuch as variational autoencoders and structural causal models. In practice,\nthis framework suggests a modified loss function that we demonstrate can\nimprove predictive performance, and further suggests an error analysis that\ninforms batching strategies. Ultimately, since cellular regulation emerges from\ninnumerable interactions amongst largely uncharted molecular components, we\ncontend that systems-level understanding cannot be achieved through structural\nbiology alone. Instead, we argue that real progress will require a\nfirst-principles perspective on how experiments capture biological phenomena,\nhow data are generated, and how these processes can be reflected in more\nfaithful modelling architectures.\n", "date": "2025-01-31", "categories": []}
{"id": "2501.07014", "title": "AlgoRxplorers | Precision in Mutation: Enhancing Drug Design with\n  Advanced Protein Stability Prediction Tools", "abstract": "  Predicting the impact of single-point amino acid mutations on protein\nstability is essential for understanding disease mechanisms and advancing drug\ndevelopment. Protein stability, quantified by changes in Gibbs free energy\n($\\Delta\\Delta G$), is influenced by these mutations. However, the scarcity of\ndata and the complexity of model interpretation pose challenges in accurately\npredicting stability changes. This study proposes the application of deep\nneural networks, leveraging transfer learning and fusing complementary\ninformation from different models, to create a feature-rich representation of\nthe protein stability landscape. We developed four models, with our third\nmodel, ThermoMPNN+, demonstrating the best performance in predicting\n$\\Delta\\Delta G$ values. This approach, which integrates diverse feature sets\nand embeddings through latent transfusion techniques, aims to refine\n$\\Delta\\Delta G$ predictions and contribute to a deeper understanding of\nprotein dynamics, potentially leading to advancements in disease research and\ndrug discovery.\n", "date": "2025-01-13", "categories": []}
{"id": "2501.17589", "title": "Extracting Inter-Protein Interactions Via Multitasking Graph Structure\n  Learning", "abstract": "  Identifying protein-protein interactions (PPI) is crucial for gaining\nin-depth insights into numerous biological processes within cells and holds\nsignificant guiding value in areas such as drug development and disease\ntreatment. Currently, most PPI prediction methods focus primarily on the study\nof protein sequences, neglecting the critical role of the internal structure of\nproteins. This paper proposes a novel PPI prediction method named MgslaPPI,\nwhich utilizes graph attention to mine protein structural information and\nenhances the expressive power of the protein encoder through multitask learning\nstrategy. Specifically, we decompose the end-to-end PPI prediction process into\ntwo stages: amino acid residue reconstruction (A2RR) and protein interaction\nprediction (PIP). In the A2RR stage, we employ a graph attention-based residue\nreconstruction method to explore the internal relationships and features of\nproteins. In the PIP stage, in addition to the basic interaction prediction\ntask, we introduce two auxiliary tasks, i.e., protein feature reconstruction\n(PFR) and masked interaction prediction (MIP). The PFR task aims to reconstruct\nthe representation of proteins in the PIP stage, while the MIP task uses\npartially masked protein features for PPI prediction, with both working in\nconcert to prompt MgslaPPI to capture more useful information. Experimental\nresults demonstrate that MgslaPPI significantly outperforms existing\nstate-of-the-art methods under various data partitioning schemes.\n", "date": "2025-01-29", "categories": []}
{"id": "2501.16382", "title": "GraPPI: A Retrieve-Divide-Solve GraphRAG Framework for Large-scale\n  Protein-protein Interaction Exploration", "abstract": "  Drug discovery (DD) has tremendously contributed to maintaining and improving\npublic health. Hypothesizing that inhibiting protein misfolding can slow\ndisease progression, researchers focus on target identification (Target ID) to\nfind protein structures for drug binding. While Large Language Models (LLMs)\nand Retrieval-Augmented Generation (RAG) frameworks have accelerated drug\ndiscovery, integrating models into cohesive workflows remains challenging. We\nconducted a user study with drug discovery researchers to identify the\napplicability of LLMs and RAGs in Target ID. We identified two main findings:\n1) an LLM should provide multiple Protein-Protein Interactions (PPIs) based on\nan initial protein and protein candidates that have a therapeutic impact; 2)\nthe model must provide the PPI and relevant explanations for better\nunderstanding. Based on these observations, we identified three limitations in\nprevious approaches for Target ID: 1) semantic ambiguity, 2) lack of\nexplainability, and 3) short retrieval units. To address these issues, we\npropose GraPPI, a large-scale knowledge graph (KG)-based retrieve-divide-solve\nagent pipeline RAG framework to support large-scale PPI signaling pathway\nexploration in understanding therapeutic impacts by decomposing the analysis of\nentire PPI pathways into sub-tasks focused on the analysis of PPI edges.\n", "date": "2025-01-24", "categories": []}
{"id": "2501.16405", "title": "DepoRanker: A Web Tool to predict Klebsiella Depolymerases using Machine\n  Learning", "abstract": "  Background: Phage therapy shows promise for treating antibiotic-resistant\nKlebsiella infections. Identifying phage depolymerases that target Klebsiella\ncapsular polysaccharides is crucial, as these capsules contribute to biofilm\nformation and virulence. However, homology-based searches have limitations in\nnovel depolymerase discovery.\n  Objective: To develop a machine learning model for identifying and ranking\npotential phage depolymerases targeting Klebsiella.\n  Methods: We developed DepoRanker, a machine learning algorithm to rank\nproteins by their likelihood of being depolymerases. The model was\nexperimentally validated on 5 newly characterized proteins and compared to\nBLAST.\n  Results: DepoRanker demonstrated superior performance to BLAST in identifying\npotential depolymerases. Experimental validation confirmed its predictive\nability on novel proteins.\n  Conclusions: DepoRanker provides an accurate and functional tool to expedite\ndepolymerase discovery for phage therapy against Klebsiella. It is available as\na webserver and open-source software.\n  Availability: Webserver: https://deporanker.dcs.warwick.ac.uk/ Source code:\nhttps://github.com/wgrgwrght/deporanker\n", "date": "2025-01-27", "categories": []}
{"id": "2501.10282", "title": "Computational Protein Science in the Era of Large Language Models (LLMs)", "abstract": "  Considering the significance of proteins, computational protein science has\nalways been a critical scientific field, dedicated to revealing knowledge and\ndeveloping applications within the protein sequence-structure-function\nparadigm. In the last few decades, Artificial Intelligence (AI) has made\nsignificant impacts in computational protein science, leading to notable\nsuccesses in specific protein modeling tasks. However, those previous AI models\nstill meet limitations, such as the difficulty in comprehending the semantics\nof protein sequences, and the inability to generalize across a wide range of\nprotein modeling tasks. Recently, LLMs have emerged as a milestone in AI due to\ntheir unprecedented language processing & generalization capability. They can\npromote comprehensive progress in fields rather than solving individual tasks.\nAs a result, researchers have actively introduced LLM techniques in\ncomputational protein science, developing protein Language Models (pLMs) that\nskillfully grasp the foundational knowledge of proteins and can be effectively\ngeneralized to solve a diversity of sequence-structure-function reasoning\nproblems. While witnessing prosperous developments, it's necessary to present a\nsystematic overview of computational protein science empowered by LLM\ntechniques. First, we summarize existing pLMs into categories based on their\nmastered protein knowledge, i.e., underlying sequence patterns, explicit\nstructural and functional information, and external scientific languages.\nSecond, we introduce the utilization and adaptation of pLMs, highlighting their\nremarkable achievements in promoting protein structure prediction, protein\nfunction prediction, and protein design studies. Then, we describe the\npractical application of pLMs in antibody design, enzyme design, and drug\ndiscovery. Finally, we specifically discuss the promising future directions in\nthis fast-growing field.\n", "date": "2025-01-17", "categories": []}
{"id": "2501.14746", "title": "Neuromorphic Spiking Neural Network Based Classification of COVID-19\n  Spike Sequences", "abstract": "  The availability of SARS-CoV-2 (severe acute respiratory syndrome coronavirus\n2) virus data post-COVID has reached exponentially to an enormous magnitude,\nopening research doors to analyze its behavior. Various studies are conducted\nby researchers to gain a deeper understanding of the virus, like genomic\nsurveillance, etc, so that efficient prevention mechanisms can be developed.\nHowever, the unstable nature of the virus (rapid mutations, multiple hosts,\netc) creates challenges in designing analytical systems for it. Therefore, we\npropose a neural network-based (NN) mechanism to perform an efficient analysis\nof the SARS-CoV-2 data, as NN portrays generalized behavior upon training.\nMoreover, rather than using the full-length genome of the virus, we apply our\nmethod to its spike region, as this region is known to have predominant\nmutations and is used to attach to the host cell membrane. In this paper, we\nintroduce a pipeline that first converts the spike protein sequences into a\nfixed-length numerical representation and then uses Neuromorphic Spiking Neural\nNetwork to classify those sequences. We compare the performance of our method\nwith various baselines using real-world SARS-CoV-2 spike sequence data and show\nthat our method is able to achieve higher predictive accuracy compared to the\nrecent baselines.\n", "date": "2024-12-19", "categories": []}
{"id": "2501.15007", "title": "Controllable Protein Sequence Generation with LLM Preference\n  Optimization", "abstract": "  Designing proteins with specific attributes offers an important solution to\naddress biomedical challenges. Pre-trained protein large language models (LLMs)\nhave shown promising results on protein sequence generation. However, to\ncontrol sequence generation for specific attributes, existing work still\nexhibits poor functionality and structural stability. In this paper, we propose\na novel controllable protein design method called CtrlProt. We finetune a\nprotein LLM with a new multi-listwise preference optimization strategy to\nimprove generation quality and support multi-attribute controllable generation.\nExperiments demonstrate that CtrlProt can meet functionality and structural\nstability requirements effectively, achieving state-of-the-art performance in\nboth single-attribute and multi-attribute protein sequence generation.\n", "date": "2025-01-25", "categories": []}
{"id": "2501.15055", "title": "Group Ligands Docking to Protein Pockets", "abstract": "  Molecular docking is a key task in computational biology that has attracted\nincreasing interest from the machine learning community. While existing methods\nhave achieved success, they generally treat each protein-ligand pair in\nisolation. Inspired by the biochemical observation that ligands binding to the\nsame target protein tend to adopt similar poses, we propose \\textsc{GroupBind},\na novel molecular docking framework that simultaneously considers multiple\nligands docking to a protein. This is achieved by introducing an interaction\nlayer for the group of ligands and a triangle attention module for embedding\nprotein-ligand and group-ligand pairs. By integrating our approach with\ndiffusion-based docking model, we set a new S performance on the PDBBind blind\ndocking benchmark, demonstrating the effectiveness of our proposed molecular\ndocking paradigm.\n", "date": "2025-01-25", "categories": []}
{"id": "2501.15631", "title": "BoKDiff: Best-of-K Diffusion Alignment for Target-Specific 3D Molecule\n  Generation", "abstract": "  Structure-based drug design (SBDD) leverages the 3D structure of biomolecular\ntargets to guide the creation of new therapeutic agents. Recent advances in\ngenerative models, including diffusion models and geometric deep learning, have\ndemonstrated promise in optimizing ligand generation. However, the scarcity of\nhigh-quality protein-ligand complex data and the inherent challenges in\naligning generated ligands with target proteins limit the effectiveness of\nthese methods. We propose BoKDiff, a novel framework that enhances ligand\ngeneration by combining multi-objective optimization and Best-of-K alignment\nmethodologies. Built upon the DecompDiff model, BoKDiff generates diverse\ncandidates and ranks them using a weighted evaluation of molecular properties\nsuch as QED, SA, and docking scores. To address alignment challenges, we\nintroduce a method that relocates the center of mass of generated ligands to\ntheir docking poses, enabling accurate sub-component extraction. Additionally,\nwe integrate a Best-of-N (BoN) sampling approach, which selects the optimal\nligand from multiple generated candidates without requiring fine-tuning. BoN\nachieves exceptional results, with QED values exceeding 0.6, SA scores above\n0.75, and a success rate surpassing 35%, demonstrating its efficiency and\npracticality. BoKDiff achieves state-of-the-art results on the CrossDocked2020\ndataset, including a -8.58 average Vina docking score and a 26% success rate in\nmolecule generation. This study is the first to apply Best-of-K alignment and\nBest-of-N sampling to SBDD, highlighting their potential to bridge generative\nmodeling with practical drug discovery requirements. The code is provided at\nhttps://github.com/khodabandeh-ali/BoKDiff.git.\n", "date": "2025-01-26", "categories": []}
{"id": "2501.14056", "title": "Prior Knowledge Injection into Deep Learning Models Predicting Gene\n  Expression from Whole Slide Images", "abstract": "  Cancer diagnosis and prognosis primarily depend on clinical parameters such\nas age and tumor grade, and are increasingly complemented by molecular data,\nsuch as gene expression, from tumor sequencing. However, sequencing is costly\nand delays oncology workflows. Recent advances in Deep Learning allow to\npredict molecular information from morphological features within Whole Slide\nImages (WSIs), offering a cost-effective proxy of the molecular markers. While\npromising, current methods lack the robustness to fully replace direct\nsequencing. Here we aim to improve existing methods by introducing a\nmodel-agnostic framework that allows to inject prior knowledge on gene-gene\ninteractions into Deep Learning architectures, thereby increasing accuracy and\nrobustness. We design the framework to be generic and flexibly adaptable to a\nwide range of architectures. In a case study on breast cancer, our strategy\nleads to an average increase of 983 significant genes (out of 25,761) across\nall 18 experiments, with 14 generalizing to an increase on an independent\ndataset. Our findings reveal a high potential for injection of prior knowledge\nto increase gene expression prediction performance from WSIs across a wide\nrange of architectures.\n", "date": "2025-01-23", "categories": []}
{"id": "2501.14079", "title": "Enhancing Biomedical Relation Extraction with Directionality", "abstract": "  Biological relation networks contain rich information for understanding the\nbiological mechanisms behind the relationship of entities such as genes,\nproteins, diseases, and chemicals. The vast growth of biomedical literature\nposes significant challenges updating the network knowledge. The recent\nBiomedical Relation Extraction Dataset (BioRED) provides valuable manual\nannotations, facilitating the develop-ment of machine-learning and pre-trained\nlanguage model approaches for automatically identifying novel document-level\n(inter-sentence context) relationships. Nonetheless, its annotations lack\ndirectionality (subject/object) for the entity roles, essential for studying\ncomplex biological networks. Herein we annotate the entity roles of the\nrelationships in the BioRED corpus and subsequently propose a novel multi-task\nlanguage model with soft-prompt learning to jointly identify the relationship,\nnovel findings, and entity roles. Our results in-clude an enriched BioRED\ncorpus with 10,864 directionality annotations. Moreover, our proposed method\noutperforms existing large language models such as the state-of-the-art GPT-4\nand Llama-3 on two benchmarking tasks. Our source code and dataset are\navailable at https://github.com/ncbi-nlp/BioREDirect.\n", "date": "2025-01-23", "categories": []}
{"id": "2305.08929", "title": "AF2-Mutation: Adversarial Sequence Mutations against AlphaFold2 on\n  Protein Tertiary Structure Prediction", "abstract": "  Deep learning-based approaches, such as AlphaFold2 (AF2), have significantly\nadvanced protein tertiary structure prediction, achieving results comparable to\nreal biological experimental methods. While AF2 has shown limitations in\npredicting the effects of mutations, its robustness against sequence mutations\nremains to be determined. Starting with the wild-type (WT) sequence, we\ninvestigate adversarial sequences generated via an evolutionary approach, which\nAF2 predicts to be substantially different from WT. Our experiments on CASP14\nreveal that by modifying merely three residues in the protein sequence using a\ncombination of replacement, deletion, and insertion strategies, the alteration\nin AF2's predictions, as measured by the Local Distance Difference Test (lDDT),\nreaches 46.61. Moreover, when applied to a specific protein, SPNS2, our\nproposed algorithm successfully identifies biologically meaningful residues\ncritical to protein structure determination and potentially indicates\nalternative conformations, thus significantly expediting the experimental\nprocess.\n", "date": "2023-05-15", "categories": []}
{"id": "2407.06405", "title": "AI-driven multi-omics integration for multi-scale predictive modeling of\n  causal genotype-environment-phenotype relationships", "abstract": "  Despite the wealth of single-cell multi-omics data, it remains challenging to\npredict the consequences of novel genetic and chemical perturbations in the\nhuman body. It requires knowledge of molecular interactions at all biological\nlevels, encompassing disease models and humans. Current machine learning\nmethods primarily establish statistical correlations between genotypes and\nphenotypes but struggle to identify physiologically significant causal factors,\nlimiting their predictive power. Key challenges in predictive modeling include\nscarcity of labeled data, generalization across different domains, and\ndisentangling causation from correlation. In light of recent advances in\nmulti-omics data integration, we propose a new artificial intelligence\n(AI)-powered biology-inspired multi-scale modeling framework to tackle these\nissues. This framework will integrate multi-omics data across biological\nlevels, organism hierarchies, and species to predict causal\ngenotype-environment-phenotype relationships under various conditions. AI\nmodels inspired by biology may identify novel molecular targets, biomarkers,\npharmaceutical agents, and personalized medicines for presently unmet medical\nneeds.\n", "date": "2024-07-08", "categories": []}
{"id": "2411.15418", "title": "Scaling Structure Aware Virtual Screening to Billions of Molecules with\n  SPRINT", "abstract": "  Virtual screening of small molecules against protein targets can accelerate\ndrug discovery and development by predicting drug-target interactions (DTIs).\nHowever, structure-based methods like molecular docking are too slow to allow\nfor broad proteome-scale screens, limiting their application in screening for\noff-target effects or new molecular mechanisms. Recently, vector-based methods\nusing protein language models (PLMs) have emerged as a complementary approach\nthat bypasses explicit 3D structure modeling. Here, we develop SPRINT, a\nvector-based approach for screening entire chemical libraries against whole\nproteomes for DTIs and novel mechanisms of action. SPRINT improves on prior\nwork by using a self-attention based architecture and structure-aware PLMs to\nlearn drug-target co-embeddings for binder prediction, search, and retrieval.\nSPRINT achieves SOTA enrichment factors in virtual screening on LIT-PCBA, DTI\nclassification benchmarks, and binding affinity prediction benchmarks, while\nproviding interpretability in the form of residue-level attention maps. In\naddition to being both accurate and interpretable, SPRINT is ultra-fast:\nquerying the whole human proteome against the ENAMINE Real Database (6.7B\ndrugs) for the 100 most likely binders per protein takes 16 minutes. SPRINT\npromises to enable virtual screening at an unprecedented scale, opening up new\nopportunities for in silico drug repurposing and development. SPRINT is\navailable on the web as ColabScreen: https://bit.ly/colab-screen\n", "date": "2024-11-23", "categories": []}
{"id": "2402.05982", "title": "Decoupled Sequence and Structure Generation for Realistic Antibody\n  Design", "abstract": "  Recently, deep learning has made rapid progress in antibody design, which\nplays a key role in the advancement of therapeutics. A dominant paradigm is to\ntrain a model to jointly generate the antibody sequence and the structure as a\ncandidate. However, the joint generation requires the model to generate both\nthe discrete amino acid categories and the continuous 3D coordinates; this\nlimits the space of possible architectures and may lead to suboptimal\nperformance. In response, we propose an antibody sequence-structure decoupling\n(ASSD) framework, which separates sequence generation and structure prediction.\nAlthough our approach is simple, our idea allows the use of powerful neural\narchitectures and demonstrates notable performance improvements. We also find\nthat the widely used non-autoregressive generators promote sequences with\noverly repeating tokens. Such sequences are both out-of-distribution and prone\nto undesirable developability properties that can trigger harmful immune\nresponses in patients. To resolve this, we introduce a composition-based\nobjective that allows an efficient trade-off between high performance and low\ntoken repetition. ASSD shows improved performance in various antibody design\nexperiments, while the composition-based objective successfully mitigates token\nrepetition of non-autoregressive models.\n", "date": "2024-02-08", "categories": []}
{"id": "2410.18864", "title": "Omics-driven hybrid dynamic modeling of bioprocesses with uncertainty\n  estimation", "abstract": "  This work presents an omics-driven modeling pipeline that integrates\nmachine-learning tools to facilitate the dynamic modeling of multiscale\nbiological systems. Random forests and permutation feature importance are\nproposed to mine omics datasets, guiding feature selection and dimensionality\nreduction for dynamic modeling. Continuous and differentiable machine-learning\nfunctions can be trained to link the reduced omics feature set to key\ncomponents of the dynamic model, resulting in a hybrid model. As proof of\nconcept, we apply this framework to a high-dimensional proteomics dataset of\n$\\textit{Saccharomyces cerevisiae}$. After identifying key intracellular\nproteins that correlate with cell growth, targeted dynamic experiments are\ndesigned, and key model parameters are captured as functions of the selected\nproteins using Gaussian processes. This approach captures the dynamic behavior\nof yeast strains under varying proteome profiles while estimating the\nuncertainty in the hybrid model's predictions. The outlined modeling framework\nis adaptable to other scenarios, such as integrating additional layers of omics\ndata for more advanced multiscale biological systems, or employing alternative\nmachine-learning methods to handle larger datasets. Overall, this study\noutlines a strategy for leveraging omics data to inform multiscale dynamic\nmodeling in systems biology and bioprocess engineering.\n", "date": "2024-10-24", "categories": []}
{"id": "2412.07051", "title": "A Misclassification Network-Based Method for Comparative Genomic\n  Analysis", "abstract": "  Classifying genome sequences based on metadata has been an active area of\nresearch in comparative genomics for decades with many important applications\nacross the life sciences. Established methods for classifying genomes can be\nbroadly grouped into sequence alignment-based and alignment-free models.\nConventional alignment-based models rely on genome similarity measures\ncalculated based on local sequence alignments or consistent ordering among\nsequences. However, such methods are computationally expensive when dealing\nwith large ensembles of even moderately sized genomes. In contrast,\nalignment-free (AF) approaches measure genome similarity based on summary\nstatistics in an unsupervised setting and are efficient enough to analyze large\ndatasets. However, both alignment-based and AF methods typically assume fixed\nscoring rubrics that lack the flexibility to assign varying importance to\ndifferent parts of the sequences based on prior knowledge. In this study, we\nintegrate AI and network science approaches to develop a comparative genomic\nanalysis framework that addresses these limitations. Our approach, termed the\nGenome Misclassification Network Analysis (GMNA), simultaneously leverages\nmisclassified instances, a learned scoring rubric, and label information to\nclassify genomes based on associated metadata and better understand potential\ndrivers of misclassification. We evaluate the utility of the GMNA using Naive\nBayes and convolutional neural network models, supplemented by additional\nexperiments with transformer-based models, to construct SARS-CoV-2 sampling\nlocation classifiers using over 500,000 viral genome sequences and study the\nresulting network of misclassifications. We demonstrate the global health\npotential of the GMNA by leveraging the SARS-CoV-2 genome misclassification\nnetworks to investigate the role human mobility played in structuring\ngeographic clustering of SARS-CoV-2.\n", "date": "2024-12-09", "categories": []}
{"id": "2412.19228", "title": "Learning Cross-Domain Representations for Transferable Drug\n  Perturbations on Single-Cell Transcriptional Responses", "abstract": "  Phenotypic drug discovery has attracted widespread attention because of its\npotential to identify bioactive molecules. Transcriptomic profiling provides a\ncomprehensive reflection of phenotypic changes in cellular responses to\nexternal perturbations. In this paper, we propose XTransferCDR, a novel\ngenerative framework designed for feature decoupling and transferable\nrepresentation learning across domains. Given a pair of perturbed expression\nprofiles, our approach decouples the perturbation representations from basal\nstates through domain separation encoders and then cross-transfers them in the\nlatent space. The transferred representations are then used to reconstruct the\ncorresponding perturbed expression profiles via a shared decoder. This\ncross-transfer constraint effectively promotes the learning of transferable\ndrug perturbation representations. We conducted extensive evaluations of our\nmodel on multiple datasets, including single-cell transcriptional responses to\ndrugs and single- and combinatorial genetic perturbations. The experimental\nresults show that XTransferCDR achieved better performance than current\nstate-of-the-art methods, showcasing its potential to advance phenotypic drug\ndiscovery.\n", "date": "2024-12-26", "categories": []}
{"id": "2501.08187", "title": "A Multi-Modal AI Copilot for Single-Cell Analysis with Instruction\n  Following", "abstract": "  Large language models excel at interpreting complex natural language\ninstructions, enabling them to perform a wide range of tasks. In the life\nsciences, single-cell RNA sequencing (scRNA-seq) data serves as the \"language\nof cellular biology\", capturing intricate gene expression patterns at the\nsingle-cell level. However, interacting with this \"language\" through\nconventional tools is often inefficient and unintuitive, posing challenges for\nresearchers. To address these limitations, we present InstructCell, a\nmulti-modal AI copilot that leverages natural language as a medium for more\ndirect and flexible single-cell analysis. We construct a comprehensive\nmulti-modal instruction dataset that pairs text-based instructions with\nscRNA-seq profiles from diverse tissues and species. Building on this, we\ndevelop a multi-modal cell language architecture capable of simultaneously\ninterpreting and processing both modalities. InstructCell empowers researchers\nto accomplish critical tasks-such as cell type annotation, conditional\npseudo-cell generation, and drug sensitivity prediction-using straightforward\nnatural language commands. Extensive evaluations demonstrate that InstructCell\nconsistently meets or exceeds the performance of existing single-cell\nfoundation models, while adapting to diverse experimental conditions. More\nimportantly, InstructCell provides an accessible and intuitive tool for\nexploring complex single-cell data, lowering technical barriers and enabling\ndeeper biological insights.\n", "date": "2025-01-14", "categories": []}
{"id": "2501.07737", "title": "Multi-megabase scale genome interpretation with genetic language models", "abstract": "  Understanding how molecular changes caused by genetic variation drive disease\nrisk is crucial for deciphering disease mechanisms. However, interpreting\ngenome sequences is challenging because of the vast size of the human genome,\nand because its consequences manifest across a wide range of cells, tissues and\nscales -- spanning from molecular to whole organism level. Here, we present\nPhenformer, a multi-scale genetic language model that learns to generate\nmechanistic hypotheses as to how differences in genome sequence lead to\ndisease-relevant changes in expression across cell types and tissues directly\nfrom DNA sequences of up to 88 million base pairs. Using whole genome\nsequencing data from more than 150 000 individuals, we show that Phenformer\ngenerates mechanistic hypotheses about disease-relevant cell and tissue types\nthat match literature better than existing state-of-the-art methods, while\nusing only sequence data. Furthermore, disease risk predictors enriched by\nPhenformer show improved prediction performance and generalisation to diverse\npopulations. Accurate multi-megabase scale interpretation of whole genomes\nwithout additional experimental data enables both a deeper understanding of\nmolecular mechanisms involved in disease and improved disease risk prediction\nat the level of individuals.\n", "date": "2025-01-13", "categories": []}
{"id": "2501.07747", "title": "Scaling Up ESM2 Architectures for Long Protein Sequences Analysis: Long\n  and Quantized Approaches", "abstract": "  Various approaches utilizing Transformer architectures have achieved\nstate-of-the-art results in Natural Language Processing (NLP). Based on this\nsuccess, numerous architectures have been proposed for other types of data,\nsuch as in biology, particularly for protein sequences. Notably among these are\nthe ESM2 architectures, pre-trained on billions of proteins, which form the\nbasis of various state-of-the-art approaches in the field. However, the ESM2\narchitectures have a limitation regarding input size, restricting it to 1,022\namino acids, which necessitates the use of preprocessing techniques to handle\nsequences longer than this limit. In this paper, we present the long and\nquantized versions of the ESM2 architectures, doubling the input size limit to\n2,048 amino acids.\n", "date": "2025-01-13", "categories": []}
{"id": "2501.07884", "title": "MD-Syn: Synergistic drug combination prediction based on the\n  multidimensional feature fusion method and attention mechanisms", "abstract": "  Drug combination therapies have shown promising therapeutic efficacy in\ncomplex diseases and have demonstrated the potential to reduce drug resistance.\nHowever, the huge number of possible drug combinations makes it difficult to\nscreen them all in traditional experiments. In this study, we proposed MD-Syn,\na computational framework, which is based on the multidimensional feature\nfusion method and multi-head attention mechanisms. Given drug pair-cell line\ntriplets, MD-Syn considers one-dimensional and two-dimensional feature spaces\nsimultaneously. It consists of a one-dimensional feature embedding module\n(1D-FEM), a two-dimensional feature embedding module (2D-FEM), and a deep\nneural network-based classifier for synergistic drug combination prediction.\nMD-Syn achieved the AUROC of 0.919 in 5-fold cross-validation, outperforming\nthe state-of-the-art methods. Further, MD-Syn showed comparable results over\ntwo independent datasets. In addition, the multi-head attention mechanisms not\nonly learn embeddings from different feature aspects but also focus on\nessential interactive feature elements, improving the interpretability of\nMD-Syn. In summary, MD-Syn is an interpretable framework to prioritize\nsynergistic drug combination pairs with chemicals and cancer cell line gene\nexpression profiles. To facilitate broader community access to this model, we\nhave developed a web portal (https://labyeh104-2.life.nthu.edu.tw/) that\nenables customized predictions of drug combination synergy effects based on\nuser-specified compounds.\n", "date": "2025-01-14", "categories": []}
{"id": "2501.07970", "title": "Comprehensive Metapath-based Heterogeneous Graph Transformer for\n  Gene-Disease Association Prediction", "abstract": "  Discovering gene-disease associations is crucial for understanding disease\nmechanisms, yet identifying these associations remains challenging due to the\ntime and cost of biological experiments. Computational methods are increasingly\nvital for efficient and scalable gene-disease association prediction.\nGraph-based learning models, which leverage node features and network\nrelationships, are commonly employed for biomolecular predictions. However,\nexisting methods often struggle to effectively integrate node features,\nheterogeneous structures, and semantic information. To address these\nchallenges, we propose COmprehensive MEtapath-based heterogeneous graph\nTransformer(COMET) for predicting gene-disease associations. COMET integrates\ndiverse datasets to construct comprehensive heterogeneous networks,\ninitializing node features with BioGPT. We define seven Metapaths and utilize a\ntransformer framework to aggregate Metapath instances, capturing global\ncontexts and long-distance dependencies. Through intra- and inter-metapath\naggregation using attention mechanisms, COMET fuses latent vectors from\nmultiple Metapaths to enhance GDA prediction accuracy. Our method demonstrates\nsuperior robustness compared to state-of-the-art approaches. Ablation studies\nand visualizations validate COMET's effectiveness, providing valuable insights\nfor advancing human health research.\n", "date": "2025-01-14", "categories": []}
{"id": "2501.06314", "title": "BioAgents: Democratizing Bioinformatics Analysis with Multi-Agent\n  Systems", "abstract": "  Creating end-to-end bioinformatics workflows requires diverse domain\nexpertise, which poses challenges for both junior and senior researchers as it\ndemands a deep understanding of both genomics concepts and computational\ntechniques. While large language models (LLMs) provide some assistance, they\noften fall short in providing the nuanced guidance needed to execute complex\nbioinformatics tasks, and require expensive computing resources to achieve high\nperformance. We thus propose a multi-agent system built on small language\nmodels, fine-tuned on bioinformatics data, and enhanced with retrieval\naugmented generation (RAG). Our system, BioAgents, enables local operation and\npersonalization using proprietary data. We observe performance comparable to\nhuman experts on conceptual genomics tasks, and suggest next steps to enhance\ncode generation capabilities.\n", "date": "2025-01-10", "categories": []}
{"id": "2501.06805", "title": "A Pan-cancer Classification Model using Multi-view Feature Selection\n  Method and Ensemble Classifier", "abstract": "  Accurately identifying cancer samples is crucial for precise diagnosis and\neffective patient treatment. Traditional methods falter with high-dimensional\nand high feature-to-sample count ratios, which are critical for classifying\ncancer samples. This study aims to develop a novel feature selection framework\nspecifically for transcriptome data and propose two ensemble classifiers. For\nfeature selection, we partition the transcriptome dataset vertically based on\nfeature types. Then apply the Boruta feature selection process on each of the\npartitions, combine the results, and apply Boruta again on the combined result.\nWe repeat the process with different parameters of Boruta and prepare the final\nfeature set. Finally, we constructed two ensemble ML models based on LR, SVM\nand XGBoost classifiers with max voting and averaging probability approach. We\nused 10-fold cross-validation to ensure robust and reliable classification\nperformance. With 97.11\\% accuracy and 0.9996 AUC value, our approach performs\nbetter compared to existing state-of-the-art methods to classify 33 types of\ncancers. A set of 12 types of cancer is traditionally challenging to\ndifferentiate between each other due to their similarity in tissue of origin.\nOur method accurately identifies over 90\\% of samples from these 12 types of\ncancers, which outperforms all known methods presented in existing literature.\nThe gene set enrichment analysis reveals that our framework's selected features\nhave enriched the pathways highly related to cancers. This study develops a\nfeature selection framework to select features highly related to cancer\ndevelopment and leads to identifying different types of cancer samples with\nhigher accuracy.\n", "date": "2025-01-12", "categories": []}
{"id": "2501.07405", "title": "PROTECT: Protein circadian time prediction using unsupervised learning", "abstract": "  Circadian rhythms regulate the physiology and behavior of humans and animals.\nDespite advancements in understanding these rhythms and predicting circadian\nphases at the transcriptional level, predicting circadian phases from proteomic\ndata remains elusive. This challenge is largely due to the scarcity of time\nlabels in proteomic datasets, which are often characterized by small sample\nsizes, high dimensionality, and significant noise. Furthermore, existing\nmethods for predicting circadian phases from transcriptomic data typically rely\non prior knowledge of known rhythmic genes, making them unsuitable for\nproteomic datasets. To address this gap, we developed a novel computational\nmethod using unsupervised deep learning techniques to predict circadian sample\nphases from proteomic data without requiring time labels or prior knowledge of\nproteins or genes. Our model involves a two-stage training process optimized\nfor robust circadian phase prediction: an initial greedy one-layer-at-a-time\npre-training which generates informative initial parameters followed by\nfine-tuning. During fine-tuning, a specialized loss function guides the model\nto align protein expression levels with circadian patterns, enabling it to\naccurately capture the underlying rhythmic structure within the data. We tested\nour method on both time-labeled and unlabeled proteomic data. For labeled data,\nwe compared our predictions to the known time labels, achieving high accuracy,\nwhile for unlabeled human datasets, including postmortem brain regions and\nurine samples, we explored circadian disruptions. Notably, our analysis\nidentified disruptions in rhythmic proteins between Alzheimer's disease and\ncontrol subjects across these samples.\n", "date": "2025-01-13", "categories": []}
{"id": "2405.06653", "title": "A unified cross-attention model for predicting antigen binding\n  specificity to both HLA and TCR molecules", "abstract": "  The immune checkpoint inhibitors have demonstrated promising clinical\nefficacy across various tumor types, yet the percentage of patients who benefit\nfrom them remains low. The bindings between tumor antigens and HLA-I/TCR\nmolecules determine the antigen presentation and T-cell activation, thereby\nplaying an important role in the immunotherapy response. In this paper, we\npropose UnifyImmun, a unified cross-attention transformer model designed to\nsimultaneously predict the bindings of peptides to both receptors, providing\nmore comprehensive evaluation of antigen immunogenicity. We devise a two-phase\nstrategy using virtual adversarial training that enables these two tasks to\nreinforce each other mutually, by compelling the encoders to extract more\nexpressive features. Our method demonstrates superior performance in predicting\nboth pHLA and pTCR binding on multiple independent and external test sets.\nNotably, on a large-scale COVID-19 pTCR binding test set without any seen\npeptide in training set, our method outperforms the current state-of-the-art\nmethods by more than 10\\%. The predicted binding scores significantly correlate\nwith the immunotherapy response and clinical outcomes on two clinical cohorts.\nFurthermore, the cross-attention scores and integrated gradients reveal the\namino-acid sites critical for peptide binding to receptors. In essence, our\napproach marks a significant step toward comprehensive evaluation of antigen\nimmunogenicity.\n", "date": "2024-04-08", "categories": []}
{"id": "2412.16220", "title": "Cross-Attention Graph Neural Networks for Inferring Gene Regulatory\n  Networks with Skewed Degree Distribution", "abstract": "  Inferencing Gene Regulatory Networks (GRNs) from gene expression data is a\npivotal challenge in systems biology, and several innovative computational\nmethods have been introduced. However, most of these studies have not\nconsidered the skewed degree distribution of genes. Specifically, some genes\nmay regulate multiple target genes while some genes may be regulated by\nmultiple regulator genes. Such a skewed degree distribution issue significantly\ncomplicates the application of directed graph embedding methods. To tackle this\nissue, we propose the Cross-Attention Complex Dual Graph Embedding Model\n(XATGRN). Our XATGRN employs a cross-attention mechanism to effectively capture\nintricate gene interactions from gene expression profiles. Additionally, it\nuses a Dual Complex Graph Embedding approach to manage the skewed degree\ndistribution, thereby ensuring precise prediction of regulatory relationships\nand their directionality. Our model consistently outperforms existing\nstate-of-the-art methods across various datasets, underscoring its efficacy in\nelucidating complex gene regulatory mechanisms. Our codes used in this paper\nare publicly available at: https://github.com/kikixiong/XATGRN.\n", "date": "2024-12-18", "categories": []}
{"id": "2403.05260", "title": "Towards generalization of drug response prediction to single cells and\n  patients utilizing importance-aware multi-source domain transfer learning", "abstract": "  The advancement of single-cell sequencing technology has promoted the\ngeneration of a large amount of single-cell transcriptional profiles, providing\nunprecedented opportunities to identify drug-resistant cell subpopulations\nwithin a tumor. However, few studies have focused on drug response prediction\nat single-cell level, and their performance remains suboptimal. This paper\nproposed scAdaDrug, a novel multi-source domain adaptation model powered by\nadaptive importance-aware representation learning to predict drug response of\nindividual cells. We used a shared encoder to extract domain-invariant features\nrelated to drug response from multiple source domains by utilizing adversarial\ndomain adaptation. Particularly, we introduced a plug-and-play module to\ngenerate importance-aware and mutually independent weights, which could\nadaptively modulate the latent representation of each sample in element-wise\nmanner between source and target domains. Extensive experimental results showed\nthat our model achieved state-of-the-art performance in predicting drug\nresponse on multiple independent datasets, including single-cell datasets\nderived from both cell lines and patient-derived xenografts (PDX) models, as\nwell as clinical tumor patient cohorts. Moreover, the ablation experiments\ndemonstrated our model effectively captured the underlying patterns determining\ndrug response from multiple source domains.\n", "date": "2024-03-08", "categories": []}
{"id": "2501.03923", "title": "Explainable AI model reveals disease-related mechanisms in single-cell\n  RNA-seq data", "abstract": "  Neurodegenerative diseases (NDDs) are complex and lack effective treatment\ndue to their poorly understood mechanism. The increasingly used data analysis\nfrom Single nucleus RNA Sequencing (snRNA-seq) allows to explore transcriptomic\nevents at a single cell level, yet face challenges in interpreting the\nmechanisms underlying a disease. On the other hand, Neural Network (NN) models\ncan handle complex data to offer insights but can be seen as black boxes with\npoor interpretability. In this context, explainable AI (XAI) emerges as a\nsolution that could help to understand disease-associated mechanisms when\ncombined with efficient NN models. However, limited research explores XAI in\nsingle-cell data. In this work, we implement a method for identifying\ndisease-related genes and the mechanistic explanation of disease progression\nbased on NN model combined with SHAP. We analyze available Huntington's disease\n(HD) data to identify both HD-altered genes and mechanisms by adding Gene Set\nEnrichment Analysis (GSEA) comparing two methods, differential gene expression\nanalysis (DGE) and NN combined with SHAP approach. Our results show that DGE\nand SHAP approaches offer both common and differential sets of altered genes\nand pathways, reinforcing the usefulness of XAI methods for a broader\nperspective of disease.\n", "date": "2025-01-07", "categories": []}
{"id": "2501.02045", "title": "METAGENE-1: Metagenomic Foundation Model for Pandemic Monitoring", "abstract": "  We pretrain METAGENE-1, a 7-billion-parameter autoregressive transformer\nmodel, which we refer to as a metagenomic foundation model, on a novel corpus\nof diverse metagenomic DNA and RNA sequences comprising over 1.5 trillion base\npairs. This dataset is sourced from a large collection of human wastewater\nsamples, processed and sequenced using deep metagenomic (next-generation)\nsequencing methods. Unlike genomic models that focus on individual genomes or\ncurated sets of specific species, the aim of METAGENE-1 is to capture the full\ndistribution of genomic information present within this wastewater, to aid in\ntasks relevant to pandemic monitoring and pathogen detection. We carry out\nbyte-pair encoding (BPE) tokenization on our dataset, tailored for metagenomic\nsequences, and then pretrain our model. In this paper, we first detail the\npretraining dataset, tokenization strategy, and model architecture,\nhighlighting the considerations and design choices that enable the effective\nmodeling of metagenomic data. We then show results of pretraining this model on\nour metagenomic dataset, providing details about our losses, system metrics,\nand training stability over the course of pretraining. Finally, we demonstrate\nthe performance of METAGENE-1, which achieves state-of-the-art results on a set\nof genomic benchmarks and new evaluations focused on human-pathogen detection\nand genomic sequence embedding, showcasing its potential for public health\napplications in pandemic monitoring, biosurveillance, and early detection of\nemerging health threats.\n", "date": "2025-01-03", "categories": []}
{"id": "2501.02176", "title": "Molecule-dynamic-based Aging Clock and Aging Roadmap Forecast with\n  Sundial", "abstract": "  Addressing the unavoidable bias inherent in supervised aging clocks, we\nintroduce Sundial, a novel framework that models molecular dynamics through a\ndiffusion field, capturing both the population-level aging process and the\nindividual-level relative aging order. Sundial enables unbiasedestimation of\nbiological age and the forecast of aging roadmap. Fasteraging individuals from\nSundial exhibit a higher disease risk compared to those identified from\nsupervised aging clocks. This framework opens new avenues for exploring key\ntopics, including age- and sex-specific aging dynamics and faster yet healthy\naging paths.\n", "date": "2025-01-04", "categories": []}
{"id": "2501.02401", "title": "iTARGET: Interpretable Tailored Age Regression for Grouped Epigenetic\n  Traits", "abstract": "  Accurately predicting chronological age from DNA methylation patterns is\ncrucial for advancing biological age estimation. However, this task is made\nchallenging by Epigenetic Correlation Drift (ECD) and Heterogeneity Among CpGs\n(HAC), which reflect the dynamic relationship between methylation and age\nacross different life stages. To address these issues, we propose a novel\ntwo-phase algorithm. The first phase employs similarity searching to cluster\nmethylation profiles by age group, while the second phase uses Explainable\nBoosting Machines (EBM) for precise, group-specific prediction. Our method not\nonly improves prediction accuracy but also reveals key age-related CpG sites,\ndetects age-specific changes in aging rates, and identifies pairwise\ninteractions between CpG sites. Experimental results show that our approach\noutperforms traditional epigenetic clocks and machine learning models, offering\na more accurate and interpretable solution for biological age estimation with\nsignificant implications for aging research.\n", "date": "2025-01-04", "categories": []}
{"id": "2501.02824", "title": "Proteomic Learning of Gamma-Aminobutyric Acid (GABA) Receptor-Mediated\n  Anesthesia", "abstract": "  Anesthetics are crucial in surgical procedures and therapeutic interventions,\nbut they come with side effects and varying levels of effectiveness, calling\nfor novel anesthetic agents that offer more precise and controllable effects.\nTargeting Gamma-aminobutyric acid (GABA) receptors, the primary inhibitory\nreceptors in the central nervous system, could enhance their inhibitory action,\npotentially reducing side effects while improving the potency of anesthetics.\nIn this study, we introduce a proteomic learning of GABA receptor-mediated\nanesthesia based on 24 GABA receptor subtypes by considering over 4000 proteins\nin protein-protein interaction (PPI) networks and over 1.5 millions known\nbinding compounds. We develop a corresponding drug-target interaction network\nto identify potential lead compounds for novel anesthetic design. To ensure\nrobust proteomic learning predictions, we curated a dataset comprising 136\ntargets from a pool of 980 targets within the PPI networks. We employed three\nmachine learning algorithms, integrating advanced natural language processing\n(NLP) models such as pretrained transformer and autoencoder embeddings. Through\na comprehensive screening process, we evaluated the side effects and\nrepurposing potential of over 180,000 drug candidates targeting the GABRA5\nreceptor. Additionally, we assessed the ADMET (absorption, distribution,\nmetabolism, excretion, and toxicity) properties of these candidates to identify\nthose with near-optimal characteristics. This approach also involved optimizing\nthe structures of existing anesthetics. Our work presents an innovative\nstrategy for the development of new anesthetic drugs, optimization of\nanesthetic use, and deeper understanding of potential anesthesia-related side\neffects.\n", "date": "2025-01-06", "categories": []}
{"id": "2501.02855", "title": "Synthetic Fungi Datasets: A Time-Aligned Approach", "abstract": "  Fungi undergo dynamic morphological transformations throughout their\nlifecycle, forming intricate networks as they transition from spores to mature\nmycelium structures. To support the study of these time-dependent processes, we\npresent a synthetic, time-aligned image dataset that models key stages of\nfungal growth. This dataset systematically captures phenomena such as spore\nsize reduction, branching dynamics, and the emergence of complex mycelium\nnetworks. The controlled generation process ensures temporal consistency,\nscalability, and structural alignment, addressing the limitations of real-world\nfungal datasets. Optimized for deep learning (DL) applications, this dataset\nfacilitates the development of models for classifying growth stages, predicting\nfungal development, and analyzing morphological patterns over time. With\napplications spanning agriculture, medicine, and industrial mycology, this\nresource provides a robust foundation for automating fungal analysis, enhancing\ndisease monitoring, and advancing fungal biology research through artificial\nintelligence.\n", "date": "2025-01-06", "categories": []}
{"id": "2409.03773", "title": "CoPRA: Bridging Cross-domain Pretrained Sequence Models with Complex\n  Structures for Protein-RNA Binding Affinity Prediction", "abstract": "  Accurately measuring protein-RNA binding affinity is crucial in many\nbiological processes and drug design. Previous computational methods for\nprotein-RNA binding affinity prediction rely on either sequence or structure\nfeatures, unable to capture the binding mechanisms comprehensively. The recent\nemerging pre-trained language models trained on massive unsupervised sequences\nof protein and RNA have shown strong representation ability for various\nin-domain downstream tasks, including binding site prediction. However,\napplying different-domain language models collaboratively for complex-level\ntasks remains unexplored. In this paper, we propose CoPRA to bridge pre-trained\nlanguage models from different biological domains via Complex structure for\nProtein-RNA binding Affinity prediction. We demonstrate for the first time that\ncross-biological modal language models can collaborate to improve binding\naffinity prediction. We propose a Co-Former to combine the cross-modal sequence\nand structure information and a bi-scope pre-training strategy for improving\nCo-Former's interaction understanding. Meanwhile, we build the largest\nprotein-RNA binding affinity dataset PRA310 for performance evaluation. We also\ntest our model on a public dataset for mutation effect prediction. CoPRA\nreaches state-of-the-art performance on all the datasets. We provide extensive\nanalyses and verify that CoPRA can (1) accurately predict the protein-RNA\nbinding affinity; (2) understand the binding affinity change caused by\nmutations; and (3) benefit from scaling data and model size.\n", "date": "2024-08-21", "categories": []}
{"id": "2501.01458", "title": "GAN-TAT: A Novel Framework Using Protein Interaction Networks in\n  Druggable Gene Identification", "abstract": "  Identifying druggable genes is essential for developing effective\npharmaceuticals. With the availability of extensive, high-quality data,\ncomputational methods have become a significant asset. Protein Interaction\nNetwork (PIN) is valuable but challenging to implement due to its high\ndimensionality and sparsity. Previous methods relied on indirect integration,\nleading to resolution loss. This study proposes GAN-TAT, a framework utilizing\nan advanced graph embedding technology, ImGAGN, to directly integrate PIN for\ndruggable gene inference work. Tested on three Pharos datasets, GAN-TAT\nachieved the highest AUC-ROC score of 0.951 on Tclin. Further evaluation shows\nthat GAN-TAT's predictions are supported by clinical evidence, highlighting its\npotential practical applications in pharmacogenomics. This research represents\na methodological attempt with the direct utilization of PIN, expanding\npotential new solutions for developing drug targets. The source code of GAN-TAT\nis available at (https://github.com/george-yuanji-wang/GAN-TAT).\n", "date": "2024-12-31", "categories": []}
{"id": "2501.01930", "title": "GoBERT: Gene Ontology Graph Informed BERT for Universal Gene Function\n  Prediction", "abstract": "  Exploring the functions of genes and gene products is crucial to a wide range\nof fields, including medical research, evolutionary biology, and environmental\nscience. However, discovering new functions largely relies on expensive and\nexhaustive wet lab experiments. Existing methods of automatic function\nannotation or prediction mainly focus on protein function prediction with\nsequence, 3D-structures or protein family information. In this study, we\npropose to tackle the gene function prediction problem by exploring Gene\nOntology graph and annotation with BERT (GoBERT) to decipher the underlying\nrelationships among gene functions. Our proposed novel function prediction task\nutilizes existing functions as inputs and generalizes the function prediction\nto gene and gene products. Specifically, two pre-train tasks are designed to\njointly train GoBERT to capture both explicit and implicit relations of\nfunctions. Neighborhood prediction is a self-supervised multi-label\nclassification task that captures the explicit function relations. Specified\nmasking and recovering task helps GoBERT in finding implicit patterns among\nfunctions. The pre-trained GoBERT possess the ability to predict novel\nfunctions for various gene and gene products based on known functional\nannotations. Extensive experiments, biological case studies, and ablation\nstudies are conducted to demonstrate the superiority of our proposed GoBERT.\n", "date": "2025-01-03", "categories": []}
{"id": "2207.01586", "title": "Accurate RNA 3D structure prediction using a language model-based deep\n  learning approach", "abstract": "  Accurate prediction of RNA three-dimensional (3D) structure remains an\nunsolved challenge. Determining RNA 3D structures is crucial for understanding\ntheir functions and informing RNA-targeting drug development and synthetic\nbiology design. The structural flexibility of RNA, which leads to scarcity of\nexperimentally determined data, complicates computational prediction efforts.\nHere, we present RhoFold+, an RNA language model-based deep learning method\nthat accurately predicts 3D structures of single-chain RNAs from sequences. By\nintegrating an RNA language model pre-trained on ~23.7 million RNA sequences\nand leveraging techniques to address data scarcity, RhoFold+ offers a fully\nautomated end-to-end pipeline for RNA 3D structure prediction. Retrospective\nevaluations on RNA-Puzzles and CASP15 natural RNA targets demonstrate\nRhoFold+'s superiority over existing methods, including human expert groups.\nIts efficacy and generalizability are further validated through cross-family\nand cross-type assessments, as well as time-censored benchmarks. Additionally,\nRhoFold+ predicts RNA secondary structures and inter-helical angles, providing\nempirically verifiable features that broaden its applicability to RNA structure\nand function studies.\n", "date": "2022-07-04", "categories": []}
{"id": "2312.16624", "title": "Dual-stage optimizer for systematic overestimation adjustment applied to\n  multi-objective genetic algorithms for biomarker selection", "abstract": "  The challenge in biomarker discovery using machine learning from omics data\nlies in the abundance of molecular features but scarcity of samples. Most\nfeature selection methods in machine learning require evaluating various sets\nof features (models) to determine the most effective combination. This process,\ntypically conducted using a validation dataset, involves testing different\nfeature sets to optimize the model's performance. Evaluations have performance\nestimation error and when the selection involves many models the best ones are\nalmost certainly overestimated. Biomarker identification with feature selection\nmethods can be addressed as a multi-objective problem with trade-offs between\npredictive ability and parsimony in the number of features. Genetic algorithms\nare a popular tool for multi-objective optimization but they evolve numerous\nsolutions thus are prone to overestimation. Methods have been proposed to\nreduce the overestimation after a model has already been selected in\nsingle-objective problems, but no algorithm existed capable of reducing the\noverestimation during the optimization, improving model selection, or applied\nin the more general multi-objective domain. We propose DOSA-MO, a novel\nmulti-objective optimization wrapper algorithm that learns how the original\nestimation, its variance, and the feature set size of the solutions predict the\noverestimation. DOSA-MO adjusts the expectation of the performance during the\noptimization, improving the composition of the solution set. We verify that\nDOSA-MO improves the performance of a state-of-the-art genetic algorithm on\nleft-out or external sample sets, when predicting cancer subtypes and/or\npatient overall survival, using three transcriptomics datasets for kidney and\nbreast cancer.\n", "date": "2023-12-27", "categories": []}
{"id": "2410.22452", "title": "Explainable convolutional neural network model provides an alternative\n  genome-wide association perspective on mutations in SARS-CoV-2", "abstract": "  Identifying mutations of SARS-CoV-2 strains associated with their phenotypic\nchanges is critical for pandemic prediction and prevention. We compared an\nexplainable convolutional neural network (CNN) approach and the traditional\ngenome-wide association study (GWAS) on the mutations associated with WHO\nlabels of SARS-CoV-2, a proxy for virulence phenotypes. We trained a CNN\nclassification model that can predict genomic sequences into Variants of\nConcern (VOCs) and then applied Shapley Additive explanations (SHAP) model to\nidentify mutations that are important for the correct predictions. For\ncomparison, we performed traditional GWAS to identify mutations associated with\nVOCs. Comparison of the two approaches shows that the explainable neural\nnetwork approach can more effectively reveal known nucleotide substitutions\nassociated with VOCs, such as those in the spike gene regions. Our results\nsuggest that explainable neural networks for genomic sequences offer a\npromising alternative to the traditional genome wide analysis approaches.\n", "date": "2024-10-29", "categories": []}
{"id": "2501.00013", "title": "Relation-Aware Equivariant Graph Networks for Epitope-Unknown Antibody\n  Design and Specificity Optimization", "abstract": "  Antibodies are Y-shaped proteins that protect the host by binding to specific\nantigens, and their binding is mainly determined by the Complementary\nDetermining Regions (CDRs) in the antibody. Despite the great progress made in\nCDR design, existing computational methods still encounter several challenges:\n1) poor capability of modeling complex CDRs with long sequences due to\ninsufficient contextual information; 2) conditioned on pre-given antigenic\nepitopes and their static interaction with the target antibody; 3) neglect of\nspecificity during antibody optimization leads to non-specific antibodies. In\nthis paper, we take into account a variety of node features, edge features, and\nedge relations to include more contextual and geometric information. We propose\na novel Relation-Aware Antibody Design (RAAD) framework, which dynamically\nmodels antigen-antibody interactions for co-designing the sequences and\nstructures of antigen-specific CDRs. Furthermore, we propose a new evaluation\nmetric to better measure antibody specificity and develop a contrasting\nspecificity-enhancing constraint to optimize the specificity of antibodies.\nExtensive experiments have demonstrated the superior capability of RAAD in\nterms of antibody modeling, generation, and optimization across different CDR\ntypes, sequence lengths, pre-training strategies, and input contexts.\n", "date": "2024-12-14", "categories": []}
{"id": "2412.20616", "title": "Hilbert Curve Based Molecular Sequence Analysis", "abstract": "  Accurate molecular sequence analysis is a key task in the field of\nbioinformatics. To apply molecular sequence classification algorithms, we first\nneed to generate the appropriate representations of the sequences. Traditional\nnumeric sequence representation techniques are mostly based on sequence\nalignment that faces limitations in the form of lack of accuracy. Although\nseveral alignment-free techniques have also been introduced, their tabular data\nform results in low performance when used with Deep Learning (DL) models\ncompared to the competitive performance observed in the case of image-based\ndata. To find a solution to this problem and to make Deep Learning (DL) models\nfunction to their maximum potential while capturing the important spatial\ninformation in the sequence data, we propose a universal Hibert curve-based\nChaos Game Representation (CGR) method. This method is a transformative\nfunction that involves a novel Alphabetic index mapping technique used in\nconstructing Hilbert curve-based image representation from molecular sequences.\nOur method can be globally applied to any type of molecular sequence data. The\nHilbert curve-based image representations can be used as input to sophisticated\nvision DL models for sequence classification. The proposed method shows\npromising results as it outperforms current state-of-the-art methods by\nachieving a high accuracy of $94.5$\\% and an F1 score of $93.9\\%$ when tested\nwith the CNN model on the lung cancer dataset. This approach opens up a new\nhorizon for exploring molecular sequence analysis using image classification\nmethods.\n", "date": "2024-12-29", "categories": []}
{"id": "2308.07136", "title": "Pairing interacting protein sequences using masked language modeling", "abstract": "  Predicting which proteins interact together from amino-acid sequences is an\nimportant task. We develop a method to pair interacting protein sequences which\nleverages the power of protein language models trained on multiple sequence\nalignments, such as MSA Transformer and the EvoFormer module of AlphaFold. We\nformulate the problem of pairing interacting partners among the paralogs of two\nprotein families in a differentiable way. We introduce a method called DiffPALM\nthat solves it by exploiting the ability of MSA Transformer to fill in masked\namino acids in multiple sequence alignments using the surrounding context. MSA\nTransformer encodes coevolution between functionally or structurally coupled\namino acids. We show that it captures inter-chain coevolution, while it was\ntrained on single-chain data, which means that it can be used\nout-of-distribution. Relying on MSA Transformer without fine-tuning, DiffPALM\noutperforms existing coevolution-based pairing methods on difficult benchmarks\nof shallow multiple sequence alignments extracted from ubiquitous prokaryotic\nprotein datasets. It also outperforms an alternative method based on a\nstate-of-the-art protein language model trained on single sequences. Paired\nalignments of interacting protein sequences are a crucial ingredient of\nsupervised deep learning methods to predict the three-dimensional structure of\nprotein complexes. DiffPALM substantially improves the structure prediction of\nsome eukaryotic protein complexes by AlphaFold-Multimer, without significantly\ndeteriorating any of those we tested. It also achieves competitive performance\nwith using orthology-based pairing.\n", "date": "2023-08-14", "categories": []}
{"id": "2402.16901", "title": "FGBERT: Function-Driven Pre-trained Gene Language Model for Metagenomics", "abstract": "  Metagenomic data, comprising mixed multi-species genomes, are prevalent in\ndiverse environments like oceans and soils, significantly impacting human\nhealth and ecological functions. However, current research relies on K-mer,\nwhich limits the capture of structurally and functionally relevant gene\ncontexts. Moreover, these approaches struggle with encoding biologically\nmeaningful genes and fail to address the One-to-Many and Many-to-One\nrelationships inherent in metagenomic data. To overcome these challenges, we\nintroduce FGBERT, a novel metagenomic pre-trained model that employs a\nprotein-based gene representation as a context-aware and structure-relevant\ntokenizer. FGBERT incorporates Masked Gene Modeling (MGM) to enhance the\nunderstanding of inter-gene contextual relationships and Triplet Enhanced\nMetagenomic Contrastive Learning (TMC) to elucidate gene sequence-function\nrelationships. Pre-trained on over 100 million metagenomic sequences, FGBERT\ndemonstrates superior performance on metagenomic datasets at four levels,\nspanning gene, functional, bacterial, and environmental levels and ranging from\n1k to 213k input sequences. Case studies of ATP Synthase and Gene Operons\nhighlight FGBERT's capability for functional recognition and its biological\nrelevance in metagenomic research.\n", "date": "2024-02-24", "categories": []}
{"id": "2412.18827", "title": "PhyloGen: Language Model-Enhanced Phylogenetic Inference via Graph\n  Structure Generation", "abstract": "  Phylogenetic trees elucidate evolutionary relationships among species, but\nphylogenetic inference remains challenging due to the complexity of combining\ncontinuous (branch lengths) and discrete parameters (tree topology).\nTraditional Markov Chain Monte Carlo methods face slow convergence and\ncomputational burdens. Existing Variational Inference methods, which require\npre-generated topologies and typically treat tree structures and branch lengths\nindependently, may overlook critical sequence features, limiting their accuracy\nand flexibility. We propose PhyloGen, a novel method leveraging a pre-trained\ngenomic language model to generate and optimize phylogenetic trees without\ndependence on evolutionary models or aligned sequence constraints. PhyloGen\nviews phylogenetic inference as a conditionally constrained tree structure\ngeneration problem, jointly optimizing tree topology and branch lengths through\nthree core modules: (i) Feature Extraction, (ii) PhyloTree Construction, and\n(iii) PhyloTree Structure Modeling. Meanwhile, we introduce a Scoring Function\nto guide the model towards a more stable gradient descent. We demonstrate the\neffectiveness and robustness of PhyloGen on eight real-world benchmark\ndatasets. Visualization results confirm PhyloGen provides deeper insights into\nphylogenetic relationships.\n", "date": "2024-12-25", "categories": []}
{"id": "2412.19589", "title": "ViDTA: Enhanced Drug-Target Affinity Prediction via Virtual Graph Nodes\n  and Attention-based Feature Fusion", "abstract": "  Drug-target interaction is fundamental in understanding how drugs affect\nbiological systems, and accurately predicting drug-target affinity (DTA) is\nvital for drug discovery. Recently, deep learning methods have emerged as a\nsignificant approach for estimating the binding strength between drugs and\ntarget proteins. However, existing methods simply utilize the drug's local\ninformation from molecular topology rather than global information.\nAdditionally, the features of drugs and proteins are usually fused with a\nsimple concatenation operation, limiting their effectiveness. To address these\nchallenges, we proposed ViDTA, an enhanced DTA prediction framework. We\nintroduce virtual nodes into the Graph Neural Network (GNN)-based drug feature\nextraction network, which acts as a global memory to exchange messages more\nefficiently. By incorporating virtual graph nodes, we seamlessly integrate\nlocal and global features of drug molecular structures, expanding the GNN's\nreceptive field. Additionally, we propose an attention-based linear feature\nfusion network for better capturing the interaction information between drugs\nand proteins. Experimental results evaluated on various benchmarks including\nDavis, Metz, and KIBA demonstrate that our proposed ViDTA outperforms the\nstate-of-the-art baselines.\n", "date": "2024-12-27", "categories": []}
{"id": "2408.17139", "title": "Flow Matching for Optimal Reaction Coordinates of Biomolecular System", "abstract": "  We present flow matching for reaction coordinates (FMRC), a novel deep\nlearning algorithm designed to identify optimal reaction coordinates (RC) in\nbiomolecular reversible dynamics. FMRC is based on the mathematical principles\nof lumpability and decomposability, which we reformulate into a conditional\nprobability framework for efficient data-driven optimization using deep\ngenerative models. While FMRC does not explicitly learn the well-established\ntransfer operator or its eigenfunctions, it can effectively encode the dynamics\nof leading eigenfunctions of the system transfer operator into its\nlow-dimensional RC space. We further quantitatively compare its performance\nwith several state-of-the-art algorithms by evaluating the quality of Markov\nstate models (MSM) constructed in their respective RC spaces, demonstrating the\nsuperiority of FMRC in three increasingly complex biomolecular systems. In\naddition, we successfully demonstrated the efficacy of FMRC for bias deposition\nin the enhanced sampling of a simple model system. Finally, we discuss its\npotential applications in downstream applications such as enhanced sampling\nmethods and MSM construction.\n", "date": "2024-08-30", "categories": []}
{"id": "2412.16262", "title": "VirusT5: Harnessing Large Language Models to Predicting SARS-CoV-2\n  Evolution", "abstract": "  During a virus's evolution,various regions of the genome are subjected to\ndistinct levels of functional constraints.Combined with factors like codon bias\nand DNA repair efficiency,these constraints contribute to unique mutation\npatterns within the genome or a specific gene. In this project, we harnessed\nthe power of Large Language Models(LLMs) to predict the evolution of\nSARS-CoV-2. By treating the mutation process from one generation to the next as\na translation task, we trained a transformer model, called VirusT5, to capture\nthe mutation patterns underlying SARS-CoV-2 evolution. We evaluated the\nVirusT5's ability to detect these mutation patterns including its ability to\nidentify mutation hotspots and explored the potential of using VirusT5 to\npredict future virus variants. Our findings demonstrate the feasibility of\nusing a large language model to model viral evolution as a translation process.\nThis study establishes the groundbreaking concept of \"mutation-as-translation,\"\npaving the way for new methodologies and tools for combating virus threats\n", "date": "2024-12-20", "categories": []}
{"id": "2408.16975", "title": "Technical Report of HelixFold3 for Biomolecular Structure Prediction", "abstract": "  The AlphaFold series has transformed protein structure prediction with\nremarkable accuracy, often matching experimental methods. AlphaFold2,\nAlphaFold-Multimer, and the latest AlphaFold3 represent significant strides in\npredicting single protein chains, protein complexes, and biomolecular\nstructures. While AlphaFold2 and AlphaFold-Multimer are open-sourced,\nfacilitating rapid and reliable predictions, AlphaFold3 remains partially\naccessible through a limited online server and has not been open-sourced,\nrestricting further development. To address these challenges, the PaddleHelix\nteam is developing HelixFold3, aiming to replicate AlphaFold3's capabilities.\nLeveraging insights from previous models and extensive datasets, HelixFold3\nachieves accuracy comparable to AlphaFold3 in predicting the structures of the\nconventional ligands, nucleic acids, and proteins. The initial release of\nHelixFold3 is available as open source on GitHub for academic research,\npromising to advance biomolecular research and accelerate discoveries. The\nlatest version will be continuously updated on the HelixFold3 web server,\nproviding both interactive visualization and API access.\n", "date": "2024-08-30", "categories": []}
{"id": "2412.16664", "title": "Transformer-based toxin-protein interaction analysis prioritizes\n  airborne particulate matter components with potential adverse health effects", "abstract": "  Air pollution, particularly airborne particulate matter (PM), poses a\nsignificant threat to public health globally. It is crucial to comprehend the\nassociation between PM-associated toxic components and their cellular targets\nin humans to understand the mechanisms by which air pollution impacts health\nand to establish causal relationships between air pollution and public health\nconsequences. Although many studies have explored the impact of PM on human\nhealth, the understanding of the association between toxins and the associated\ntargets remain limited. Leveraging cutting-edge deep learning technologies, we\ndeveloped tipFormer (toxin-protein interaction prediction based on\ntransformer), a novel deep-learning tool for identifying toxic components\ncapable of penetrating human cells and instigating pathogenic biological\nactivities and signaling cascades. Experimental results show that tipFormer\neffectively captures interactions between proteins and toxic components. It\nincorporates dual pre-trained language models to encode protein sequences and\nchemicals. It employs a convolutional encoder to assimilate the sequential\nattributes of proteins and chemicals. It then introduces a learning module with\na cross-attention mechanism to decode and elucidate the multifaceted\ninteractions pivotal for the hotspots binding proteins and chemicals.\nExperimental results show that tipFormer effectively captures interactions\nbetween proteins and toxic components. This approach offers significant value\nto air quality and toxicology researchers by allowing high-throughput\nidentification and prioritization of hazards. It supports more targeted\nlaboratory studies and field measurements, ultimately enhancing our\nunderstanding of how air pollution impacts human health.\n", "date": "2024-12-21", "categories": []}
{"id": "2408.09048", "title": "mRNA2vec: mRNA Embedding with Language Model in the 5'UTR-CDS for mRNA\n  Design", "abstract": "  Messenger RNA (mRNA)-based vaccines are accelerating the discovery of new\ndrugs and revolutionizing the pharmaceutical industry. However, selecting\nparticular mRNA sequences for vaccines and therapeutics from extensive mRNA\nlibraries is costly. Effective mRNA therapeutics require carefully designed\nsequences with optimized expression levels and stability. This paper proposes a\nnovel contextual language model (LM)-based embedding method: mRNA2vec. In\ncontrast to existing mRNA embedding approaches, our method is based on the\nself-supervised teacher-student learning framework of data2vec. We jointly use\nthe 5' untranslated region (UTR) and coding sequence (CDS) region as the input\nsequences. We adapt our LM-based approach specifically to mRNA by 1)\nconsidering the importance of location on the mRNA sequence with probabilistic\nmasking, 2) using Minimum Free Energy (MFE) prediction and Secondary Structure\n(SS) classification as additional pretext tasks. mRNA2vec demonstrates\nsignificant improvements in translation efficiency (TE) and expression level\n(EL) prediction tasks in UTR compared to SOTA methods such as UTR-LM. It also\ngives a competitive performance in mRNA stability and protein production level\ntasks in CDS such as CodonBERT.\n", "date": "2024-08-16", "categories": []}
{"id": "2409.15092", "title": "M2OST: Many-to-one Regression for Predicting Spatial Transcriptomics\n  from Digital Pathology Images", "abstract": "  The advancement of Spatial Transcriptomics (ST) has facilitated the\nspatially-aware profiling of gene expressions based on histopathology images.\nAlthough ST data offers valuable insights into the micro-environment of tumors,\nits acquisition cost remains expensive. Therefore, directly predicting the ST\nexpressions from digital pathology images is desired. Current methods usually\nadopt existing regression backbones along with patch-sampling for this task,\nwhich ignores the inherent multi-scale information embedded in the pyramidal\ndata structure of digital pathology images, and wastes the inter-spot visual\ninformation crucial for accurate gene expression prediction. To address these\nlimitations, we propose M2OST, a many-to-one regression Transformer that can\naccommodate the hierarchical structure of the pathology images via a decoupled\nmulti-scale feature extractor. Unlike traditional models that are trained with\none-to-one image-label pairs, M2OST uses multiple images from different levels\nof the digital pathology image to jointly predict the gene expressions in their\ncommon corresponding spot. Built upon our many-to-one scheme, M2OST can be\neasily scaled to fit different numbers of inputs, and its network structure\ninherently incorporates nearby inter-spot features, enhancing regression\nperformance. We have tested M2OST on three public ST datasets and the\nexperimental results show that M2OST can achieve state-of-the-art performance\nwith fewer parameters and floating-point operations (FLOPs).\n", "date": "2024-09-23", "categories": []}
{"id": "2412.10743", "title": "NeuralPLexer3: Accurate Biomolecular Complex Structure Prediction with\n  Flow Models", "abstract": "  Structure determination is essential to a mechanistic understanding of\ndiseases and the development of novel therapeutics. Machine-learning-based\nstructure prediction methods have made significant advancements by\ncomputationally predicting protein and bioassembly structures from sequences\nand molecular topology alone. Despite substantial progress in the field,\nchallenges remain to deliver structure prediction models to real-world drug\ndiscovery. Here, we present NeuralPLexer3 -- a physics-inspired flow-based\ngenerative model that achieves state-of-the-art prediction accuracy on key\nbiomolecular interaction types and improves training and sampling efficiency\ncompared to its predecessors and alternative methodologies. Examined through\nnewly developed benchmarking strategies, NeuralPLexer3 excels in vital areas\nthat are crucial to structure-based drug design, such as physical validity and\nligand-induced conformational changes.\n", "date": "2024-12-14", "categories": []}
{"id": "2412.13223", "title": "Generative modeling of protein ensembles guided by crystallographic\n  electron densities", "abstract": "  Proteins are dynamic, adopting ensembles of conformations. The nature of this\nconformational heterogenity is imprinted in the raw electron density\nmeasurements obtained from X-ray crystallography experiments. Fitting an\nensemble of protein structures to these measurements is a challenging,\nill-posed inverse problem. We propose a non-i.i.d. ensemble guidance approach\nto solve this problem using existing protein structure generative models and\ndemonstrate that it accurately recovers complicated multi-modal alternate\nprotein backbone conformations observed in certain single crystal measurements.\n", "date": "2024-12-17", "categories": []}
{"id": "2412.13716", "title": "Model Decides How to Tokenize: Adaptive DNA Sequence Tokenization with\n  MxDNA", "abstract": "  Foundation models have made significant strides in understanding the genomic\nlanguage of DNA sequences. However, previous models typically adopt the\ntokenization methods designed for natural language, which are unsuitable for\nDNA sequences due to their unique characteristics. In addition, the optimal\napproach to tokenize DNA remains largely under-explored, and may not be\nintuitively understood by humans even if discovered. To address these\nchallenges, we introduce MxDNA, a novel framework where the model autonomously\nlearns an effective DNA tokenization strategy through gradient decent. MxDNA\nemploys a sparse Mixture of Convolution Experts coupled with a deformable\nconvolution to model the tokenization process, with the discontinuous,\noverlapping, and ambiguous nature of meaningful genomic segments explicitly\nconsidered. On Nucleotide Transformer Benchmarks and Genomic Benchmarks, MxDNA\ndemonstrates superior performance to existing methods with less pretraining\ndata and time, highlighting its effectiveness. Finally, we show that MxDNA\nlearns unique tokenization strategy distinct to those of previous methods and\ncaptures genomic functionalities at a token level during self-supervised\npretraining. Our MxDNA aims to provide a new perspective on DNA tokenization,\npotentially offering broad applications in various domains and yielding\nprofound insights.\n", "date": "2024-12-18", "categories": []}
{"id": "2412.08649", "title": "Multi-modal Representation Learning Enables Accurate Protein Function\n  Prediction in Low-Data Setting", "abstract": "  In this study, we propose HOPER (HOlistic ProtEin Representation), a novel\nmultimodal learning framework designed to enhance protein function prediction\n(PFP) in low-data settings. The challenge of predicting protein functions is\ncompounded by the limited availability of labeled data. Traditional machine\nlearning models already struggle in such cases, and while deep learning models\nexcel with abundant data, they also face difficulties when data is scarce.\nHOPER addresses this issue by integrating three distinct modalities - protein\nsequences, biomedical text, and protein-protein interaction (PPI) networks - to\ncreate a comprehensive protein representation. The model utilizes autoencoders\nto generate holistic embeddings, which are then employed for PFP tasks using\ntransfer learning. HOPER outperforms existing methods on a benchmark dataset\nacross all Gene Ontology categories, i.e., molecular function, biological\nprocess, and cellular component. Additionally, we demonstrate its practical\nutility by identifying new immune-escape proteins in lung adenocarcinoma,\noffering insights into potential therapeutic targets. Our results highlight the\neffectiveness of multimodal representation learning for overcoming data\nlimitations in biological research, potentially enabling more accurate and\nscalable protein function prediction. HOPER source code and datasets are\navailable at https://github.com/kansil/HOPER\n", "date": "2024-11-22", "categories": []}
{"id": "2412.12101", "title": "InterPLM: Discovering Interpretable Features in Protein Language Models\n  via Sparse Autoencoders", "abstract": "  Protein language models (PLMs) have demonstrated remarkable success in\nprotein modeling and design, yet their internal mechanisms for predicting\nstructure and function remain poorly understood. Here we present a systematic\napproach to extract and analyze interpretable features from PLMs using sparse\nautoencoders (SAEs). By training SAEs on embeddings from the PLM ESM-2, we\nidentify up to 2,548 human-interpretable latent features per layer that\nstrongly correlate with up to 143 known biological concepts such as binding\nsites, structural motifs, and functional domains. In contrast, examining\nindividual neurons in ESM-2 reveals up to 46 neurons per layer with clear\nconceptual alignment across 15 known concepts, suggesting that PLMs represent\nmost concepts in superposition. Beyond capturing known annotations, we show\nthat ESM-2 learns coherent concepts that do not map onto existing annotations\nand propose a pipeline using language models to automatically interpret novel\nlatent features learned by the SAEs. As practical applications, we demonstrate\nhow these latent features can fill in missing annotations in protein databases\nand enable targeted steering of protein sequence generation. Our results\ndemonstrate that PLMs encode rich, interpretable representations of protein\nbiology and we propose a systematic framework to extract and analyze these\nlatent features. In the process, we recover both known biology and potentially\nnew protein motifs. As community resources, we introduce InterPLM\n(interPLM.ai), an interactive visualization platform for exploring and\nanalyzing learned PLM features, and release code for training and analysis at\ngithub.com/ElanaPearl/interPLM.\n", "date": "2024-11-13", "categories": []}
{"id": "2412.12214", "title": "DLSOM: A Deep learning-based strategy for liver cancer subtyping", "abstract": "  Liver cancer is a leading cause of cancer-related mortality worldwide, with\nits high genetic heterogeneity complicating diagnosis and treatment. This study\nintroduces DLSOM, a deep learning framework utilizing stacked autoencoders to\nanalyze the complete somatic mutation landscape of 1,139 liver cancer samples,\ncovering 20,356 protein-coding genes. By transforming high-dimensional mutation\ndata into three low-dimensional features, DLSOM enables robust clustering and\nidentifies five distinct liver cancer subtypes with unique mutational,\nfunctional, and biological profiles. Subtypes SC1 and SC2 exhibit higher\nmutational loads, while SC3 has the lowest, reflecting mutational\nheterogeneity. Novel and COSMIC-associated mutational signatures reveal\nsubtype-specific molecular mechanisms, including links to hypermutation and\nchemotherapy resistance. Functional analyses further highlight the biological\nrelevance of each subtype. This comprehensive framework advances precision\nmedicine in liver cancer by enabling the development of subtype-specific\ndiagnostics, biomarkers, and therapies, showcasing the potential of deep\nlearning in addressing cancer complexity.\n", "date": "2024-12-15", "categories": []}
{"id": "2412.10411", "title": "Pre-trained protein language model for codon optimization", "abstract": "  Motivation: Codon optimization of Open Reading Frame (ORF) sequences is\nessential for enhancing mRNA stability and expression in applications like mRNA\nvaccines, where codon choice can significantly impact protein yield which\ndirectly impacts immune strength. In this work, we investigate the use of a\npre-trained protein language model (PPLM) for getting a rich representation of\namino acids which could be utilized for codon optimization. This leaves us with\na simpler fine-tuning task over PPLM in optimizing ORF sequences.\n  Results: The ORFs generated by our proposed models outperformed their natural\ncounterparts encoding the same proteins on computational metrics for stability\nand expression. They also demonstrated enhanced performance against the\nbenchmark ORFs used in mRNA vaccines for the SARS-CoV-2 viral spike protein and\nthe varicella-zoster virus (VZV). These results highlight the potential of\nadapting PPLM for designing ORFs tailored to encode target antigens in mRNA\nvaccines.\n", "date": "2024-12-08", "categories": []}
{"id": "2412.11084", "title": "BarcodeMamba: State Space Models for Biodiversity Analysis", "abstract": "  DNA barcodes are crucial in biodiversity analysis for building automatic\nidentification systems that recognize known species and discover unseen\nspecies. Unlike human genome modeling, barcode-based invertebrate\nidentification poses challenges in the vast diversity of species and taxonomic\ncomplexity. Among Transformer-based foundation models, BarcodeBERT excelled in\nspecies-level identification of invertebrates, highlighting the effectiveness\nof self-supervised pretraining on barcode-specific datasets. Recently,\nstructured state space models (SSMs) have emerged, with a time complexity that\nscales sub-quadratically with the context length. SSMs provide an efficient\nparameterization of sequence modeling relative to attention-based\narchitectures. Given the success of Mamba and Mamba-2 in natural language, we\ndesigned BarcodeMamba, a performant and efficient foundation model for DNA\nbarcodes in biodiversity analysis. We conducted a comprehensive ablation study\non the impacts of self-supervised training and tokenization methods, and\ncompared both versions of Mamba layers in terms of expressiveness and their\ncapacity to identify \"unseen\" species held back from training. Our study shows\nthat BarcodeMamba has better performance than BarcodeBERT even when using only\n8.3% as many parameters, and improves accuracy to 99.2% on species-level\naccuracy in linear probing without fine-tuning for \"seen\" species. In our\nscaling study, BarcodeMamba with 63.6% of BarcodeBERT's parameters achieved\n70.2% genus-level accuracy in 1-nearest neighbor (1-NN) probing for unseen\nspecies. The code repository to reproduce our experiments is available at\nhttps://github.com/bioscan-ml/BarcodeMamba.\n", "date": "2024-12-15", "categories": []}
{"id": "2407.11242", "title": "Bridging Sequence-Structure Alignment in RNA Foundation Models", "abstract": "  The alignment between RNA sequences and structures in foundation models (FMs)\nhas yet to be thoroughly investigated. Existing FMs have struggled to establish\nsequence-structure alignment, hindering the free flow of genomic information\nbetween RNA sequences and structures. In this study, we introduce OmniGenome,\nan RNA FM trained to align RNA sequences with respect to secondary structures\nbased on structure-contextualised modelling. The alignment enables free and\nbidirectional mappings between sequences and structures by utilising the\nflexible RNA modelling paradigm that supports versatile input and output\nmodalities, i.e., sequence and/or structure as input/output. We implement RNA\ndesign and zero-shot secondary structure prediction as case studies to evaluate\nthe Seq2Str and Str2Seq mapping capacity of OmniGenome. Results on the EternaV2\nbenchmark show that OmniGenome solved 74% of puzzles, whereas existing FMs only\nsolved up to 3% of the puzzles due to the oversight of sequence-structure\nalignment. We leverage four comprehensive in-silico genome modelling benchmarks\nto evaluate performance across a diverse set of genome downstream tasks, where\nthe results show that OmniGenome achieves state-of-the-art performance on RNA\nand DNA benchmarks, even without any training on DNA genomes.\n", "date": "2024-07-15", "categories": []}
{"id": "2412.09661", "title": "Language model driven: a PROTAC generation pipeline with dual\n  constraints of structure and property", "abstract": "  The imperfect modeling of ternary complexes has limited the application of\ncomputer-aided drug discovery tools in PROTAC research and development. In this\nstudy, an AI-assisted approach for PROTAC molecule design pipeline named\nLM-PROTAC was developed, which stands for language model driven Proteolysis\nTargeting Chimera, by embedding a transformer-based generative model with dual\nconstraints on structure and properties, referred to as the DCT. This study\nutilized the fragmentation representation of molecules and developed a language\nmodel driven pipeline. Firstly, a language model driven affinity model for\nprotein compounds to screen molecular fragments with high affinity for the\ntarget protein. Secondly, structural and physicochemical properties of these\nfragments were constrained during the generation process to meet specific\nscenario requirements. Finally, a two-round screening of the preliminary\ngenerated molecules using a multidimensional property prediction model to\ngenerate a batch of PROTAC molecules capable of degrading disease-relevant\ntarget proteins for validation in vitro experiments, thus achieving a complete\nsolution for AI-assisted PROTAC drug generation. Taking the tumor key target\nWnt3a as an example, the LM-PROTAC pipeline successfully generated PROTAC\nmolecules capable of inhibiting Wnt3a. The results show that DCT can\nefficiently generate PROTAC that targets and hydrolyses Wnt3a.\n", "date": "2024-12-12", "categories": []}
{"id": "2412.09803", "title": "deepNoC: A deep learning system to assign the number of contributors to\n  a short tandem repeat DNA profile", "abstract": "  A common task in forensic biology is to interpret and evaluate short tandem\nrepeat DNA profiles. The first step in these interpretations is to assign a\nnumber of contributors to the profiles, a task that is most often performed\nmanually by a scientist using their knowledge of DNA profile behaviour. Studies\nusing constructed DNA profiles have shown that as DNA profiles become more\ncomplex, and the number of DNA-donating individuals increases, the ability for\nscientists to assign the target number. There have been a number of machine\nlearning algorithms developed that seek to assign the number of contributors to\na DNA profile, however due to practical limitations in being able to generate\nDNA profiles in a laboratory, the algorithms have been based on summaries of\nthe available information. In this work we develop an analysis pipeline that\nsimulates the electrophoretic signal of an STR profile, allowing virtually\nunlimited, pre-labelled training material to be generated. We show that by\nsimulating 100 000 profiles and training a number of contributors estimation\ntool using a deep neural network architecture (in an algorithm named deepNoC)\nthat a high level of performance is achieved (89% for 1 to 10 contributors).\nThe trained network can then have fine-tuning training performed with only a\nfew hundred profiles in order to achieve the same accuracy within a specific\nlaboratory. We also build into deepNoC secondary outputs that provide a level\nof explainability to a user of algorithm, and show how they can be displayed in\nan intuitive manner.\n", "date": "2024-12-13", "categories": []}
{"id": "2412.09826", "title": "Precise Antigen-Antibody Structure Predictions Enhance Antibody\n  Development with HelixFold-Multimer", "abstract": "  The accurate prediction of antigen-antibody structures is essential for\nadvancing immunology and therapeutic development, as it helps elucidate\nmolecular interactions that underlie immune responses. Despite recent progress\nwith deep learning models like AlphaFold and RoseTTAFold, accurately modeling\nantigen-antibody complexes remains a challenge due to their unique evolutionary\ncharacteristics. HelixFold-Multimer, a specialized model developed for this\npurpose, builds on the framework of AlphaFold-Multimer and demonstrates\nimproved precision for antigen-antibody structures. HelixFold-Multimer not only\nsurpasses other models in accuracy but also provides essential insights into\nantibody development, enabling more precise identification of binding sites,\nimproved interaction prediction, and enhanced design of therapeutic antibodies.\nThese advances underscore HelixFold-Multimer's potential in supporting antibody\nresearch and therapeutic innovation.\n", "date": "2024-12-13", "categories": []}
{"id": "2412.10347", "title": "COMET: Benchmark for Comprehensive Biological Multi-omics Evaluation\n  Tasks and Language Models", "abstract": "  As key elements within the central dogma, DNA, RNA, and proteins play crucial\nroles in maintaining life by guaranteeing accurate genetic expression and\nimplementation. Although research on these molecules has profoundly impacted\nfields like medicine, agriculture, and industry, the diversity of machine\nlearning approaches-from traditional statistical methods to deep learning\nmodels and large language models-poses challenges for researchers in choosing\nthe most suitable models for specific tasks, especially for cross-omics and\nmulti-omics tasks due to the lack of comprehensive benchmarks. To address this,\nwe introduce the first comprehensive multi-omics benchmark COMET (Benchmark for\nBiological COmprehensive Multi-omics Evaluation Tasks and Language Models),\ndesigned to evaluate models across single-omics, cross-omics, and multi-omics\ntasks. First, we curate and develop a diverse collection of downstream tasks\nand datasets covering key structural and functional aspects in DNA, RNA, and\nproteins, including tasks that span multiple omics levels. Then, we evaluate\nexisting foundational language models for DNA, RNA, and proteins, as well as\nthe newly proposed multi-omics method, offering valuable insights into their\nperformance in integrating and analyzing data from different biological\nmodalities. This benchmark aims to define critical issues in multi-omics\nresearch and guide future directions, ultimately promoting advancements in\nunderstanding biological processes through integrated and different omics data\nanalysis.\n", "date": "2024-12-13", "categories": []}
{"id": "2404.16866", "title": "Annotation-guided Protein Design with Multi-Level Domain Alignment", "abstract": "  The core challenge of de novo protein design lies in creating proteins with\nspecific functions or properties, guided by certain conditions. Current models\nexplore to generate protein using structural and evolutionary guidance, which\nonly provide indirect conditions concerning functions and properties. However,\ntextual annotations of proteins, especially the annotations for protein\ndomains, which directly describe the protein's high-level functionalities,\nproperties, and their correlation with target amino acid sequences, remain\nunexplored in the context of protein design tasks. In this paper, we propose\nProtein-Annotation Alignment Generation, PAAG, a multi-modality protein design\nframework that integrates the textual annotations extracted from protein\ndatabase for controllable generation in sequence space. Specifically, within a\nmulti-level alignment module, PAAG can explicitly generate proteins containing\nspecific domains conditioned on the corresponding domain annotations, and can\neven design novel proteins with flexible combinations of different kinds of\nannotations. Our experimental results underscore the superiority of the aligned\nprotein representations from PAAG over 7 prediction tasks. Furthermore, PAAG\ndemonstrates a significant increase in generation success rate (24.7% vs 4.7%\nin zinc finger, and 54.3% vs 22.0% in the immunoglobulin domain) in comparison\nto the existing model. We anticipate that PAAG will broaden the horizons of\nprotein design by leveraging the knowledge from between textual annotation and\nproteins.\n", "date": "2024-04-18", "categories": []}
{"id": "2406.10391", "title": "BEACON: Benchmark for Comprehensive RNA Tasks and Language Models", "abstract": "  RNA plays a pivotal role in translating genetic instructions into functional\noutcomes, underscoring its importance in biological processes and disease\nmechanisms. Despite the emergence of numerous deep learning approaches for RNA,\nparticularly universal RNA language models, there remains a significant lack of\nstandardized benchmarks to assess the effectiveness of these methods. In this\nstudy, we introduce the first comprehensive RNA benchmark BEACON\n(\\textbf{BE}nchm\\textbf{A}rk for \\textbf{CO}mprehensive R\\textbf{N}A Task and\nLanguage Models). First, BEACON comprises 13 distinct tasks derived from\nextensive previous work covering structural analysis, functional studies, and\nengineering applications, enabling a comprehensive assessment of the\nperformance of methods on various RNA understanding tasks. Second, we examine a\nrange of models, including traditional approaches like CNNs, as well as\nadvanced RNA foundation models based on language models, offering valuable\ninsights into the task-specific performances of these models. Third, we\ninvestigate the vital RNA language model components from the tokenizer and\npositional encoding aspects. Notably, our findings emphasize the superiority of\nsingle nucleotide tokenization and the effectiveness of Attention with Linear\nBiases (ALiBi) over traditional positional encoding methods. Based on these\ninsights, a simple yet strong baseline called BEACON-B is proposed, which can\nachieve outstanding performance with limited data and computational resources.\nThe datasets and source code of our benchmark are available at\nhttps://github.com/terry-r123/RNABenchmark.\n", "date": "2024-06-14", "categories": []}
{"id": "2412.08751", "title": "Sampling-based Continuous Optimization with Coupled Variables for RNA\n  Design", "abstract": "  The task of RNA design given a target structure aims to find a sequence that\ncan fold into that structure. It is a computationally hard problem where some\nversion(s) have been proven to be NP-hard. As a result, heuristic methods such\nas local search have been popular for this task, but by only exploring a fixed\nnumber of candidates. They can not keep up with the exponential growth of the\ndesign space, and often perform poorly on longer and harder-to-design\nstructures. We instead formulate these discrete problems as continuous\noptimization, which starts with a distribution over all possible candidate\nsequences, and uses gradient descent to improve the expectation of an objective\nfunction. We define novel distributions based on coupled variables to rule out\ninvalid sequences given the target structure and to model the correlation\nbetween nucleotides. To make it universally applicable to any objective\nfunction, we use sampling to approximate the expected objective function, to\nestimate the gradient, and to select the final candidate. Compared to the\nstate-of-the-art methods, our work consistently outperforms them in key metrics\nsuch as Boltzmann probability, ensemble defect, and energy gap, especially on\nlong and hard-to-design puzzles in the Eterna100 benchmark. Our code is\navailable at: http://github.com/weiyutang1010/ncrna_design.\n", "date": "2024-12-11", "categories": []}
{"id": "2210.17401", "title": "Towards a Better Model with Dual Transformer for Drug Response\n  Prediction", "abstract": "  GNN-based methods have achieved excellent results as a mainstream task in\ndrug response prediction tasks in recent years. Traditional GNN methods use\nonly the atoms in a drug molecule as nodes to obtain the representation of the\nmolecular graph through node information passing, whereas the method using the\ntransformer can only extract information about the nodes. However, the covalent\nbonding and chirality of a drug molecule have a great influence on the\npharmacological properties of the molecule, and these information are implied\nin the chemical bonds formed by the edges between the atoms. In addition, CNN\nmethods for modelling cell lines genomics sequences can only perceive local\nrather than global information about the sequence. In order to solve the above\nproblems, we propose the decoupled dual transformer structure with edge\nembedded for drug respond prediction (TransEDRP), which is used for the\nrepresentation of cell line genomics and drug respectively. For the drug\nbranch, we encoded the chemical bond information within the molecule as the\nembedding of the edge in the molecular graph, extracted the global structural\nand biochemical information of the drug molecule using graph transformer. For\nthe branch of cell lines genomics, we use the multi-headed attention mechanism\nto globally represent the genomics sequence. Finally, the drug and genomics\nbranches are fused to predict IC50 values through the transformer layer and the\nfully connected layer, which two branches are different modalities. Extensive\nexperiments have shown that our method is better than the current mainstream\napproach in all evaluation indicators.\n", "date": "2022-10-23", "categories": []}
{"id": "2405.20313", "title": "Sequence-Augmented SE(3)-Flow Matching For Conditional Protein Backbone\n  Generation", "abstract": "  Proteins are essential for almost all biological processes and derive their\ndiverse functions from complex 3D structures, which are in turn determined by\ntheir amino acid sequences. In this paper, we exploit the rich biological\ninductive bias of amino acid sequences and introduce FoldFlow-2, a novel\nsequence-conditioned SE(3)-equivariant flow matching model for protein\nstructure generation. FoldFlow-2 presents substantial new architectural\nfeatures over the previous FoldFlow family of models including a protein large\nlanguage model to encode sequence, a new multi-modal fusion trunk that combines\nstructure and sequence representations, and a geometric transformer based\ndecoder. To increase diversity and novelty of generated samples -- crucial for\nde-novo drug design -- we train FoldFlow-2 at scale on a new dataset that is an\norder of magnitude larger than PDB datasets of prior works, containing both\nknown proteins in PDB and high-quality synthetic structures achieved through\nfiltering. We further demonstrate the ability to align FoldFlow-2 to arbitrary\nrewards, e.g. increasing secondary structures diversity, by introducing a\nReinforced Finetuning (ReFT) objective. We empirically observe that FoldFlow-2\noutperforms previous state-of-the-art protein structure-based generative\nmodels, improving over RFDiffusion in terms of unconditional generation across\nall metrics including designability, diversity, and novelty across all protein\nlengths, as well as exhibiting generalization on the task of equilibrium\nconformation sampling. Finally, we demonstrate that a fine-tuned FoldFlow-2\nmakes progress on challenging conditional design tasks such as designing\nscaffolds for the VHH nanobody.\n", "date": "2024-05-30", "categories": []}
{"id": "2411.06798", "title": "LA4SR: illuminating the dark proteome with generative AI", "abstract": "  AI language models (LMs) show promise for biological sequence analysis. We\nre-engineered open-source LMs (GPT-2, BLOOM, DistilRoBERTa, ELECTRA, and Mamba,\nranging from 70M to 12B parameters) for microbial sequence classification. The\nmodels achieved F1 scores up to 95 and operated 16,580x faster and at 2.9x the\nrecall of BLASTP. They effectively classified the algal dark proteome -\nuncharacterized proteins comprising about 65% of total proteins - validated on\nnew data including a new, complete Hi-C/Pacbio Chlamydomonas genome. Larger\n(>1B) LA4SR models reached high accuracy (F1 > 86) when trained on less than 2%\nof available data, rapidly achieving strong generalization capacity. High\naccuracy was achieved when training data had intact or scrambled terminal\ninformation, demonstrating robust generalization to incomplete sequences.\nFinally, we provide custom AI explainability software tools for attributing\namino acid patterns to AI generative processes and interpret their outputs in\nevolutionary and biophysical contexts.\n", "date": "2024-11-11", "categories": []}
{"id": "2411.12010", "title": "Active learning for efficient discovery of optimal gene combinations in\n  the combinatorial perturbation space", "abstract": "  The advancement of novel combinatorial CRISPR screening technologies enables\nthe identification of synergistic gene combinations on a large scale. This is\ncrucial for developing novel and effective combination therapies, but the\ncombinatorial space makes exhaustive experimentation infeasible. We introduce\nNAIAD, an active learning framework that efficiently discovers optimal gene\npairs capable of driving cells toward desired cellular phenotypes. NAIAD\nleverages single-gene perturbation effects and adaptive gene embeddings that\nscale with the training data size, mitigating overfitting in small-sample\nlearning while capturing complex gene interactions as more data is collected.\nEvaluated on four CRISPR combinatorial perturbation datasets totaling over\n350,000 genetic interactions, NAIAD, trained on small datasets, outperforms\nexisting models by up to 40\\% relative to the second-best. NAIAD's\nrecommendation system prioritizes gene pairs with the maximum predicted\neffects, resulting in the highest marginal gain in each AI-experiment round and\naccelerating discovery with fewer CRISPR experimental iterations. Our NAIAD\nframework (https://github.com/NeptuneBio/NAIAD) improves the identification of\nnovel, effective gene combinations, enabling more efficient CRISPR library\ndesign and offering promising applications in genomics research and therapeutic\ndevelopment.\n", "date": "2024-11-18", "categories": []}
{"id": "2412.07778", "title": "MIN: Multi-channel Interaction Network for Drug-Target Interaction with\n  Protein Distillation", "abstract": "  Traditional drug discovery processes are both time-consuming and require\nextensive professional expertise. With the accumulation of drug-target\ninteraction (DTI) data from experimental studies, leveraging modern\nmachine-learning techniques to discern patterns between drugs and target\nproteins has become increasingly feasible. In this paper, we introduce the\nMulti-channel Interaction Network (MIN), a novel framework designed to predict\nDTIs through two primary components: a representation learning module and a\nmulti-channel interaction module. The representation learning module features a\nC-Score Predictor-assisted screening mechanism, which selects critical residues\nto enhance prediction accuracy and reduce noise. The multi-channel interaction\nmodule incorporates a structure-agnostic channel, a structure-aware channel,\nand an extended-mixture channel, facilitating the identification of interaction\npatterns at various levels for optimal complementarity. Additionally,\ncontrastive learning is utilized to harmonize the representations of diverse\ndata types. Our experimental evaluations on public datasets demonstrate that\nMIN surpasses other strong DTI prediction methods. Furthermore, the case study\nreveals a high overlap between the residues selected by the C-Score Predictor\nand those in actual binding pockets, underscoring MIN's explainability\ncapability. These findings affirm that MIN is not only a potent tool for DTI\nprediction but also offers fresh insights into the prediction of protein\nbinding sites.\n", "date": "2024-11-23", "categories": []}
{"id": "2412.06993", "title": "Toward AI-Driven Digital Organism: Multiscale Foundation Models for\n  Predicting, Simulating and Programming Biology at All Levels", "abstract": "  We present an approach of using AI to model and simulate biology and life.\nWhy is it important? Because at the core of medicine, pharmacy, public health,\nlongevity, agriculture and food security, environmental protection, and clean\nenergy, it is biology at work. Biology in the physical world is too complex to\nmanipulate and always expensive and risky to tamper with. In this perspective,\nwe layout an engineering viable approach to address this challenge by\nconstructing an AI-Driven Digital Organism (AIDO), a system of integrated\nmultiscale foundation models, in a modular, connectable, and holistic fashion\nto reflect biological scales, connectedness, and complexities. An AIDO opens up\na safe, affordable and high-throughput alternative platform for predicting,\nsimulating and programming biology at all levels from molecules to cells to\nindividuals. We envision that an AIDO is poised to trigger a new wave of\nbetter-guided wet-lab experimentation and better-informed first-principle\nreasoning, which can eventually help us better decode and improve life.\n", "date": "2024-12-09", "categories": []}
{"id": "2412.07763", "title": "Bayesian Optimization of Antibodies Informed by a Generative Model of\n  Evolving Sequences", "abstract": "  To build effective therapeutics, biologists iteratively mutate antibody\nsequences to improve binding and stability. Proposed mutations can be informed\nby previous measurements or by learning from large antibody databases to\npredict only typical antibodies. Unfortunately, the space of typical antibodies\nis enormous to search, and experiments often fail to find suitable antibodies\non a budget. We introduce Clone-informed Bayesian Optimization (CloneBO), a\nBayesian optimization procedure that efficiently optimizes antibodies in the\nlab by teaching a generative model how our immune system optimizes antibodies.\nOur immune system makes antibodies by iteratively evolving specific portions of\ntheir sequences to bind their target strongly and stably, resulting in a set of\nrelated, evolving sequences known as a clonal family. We train a large language\nmodel, CloneLM, on hundreds of thousands of clonal families and use it to\ndesign sequences with mutations that are most likely to optimize an antibody\nwithin the human immune system. We propose to guide our designs to fit previous\nmeasurements with a twisted sequential Monte Carlo procedure. We show that\nCloneBO optimizes antibodies substantially more efficiently than previous\nmethods in realistic in silico experiments and designs stronger and more stable\nbinders in in vitro wet lab experiments.\n", "date": "2024-12-10", "categories": []}
{"id": "2401.06199", "title": "xTrimoPGLM: Unified 100B-Scale Pre-trained Transformer for Deciphering\n  the Language of Protein", "abstract": "  Protein language models have shown remarkable success in learning biological\ninformation from protein sequences. However, most existing models are limited\nby either autoencoding or autoregressive pre-training objectives, which makes\nthem struggle to handle protein understanding and generation tasks\nconcurrently. We propose a unified protein language model, xTrimoPGLM, to\naddress these two types of tasks simultaneously through an innovative\npre-training framework. Our key technical contribution is an exploration of the\ncompatibility and the potential for joint optimization of the two types of\nobjectives, which has led to a strategy for training xTrimoPGLM at an\nunprecedented scale of 100 billion parameters and 1 trillion training tokens.\nOur extensive experiments reveal that 1) xTrimoPGLM significantly outperforms\nother advanced baselines in 18 protein understanding benchmarks across four\ncategories. The model also facilitates an atomic-resolution view of protein\nstructures, leading to an advanced 3D structural prediction model that\nsurpasses existing language model-based tools. 2) xTrimoPGLM not only can\ngenerate de novo protein sequences following the principles of natural ones,\nbut also can perform programmable generation after supervised fine-tuning (SFT)\non curated sequences. These results highlight the substantial capability and\nversatility of xTrimoPGLM in understanding and generating protein sequences,\ncontributing to the evolving landscape of foundation models in protein science.\n", "date": "2024-01-11", "categories": []}
{"id": "2405.06642", "title": "PPFlow: Target-aware Peptide Design with Torsional Flow Matching", "abstract": "  Therapeutic peptides have proven to have great pharmaceutical value and\npotential in recent decades. However, methods of AI-assisted peptide drug\ndiscovery are not fully explored. To fill the gap, we propose a target-aware\npeptide design method called \\textsc{PPFlow}, based on conditional flow\nmatching on torus manifolds, to model the internal geometries of torsion angles\nfor the peptide structure design. Besides, we establish a protein-peptide\nbinding dataset named PPBench2024 to fill the void of massive data for the task\nof structure-based peptide drug design and to allow the training of deep\nlearning methods. Extensive experiments show that PPFlow reaches\nstate-of-the-art performance in tasks of peptide drug generation and\noptimization in comparison with baseline models, and can be generalized to\nother tasks including docking and side-chain packing.\n", "date": "2024-03-05", "categories": []}
{"id": "2410.09667", "title": "EquiJump: Protein Dynamics Simulation via SO(3)-Equivariant Stochastic\n  Interpolants", "abstract": "  Mapping the conformational dynamics of proteins is crucial for elucidating\ntheir functional mechanisms. While Molecular Dynamics (MD) simulation enables\ndetailed time evolution of protein motion, its computational toll hinders its\nuse in practice. To address this challenge, multiple deep learning models for\nreproducing and accelerating MD have been proposed drawing on transport-based\ngenerative methods. However, existing work focuses on generation through\ntransport of samples from prior distributions, that can often be distant from\nthe data manifold. The recently proposed framework of stochastic interpolants,\ninstead, enables transport between arbitrary distribution endpoints. Building\nupon this work, we introduce EquiJump, a transferable SO(3)-equivariant model\nthat bridges all-atom protein dynamics simulation time steps directly. Our\napproach unifies diverse sampling methods and is benchmarked against existing\nmodels on trajectory data of fast folding proteins. EquiJump achieves\nstate-of-the-art results on dynamics simulation with a transferable model on\nall of the fast folding proteins.\n", "date": "2024-10-12", "categories": []}
{"id": "2412.04737", "title": "Generative Humanization for Therapeutic Antibodies", "abstract": "  Antibody therapies have been employed to address some of today's most\nchallenging diseases, but must meet many criteria during drug development\nbefore reaching a patient. Humanization is a sequence optimization strategy\nthat addresses one critical risk called immunogenicity - a patient's immune\nresponse to the drug - by making an antibody more \"human-like\" in the absence\nof a predictive lab-based test for immunogenicity. However, existing\nhumanization strategies generally yield very few humanized candidates, which\nmay have degraded biophysical properties or decreased drug efficacy. Here, we\nre-frame humanization as a conditional generative modeling task, where\nhumanizing mutations are sampled from a language model trained on human\nantibody data. We describe a sampling process that incorporates models of\ntherapeutic attributes, such as antigen binding affinity, to obtain candidate\nsequences that have both reduced immunogenicity risk and maintained or improved\ntherapeutic properties, allowing this algorithm to be readily embedded into an\niterative antibody optimization campaign. We demonstrate in silico and in lab\nvalidation that in real therapeutic programs our generative humanization method\nproduces diverse sets of antibodies that are both (1) highly-human and (2) have\nfavorable therapeutic properties, such as improved binding to target antigens.\n", "date": "2024-12-06", "categories": []}
{"id": "2309.11087", "title": "Embed-Search-Align: DNA Sequence Alignment using Transformer Models", "abstract": "  DNA sequence alignment involves assigning short DNA reads to the most\nprobable locations on an extensive reference genome. This process is crucial\nfor various genomic analyses, including variant calling, transcriptomics, and\nepigenomics. Conventional methods, refined over decades, tackle this challenge\nin 2 steps: genome indexing followed by efficient search to locate likely\npositions for given reads. Building on the success of Large Language Models in\nencoding text into embeddings, where the distance metric captures semantic\nsimilarity, recent efforts have explored whether the same Transformer\narchitecture can produce embeddings for DNA sequences. Such models have shown\nearly promise in classifying short DNA sequences, such as detecting\ncoding/non-coding regions, and enhancer, promoter sequences. However,\nperformance at sequence classification tasks does not translate to sequence\nalignment, where it is necessary to search across the genome to align each\nread, a significantly longer-range task. We bridge this gap by framing the\nSequence Alignment task for Transformer models as an \"Embed-Search-Align\" task.\nIn this framework, a novel Reference-Free DNA Embedding model generates\nembeddings of reads and reference fragments, which are projected into a shared\nvector space where the read-fragment distance is used as a surrogate for\nalignment. Technical contributions include: (1) Contrastive loss for\nself-supervised training of DNA sequence representations, facilitating rich\nreference-free, sequence-level embeddings, and (2) a DNA vector store to enable\nsearch across fragments on a global scale. DNA-ESA is 99% accurate when\naligning 250-length reads onto a human genome (3gb), rivaling conventional\nmethods such as Bowtie and BWA-Mem. DNA-ESA exceeds the performance of 6\nTransformer model baselines such as Nucleotide Transformer, Hyena-DNA, and\nshows task transfer across chromosomes and species.\n", "date": "2023-09-20", "categories": []}
{"id": "2411.06740", "title": "Dockformer: A transformer-based molecular docking paradigm for\n  large-scale virtual screening", "abstract": "  Molecular docking is a crucial step in drug development, which enables the\nvirtual screening of compound libraries to identify potential ligands that\ntarget proteins of interest. However, the computational complexity of\ntraditional docking models increases as the size of the compound library\nincreases. Recently, deep learning algorithms can provide data-driven research\nand development models to increase the speed of the docking process.\nUnfortunately, few models can achieve superior screening performance compared\nto that of traditional models. Therefore, a novel deep learning-based docking\napproach named Dockformer is introduced in this study. Dockformer leverages\nmultimodal information to capture the geometric topology and structural\nknowledge of molecules and can directly generate binding conformations with the\ncorresponding confidence measures in an end-to-end manner. The experimental\nresults show that Dockformer achieves success rates of 90.53% and 82.71% on the\nPDBbind core set and PoseBusters benchmarks, respectively, and more than a\n100-fold increase in the inference process speed, outperforming almost all\nstate-of-the-art docking methods. In addition, the ability of Dockformer to\nidentify the main protease inhibitors of coronaviruses is demonstrated in a\nreal-world virtual screening scenario. Considering its high docking accuracy\nand screening efficiency, Dockformer can be regarded as a powerful and robust\ntool in the field of drug design.\n", "date": "2024-11-11", "categories": []}
{"id": "2412.04011", "title": "A Note on Spectral Map", "abstract": "  In molecular dynamics (MD) simulations, transitions between states are often\nrare events due to energy barriers that exceed the thermal temperature. Because\nof their infrequent occurrence and the huge number of degrees of freedom in\nmolecular systems, understanding the physical properties that drive rare events\nis immensely difficult. A common approach to this problem is to propose a\ncollective variable (CV) that describes this process by a simplified\nrepresentation. However, choosing CVs is not easy, as it often relies on\nphysical intuition. Machine learning (ML) techniques provide a promising\napproach for effectively extracting optimal CVs from MD data. Here, we provide\na note on a recent unsupervised ML method called spectral map, which constructs\nCVs by maximizing the timescale separation between slow and fast variables in\nthe system.\n", "date": "2024-12-05", "categories": []}
{"id": "2412.04069", "title": "ProtDAT: A Unified Framework for Protein Sequence Design from Any\n  Protein Text Description", "abstract": "  Protein design has become a critical method in advancing significant\npotential for various applications such as drug development and enzyme\nengineering. However, protein design methods utilizing large language models\nwith solely pretraining and fine-tuning struggle to capture relationships in\nmulti-modal protein data. To address this, we propose ProtDAT, a de novo\nfine-grained framework capable of designing proteins from any descriptive\nprotein text input. ProtDAT builds upon the inherent characteristics of protein\ndata to unify sequences and text as a cohesive whole rather than separate\nentities. It leverages an innovative multi-modal cross-attention, integrating\nprotein sequences and textual information for a foundational level and seamless\nintegration. Experimental results demonstrate that ProtDAT achieves the\nstate-of-the-art performance in protein sequence generation, excelling in\nrationality, functionality, structural similarity, and validity. On 20,000\ntext-sequence pairs from Swiss-Prot, it improves pLDDT by 6%, TM-score by 0.26,\nand reduces RMSD by 1.2 {\\AA}, highlighting its potential to advance protein\ndesign.\n", "date": "2024-12-05", "categories": []}
{"id": "2412.04075", "title": "Does your model understand genes? A benchmark of gene properties for\n  biological and text models", "abstract": "  The application of deep learning methods, particularly foundation models, in\nbiological research has surged in recent years. These models can be text-based\nor trained on underlying biological data, especially omics data of various\ntypes. However, comparing the performance of these models consistently has\nproven to be a challenge due to differences in training data and downstream\ntasks. To tackle this problem, we developed an architecture-agnostic\nbenchmarking approach that, instead of evaluating the models directly,\nleverages entity representation vectors from each model and trains simple\npredictive models for each benchmarking task. This ensures that all types of\nmodels are evaluated using the same input and output types. Here we focus on\ngene properties collected from professionally curated bioinformatics databases.\nThese gene properties are categorized into five major groups: genomic\nproperties, regulatory functions, localization, biological processes, and\nprotein properties. Overall, we define hundreds of tasks based on these\ndatabases, which include binary, multi-label, and multi-class classification\ntasks. We apply these benchmark tasks to evaluate expression-based models,\nlarge language models, protein language models, DNA-based models, and\ntraditional baselines. Our findings suggest that text-based models and protein\nlanguage models generally outperform expression-based models in genomic\nproperties and regulatory functions tasks, whereas expression-based models\ndemonstrate superior performance in localization tasks. These results should\naid in the development of more informed artificial intelligence strategies for\nbiological understanding and therapeutic discovery. To ensure the\nreproducibility and transparency of our findings, we have made the source code\nand benchmark data publicly accessible for further investigation and expansion\nat github.com/BiomedSciAI/gene-benchmark.\n", "date": "2024-12-05", "categories": []}
{"id": "2410.08631", "title": "CryoFM: A Flow-based Foundation Model for Cryo-EM Densities", "abstract": "  Cryo-electron microscopy (cryo-EM) is a powerful technique in structural\nbiology and drug discovery, enabling the study of biomolecules at high\nresolution. Significant advancements by structural biologists using cryo-EM\nhave led to the production of over 38,626 protein density maps at various\nresolutions1. However, cryo-EM data processing algorithms have yet to fully\nbenefit from our knowledge of biomolecular density maps, with only a few recent\nmodels being data-driven but limited to specific tasks. In this study, we\npresent CryoFM, a foundation model designed as a generative model, learning the\ndistribution of high-quality density maps and generalizing effectively to\ndownstream tasks. Built on flow matching, CryoFM is trained to accurately\ncapture the prior distribution of biomolecular density maps. Furthermore, we\nintroduce a flow posterior sampling method that leverages CRYOFM as a flexible\nprior for several downstream tasks in cryo-EM and cryo-electron tomography\n(cryo-ET) without the need for fine-tuning, achieving state-of-the-art\nperformance on most tasks and demonstrating its potential as a foundational\nmodel for broader applications in these fields.\n", "date": "2024-10-11", "categories": []}
{"id": "2410.14946", "title": "DEL-Ranking: Ranking-Correction Denoising Framework for Elucidating\n  Molecular Affinities in DNA-Encoded Libraries", "abstract": "  DNA-encoded library (DEL) screening has revolutionized the detection of\nprotein-ligand interactions through read counts, enabling rapid exploration of\nvast chemical spaces. However, noise in read counts, stemming from nonspecific\ninteractions, can mislead this exploration process. We present DEL-Ranking, a\nnovel distribution-correction denoising framework that addresses these\nchallenges. Our approach introduces two key innovations: (1) a novel ranking\nloss that rectifies relative magnitude relationships between read counts,\nenabling the learning of causal features determining activity levels, and (2)\nan iterative algorithm employing self-training and consistency loss to\nestablish model coherence between activity label and read count predictions.\nFurthermore, we contribute three new DEL screening datasets, the first to\ncomprehensively include multi-dimensional molecular representations,\nprotein-ligand enrichment values, and their activity labels. These datasets\nmitigate data scarcity issues in AI-driven DEL screening research. Rigorous\nevaluation on diverse DEL datasets demonstrates DEL-Ranking's superior\nperformance across multiple correlation metrics, with significant improvements\nin binding affinity prediction accuracy. Our model exhibits zero-shot\ngeneralization ability across different protein targets and successfully\nidentifies potential motifs determining compound binding affinity. This work\nadvances DEL screening analysis and provides valuable resources for future\nresearch in this area.\n", "date": "2024-10-19", "categories": []}
{"id": "2412.02915", "title": "Single-Cell Omics Arena: A Benchmark Study for Large Language Models on\n  Cell Type Annotation Using Single-Cell Data", "abstract": "  Over the past decade, the revolution in single-cell sequencing has enabled\nthe simultaneous molecular profiling of various modalities across thousands of\nindividual cells, allowing scientists to investigate the diverse functions of\ncomplex tissues and uncover underlying disease mechanisms. Among all the\nanalytical steps, assigning individual cells to specific types is fundamental\nfor understanding cellular heterogeneity. However, this process is usually\nlabor-intensive and requires extensive expert knowledge. Recent advances in\nlarge language models (LLMs) have demonstrated their ability to efficiently\nprocess and synthesize vast corpora of text to automatically extract essential\nbiological knowledge, such as marker genes, potentially promoting more\nefficient and automated cell type annotations. To thoroughly evaluate the\ncapability of modern instruction-tuned LLMs in automating the cell type\nidentification process, we introduce SOAR, a comprehensive benchmarking study\nof LLMs for cell type annotation tasks in single-cell genomics. Specifically,\nwe assess the performance of 8 instruction-tuned LLMs across 11 datasets,\nspanning multiple cell types and species. Our study explores the potential of\nLLMs to accurately classify and annotate cell types in single-cell RNA\nsequencing (scRNA-seq) data, while extending their application to multiomics\ndata through cross-modality translation. Additionally, we evaluate the\neffectiveness of chain-of-thought (CoT) prompting techniques in generating\ndetailed biological insights during the annotation process. The results\ndemonstrate that LLMs can provide robust interpretations of single-cell data\nwithout requiring additional fine-tuning, advancing the automation of cell type\nannotation in genomics research.\n", "date": "2024-12-03", "categories": []}
{"id": "2212.07492", "title": "Machine Learning Coarse-Grained Potentials of Protein Thermodynamics", "abstract": "  A generalized understanding of protein dynamics is an unsolved scientific\nproblem, the solution of which is critical to the interpretation of the\nstructure-function relationships that govern essential biological processes.\nHere, we approach this problem by constructing coarse-grained molecular\npotentials based on artificial neural networks and grounded in statistical\nmechanics. For training, we build a unique dataset of unbiased all-atom\nmolecular dynamics simulations of approximately 9 ms for twelve different\nproteins with multiple secondary structure arrangements. The coarse-grained\nmodels are capable of accelerating the dynamics by more than three orders of\nmagnitude while preserving the thermodynamics of the systems. Coarse-grained\nsimulations identify relevant structural states in the ensemble with comparable\nenergetics to the all-atom systems. Furthermore, we show that a single\ncoarse-grained potential can integrate all twelve proteins and can capture\nexperimental structural features of mutated proteins. These results indicate\nthat machine learning coarse-grained potentials could provide a feasible\napproach to simulate and understand protein dynamics.\n", "date": "2022-12-14", "categories": []}
{"id": "2305.00386", "title": "Importance Weighted Expectation-Maximization for Protein Sequence Design", "abstract": "  Designing protein sequences with desired biological function is crucial in\nbiology and chemistry. Recent machine learning methods use a surrogate\nsequence-function model to replace the expensive wet-lab validation. How can we\nefficiently generate diverse and novel protein sequences with high fitness? In\nthis paper, we propose IsEM-Pro, an approach to generate protein sequences\ntowards a given fitness criterion. At its core, IsEM-Pro is a latent generative\nmodel, augmented by combinatorial structure features from a separately learned\nMarkov random fields (MRFs). We develop an Monte Carlo Expectation-Maximization\nmethod (MCEM) to learn the model. During inference, sampling from its latent\nspace enhances diversity while its MRFs features guide the exploration in high\nfitness regions. Experiments on eight protein sequence design tasks show that\nour IsEM-Pro outperforms the previous best methods by at least 55% on average\nfitness score and generates more diverse and novel protein sequences.\n", "date": "2023-04-30", "categories": []}
{"id": "2409.12816", "title": "Hierarchical Gradient-Based Genetic Sampling for Accurate Prediction of\n  Biological Oscillations", "abstract": "  Biological oscillations are periodic changes in various signaling processes\ncrucial for the proper functioning of living organisms. These oscillations are\nmodeled by ordinary differential equations, with coefficient variations leading\nto diverse periodic behaviors, typically measured by oscillatory frequencies.\nThis paper explores sampling techniques for neural networks to model the\nrelationship between system coefficients and oscillatory frequency. However,\nthe scarcity of oscillations in the vast coefficient space results in many\nsamples exhibiting non-periodic behaviors, and small coefficient changes near\noscillation boundaries can significantly alter oscillatory properties. This\nleads to non-oscillatory bias and boundary sensitivity, making accurate\npredictions difficult. While existing importance and uncertainty sampling\napproaches partially mitigate these challenges, they either fail to resolve the\nsensitivity problem or result in redundant sampling. To address these\nlimitations, we propose the Hierarchical Gradient-based Genetic Sampling (HGGS)\nframework, which improves the accuracy of neural network predictions for\nbiological oscillations. The first layer, Gradient-based Filtering, extracts\nsensitive oscillation boundaries and removes redundant non-oscillatory samples,\ncreating a balanced coarse dataset. The second layer, Multigrid Genetic\nSampling, utilizes residual information to refine these boundaries and explore\nnew high-residual regions, increasing data diversity for model training.\nExperimental results demonstrate that HGGS outperforms seven comparative\nsampling methods across four biological systems, highlighting its effectiveness\nin enhancing sampling and prediction accuracy.\n", "date": "2024-09-19", "categories": []}
{"id": "2309.04019", "title": "Evaluation of large language models for discovery of gene set function", "abstract": "  Gene set analysis is a mainstay of functional genomics, but it relies on\ncurated databases of gene functions that are incomplete. Here we evaluate five\nLarge Language Models (LLMs) for their ability to discover the common\nbiological functions represented by a gene set, substantiated by supporting\nrationale, citations and a confidence assessment. Benchmarking against\ncanonical gene sets from the Gene Ontology, GPT-4 confidently recovered the\ncurated name or a more general concept (73% of cases), while benchmarking\nagainst random gene sets correctly yielded zero confidence. Gemini-Pro and\nMixtral-Instruct showed ability in naming but were falsely confident for random\nsets, whereas Llama2-70b had poor performance overall. In gene sets derived\nfrom 'omics data, GPT-4 identified novel functions not reported by classical\nfunctional enrichment (32% of cases), which independent review indicated were\nlargely verifiable and not hallucinations. The ability to rapidly synthesize\ncommon gene functions positions LLMs as valuable 'omics assistants.\n", "date": "2023-09-07", "categories": []}
{"id": "2405.15158", "title": "ProtFAD: Introducing function-aware domains as implicit modality towards\n  protein function prediction", "abstract": "  Protein function prediction is currently achieved by encoding its sequence or\nstructure, where the sequence-to-function transcendence and high-quality\nstructural data scarcity lead to obvious performance bottlenecks. Protein\ndomains are \"building blocks\" of proteins that are functionally independent,\nand their combinations determine the diverse biological functions. However,\nmost existing studies have yet to thoroughly explore the intricate functional\ninformation contained in the protein domains. To fill this gap, we propose a\nsynergistic integration approach for a function-aware domain representation,\nand a domain-joint contrastive learning strategy to distinguish different\nprotein functions while aligning the modalities. Specifically, we align the\ndomain semantics with GO terms and text description to pre-train domain\nembeddings. Furthermore, we partition proteins into multiple sub-views based on\ncontinuous joint domains for contrastive training under the supervision of a\nnovel triplet InfoNCE loss. Our approach significantly and comprehensively\noutperforms the state-of-the-art methods on various benchmarks, and clearly\ndifferentiates proteins carrying distinct functions compared to the competitor.\nOur implementation is available at\nhttps://github.com/AI-HPC-Research-Team/ProtFAD.\n", "date": "2024-05-24", "categories": []}
{"id": "2412.00109", "title": "Deep Neural Network-Based Prediction of B-Cell Epitopes for SARS-CoV and\n  SARS-CoV-2: Enhancing Vaccine Design through Machine Learning", "abstract": "  The accurate prediction of B-cell epitopes is critical for guiding vaccine\ndevelopment against infectious diseases, including SARS and COVID-19. This\nstudy explores the use of a deep neural network (DNN) model to predict B-cell\nepitopes for SARS-CoVandSARS-CoV-2,leveraging a dataset that incorporates\nessential protein and peptide features. Traditional sequence-based methods\noften struggle with large, complex datasets, but deep learning offers promising\nimprovements in predictive accuracy. Our model employs regularization\ntechniques, such as dropout and early stopping, to enhance generalization,\nwhile also analyzing key features, including isoelectric point and aromaticity,\nthat influence epitope recognition. Results indicate an overall accuracy of 82%\nin predicting COVID-19 negative and positive cases, with room for improvement\nin detecting positive samples. This research demonstrates the applicability of\ndeep learning in epitope mapping, suggesting that such approaches can enhance\nthe speed and precision of vaccine design for emerging pathogens. Future work\ncould incorporate structural data and diverse viral strains to further refine\nprediction capabilities.\n", "date": "2024-11-28", "categories": []}
{"id": "2412.00807", "title": "Generative Model for Synthesizing Ionizable Lipids: A Monte Carlo Tree\n  Search Approach", "abstract": "  Ionizable lipids are essential in developing lipid nanoparticles (LNPs) for\neffective messenger RNA (mRNA) delivery. While traditional methods for\ndesigning new ionizable lipids are typically time-consuming, deep generative\nmodels have emerged as a powerful solution, significantly accelerating the\nmolecular discovery process. However, a practical challenge arises as the\nmolecular structures generated can often be difficult or infeasible to\nsynthesize. This project explores Monte Carlo tree search (MCTS)-based\ngenerative models for synthesizable ionizable lipids. Leveraging a\nsynthetically accessible lipid building block dataset and two specialized\npredictors to guide the search through chemical space, we introduce a policy\nnetwork guided MCTS generative model capable of producing new ionizable lipids\nwith available synthesis pathways.\n", "date": "2024-12-01", "categories": []}
{"id": "2411.16686", "title": "ProteinWeaver: A Divide-and-Assembly Approach for Protein Backbone\n  Design", "abstract": "  Nature creates diverse proteins through a 'divide and assembly' strategy.\nInspired by this idea, we introduce ProteinWeaver, a two-stage framework for\nprotein backbone design. Our method first generates individual protein domains\nand then employs an SE(3) diffusion model to flexibly assemble these domains. A\nkey challenge lies in the assembling step, given the complex and rugged nature\nof the inter-domain interaction landscape. To address this challenge, we employ\npreference alignment to discern complex relationships between structure and\ninteraction landscapes through comparative analysis of generated samples.\nComprehensive experiments demonstrate that ProteinWeaver: (1) generates\nhigh-quality, novel protein backbones through versatile domain assembly; (2)\noutperforms RFdiffusion, the current state-of-the-art in backbone design, by\n13\\% and 39\\% for long-chain proteins; (3) shows the potential for cooperative\nfunction design through illustrative case studies. To sum up, by introducing a\n`divide-and-assembly' paradigm, ProteinWeaver advances protein engineering and\nopens new avenues for functional protein design.\n", "date": "2024-11-08", "categories": []}
{"id": "2411.17795", "title": "Pan-protein Design Learning Enables Task-adaptive Generalization for\n  Low-resource Enzyme Design", "abstract": "  Computational protein design (CPD) offers transformative potential for\nbioengineering, but current deep CPD models, focused on universal domains,\nstruggle with function-specific designs. This work introduces a novel CPD\nparadigm tailored for functional design tasks, particularly for enzymes-a key\nprotein class often lacking specific application efficiency. To address\nstructural data scarcity, we present CrossDesign, a domain-adaptive framework\nthat leverages pretrained protein language models (PPLMs). By aligning protein\nstructures with sequences, CrossDesign transfers pretrained knowledge to\nstructure models, overcoming the limitations of limited structural data. The\nframework combines autoregressive (AR) and non-autoregressive (NAR) states in\nits encoder-decoder architecture, applying it to enzyme datasets and\npan-proteins. Experimental results highlight CrossDesign's superior performance\nand robustness, especially with out-of-domain enzymes. Additionally, the model\nexcels in fitness prediction when tested on large-scale mutation data,\nshowcasing its stability.\n", "date": "2024-11-26", "categories": []}
{"id": "2411.17798", "title": "DapPep: Domain Adaptive Peptide-agnostic Learning for Universal T-cell\n  Receptor-antigen Binding Affinity Prediction", "abstract": "  Identifying T-cell receptors (TCRs) that interact with antigenic peptides\nprovides the technical basis for developing vaccines and immunotherapies. The\nemergent deep learning methods excel at learning antigen binding patterns from\nknown TCRs but struggle with novel or sparsely represented antigens. However,\nbinding specificity for unseen antigens or exogenous peptides is critical. We\nintroduce a domain-adaptive peptide-agnostic learning framework DapPep for\nuniversal TCR-antigen binding affinity prediction to address this challenge.\nThe lightweight self-attention architecture combines a pre-trained protein\nlanguage model with an inner-loop self-supervised regime to enable robust\nTCR-peptide representations. Extensive experiments on various benchmarks\ndemonstrate that DapPep consistently outperforms existing tools, showcasing\nrobust generalization capability, especially for data-scarce settings and\nunseen peptides. Moreover, DapPep proves effective in challenging clinical\ntasks such as sorting reactive T cells in tumor neoantigen therapy and\nidentifying key positions in 3D structures.\n", "date": "2024-11-26", "categories": []}
{"id": "2411.18391", "title": "GeneQuery: A General QA-based Framework for Spatial Gene Expression\n  Predictions from Histology Images", "abstract": "  Gene expression profiling provides profound insights into molecular\nmechanisms, but its time-consuming and costly nature often presents significant\nchallenges. In contrast, whole-slide hematoxylin and eosin (H&E) stained\nhistological images are readily accessible and allow for detailed examinations\nof tissue structure and composition at the microscopic level. Recent\nadvancements have utilized these histological images to predict spatially\nresolved gene expression profiles. However, state-of-the-art works treat gene\nexpression prediction as a multi-output regression problem, where each gene is\nlearned independently with its own weights, failing to capture the shared\ndependencies and co-expression patterns between genes. Besides, existing works\ncan only predict gene expression values for genes seen during training,\nlimiting their ability to generalize to new, unseen genes.\n  To address the above limitations, this paper presents GeneQuery, which aims\nto solve this gene expression prediction task in a question-answering (QA)\nmanner for better generality and flexibility. Specifically, GeneQuery takes\ngene-related texts as queries and whole-slide images as contexts and then\npredicts the queried gene expression values. With such a transformation,\nGeneQuery can implicitly estimate the gene distribution by introducing the gene\nrandom variable. Besides, the proposed GeneQuery consists of two architecture\nimplementations, i.e., spot-aware GeneQuery for capturing patterns between\nimages and gene-aware GeneQuery for capturing patterns between genes.\nComprehensive experiments on spatial transcriptomics datasets show that the\nproposed GeneQuery outperforms existing state-of-the-art methods on known and\nunseen genes. More results also demonstrate that GeneQuery can potentially\nanalyze the tissue structure.\n", "date": "2024-11-27", "categories": []}
{"id": "2401.15122", "title": "A Multi-Grained Symmetric Differential Equation Model for Learning\n  Protein-Ligand Binding Dynamics", "abstract": "  In drug discovery, molecular dynamics (MD) simulation for protein-ligand\nbinding provides a powerful tool for predicting binding affinities, estimating\ntransport properties, and exploring pocket sites. There has been a long history\nof improving the efficiency of MD simulations through better numerical methods\nand, more recently, by utilizing machine learning (ML) methods. Yet, challenges\nremain, such as accurate modeling of extended-timescale simulations. To address\nthis issue, we propose NeuralMD, the first ML surrogate that can facilitate\nnumerical MD and provide accurate simulations in protein-ligand binding\ndynamics. We propose a principled approach that incorporates a novel\nphysics-informed multi-grained group symmetric framework. Specifically, we\npropose (1) the BindingNet model that satisfies group symmetry using vector\nframes and captures the multi-level protein-ligand interactions, and (2) an\naugmented neural differential equation solver that learns the trajectory under\nNewtonian mechanics. For the experiment, we design ten single-trajectory and\nthree multi-trajectory binding simulation tasks. We demonstrate the efficiency\nand effectiveness of NeuralMD, achieving over 1K$\\times$ speedup compared to\nstandard numerical MD simulations. NeuralMD also outperforms all other ML\napproaches, achieving up to 15$\\times$ reduction in reconstruction error and\n70% increase in validity. Additionally, we qualitatively illustrate that the\noscillations in the predicted trajectories align more closely with ground-truth\ndynamics than those of other machine-learning methods. We believe NeuralMD\npaves the foundation for a new research paradigm in simulating protein-ligand\ndynamics.\n", "date": "2024-01-26", "categories": []}
{"id": "2410.20688", "title": "Reprogramming Pretrained Target-Specific Diffusion Models for\n  Dual-Target Drug Design", "abstract": "  Dual-target therapeutic strategies have become a compelling approach and\nattracted significant attention due to various benefits, such as their\npotential in overcoming drug resistance in cancer therapy. Considering the\ntremendous success that deep generative models have achieved in structure-based\ndrug design in recent years, we formulate dual-target drug design as a\ngenerative task and curate a novel dataset of potential target pairs based on\nsynergistic drug combinations. We propose to design dual-target drugs with\ndiffusion models that are trained on single-target protein-ligand complex\npairs. Specifically, we align two pockets in 3D space with protein-ligand\nbinding priors and build two complex graphs with shared ligand nodes for\nSE(3)-equivariant composed message passing, based on which we derive a composed\ndrift in both 3D and categorical probability space in the generative process.\nOur algorithm can well transfer the knowledge gained in single-target\npretraining to dual-target scenarios in a zero-shot manner. We also repurpose\nlinker design methods as strong baselines for this task. Extensive experiments\ndemonstrate the effectiveness of our method compared with various baselines.\n", "date": "2024-10-28", "categories": []}
{"id": "2411.17154", "title": "Emergenet: A Digital Twin of Sequence Evolution for Scalable Emergence\n  Risk Assessment of Animal Influenza A Strains", "abstract": "  Despite having triggered devastating pandemics in the past, our ability to\nquantitatively assess the emergence potential of individual strains of animal\ninfluenza viruses remains limited. This study introduces Emergenet, a tool to\ninfer a digital twin of sequence evolution to chart how new variants might\nemerge in the wild. Our predictions based on Emergenets built only using\n220,151 Hemagglutinnin (HA) sequences consistently outperform WHO seasonal\nvaccine recommendations for H1N1/H3N2 subtypes over two decades (average\nmatch-improvement: 3.73 AAs, 28.40\\%), and are at par with state-of-the-art\napproaches that use more detailed phenotypic annotations. Finally, our\ngenerative models are used to scalably calculate the current odds of emergence\nof animal strains not yet in human circulation, which strongly correlates with\nCDC's expert-assessed Influenza Risk Assessment Tool (IRAT) scores (Pearson's\n$r = 0.721, p = 10^{-4}$). A minimum five orders of magnitude speedup over\nCDC's assessment (seconds vs months) then enabled us to analyze 6,354 animal\nstrains collected post-2020 to identify 35 strains with high emergence scores\n($> 7.7$). The Emergenet framework opens the door to preemptive pandemic\nmitigation through targeted inoculation of animal hosts before the first human\ninfection.\n", "date": "2024-11-26", "categories": []}
{"id": "2406.08638", "title": "Conditional Similarity Triplets Enable Covariate-Informed\n  Representations of Single-Cell Data", "abstract": "  Single-cell technologies enable comprehensive profiling of diverse immune\ncell-types through the measurement of multiple genes or proteins per individual\ncell. In order to translate immune signatures assayed from blood or tissue into\npowerful diagnostics, machine learning approaches are often employed to compute\nimmunological summaries or per-sample featurizations, which can be used as\ninputs to models for outcomes of interest. Current supervised learning\napproaches for computing per-sample representations are trained only to\naccurately predict a single outcome and do not take into account relevant\nadditional clinical features or covariates that are likely to also be measured\nfor each sample. Here, we introduce a novel approach for incorporating measured\ncovariates in optimizing model parameters to ultimately specify per-sample\nencodings that accurately affect both immune signatures and additional clinical\ninformation. Our introduced method CytoCoSet is a set-based encoding method for\nlearning per-sample featurizations, which formulates a loss function with an\nadditional triplet term penalizing samples with similar covariates from having\ndisparate embedding results in per-sample representations. Overall,\nincorporating clinical covariates enables the learning of encodings for each\nindividual sample that ultimately improve prediction of clinical outcome.\n", "date": "2024-06-12", "categories": []}
{"id": "2411.15180", "title": "Multi-layer matrix factorization for cancer subtyping using full and\n  partial multi-omics dataset", "abstract": "  Cancer, with its inherent heterogeneity, is commonly categorized into\ndistinct subtypes based on unique traits, cellular origins, and molecular\nmarkers specific to each type. However, current studies primarily rely on\ncomplete multi-omics datasets for predicting cancer subtypes, often overlooking\npredictive performance in cases where some omics data may be missing and\nneglecting implicit relationships across multiple layers of omics data\nintegration. This paper introduces Multi-Layer Matrix Factorization (MLMF), a\nnovel approach for cancer subtyping that employs multi-omics data clustering.\nMLMF initially processes multi-omics feature matrices by performing multi-layer\nlinear or nonlinear factorization, decomposing the original data into latent\nfeature representations unique to each omics type. These latent representations\nare subsequently fused into a consensus form, on which spectral clustering is\nperformed to determine subtypes. Additionally, MLMF incorporates a class\nindicator matrix to handle missing omics data, creating a unified framework\nthat can manage both complete and incomplete multi-omics data. Extensive\nexperiments conducted on 10 multi-omics cancer datasets, both complete and with\nmissing values, demonstrate that MLMF achieves results that are comparable to\nor surpass the performance of several state-of-the-art approaches.\n", "date": "2024-11-18", "categories": []}
{"id": "2411.15208", "title": "M2oE: Multimodal Collaborative Expert Peptide Model", "abstract": "  Peptides are biomolecules comprised of amino acids that play an important\nrole in our body. In recent years, peptides have received extensive attention\nin drug design and synthesis, and peptide prediction tasks help us better\nsearch for functional peptides. Typically, we use the primary sequence and\nstructural information of peptides for model encoding. However, recent studies\nhave focused more on single-modal information (structure or sequence) for\nprediction without multi-modal approaches. We found that single-modal models\nare not good at handling datasets with less information in that particular\nmodality. Therefore, this paper proposes the M2oE multi-modal collaborative\nexpert peptide model. Based on previous work, by integrating sequence and\nspatial structural information, employing expert model and Cross-Attention\nMechanism, the model's capabilities are balanced and improved. Experimental\nresults indicate that the M2oE model performs excellently in complex task\npredictions.\n", "date": "2024-11-20", "categories": []}
{"id": "2411.15215", "title": "S$^2$ALM: Sequence-Structure Pre-trained Large Language Model for\n  Comprehensive Antibody Representation Learning", "abstract": "  Antibodies safeguard our health through their precise and potent binding to\nspecific antigens, demonstrating promising therapeutic efficacy in the\ntreatment of numerous diseases, including COVID-19. Recent advancements in\nbiomedical language models have shown the great potential to interpret complex\nbiological structures and functions. However, existing antibody specific models\nhave a notable limitation that they lack explicit consideration for antibody\nstructural information, despite the fact that both 1D sequence and 3D structure\ncarry unique and complementary insights into antibody behavior and\nfunctionality. This paper proposes Sequence-Structure multi-level pre-trained\nAntibody Language Model (S$^2$ALM), combining holistic sequential and\nstructural information in one unified, generic antibody foundation model. We\nconstruct a hierarchical pre-training paradigm incorporated with two customized\nmulti-level training objectives to facilitate the modeling of comprehensive\nantibody representations. S$^2$ALM's representation space uncovers inherent\nfunctional binding mechanisms, biological evolution properties and structural\ninteraction patterns. Pre-trained over 75 million sequences and 11.7 million\nstructures, S$^2$ALM can be adopted for diverse downstream tasks: accurately\npredicting antigen-antibody binding affinities, precisely distinguishing B cell\nmaturation stages, identifying antibody crucial binding positions, and\nspecifically designing novel coronavirus-binding antibodies. Remarkably,\nS$^2$ALM outperforms well-established and renowned baselines and sets new\nstate-of-the-art performance across extensive antibody specific understanding\nand generation tasks. S$^2$ALM's ability to model comprehensive and generalized\nrepresentations further positions its potential to advance real-world\ntherapeutic antibody development, potentially addressing unmet academic,\nindustrial, and clinical needs.\n", "date": "2024-11-20", "categories": []}
{"id": "2411.15618", "title": "Accelerated Hydration Site Localization and Thermodynamic Profiling", "abstract": "  Water plays a fundamental role in the structure and function of proteins and\nother biomolecules. The thermodynamic profile of water molecules surrounding a\nprotein are critical for ligand binding and recognition. Therefore, identifying\nthe location and thermodynamic behavior of relevant water molecules is\nimportant for generating and optimizing lead compounds for affinity and\nselectivity to a given target. Computational methods have been developed to\nidentify these hydration sites, but are largely limited to simplified models\nthat fail to capture multi-body interactions, or dynamics-based methods that\nrely on extensive sampling. Here we present a method for fast and accurate\nlocalization and thermodynamic profiling of hydration sites for protein\nstructures. The method is based on a geometric deep neural network trained on a\nlarge, novel dataset of explicit water molecular dynamics simulations. We\nconfirm the accuracy and robustness of our model on experimental data and\ndemonstrate it's utility on several case studies.\n", "date": "2024-11-23", "categories": []}
{"id": "2402.13653", "title": "PQA: Zero-shot Protein Question Answering for Free-form Scientific\n  Enquiry with Large Language Models", "abstract": "  Understanding protein structure and function is crucial in biology. However,\ncurrent computational methods are often task-specific and resource-intensive.\nTo address this, we propose zero-shot Protein Question Answering (PQA), a task\ndesigned to answer a wide range of protein-related queries without\ntask-specific training. The success of PQA hinges on high-quality datasets and\nrobust evaluation strategies, both of which are lacking in current research.\nExisting datasets suffer from biases, noise, and lack of evolutionary context,\nwhile current evaluation methods fail to accurately assess model performance.\nWe introduce the Pika framework to overcome these limitations. Pika comprises a\ncurated, debiased dataset tailored for PQA and a biochemically relevant\nbenchmarking strategy. We also propose multimodal large language models as a\nstrong baseline for PQA, leveraging their natural language processing and\nknowledge. This approach promises a more flexible and efficient way to explore\nprotein properties, advancing protein research. Our comprehensive PQA\nframework, Pika, including dataset, code, and model checkpoints, is openly\naccessible on github.com/EMCarrami/Pika, promoting wider research in the field.\n", "date": "2024-02-21", "categories": []}
{"id": "2411.08992", "title": "IDCIA: Immunocytochemistry Dataset for Cellular Image Analysis", "abstract": "  We present a new annotated microscopic cellular image dataset to improve the\neffectiveness of machine learning methods for cellular image analysis. Cell\ncounting is an important step in cell analysis. Typically, domain experts\nmanually count cells in a microscopic image. Automated cell counting can\npotentially eliminate this tedious, time-consuming process. However, a good,\nlabeled dataset is required for training an accurate machine learning model.\nOur dataset includes microscopic images of cells, and for each image, the cell\ncount and the location of individual cells. The data were collected as part of\nan ongoing study investigating the potential of electrical stimulation to\nmodulate stem cell differentiation and possible applications for neural repair.\nCompared to existing publicly available datasets, our dataset has more images\nof cells stained with more variety of antibodies (protein components of immune\nresponses against invaders) typically used for cell analysis. The experimental\nresults on this dataset indicate that none of the five existing models under\nthis study are able to achieve sufficiently accurate count to replace the\nmanual methods. The dataset is available at\nhttps://figshare.com/articles/dataset/Dataset/21970604.\n", "date": "2024-11-13", "categories": []}
{"id": "2411.11915", "title": "Phenome-wide causal proteomics enhance systemic lupus erythematosus\n  flare prediction: A study in Asian populations", "abstract": "  Objective: Systemic lupus erythematosus (SLE) is a complex autoimmune disease\ncharacterized by unpredictable flares. This study aimed to develop a novel\nproteomics-based risk prediction model specifically for Asian SLE populations\nto enhance personalized disease management and early intervention. Methods: A\nlongitudinal cohort study was conducted over 48 weeks, including 139 SLE\npatients monitored every 12 weeks. Patients were classified into flare (n = 53)\nand non-flare (n = 86) groups. Baseline plasma samples underwent\ndata-independent acquisition (DIA) proteomics analysis, and phenome-wide\nMendelian randomization (PheWAS) was performed to evaluate causal relationships\nbetween proteins and clinical predictors. Logistic regression (LR) and random\nforest (RF) models were used to integrate proteomic and clinical data for flare\nrisk prediction. Results: Five proteins (SAA1, B4GALT5, GIT2, NAA15, and RPIA)\nwere significantly associated with SLE Disease Activity Index-2K (SLEDAI-2K)\nscores and 1-year flare risk, implicating key pathways such as B-cell receptor\nsignaling and platelet degranulation. SAA1 demonstrated causal effects on\nflare-related clinical markers, including hemoglobin and red blood cell counts.\nA combined model integrating clinical and proteomic data achieved the highest\npredictive accuracy (AUC = 0.769), surpassing individual models. SAA1 was\nhighlighted as a priority biomarker for rapid flare discrimination. Conclusion:\nThe integration of proteomic and clinical data significantly improves flare\nprediction in Asian SLE patients. The identification of key proteins and their\ncausal relationships with flare-related clinical markers provides valuable\ninsights for proactive SLE management and personalized therapeutic approaches.\n", "date": "2024-11-18", "categories": []}
{"id": "2411.12078", "title": "Molecule Generation with Fragment Retrieval Augmentation", "abstract": "  Fragment-based drug discovery, in which molecular fragments are assembled\ninto new molecules with desirable biochemical properties, has achieved great\nsuccess. However, many fragment-based molecule generation methods show limited\nexploration beyond the existing fragments in the database as they only\nreassemble or slightly modify the given ones. To tackle this problem, we\npropose a new fragment-based molecule generation framework with retrieval\naugmentation, namely Fragment Retrieval-Augmented Generation (f-RAG). f-RAG is\nbased on a pre-trained molecular generative model that proposes additional\nfragments from input fragments to complete and generate a new molecule. Given a\nfragment vocabulary, f-RAG retrieves two types of fragments: (1) hard\nfragments, which serve as building blocks that will be explicitly included in\nthe newly generated molecule, and (2) soft fragments, which serve as reference\nto guide the generation of new fragments through a trainable fragment injection\nmodule. To extrapolate beyond the existing fragments, f-RAG updates the\nfragment vocabulary with generated fragments via an iterative refinement\nprocess which is further enhanced with post-hoc genetic fragment modification.\nf-RAG can achieve an improved exploration-exploitation trade-off by maintaining\na pool of fragments and expanding it with novel and high-quality fragments\nthrough a strong generative prior.\n", "date": "2024-11-18", "categories": []}
{"id": "2306.11768", "title": "Geometric Deep Learning for Structure-Based Drug Design: A Survey", "abstract": "  Structure-based drug design (SBDD) leverages the three-dimensional geometry\nof proteins to identify potential drug candidates. Traditional approaches,\nrooted in physicochemical modeling and domain expertise, are often\nresource-intensive. Recent advancements in geometric deep learning, which\neffectively integrate and process 3D geometric data, alongside breakthroughs in\naccurate protein structure predictions from tools like AlphaFold, have\nsignificantly propelled the field forward. This paper systematically reviews\nthe state-of-the-art in geometric deep learning for SBDD. We begin by outlining\nfoundational tasks in SBDD, discussing prevalent 3D protein representations,\nand highlighting representative predictive and generative models. Next, we\nprovide an in-depth review of key tasks, including binding site prediction,\nbinding pose generation, de novo molecule generation, linker design, protein\npocket generation, and binding affinity prediction. For each task, we present\nformal problem definitions, key methods, datasets, evaluation metrics, and\nperformance benchmarks. Lastly, we explore current challenges and future\nopportunities in SBDD. Challenges include oversimplified problem formulations,\nlimited out-of-distribution generalization, biosecurity concerns related to the\nmisuse of structural data, insufficient evaluation metrics and large-scale\nbenchmarks, and the need for experimental validation and enhanced model\ninterpretability. Opportunities lie in leveraging multimodal datasets,\nintegrating domain knowledge, developing comprehensive benchmarks, establishing\ncriteria aligned with clinical outcomes, and designing foundation models to\nexpand the scope of design tasks. We also curate\n\\url{https://github.com/zaixizhang/Awesome-SBDD}, reflecting ongoing\ncontributions and new datasets in SBDD.\n", "date": "2023-06-20", "categories": []}
{"id": "2402.02164", "title": "Hierarchical Structure Enhances the Convergence and Generalizability of\n  Linear Molecular Representation", "abstract": "  Language models demonstrate fundamental abilities in syntax, semantics, and\nreasoning, though their performance often depends significantly on the inputs\nthey process. This study introduces TSIS (Simplified TSID) and its\nvariants:TSISD (TSIS with Depth-First Search), TSISO (TSIS in Order), and TSISR\n(TSIS in Random), as integral components of the t-SMILES framework. These\nadditions complete the framework's design, providing diverse approaches to\nmolecular representation. Through comprehensive analysis and experiments\nemploying deep generative models, including GPT, diffusion models, and\nreinforcement learning, the findings reveal that the hierarchical structure of\nt-SMILES is more straightforward to parse than initially anticipated.\nFurthermore, t-SMILES consistently outperforms other linear representations\nsuch as SMILES, SELFIES, and SAFE, demonstrating superior convergence speed and\nenhanced generalization capabilities.\n", "date": "2024-02-03", "categories": []}
{"id": "2407.03154", "title": "Reinforcement Learning for Sequence Design Leveraging Protein Language\n  Models", "abstract": "  Protein sequence design, determined by amino acid sequences, are essential to\nprotein engineering problems in drug discovery. Prior approaches have resorted\nto evolutionary strategies or Monte-Carlo methods for protein design, but often\nfail to exploit the structure of the combinatorial search space, to generalize\nto unseen sequences. In the context of discrete black box optimization over\nlarge search spaces, learning a mutation policy to generate novel sequences\nwith reinforcement learning is appealing. Recent advances in protein language\nmodels (PLMs) trained on large corpora of protein sequences offer a potential\nsolution to this problem by scoring proteins according to their biological\nplausibility (such as the TM-score). In this work, we propose to use PLMs as a\nreward function to generate new sequences. Yet the PLM can be computationally\nexpensive to query due to its large size. To this end, we propose an\nalternative paradigm where optimization can be performed on scores from a\nsmaller proxy model that is periodically finetuned, jointly while learning the\nmutation policy. We perform extensive experiments on various sequence lengths\nto benchmark RL-based approaches, and provide comprehensive evaluations along\nbiological plausibility and diversity of the protein. Our experimental results\ninclude favorable evaluations of the proposed sequences, along with high\ndiversity scores, demonstrating that RL is a strong candidate for biological\nsequence design. Finally, we provide a modular open source implementation can\nbe easily integrated in most RL training loops, with support for replacing the\nreward model with other PLMs, to spur further research in this domain. The code\nfor all experiments is provided in the supplementary material.\n", "date": "2024-07-03", "categories": []}
{"id": "2407.19296", "title": "Multi-Modal CLIP-Informed Protein Editing", "abstract": "  Proteins govern most biological functions essential for life, but achieving\ncontrollable protein discovery and optimization remains challenging. Recently,\nmachine learning-assisted protein editing (MLPE) has shown promise in\naccelerating optimization cycles and reducing experimental workloads. However,\ncurrent methods struggle with the vast combinatorial space of potential protein\nedits and cannot explicitly conduct protein editing using biotext instructions,\nlimiting their interactivity with human feedback. To fill these gaps, we\npropose a novel method called ProtET for efficient CLIP-informed protein\nediting through multi-modality learning. Our approach comprises two stages: in\nthe pretraining stage, contrastive learning aligns protein-biotext\nrepresentations encoded by two large language models (LLMs), respectively.\nSubsequently, during the protein editing stage, the fused features from editing\ninstruction texts and original protein sequences serve as the final editing\ncondition for generating target protein sequences. Comprehensive experiments\ndemonstrated the superiority of ProtET in editing proteins to enhance\nhuman-expected functionality across multiple attribute domains, including\nenzyme catalytic activity, protein stability and antibody specific binding\nability. And ProtET improves the state-of-the-art results by a large margin,\nleading to significant stability improvements of 16.67% and 16.90%. This\ncapability positions ProtET to advance real-world artificial protein editing,\npotentially addressing unmet academic, industrial, and clinical needs.\n", "date": "2024-07-27", "categories": []}
{"id": "2411.10720", "title": "Multi Scale Graph Neural Network for Alzheimer's Disease", "abstract": "  Alzheimer's disease (AD) is a complex, progressive neurodegenerative disorder\ncharacterized by extracellular A\\b{eta} plaques, neurofibrillary tau tangles,\nglial activation, and neuronal degeneration, involving multiple cell types and\npathways. Current models often overlook the cellular context of these pathways.\nTo address this, we developed a multiscale graph neural network (GNN) model,\nALZ PINNACLE, using brain omics data from donors spanning the entire aging to\nAD spectrum. ALZ PINNACLE is based on the PINNACLE GNN framework, which learns\ncontext-aware protein, cell type, and tissue representations within a unified\nlatent space. ALZ PINNACLE was trained on 14,951 proteins, 206,850 protein\ninteractions, 7 cell types, and 48 cell subtypes or states. After pretraining,\nwe investigated the learned embedding of APOE, the largest genetic risk factor\nfor AD, across different cell types. Notably, APOE embeddings showed high\nsimilarity in microglial, neuronal, and CD8 cells, suggesting a similar role of\nAPOE in these cell types. Fine tuning the model on AD risk genes revealed cell\ntype contexts predictive of the role of APOE in AD. Our results suggest that\nALZ PINNACLE may provide a valuable framework for uncovering novel insights\ninto AD neurobiology.\n", "date": "2024-11-16", "categories": []}
{"id": "2411.11265", "title": "GROOT: Effective Design of Biological Sequences with Limited\n  Experimental Data", "abstract": "  Latent space optimization (LSO) is a powerful method for designing discrete,\nhigh-dimensional biological sequences that maximize expensive black-box\nfunctions, such as wet lab experiments. This is accomplished by learning a\nlatent space from available data and using a surrogate model to guide\noptimization algorithms toward optimal outputs. However, existing methods\nstruggle when labeled data is limited, as training the surrogate model with few\nlabeled data points can lead to subpar outputs, offering no advantage over the\ntraining data itself. We address this challenge by introducing GROOT, a\nGraph-based Latent Smoothing for Biological Sequence Optimization. In\nparticular, GROOT generates pseudo-labels for neighbors sampled around the\ntraining latent embeddings. These pseudo-labels are then refined and smoothed\nby Label Propagation. Additionally, we theoretically and empirically justify\nour approach, demonstrate GROOT's ability to extrapolate to regions beyond the\ntraining set while maintaining reliability within an upper bound of their\nexpected distances from the training regions. We evaluate GROOT on various\nbiological sequence design tasks, including protein optimization (GFP and AAV)\nand three tasks with exact oracles from Design-Bench. The results demonstrate\nthat GROOT equalizes and surpasses existing methods without requiring access to\nblack-box oracles or vast amounts of labeled data, highlighting its\npracticality and effectiveness. We release our code at\nhttps://anonymous.4open.science/r/GROOT-D554\n", "date": "2024-11-18", "categories": []}
{"id": "2312.14249", "title": "GenoCraft: A Comprehensive, User-Friendly Web-Based Platform for\n  High-Throughput Omics Data Analysis and Visualization", "abstract": "  The surge in high-throughput omics data has reshaped the landscape of\nbiological research, underlining the need for powerful, user-friendly data\nanalysis and interpretation tools. This paper presents GenoCraft, a web-based\ncomprehensive software solution designed to handle the entire pipeline of omics\ndata processing. GenoCraft offers a unified platform featuring advanced\nbioinformatics tools, covering all aspects of omics data analysis. It\nencompasses a range of functionalities, such as normalization, quality control,\ndifferential analysis, network analysis, pathway analysis, and diverse\nvisualization techniques. This software makes state-of-the-art omics data\nanalysis more accessible to a wider range of users. With GenoCraft, researchers\nand data scientists have access to an array of cutting-edge bioinformatics\ntools under a user-friendly interface, making it a valuable resource for\nmanaging and analyzing large-scale omics data. The API with an interactive web\ninterface is publicly available at https://genocraft.stanford. edu/. We also\nrelease all the codes in https://github.com/futianfan/GenoCraft.\n", "date": "2023-12-21", "categories": []}
{"id": "2411.08900", "title": "RNA-GPT: Multimodal Generative System for RNA Sequence Understanding", "abstract": "  RNAs are essential molecules that carry genetic information vital for life,\nwith profound implications for drug development and biotechnology. Despite this\nimportance, RNA research is often hindered by the vast literature available on\nthe topic. To streamline this process, we introduce RNA-GPT, a multi-modal RNA\nchat model designed to simplify RNA discovery by leveraging extensive RNA\nliterature. RNA-GPT integrates RNA sequence encoders with linear projection\nlayers and state-of-the-art large language models (LLMs) for precise\nrepresentation alignment, enabling it to process user-uploaded RNA sequences\nand deliver concise, accurate responses. Built on a scalable training pipeline,\nRNA-GPT utilizes RNA-QA, an automated system that gathers RNA annotations from\nRNACentral using a divide-and-conquer approach with GPT-4o and latent Dirichlet\nallocation (LDA) to efficiently handle large datasets and generate\ninstruction-tuning samples. Our experiments indicate that RNA-GPT effectively\naddresses complex RNA queries, thereby facilitating RNA research. Additionally,\nwe present RNA-QA, a dataset of 407,616 RNA samples for modality alignment and\ninstruction tuning, further advancing the potential of RNA research tools.\n", "date": "2024-10-29", "categories": []}
{"id": "2411.08072", "title": "Modeling variable guide efficiency in pooled CRISPR screens with\n  ContrastiveVI+", "abstract": "  Genetic screens mediated via CRISPR-Cas9 combined with high-content readouts\nhave emerged as powerful tools for biological discovery. However, computational\nanalyses of these screens come with additional challenges beyond those found\nwith standard scRNA-seq analyses. For example, perturbation-induced variations\nof interest may be subtle and masked by other dominant source of variation\nshared with controls, and variable guide efficiency results in some cells not\nundergoing genetic perturbation despite expressing a guide RNA. While a number\nof methods have been developed to address the former problem by explicitly\ndisentangling perturbation-induced variations from those shared with controls,\nless attention has been paid to the latter problem of noisy perturbation\nlabels. To address this issue, here we propose ContrastiveVI+, a generative\nmodeling framework that both disentangles perturbation-induced from\nnon-perturbation-related variations while also inferring whether cells truly\nunderwent genomic edits. Applied to three large-scale Perturb-seq datasets, we\nfind that ContrastiveVI+ better recovers known perturbation-induced variations\ncompared to previous methods while successfully identifying cells that escaped\nthe functional consequences of guide RNA expression. An open-source\nimplementation of our model is available at\n\\url{https://github.com/insitro/contrastive_vi_plus}.\n", "date": "2024-11-11", "categories": []}
{"id": "2411.08073", "title": "LoRA-BERT: a Natural Language Processing Model for Robust and Accurate\n  Prediction of long non-coding RNAs", "abstract": "  Long non-coding RNAs (lncRNAs) serve as crucial regulators in numerous\nbiological processes. Although they share sequence similarities with messenger\nRNAs (mRNAs), lncRNAs perform entirely different roles, providing new avenues\nfor biological research. The emergence of next-generation sequencing\ntechnologies has greatly advanced the detection and identification of lncRNA\ntranscripts and deep learning-based approaches have been introduced to classify\nlong non-coding RNAs (lncRNAs). These advanced methods have significantly\nenhanced the efficiency of identifying lncRNAs. However, many of these methods\nare devoid of robustness and accuracy due to the extended length of the\nsequences involved. To tackle this issue, we have introduced a novel\npre-trained bidirectional encoder representation called LoRA-BERT. LoRA-BERT is\ndesigned to capture the importance of nucleotide-level information during\nsequence classification, leading to more robust and satisfactory outcomes. In a\ncomprehensive comparison with commonly used sequence prediction tools, we have\ndemonstrated that LoRA-BERT outperforms them in terms of accuracy and\nefficiency. Our results indicate that, when utilizing the transformer model,\nLoRA-BERT achieves state-of-the-art performance in predicting both lncRNAs and\nmRNAs for human and mouse species. Through the utilization of LoRA-BERT, we\nacquire valuable insights into the traits of lncRNAs and mRNAs, offering the\npotential to aid in the comprehension and detection of diseases linked to\nlncRNAs in humans.\n", "date": "2024-11-11", "categories": []}
{"id": "2411.08286", "title": "Hashing for Protein Structure Similarity Search", "abstract": "  Protein structure similarity search (PSSS), which tries to search proteins\nwith similar structures, plays a crucial role across diverse domains from drug\ndesign to protein function prediction and molecular evolution. Traditional\nalignment-based PSSS methods, which directly calculate alignment on the protein\nstructures, are highly time-consuming with high memory cost. Recently,\nalignment-free methods, which represent protein structures as fixed-length\nreal-valued vectors, are proposed for PSSS. Although these methods have lower\ntime and memory cost than alignment-based methods, their time and memory cost\nis still too high for large-scale PSSS, and their accuracy is unsatisfactory.\nIn this paper, we propose a novel method, called\n$\\underline{\\text{p}}$r$\\underline{\\text{o}}$tein\n$\\underline{\\text{s}}$tructure $\\underline{\\text{h}}$ashing (POSH), for PSSS.\nPOSH learns a binary vector representation for each protein structure, which\ncan dramatically reduce the time and memory cost for PSSS compared with\nreal-valued vector representation based methods. Furthermore, in POSH we also\npropose expressive hand-crafted features and a structure encoder to well model\nboth node and edge interactions in proteins. Experimental results on real\ndatasets show that POSH can outperform other methods to achieve\nstate-of-the-art accuracy. Furthermore, POSH achieves a memory saving of more\nthan six times and speed improvement of more than four times, compared with\nother methods.\n", "date": "2024-11-13", "categories": []}
{"id": "2410.20354", "title": "FoldMark: Protecting Protein Generative Models with Watermarking", "abstract": "  Protein structure is key to understanding protein function and is essential\nfor progress in bioengineering, drug discovery, and molecular biology.\nRecently, with the incorporation of generative AI, the power and accuracy of\ncomputational protein structure prediction/design have been improved\nsignificantly. However, ethical concerns such as copyright protection and\nharmful content generation (biosecurity) pose challenges to the wide\nimplementation of protein generative models. Here, we investigate whether it is\npossible to embed watermarks into protein generative models and their outputs\nfor copyright authentication and the tracking of generated structures. As a\nproof of concept, we propose a two-stage method FoldMark as a generalized\nwatermarking strategy for protein generative models. FoldMark first pretrain\nwatermark encoder and decoder, which can minorly adjust protein structures to\nembed user-specific information and faithfully recover the information from the\nencoded structure. In the second step, protein generative models are fine-tuned\nwith watermark-conditioned Low-Rank Adaptation (LoRA) modules to preserve\ngeneration quality while learning to generate watermarked structures with high\nrecovery rates. Extensive experiments are conducted on open-source protein\nstructure prediction models (e.g., ESMFold and MultiFlow) and de novo structure\ndesign models (e.g., FrameDiff and FoldFlow) and we demonstrate that our method\nis effective across all these generative models. Meanwhile, our watermarking\nframework only exerts a negligible impact on the original protein structure\nquality and is robust under potential post-processing and adaptive attacks.\n", "date": "2024-10-27", "categories": []}
{"id": "2407.18184", "title": "AsEP: Benchmarking Deep Learning Methods for Antibody-specific Epitope\n  Prediction", "abstract": "  Epitope identification is vital for antibody design yet challenging due to\nthe inherent variability in antibodies. While many deep learning methods have\nbeen developed for general protein binding site prediction tasks, whether they\nwork for epitope prediction remains an understudied research question. The\nchallenge is also heightened by the lack of a consistent evaluation pipeline\nwith sufficient dataset size and epitope diversity. We introduce a filtered\nantibody-antigen complex structure dataset, AsEP (Antibody-specific Epitope\nPrediction). AsEP is the largest of its kind and provides clustered epitope\ngroups, allowing the community to develop and test novel epitope prediction\nmethods and evaluate their generalisability. AsEP comes with an easy-to-use\ninterface in Python and pre-built graph representations of each\nantibody-antigen complex while also supporting customizable embedding methods.\nUsing this new dataset, we benchmark several representative general\nprotein-binding site prediction methods and find that their performances fall\nshort of expectations for epitope prediction. To address this, we propose a\nnovel method, WALLE, which leverages both unstructured modeling from protein\nlanguage models and structural modeling from graph neural networks. WALLE\ndemonstrate up to 3-10X performance improvement over the baseline methods. Our\nempirical findings suggest that epitope prediction benefits from combining\nsequential features provided by language models with geometrical information\nfrom graph representations. This provides a guideline for future epitope\nprediction method design. In addition, we reformulate the task as bipartite\nlink prediction, allowing convenient model performance attribution and\ninterpretability. We open source our data and code at\nhttps://github.com/biochunan/AsEP-dataset.\n", "date": "2024-07-25", "categories": []}
{"id": "2411.05055", "title": "Integrating Large Language Models for Genetic Variant Classification", "abstract": "  The classification of genetic variants, particularly Variants of Uncertain\nSignificance (VUS), poses a significant challenge in clinical genetics and\nprecision medicine. Large Language Models (LLMs) have emerged as transformative\ntools in this realm. These models can uncover intricate patterns and predictive\ninsights that traditional methods might miss, thus enhancing the predictive\naccuracy of genetic variant pathogenicity.\n  This study investigates the integration of state-of-the-art LLMs, including\nGPN-MSA, ESM1b, and AlphaMissense, which leverage DNA and protein sequence data\nalongside structural insights to form a comprehensive analytical framework for\nvariant classification. Our approach evaluates these integrated models using\nthe well-annotated ProteinGym and ClinVar datasets, setting new benchmarks in\nclassification performance. The models were rigorously tested on a set of\nchallenging variants, demonstrating substantial improvements over existing\nstate-of-the-art tools, especially in handling ambiguous and clinically\nuncertain variants.\n  The results of this research underline the efficacy of combining multiple\nmodeling approaches to significantly refine the accuracy and reliability of\ngenetic variant classification systems. These findings support the deployment\nof these advanced computational models in clinical environments, where they can\nsignificantly enhance the diagnostic processes for genetic disorders,\nultimately pushing the boundaries of personalized medicine by offering more\ndetailed and actionable genetic insights.\n", "date": "2024-11-07", "categories": []}
{"id": "2411.05238", "title": "Generating Highly Designable Proteins with Geometric Algebra Flow\n  Matching", "abstract": "  We introduce a generative model for protein backbone design utilizing\ngeometric products and higher order message passing. In particular, we propose\nClifford Frame Attention (CFA), an extension of the invariant point attention\n(IPA) architecture from AlphaFold2, in which the backbone residue frames and\ngeometric features are represented in the projective geometric algebra. This\nenables to construct geometrically expressive messages between residues,\nincluding higher order terms, using the bilinear operations of the algebra. We\nevaluate our architecture by incorporating it into the framework of FrameFlow,\na state-of-the-art flow matching model for protein backbone generation. The\nproposed model achieves high designability, diversity and novelty, while also\nsampling protein backbones that follow the statistical distribution of\nsecondary structure elements found in naturally occurring proteins, a property\nso far only insufficiently achieved by many state-of-the-art generative models.\n", "date": "2024-11-07", "categories": []}
{"id": "2411.05421", "title": "Learning the rules of peptide self-assembly through data mining with\n  large language models", "abstract": "  Peptides are ubiquitous and important biologically derived molecules, that\nhave been found to self-assemble to form a wide array of structures. Extensive\nresearch has explored the impacts of both internal chemical composition and\nexternal environmental stimuli on the self-assembly behaviour of these systems.\nHowever, there is yet to be a systematic study that gathers this rich\nliterature data and collectively examines these experimental factors to provide\na global picture of the fundamental rules that govern protein self-assembly\nbehavior. In this work, we curate a peptide assembly database through a\ncombination of manual processing by human experts and literature mining\nfacilitated by a large language model. As a result, we collect more than 1,000\nexperimental data entries with information about peptide sequence, experimental\nconditions and corresponding self-assembly phases. Utilizing the collected\ndata, ML models are trained and evaluated, demonstrating excellent accuracy\n(>80\\%) and efficiency in peptide assembly phase classification. Moreover, we\nfine-tune our GPT model for peptide literature mining with the developed\ndataset, which exhibits markedly superior performance in extracting information\nfrom academic publications relative to the pre-trained model. We find that this\nworkflow can substantially improve efficiency when exploring potential\nself-assembling peptide candidates, through guiding experimental work, while\nalso deepening our understanding of the mechanisms governing peptide\nself-assembly. In doing so, novel structures can be accessed for a range of\napplications including sensing, catalysis and biomaterials.\n", "date": "2024-11-08", "categories": []}
{"id": "2310.09254", "title": "GENOT: Entropic (Gromov) Wasserstein Flow Matching with Applications to\n  Single-Cell Genomics", "abstract": "  Single-cell genomics has significantly advanced our understanding of cellular\nbehavior, catalyzing innovations in treatments and precision medicine. However,\nsingle-cell sequencing technologies are inherently destructive and can only\nmeasure a limited array of data modalities simultaneously. This limitation\nunderscores the need for new methods capable of realigning cells. Optimal\ntransport (OT) has emerged as a potent solution, but traditional discrete\nsolvers are hampered by scalability, privacy, and out-of-sample estimation\nissues. These challenges have spurred the development of neural network-based\nsolvers, known as neural OT solvers, that parameterize OT maps. Yet, these\nmodels often lack the flexibility needed for broader life science applications.\nTo address these deficiencies, our approach learns stochastic maps (i.e.\ntransport plans), allows for any cost function, relaxes mass conservation\nconstraints and integrates quadratic solvers to tackle the complex challenges\nposed by the (Fused) Gromov-Wasserstein problem. Utilizing flow matching as a\nbackbone, our method offers a flexible and effective framework. We demonstrate\nits versatility and robustness through applications in cell development\nstudies, cellular drug response modeling, and cross-modality cell translation,\nillustrating significant potential for enhancing therapeutic strategies.\n", "date": "2023-10-13", "categories": []}
{"id": "2411.04150", "title": "BAPULM: Binding Affinity Prediction using Language Models", "abstract": "  Identifying drug-target interactions is essential for developing effective\ntherapeutics. Binding affinity quantifies these interactions, and traditional\napproaches rely on computationally intensive 3D structural data. In contrast,\nlanguage models can efficiently process sequential data, offering an\nalternative approach to molecular representation. In the current study, we\nintroduce BAPULM, an innovative sequence-based framework that leverages the\nchemical latent representations of proteins via ProtT5-XL-U50 and ligands\nthrough MolFormer, eliminating reliance on complex 3D configurations. Our\napproach was validated extensively on benchmark datasets, achieving scoring\npower (R) values of 0.925 $\\pm$ 0.043, 0.914 $\\pm$ 0.004, and 0.8132 $\\pm$\n0.001 on benchmark1k2101, Test2016_290, and CSAR-HiQ_36, respectively. These\nfindings indicate the robustness and accuracy of BAPULM across diverse datasets\nand underscore the potential of sequence-based models in-silico drug discovery,\noffering a scalable alternative to 3D-centric methods for screening potential\nligands.\n", "date": "2024-11-06", "categories": []}
{"id": "2411.04165", "title": "Bio-xLSTM: Generative modeling, representation and in-context learning\n  of biological and chemical sequences", "abstract": "  Language models for biological and chemical sequences enable crucial\napplications such as drug discovery, protein engineering, and precision\nmedicine. Currently, these language models are predominantly based on\nTransformer architectures. While Transformers have yielded impressive results,\ntheir quadratic runtime dependency on the sequence length complicates their use\nfor long genomic sequences and in-context learning on proteins and chemical\nsequences. Recently, the recurrent xLSTM architecture has been shown to perform\nfavorably compared to Transformers and modern state-space model (SSM)\narchitectures in the natural language domain. Similar to SSMs, xLSTMs have a\nlinear runtime dependency on the sequence length and allow for constant-memory\ndecoding at inference time, which makes them prime candidates for modeling\nlong-range dependencies in biological and chemical sequences. In this work, we\ntailor xLSTM towards these domains and propose a suite of architectural\nvariants called Bio-xLSTM. Extensive experiments in three large domains,\ngenomics, proteins, and chemistry, were performed to assess xLSTM's ability to\nmodel biological and chemical sequences. The results show that models based on\nBio-xLSTM a) can serve as proficient generative models for DNA, protein, and\nchemical sequences, b) learn rich representations for those modalities, and c)\ncan perform in-context learning for proteins and small molecules.\n", "date": "2024-11-06", "categories": []}
{"id": "2411.03460", "title": "Pathway-Guided Optimization of Deep Generative Molecular Design Models\n  for Cancer Therapy", "abstract": "  The data-driven drug design problem can be formulated as an optimization task\nof a potentially expensive black-box objective function over a huge\nhigh-dimensional and structured molecular space. The junction tree variational\nautoencoder (JTVAE) has been shown to be an efficient generative model that can\nbe used for suggesting legitimate novel drug-like small molecules with improved\nproperties. While the performance of the generative molecular design (GMD)\nscheme strongly depends on the initial training data, one can improve its\nsampling efficiency for suggesting better molecules with enhanced properties by\noptimizing the latent space. In this work, we propose how mechanistic models -\nsuch as pathway models described by differential equations - can be used for\neffective latent space optimization(LSO) of JTVAEs and other similar models for\nGMD. To demonstrate the potential of our proposed approach, we show how a\npharmacodynamic model, assessing the therapeutic efficacy of a drug-like small\nmolecule by predicting how it modulates a cancer pathway, can be incorporated\nfor effective LSO of data-driven models for GMD.\n", "date": "2024-11-05", "categories": []}
{"id": "2411.03522", "title": "Exploring the Potentials and Challenges of Using Large Language Models\n  for the Analysis of Transcriptional Regulation of Long Non-coding RNAs", "abstract": "  Research on long non-coding RNAs (lncRNAs) has garnered significant attention\ndue to their critical roles in gene regulation and disease mechanisms. However,\nthe complexity and diversity of lncRNA sequences, along with the limited\nknowledge of their functional mechanisms and the regulation of their\nexpressions, pose significant challenges to lncRNA studies. Given the\ntremendous success of large language models (LLMs) in capturing complex\ndependencies in sequential data, this study aims to systematically explore the\npotential and limitations of LLMs in the sequence analysis related to the\ntranscriptional regulation of lncRNA genes. Our extensive experiments\ndemonstrated promising performance of fine-tuned genome foundation models on\nprogressively complex tasks. Furthermore, we conducted an insightful analysis\nof the critical impact of task complexity, model selection, data quality, and\nbiological interpretability for the studies of the regulation of lncRNA gene\nexpression.\n", "date": "2024-11-05", "categories": []}
{"id": "2411.03743", "title": "Automating Exploratory Proteomics Research via Language Models", "abstract": "  With the development of artificial intelligence, its contribution to science\nis evolving from simulating a complex problem to automating entire research\nprocesses and producing novel discoveries. Achieving this advancement requires\nboth specialized general models grounded in real-world scientific data and\niterative, exploratory frameworks that mirror human scientific methodologies.\nIn this paper, we present PROTEUS, a fully automated system for scientific\ndiscovery from raw proteomics data. PROTEUS uses large language models (LLMs)\nto perform hierarchical planning, execute specialized bioinformatics tools, and\niteratively refine analysis workflows to generate high-quality scientific\nhypotheses. The system takes proteomics datasets as input and produces a\ncomprehensive set of research objectives, analysis results, and novel\nbiological hypotheses without human intervention. We evaluated PROTEUS on 12\nproteomics datasets collected from various biological samples (e.g. immune\ncells, tumors) and different sample types (single-cell and bulk), generating\n191 scientific hypotheses. These were assessed using both automatic LLM-based\nscoring on 5 metrics and detailed reviews from human experts. Results\ndemonstrate that PROTEUS consistently produces reliable, logically coherent\nresults that align well with existing literature while also proposing novel,\nevaluable hypotheses. The system's flexible architecture facilitates seamless\nintegration of diverse analysis tools and adaptation to different proteomics\ndata types. By automating complex proteomics analysis workflows and hypothesis\ngeneration, PROTEUS has the potential to considerably accelerate the pace of\nscientific discovery in proteomics research, enabling researchers to\nefficiently explore large-scale datasets and uncover biological insights.\n", "date": "2024-11-06", "categories": []}
{"id": "2411.02534", "title": "Multi-modal Spatial Clustering for Spatial Transcriptomics Utilizing\n  High-resolution Histology Images", "abstract": "  Understanding the intricate cellular environment within biological tissues is\ncrucial for uncovering insights into complex biological functions. While\nsingle-cell RNA sequencing has significantly enhanced our understanding of\ncellular states, it lacks the spatial context necessary to fully comprehend the\ncellular environment. Spatial transcriptomics (ST) addresses this limitation by\nenabling transcriptome-wide gene expression profiling while preserving spatial\ncontext. One of the principal challenges in ST data analysis is spatial\nclustering, which reveals spatial domains based on the spots within a tissue.\nModern ST sequencing procedures typically include a high-resolution histology\nimage, which has been shown in previous studies to be closely connected to gene\nexpression profiles. However, current spatial clustering methods often fail to\nfully integrate high-resolution histology image features with gene expression\ndata, limiting their ability to capture critical spatial and cellular\ninteractions.\n  In this study, we propose the spatial transcriptomics multi-modal clustering\n(stMMC) model, a novel contrastive learning-based deep learning approach that\nintegrates gene expression data with histology image features through a\nmulti-modal parallel graph autoencoder. We tested stMMC against four\nstate-of-the-art baseline models: Leiden, GraphST, SpaGCN, and stLearn on two\npublic ST datasets with 13 sample slices in total. The experiments demonstrated\nthat stMMC outperforms all the baseline models in terms of ARI and NMI. An\nablation study further validated the contributions of contrastive learning and\nthe incorporation of histology image features.\n", "date": "2024-10-31", "categories": []}
{"id": "2404.09738", "title": "AMPCliff: quantitative definition and benchmarking of activity cliffs in\n  antimicrobial peptides", "abstract": "  Since the mechanism of action of drug molecules in the human body is\ndifficult to reproduce in the in vitro environment, it becomes difficult to\nreveal the causes of the activity cliff phenomenon of drug molecules. We found\nout the AC of small molecules has been extensively investigated but limited\nknowledge is accumulated about the AC phenomenon in peptides with canonical\namino acids. Understanding the mechanism of AC in canonical amino acids might\nhelp understand the one in drug molecules. This study introduces a quantitative\ndefinition and benchmarking framework AMPCliff for the AC phenomenon in\nantimicrobial peptides (AMPs) composed by canonical amino acids. A\ncomprehensive analysis of the existing AMP dataset reveals a significant\nprevalence of AC within AMPs. AMPCliff quantifies the activities of AMPs by the\nMIC, and defines 0.9 as the minimum threshold for the normalized BLOSUM62\nsimilarity score between a pair of aligned peptides with at least two-fold MIC\nchanges. This study establishes a benchmark dataset of paired AMPs in\nStaphylococcus aureus from the publicly available AMP dataset GRAMPA, and\nconducts a rigorous procedure to evaluate various AMP AC prediction models,\nincluding nine machine learning, four deep learning algorithms, four masked\nlanguage models, and four generative language models. Our analysis reveals that\nthese models are capable of detecting AMP AC events and the pre-trained protein\nlanguage model ESM2 demonstrates superior performance across the evaluations.\nThe predictive performance of AMP activity cliffs remains to be further\nimproved, considering that ESM2 with 33 layers only achieves the Spearman\ncorrelation coefficient 0.4669 for the regression task of the MIC values on the\nbenchmark dataset. Source code and additional resources are available at\nhttps://www.healthinformaticslab.org/supp/ or\nhttps://github.com/Kewei2023/AMPCliff-generation.\n", "date": "2024-04-15", "categories": []}
{"id": "2405.08226", "title": "Self-Normalizing Foundation Model for Enhanced Multi-Omics Data Analysis\n  in Oncology", "abstract": "  Multi-omics research has enhanced our understanding of cancer heterogeneity\nand progression. Investigating molecular data through multi-omics approaches is\ncrucial for unraveling the complex biological mechanisms underlying cancer,\nthereby enabling more effective diagnosis, treatment, and prevention\nstrategies. However, predicting patient outcomes through the integration of all\navailable multi-omics data is still an under-study research direction. Here, we\npresent SeNMo, a foundation model that has been trained on multi-omics data\nacross 33 cancer types. SeNMo is particularly efficient in handling multi-omics\ndata characterized by high-width and low-length attributes. We trained SeNMo\nfor the task of overall survival of patients using pan-cancer multi-omics data\ninvolving 33 cancer sites from the GDC. The training multi-omics data includes\ngene expression, DNA methylation, miRNA expression, DNA mutations, protein\nexpression modalities, and clinical data. SeNMo was validated on two\nindependent cohorts: Moffitt Cancer Center and CPTAC lung squamous cell\ncarcinoma. We evaluated the model's performance in predicting patient's overall\nsurvival using the C-Index. SeNMo performed consistently well in the training\nregime, reflected by the validation C-Index of 0.76 on GDC's public data. In\nthe testing regime, SeNMo performed with a C-Index of 0.758 on a held-out test\nset. The model showed an average accuracy of 99.8% on the task of classifying\nthe primary cancer type on the pan-cancer test cohort. SeNMo demonstrated\nrobust performance on the classification task of predicting the primary cancer\ntype of patients. SeNMo further demonstrated significant performance in\npredicting tertiary lymph structures from multi-omics data, showing\ngeneralizability across cancer types, molecular data types, and clinical\nendpoints.\n", "date": "2024-05-13", "categories": []}
{"id": "2406.16192", "title": "HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image\n  Analysis", "abstract": "  Spatial transcriptomics enables interrogating the molecular composition of\ntissue with ever-increasing resolution and sensitivity. However, costs, rapidly\nevolving technology, and lack of standards have constrained computational\nmethods in ST to narrow tasks and small cohorts. In addition, the underlying\ntissue morphology, as reflected by H&E-stained whole slide images (WSIs),\nencodes rich information often overlooked in ST studies. Here, we introduce\nHEST-1k, a collection of 1,229 spatial transcriptomic profiles, each linked to\na WSI and extensive metadata. HEST-1k was assembled from 153 public and\ninternal cohorts encompassing 26 organs, two species (Homo Sapiens and Mus\nMusculus), and 367 cancer samples from 25 cancer types. HEST-1k processing\nenabled the identification of 2.1 million expression--morphology pairs and over\n76 million nuclei. To support its development, we additionally introduce the\nHEST-Library, a Python package designed to perform a range of actions with HEST\nsamples. We test HEST-1k and Library on three use cases: (1) benchmarking\nfoundation models for pathology (HEST-Benchmark), (2) biomarker exploration,\nand (3) multimodal representation learning. HEST-1k, HEST-Library, and\nHEST-Benchmark can be freely accessed at https://github.com/mahmoodlab/hest.\n", "date": "2024-06-23", "categories": []}
{"id": "2410.13956", "title": "Benchmarking Transcriptomics Foundation Models for Perturbation Analysis\n  : one PCA still rules them all", "abstract": "  Understanding the relationships among genes, compounds, and their\ninteractions in living organisms remains limited due to technological\nconstraints and the complexity of biological data. Deep learning has shown\npromise in exploring these relationships using various data types. However,\ntranscriptomics, which provides detailed insights into cellular states, is\nstill underused due to its high noise levels and limited data availability.\nRecent advancements in transcriptomics sequencing provide new opportunities to\nuncover valuable insights, especially with the rise of many new foundation\nmodels for transcriptomics, yet no benchmark has been made to robustly evaluate\nthe effectiveness of these rising models for perturbation analysis. This\narticle presents a novel biologically motivated evaluation framework and a\nhierarchy of perturbation analysis tasks for comparing the performance of\npretrained foundation models to each other and to more classical techniques of\nlearning from transcriptomics data. We compile diverse public datasets from\ndifferent sequencing techniques and cell lines to assess models performance.\nOur approach identifies scVI and PCA to be far better suited models for\nunderstanding biological perturbations in comparison to existing foundation\nmodels, especially in their application in real-world scenarios.\n", "date": "2024-10-17", "categories": []}
{"id": "2411.00885", "title": "Revolutionizing Personalized Cancer Vaccines with NEO: Novel Epitope\n  Optimization Using an Aggregated Feed Forward and Recurrent Neural Network\n  with LSTM Architecture", "abstract": "  As cancer cases continue to rise, with a 2023 study from Zhejiang and Harvard\npredicting a 31 percent increase in cases and a 21 percent increase in deaths\nby 2030, the need to find more effective treatments for cancer is greater than\never before. Traditional approaches to treating cancer, such as chemotherapy,\noften kill healthy cells because of their lack of targetability. In contrast,\npersonalized cancer vaccines can utilize neoepitopes - distinctive peptides on\ncancer cells that are often missed by the body's immune system - that have\nstrong binding affinities to a patient's MHC to provide a more targeted\ntreatment approach. The selection of optimal neoepitopes that elicit an immune\nresponse is a time-consuming and costly process due to the required inputs of\nmodern predictive methods. This project aims to facilitate faster, cheaper, and\nmore accurate neoepitope binding predictions using Feed Forward Neural Networks\n(FFNN) and Recurrent Neural Networks (RNN).\n  To address this, NEO was created. NEO requires next-generation sequencing\ndata and uses a stacking ensemble method by calculating scores from\nstate-of-the-art models (MHCFlurry 1.6, NetMHCstabpan 1.0, and IEDB). The\nmodel's architecture includes an FFNN and an RNN with LSTM layers capable of\nanalyzing both sequential and non-sequential data. The results from both models\nare aggregated to produce predictions. Using this model, personalized cancer\nvaccines can be produced with improved results (AUC = 0.9166, recall = 91.67\npercent).\n", "date": "2024-10-31", "categories": []}
{"id": "2411.01422", "title": "GramSeq-DTA: A grammar-based drug-target affinity prediction approach\n  fusing gene expression information", "abstract": "  Drug-target affinity (DTA) prediction is a critical aspect of drug discovery.\nThe meaningful representation of drugs and targets is crucial for accurate\nprediction. Using 1D string-based representations for drugs and targets is a\ncommon approach that has demonstrated good results in drug-target affinity\nprediction. However, these approach lacks information on the relative position\nof the atoms and bonds. To address this limitation, graph-based representations\nhave been used to some extent. However, solely considering the structural\naspect of drugs and targets may be insufficient for accurate DTA prediction.\nIntegrating the functional aspect of these drugs at the genetic level can\nenhance the prediction capability of the models. To fill this gap, we propose\nGramSeq-DTA, which integrates chemical perturbation information with the\nstructural information of drugs and targets. We applied a Grammar Variational\nAutoencoder (GVAE) for drug feature extraction and utilized two different\napproaches for protein feature extraction: Convolutional Neural Network (CNN)\nand Recurrent Neural Network (RNN). The chemical perturbation data is obtained\nfrom the L1000 project, which provides information on the upregulation and\ndownregulation of genes caused by selected drugs. This chemical perturbation\ninformation is processed, and a compact dataset is prepared, serving as the\nfunctional feature set of the drugs. By integrating the drug, gene, and target\nfeatures in the model, our approach outperforms the current state-of-the-art\nDTA prediction models when validated on widely used DTA datasets (BindingDB,\nDavis, and KIBA). This work provides a novel and practical approach to DTA\nprediction by merging the structural and functional aspects of biological\nentities, and it encourages further research in multi-modal DTA prediction.\n", "date": "2024-11-03", "categories": []}
{"id": "2411.01856", "title": "MeToken: Uniform Micro-environment Token Boosts Post-Translational\n  Modification Prediction", "abstract": "  Post-translational modifications (PTMs) profoundly expand the complexity and\nfunctionality of the proteome, regulating protein attributes and interactions\nthat are crucial for biological processes. Accurately predicting PTM sites and\ntheir specific types is therefore essential for elucidating protein function\nand understanding disease mechanisms. Existing computational approaches\npredominantly focus on protein sequences to predict PTM sites, driven by the\nrecognition of sequence-dependent motifs. However, these approaches often\noverlook protein structural contexts. In this work, we first compile a\nlarge-scale sequence-structure PTM dataset, which serves as the foundation for\nfair comparison. We introduce the MeToken model, which tokenizes the\nmicro-environment of each amino acid, integrating both sequence and structural\ninformation into unified discrete tokens. This model not only captures the\ntypical sequence motifs associated with PTMs but also leverages the spatial\narrangements dictated by protein tertiary structures, thus providing a holistic\nview of the factors influencing PTM sites. Designed to address the long-tail\ndistribution of PTM types, MeToken employs uniform sub-codebooks that ensure\neven the rarest PTMs are adequately represented and distinguished. We validate\nthe effectiveness and generalizability of MeToken across multiple datasets,\ndemonstrating its superior performance in accurately identifying PTM types. The\nresults underscore the importance of incorporating structural data and\nhighlight MeToken's potential in facilitating accurate and comprehensive PTM\npredictions, which could significantly impact proteomics research. The code and\ndatasets are available at https://github.com/A4Bio/MeToken.\n", "date": "2024-11-04", "categories": []}
{"id": "2411.02120", "title": "Bridge-IF: Learning Inverse Protein Folding with Markov Bridges", "abstract": "  Inverse protein folding is a fundamental task in computational protein\ndesign, which aims to design protein sequences that fold into the desired\nbackbone structures. While the development of machine learning algorithms for\nthis task has seen significant success, the prevailing approaches, which\npredominantly employ a discriminative formulation, frequently encounter the\nerror accumulation issue and often fail to capture the extensive variety of\nplausible sequences. To fill these gaps, we propose Bridge-IF, a generative\ndiffusion bridge model for inverse folding, which is designed to learn the\nprobabilistic dependency between the distributions of backbone structures and\nprotein sequences. Specifically, we harness an expressive structure encoder to\npropose a discrete, informative prior derived from structures, and establish a\nMarkov bridge to connect this prior with native sequences. During the inference\nstage, Bridge-IF progressively refines the prior sequence, culminating in a\nmore plausible design. Moreover, we introduce a reparameterization perspective\non Markov bridge models, from which we derive a simplified loss function that\nfacilitates more effective training. We also modulate protein language models\n(PLMs) with structural conditions to precisely approximate the Markov bridge\nprocess, thereby significantly enhancing generation performance while\nmaintaining parameter-efficient training. Extensive experiments on\nwell-established benchmarks demonstrate that Bridge-IF predominantly surpasses\nexisting baselines in sequence recovery and excels in the design of plausible\nproteins with high foldability. The code is available at\nhttps://github.com/violet-sto/Bridge-IF.\n", "date": "2024-11-04", "categories": []}
{"id": "2411.02125", "title": "Revisiting K-mer Profile for Effective and Scalable Genome\n  Representation Learning", "abstract": "  Obtaining effective representations of DNA sequences is crucial for genome\nanalysis. Metagenomic binning, for instance, relies on genome representations\nto cluster complex mixtures of DNA fragments from biological samples with the\naim of determining their microbial compositions. In this paper, we revisit\nk-mer-based representations of genomes and provide a theoretical analysis of\ntheir use in representation learning. Based on the analysis, we propose a\nlightweight and scalable model for performing metagenomic binning at the genome\nread level, relying only on the k-mer compositions of the DNA fragments. We\ncompare the model to recent genome foundation models and demonstrate that while\nthe models are comparable in performance, the proposed model is significantly\nmore effective in terms of scalability, a crucial aspect for performing\nmetagenomic binning of real-world datasets.\n", "date": "2024-11-04", "categories": []}
{"id": "2410.21335", "title": "E(3)-invariant diffusion model for pocket-aware peptide generation", "abstract": "  Biologists frequently desire protein inhibitors for a variety of reasons,\nincluding use as research tools for understanding biological processes and\napplication to societal problems in agriculture, healthcare, etc.\nImmunotherapy, for instance, relies on immune checkpoint inhibitors to block\ncheckpoint proteins, preventing their binding with partner proteins and\nboosting immune cell function against abnormal cells. Inhibitor discovery has\nlong been a tedious process, which in recent years has been accelerated by\ncomputational approaches. Advances in artificial intelligence now provide an\nopportunity to make inhibitor discovery smarter than ever before. While\nextensive research has been conducted on computer-aided inhibitor discovery, it\nhas mainly focused on either sequence-to-structure mapping, reverse mapping, or\nbio-activity prediction, making it unrealistic for biologists to utilize such\ntools. Instead, our work proposes a new method of computer-assisted inhibitor\ndiscovery: de novo pocket-aware peptide structure and sequence generation\nnetwork. Our approach consists of two sequential diffusion models for\nend-to-end structure generation and sequence prediction. By leveraging angle\nand dihedral relationships between backbone atoms, we ensure an E(3)-invariant\nrepresentation of peptide structures. Our results demonstrate that our method\nachieves comparable performance to state-of-the-art models, highlighting its\npotential in pocket-aware peptide design. This work offers a new approach for\nprecise drug discovery using receptor-specific peptide generation.\n", "date": "2024-10-27", "categories": []}
{"id": "2411.00004", "title": "RapidDock: Unlocking Proteome-scale Molecular Docking", "abstract": "  Accelerating molecular docking -- the process of predicting how molecules\nbind to protein targets -- could boost small-molecule drug discovery and\nrevolutionize medicine. Unfortunately, current molecular docking tools are too\nslow to screen potential drugs against all relevant proteins, which often\nresults in missed drug candidates or unexpected side effects occurring in\nclinical trials. To address this gap, we introduce RapidDock, an efficient\ntransformer-based model for blind molecular docking. RapidDock achieves at\nleast a $100 \\times$ speed advantage over existing methods without compromising\naccuracy. On the Posebusters and DockGen benchmarks, our method achieves\n$52.1\\%$ and $44.0\\%$ success rates ($\\text{RMSD}<2$\\r{A}), respectively. The\naverage inference time is $0.04$ seconds on a single GPU, highlighting\nRapidDock's potential for large-scale docking studies. We examine the key\nfeatures of RapidDock that enable leveraging the transformer architecture for\nmolecular docking, including the use of relative distance embeddings of $3$D\nstructures in attention matrices, pre-training on protein folding, and a custom\nloss function invariant to molecular symmetries.\n", "date": "2024-10-16", "categories": []}
{"id": "2411.00054", "title": "eDOC: Explainable Decoding Out-of-domain Cell Types with Evidential\n  Learning", "abstract": "  Single-cell RNA-seq (scRNA-seq) technology is a powerful tool for unraveling\nthe complexity of biological systems. One of essential and fundamental tasks in\nscRNA-seq data analysis is Cell Type Annotation (CTA). In spite of tremendous\nefforts in developing machine learning methods for this problem, several\nchallenges remains. They include identifying Out-of-Domain (OOD) cell types,\nquantifying the uncertainty of unseen cell type annotations, and determining\ninterpretable cell type-specific gene drivers for an OOD case. OOD cell types\nare often associated with therapeutic responses and disease origins, making\nthem critical for precision medicine and early disease diagnosis. Additionally,\nscRNA-seq data contains tens thousands of gene expressions. Pinpointing gene\ndrivers underlying CTA can provide deep insight into gene regulatory mechanisms\nand serve as disease biomarkers. In this study, we develop a new method, eDOC,\nto address aforementioned challenges. eDOC leverages a transformer architecture\nwith evidential learning to annotate In-Domain (IND) and OOD cell types as well\nas to highlight genes that contribute both IND cells and OOD cells in a single\ncell resolution. Rigorous experiments demonstrate that eDOC significantly\nimproves the efficiency and effectiveness of OOD cell type and gene driver\nidentification compared to other state-of-the-art methods. Our findings suggest\nthat eDOC may provide new insights into single-cell biology.\n", "date": "2024-10-30", "categories": []}
{"id": "2406.11906", "title": "NovoBench: Benchmarking Deep Learning-based De Novo Peptide Sequencing\n  Methods in Proteomics", "abstract": "  Tandem mass spectrometry has played a pivotal role in advancing proteomics,\nenabling the high-throughput analysis of protein composition in biological\ntissues. Many deep learning methods have been developed for \\emph{de novo}\npeptide sequencing task, i.e., predicting the peptide sequence for the observed\nmass spectrum. However, two key challenges seriously hinder the further\nadvancement of this important task. Firstly, since there is no consensus for\nthe evaluation datasets, the empirical results in different research papers are\noften not comparable, leading to unfair comparison. Secondly, the current\nmethods are usually limited to amino acid-level or peptide-level precision and\nrecall metrics. In this work, we present the first unified benchmark NovoBench\nfor \\emph{de novo} peptide sequencing, which comprises diverse mass spectrum\ndata, integrated models, and comprehensive evaluation metrics. Recent\nimpressive methods, including DeepNovo, PointNovo, Casanovo, InstaNovo, AdaNovo\nand $\\pi$-HelixNovo are integrated into our framework. In addition to amino\nacid-level and peptide-level precision and recall, we evaluate the models'\nperformance in terms of identifying post-tranlational modifications (PTMs),\nefficiency and robustness to peptide length, noise peaks and missing fragment\nratio, which are important influencing factors while seldom be considered.\nLeveraging this benchmark, we conduct a large-scale study of current methods,\nreport many insightful findings that open up new possibilities for future\ndevelopment.\n", "date": "2024-06-16", "categories": []}
{"id": "2407.00002", "title": "Kermut: Composite kernel regression for protein variant effects", "abstract": "  Reliable prediction of protein variant effects is crucial for both protein\noptimization and for advancing biological understanding. For practical use in\nprotein engineering, it is important that we can also provide reliable\nuncertainty estimates for our predictions, and while prediction accuracy has\nseen much progress in recent years, uncertainty metrics are rarely reported. We\nhere provide a Gaussian process regression model, Kermut, with a novel\ncomposite kernel for modeling mutation similarity, which obtains\nstate-of-the-art performance for supervised protein variant effect prediction\nwhile also offering estimates of uncertainty through its posterior. An analysis\nof the quality of the uncertainty estimates demonstrates that our model\nprovides meaningful levels of overall calibration, but that instance-specific\nuncertainty calibration remains more challenging.\n", "date": "2024-04-09", "categories": []}
{"id": "2410.18475", "title": "Gene-Metabolite Association Prediction with Interactive Knowledge\n  Transfer Enhanced Graph for Metabolite Production", "abstract": "  In the rapidly evolving field of metabolic engineering, the quest for\nefficient and precise gene target identification for metabolite production\nenhancement presents significant challenges. Traditional approaches, whether\nknowledge-based or model-based, are notably time-consuming and labor-intensive,\ndue to the vast scale of research literature and the approximation nature of\ngenome-scale metabolic model (GEM) simulations. Therefore, we propose a new\ntask, Gene-Metabolite Association Prediction based on metabolic graphs, to\nautomate the process of candidate gene discovery for a given pair of metabolite\nand candidate-associated genes, as well as presenting the first benchmark\ncontaining 2474 metabolites and 1947 genes of two commonly used microorganisms\nSaccharomyces cerevisiae (SC) and Issatchenkia orientalis (IO). This task is\nchallenging due to the incompleteness of the metabolic graphs and the\nheterogeneity among distinct metabolisms. To overcome these limitations, we\npropose an Interactive Knowledge Transfer mechanism based on Metabolism Graph\n(IKT4Meta), which improves the association prediction accuracy by integrating\nthe knowledge from different metabolism graphs. First, to build a bridge\nbetween two graphs for knowledge transfer, we utilize Pretrained Language\nModels (PLMs) with external knowledge of genes and metabolites to help generate\ninter-graph links, significantly alleviating the impact of heterogeneity.\nSecond, we propagate intra-graph links from different metabolic graphs using\ninter-graph links as anchors. Finally, we conduct the gene-metabolite\nassociation prediction based on the enriched metabolism graphs, which integrate\nthe knowledge from multiple microorganisms. Experiments on both types of\norganisms demonstrate that our proposed methodology outperforms baselines by up\nto 12.3% across various link prediction frameworks.\n", "date": "2024-10-24", "categories": []}
{"id": "2402.00077", "title": "Unlocking the Power of Multi-institutional Data: Integrating and\n  Harmonizing Genomic Data Across Institutions", "abstract": "  Cancer is a complex disease driven by genomic alterations, and tumor\nsequencing is becoming a mainstay of clinical care for cancer patients. The\nemergence of multi-institution sequencing data presents a powerful resource for\nlearning real-world evidence to enhance precision oncology. GENIE BPC, led by\nthe American Association for Cancer Research, establishes a unique database\nlinking genomic data with clinical information for patients treated at multiple\ncancer centers. However, leveraging such multi-institutional sequencing data\npresents significant challenges. Variations in gene panels result in loss of\ninformation when the analysis is conducted on common gene sets. Additionally,\ndifferences in sequencing techniques and patient heterogeneity across\ninstitutions add complexity. High data dimensionality, sparse gene mutation\npatterns, and weak signals at the individual gene level further complicate\nmatters. Motivated by these real-world challenges, we introduce the Bridge\nmodel. It uses a quantile-matched latent variable approach to derive integrated\nfeatures to preserve information beyond common genes and maximize the\nutilization of all available data while leveraging information sharing to\nenhance both learning efficiency and the model's capacity to generalize. By\nextracting harmonized and noise-reduced lower-dimensional latent variables, the\ntrue mutation pattern unique to each individual is captured. We assess the\nmodel's performance and parameter estimation through extensive simulation\nstudies. The extracted latent features from the Bridge model consistently excel\nin predicting patient survival across six cancer types in GENIE BPC data.\n", "date": "2024-01-30", "categories": []}
{"id": "2410.22949", "title": "MutaPLM: Protein Language Modeling for Mutation Explanation and\n  Engineering", "abstract": "  Studying protein mutations within amino acid sequences holds tremendous\nsignificance in life sciences. Protein language models (PLMs) have demonstrated\nstrong capabilities in broad biological applications. However, due to\narchitectural design and lack of supervision, PLMs model mutations implicitly\nwith evolutionary plausibility, which is not satisfactory to serve as\nexplainable and engineerable tools in real-world studies. To address these\nissues, we present MutaPLM, a unified framework for interpreting and navigating\nprotein mutations with protein language models. MutaPLM introduces a protein\ndelta network that captures explicit protein mutation representations within a\nunified feature space, and a transfer learning pipeline with a chain-of-thought\n(CoT) strategy to harvest protein mutation knowledge from biomedical texts. We\nalso construct MutaDescribe, the first large-scale protein mutation dataset\nwith rich textual annotations, which provides cross-modal supervision signals.\nThrough comprehensive experiments, we demonstrate that MutaPLM excels at\nproviding human-understandable explanations for mutational effects and\nprioritizing novel mutations with desirable properties. Our code, model, and\ndata are open-sourced at https://github.com/PharMolix/MutaPLM.\n", "date": "2024-10-30", "categories": []}
{"id": "2410.21069", "title": "EMOCPD: Efficient Attention-based Models for Computational Protein\n  Design Using Amino Acid Microenvironment", "abstract": "  Computational protein design (CPD) refers to the use of computational methods\nto design proteins. Traditional methods relying on energy functions and\nheuristic algorithms for sequence design are inefficient and do not meet the\ndemands of the big data era in biomolecules, with their accuracy limited by the\nenergy functions and search algorithms. Existing deep learning methods are\nconstrained by the learning capabilities of the networks, failing to extract\neffective information from sparse protein structures, which limits the accuracy\nof protein design. To address these shortcomings, we developed an Efficient\nattention-based Models for Computational Protein Design using amino acid\nmicroenvironment (EMOCPD). It aims to predict the category of each amino acid\nin a protein by analyzing the three-dimensional atomic environment surrounding\nthe amino acids, and optimize the protein based on the predicted\nhigh-probability potential amino acid categories. EMOCPD employs a multi-head\nattention mechanism to focus on important features in the sparse protein\nmicroenvironment and utilizes an inverse residual structure to optimize the\nnetwork architecture. The proposed EMOCPD achieves over 80% accuracy on the\ntraining set and 68.33% and 62.32% accuracy on two independent test sets,\nrespectively, surpassing the best comparative methods by over 10%. In protein\ndesign, the thermal stability and protein expression of the predicted mutants\nfrom EMOCPD show significant improvements compared to the wild type,\neffectively validating EMOCPD's potential in designing superior proteins.\nFurthermore, the predictions of EMOCPD are influenced positively, negatively,\nor have minimal impact based on the content of the 20 amino acids, categorizing\namino acids as positive, negative, or neutral. Research findings indicate that\nEMOCPD is more suitable for designing proteins with lower contents of negative\namino acids.\n", "date": "2024-10-28", "categories": []}
{"id": "2410.21345", "title": "Absorb & Escape: Overcoming Single Model Limitations in Generating\n  Genomic Sequences", "abstract": "  Abstract Recent advances in immunology and synthetic biology have accelerated\nthe development of deep generative methods for DNA sequence design. Two\ndominant approaches in this field are AutoRegressive (AR) models and Diffusion\nModels (DMs). However, genomic sequences are functionally heterogeneous,\nconsisting of multiple connected regions (e.g., Promoter Regions, Exons, and\nIntrons) where elements within each region come from the same probability\ndistribution, but the overall sequence is non-homogeneous. This heterogeneous\nnature presents challenges for a single model to accurately generate genomic\nsequences. In this paper, we analyze the properties of AR models and DMs in\nheterogeneous genomic sequence generation, pointing out crucial limitations in\nboth methods: (i) AR models capture the underlying distribution of data by\nfactorizing and learning the transition probability but fail to capture the\nglobal property of DNA sequences. (ii) DMs learn to recover the global\ndistribution but tend to produce errors at the base pair level. To overcome the\nlimitations of both approaches, we propose a post-training sampling method,\ntermed Absorb & Escape (A&E) to perform compositional generation from AR models\nand DMs. This approach starts with samples generated by DMs and refines the\nsample quality using an AR model through the alternation of the Absorb and\nEscape steps. To assess the quality of generated sequences, we conduct\nextensive experiments on 15 species for conditional and unconditional DNA\ngeneration. The experiment results from motif distribution, diversity checks,\nand genome integration tests unequivocally show that A&E outperforms\nstate-of-the-art AR models and DMs in genomic sequence generation.\n", "date": "2024-10-28", "categories": []}
{"id": "2403.16576", "title": "Antigen-Specific Antibody Design via Direct Energy-based Preference\n  Optimization", "abstract": "  Antibody design, a crucial task with significant implications across various\ndisciplines such as therapeutics and biology, presents considerable challenges\ndue to its intricate nature. In this paper, we tackle antigen-specific antibody\nsequence-structure co-design as an optimization problem towards specific\npreferences, considering both rationality and functionality. Leveraging a\npre-trained conditional diffusion model that jointly models sequences and\nstructures of antibodies with equivariant neural networks, we propose direct\nenergy-based preference optimization to guide the generation of antibodies with\nboth rational structures and considerable binding affinities to given antigens.\nOur method involves fine-tuning the pre-trained diffusion model using a\nresidue-level decomposed energy preference. Additionally, we employ gradient\nsurgery to address conflicts between various types of energy, such as\nattraction and repulsion. Experiments on RAbD benchmark show that our approach\neffectively optimizes the energy of generated antibodies and achieves\nstate-of-the-art performance in designing high-quality antibodies with low\ntotal energy and high binding affinity simultaneously, demonstrating the\nsuperiority of our approach.\n", "date": "2024-03-25", "categories": []}
{"id": "2406.05347", "title": "MSAGPT: Neural Prompting Protein Structure Prediction via MSA Generative\n  Pre-Training", "abstract": "  Multiple Sequence Alignment (MSA) plays a pivotal role in unveiling the\nevolutionary trajectories of protein families. The accuracy of protein\nstructure predictions is often compromised for protein sequences that lack\nsufficient homologous information to construct high quality MSA. Although\nvarious methods have been proposed to generate virtual MSA under these\nconditions, they fall short in comprehensively capturing the intricate\ncoevolutionary patterns within MSA or require guidance from external oracle\nmodels. Here we introduce MSAGPT, a novel approach to prompt protein structure\npredictions via MSA generative pretraining in the low MSA regime. MSAGPT\nemploys a simple yet effective 2D evolutionary positional encoding scheme to\nmodel complex evolutionary patterns. Endowed by this, its flexible 1D MSA\ndecoding framework facilitates zero or few shot learning. Moreover, we\ndemonstrate that leveraging the feedback from AlphaFold2 can further enhance\nthe model capacity via Rejective Fine tuning (RFT) and Reinforcement Learning\nfrom AF2 Feedback (RLAF). Extensive experiments confirm the efficacy of MSAGPT\nin generating faithful virtual MSA to enhance the structure prediction\naccuracy. The transfer learning capabilities also highlight its great potential\nfor facilitating other protein tasks.\n", "date": "2024-06-08", "categories": []}
{"id": "2407.01648", "title": "Aligning Target-Aware Molecule Diffusion Models with Exact Energy\n  Optimization", "abstract": "  Generating ligand molecules for specific protein targets, known as\nstructure-based drug design, is a fundamental problem in therapeutics\ndevelopment and biological discovery. Recently, target-aware generative models,\nespecially diffusion models, have shown great promise in modeling\nprotein-ligand interactions and generating candidate drugs. However, existing\nmodels primarily focus on learning the chemical distribution of all drug\ncandidates, which lacks effective steerability on the chemical quality of model\ngenerations. In this paper, we propose a novel and general alignment framework\nto align pretrained target diffusion models with preferred functional\nproperties, named AliDiff. AliDiff shifts the target-conditioned chemical\ndistribution towards regions with higher binding affinity and structural\nrationality, specified by user-defined reward functions, via the preference\noptimization approach. To avoid the overfitting problem in common preference\noptimization objectives, we further develop an improved Exact Energy Preference\nOptimization method to yield an exact and efficient alignment of the diffusion\nmodels, and provide the closed-form expression for the converged distribution.\nEmpirical studies on the CrossDocked2020 benchmark show that AliDiff can\ngenerate molecules with state-of-the-art binding energies with up to -7.07 Avg.\nVina Score, while maintaining strong molecular properties. Code is available at\nhttps://github.com/MinkaiXu/AliDiff.\n", "date": "2024-07-01", "categories": []}
{"id": "2410.19922", "title": "Disentangling Genotype and Environment Specific Latent Features for\n  Improved Trait Prediction using a Compositional Autoencoder", "abstract": "  This study introduces a compositional autoencoder (CAE) framework designed to\ndisentangle the complex interplay between genotypic and environmental factors\nin high-dimensional phenotype data to improve trait prediction in plant\nbreeding and genetics programs. Traditional predictive methods, which use\ncompact representations of high-dimensional data through handcrafted features\nor latent features like PCA or more recently autoencoders, do not separate\ngenotype-specific and environment-specific factors. We hypothesize that\ndisentangling these features into genotype-specific and environment-specific\ncomponents can enhance predictive models. To test this, we developed a\ncompositional autoencoder (CAE) that decomposes high-dimensional data into\ndistinct genotype-specific and environment-specific latent features.\n  Our CAE framework employs a hierarchical architecture within an autoencoder\nto effectively separate these entangled latent features. Applied to a maize\ndiversity panel dataset, the CAE demonstrates superior modeling of\nenvironmental influences and 5-10 times improved predictive performance for key\ntraits like Days to Pollen and Yield, compared to the traditional methods,\nincluding standard autoencoders, PCA with regression, and Partial Least Squares\nRegression (PLSR). By disentangling latent features, the CAE provides powerful\ntool for precision breeding and genetic research. This work significantly\nenhances trait prediction models, advancing agricultural and biological\nsciences.\n", "date": "2024-10-25", "categories": []}
{"id": "2410.20232", "title": "SAFE setup for generative molecular design", "abstract": "  SMILES-based molecular generative models have been pivotal in drug design but\nface challenges in fragment-constrained tasks. To address this, the Sequential\nAttachment-based Fragment Embedding (SAFE) representation was recently\nintroduced as an alternative that streamlines those tasks. In this study, we\ninvestigate the optimal setups for training SAFE generative models, focusing on\ndataset size, data augmentation through randomization, model architecture, and\nbond disconnection algorithms. We found that larger, more diverse datasets\nimprove performance, with the LLaMA architecture using Rotary Positional\nEmbedding proving most robust. SAFE-based models also consistently outperform\nSMILES-based approaches in scaffold decoration and linker design, particularly\nwith BRICS decomposition yielding the best results. These insights highlight\nkey factors that significantly impact the efficacy of SAFE-based generative\nmodels.\n", "date": "2024-10-26", "categories": []}
{"id": "2410.07981", "title": "MolMix: A Simple Yet Effective Baseline for Multimodal Molecular\n  Representation Learning", "abstract": "  In this work, we propose a simple transformer-based baseline for multimodal\nmolecular representation learning, integrating three distinct modalities:\nSMILES strings, 2D graph representations, and 3D conformers of molecules. A key\naspect of our approach is the aggregation of 3D conformers, allowing the model\nto account for the fact that molecules can adopt multiple conformations-an\nimportant factor for accurate molecular representation. The tokens for each\nmodality are extracted using modality-specific encoders: a transformer for\nSMILES strings, a message-passing neural network for 2D graphs, and an\nequivariant neural network for 3D conformers. The flexibility and modularity of\nthis framework enable easy adaptation and replacement of these encoders, making\nthe model highly versatile for different molecular tasks. The extracted tokens\nare then combined into a unified multimodal sequence, which is processed by a\ndownstream transformer for prediction tasks. To efficiently scale our model for\nlarge multimodal datasets, we utilize Flash Attention 2 and bfloat16 precision.\nDespite its simplicity, our approach achieves state-of-the-art results across\nmultiple datasets, demonstrating its effectiveness as a strong baseline for\nmultimodal molecular representation learning.\n", "date": "2024-10-10", "categories": []}
{"id": "2211.03054", "title": "ODBAE: a high-performance model identifying complex phenotypes in\n  high-dimensional biological datasets", "abstract": "  Identifying complex phenotypes from high-dimensional biological data is\nchallenging due to the intricate interdependencies among different\nphysiological indicators. Traditional approaches often focus on detecting\noutliers in single variables, overlooking the broader network of interactions\nthat contribute to phenotype emergence. Here, we introduce ODBAE (Outlier\nDetection using Balanced Autoencoders), a machine learning method designed to\nuncover both subtle and extreme outliers by capturing latent relationships\namong multiple physiological parameters. ODBAE's revised loss function enhances\nits ability to detect two key types of outliers: influential points (IP), which\ndisrupt latent correlations between dimensions, and high leverage points (HLP),\nwhich deviate from the norm but go undetected by traditional autoencoder-based\nmethods. Using data from the International Mouse Phenotyping Consortium (IMPC),\nwe show that ODBAE can identify knockout mice with complex, multi-indicator\nphenotypes - normal in individual traits, but abnormal when considered\ntogether. In addition, this method reveals novel metabolism-related genes and\nuncovers coordinated abnormalities across metabolic indicators. Our results\nhighlight the utility of ODBAE in detecting joint abnormalities and advancing\nour understanding of homeostatic perturbations in biological systems.\n", "date": "2022-11-06", "categories": []}
{"id": "2406.08649", "title": "MOTIVE: A Drug-Target Interaction Graph For Inductive Link Prediction", "abstract": "  Drug-target interaction (DTI) prediction is crucial for identifying new\ntherapeutics and detecting mechanisms of action. While structure-based methods\naccurately model physical interactions between a drug and its protein target,\ncell-based assays such as Cell Painting can better capture complex DTI\ninteractions. This paper introduces MOTIVE, a Morphological cOmpound Target\nInteraction Graph dataset comprising Cell Painting features for 11,000 genes\nand 3,600 compounds, along with their relationships extracted from seven\npublicly available databases. We provide random, cold-source (new drugs), and\ncold-target (new genes) data splits to enable rigorous evaluation under\nrealistic use cases. Our benchmark results show that graph neural networks that\nuse Cell Painting features consistently outperform those that learn from graph\nstructure alone, feature-based models, and topological heuristics. MOTIVE\naccelerates both graph ML research and drug discovery by promoting the\ndevelopment of more reliable DTI prediction models. MOTIVE resources are\navailable at https://github.com/carpenter-singh-lab/motive.\n", "date": "2024-06-12", "categories": []}
{"id": "2410.15592", "title": "CPE-Pro: A Structure-Sensitive Deep Learning Method for Protein\n  Representation and Origin Evaluation", "abstract": "  Protein structures are important for understanding their functions and\ninteractions. Currently, many protein structure prediction methods are\nenriching the structure database. Discriminating the origin of structures is\ncrucial for distinguishing between experimentally resolved and computationally\npredicted structures, evaluating the reliability of prediction methods, and\nguiding downstream biological studies. Building on works in structure\nprediction, We developed a structure-sensitive supervised deep learning model,\nCrystal vs Predicted Evaluator for Protein Structure (CPE-Pro), to represent\nand discriminate the origin of protein structures. CPE-Pro learns the\nstructural information of proteins and captures inter-structural differences to\nachieve accurate traceability on four data classes, and is expected to be\nextended to more. Simultaneously, we utilized Foldseek to encode protein\nstructures into \"structure-sequences\" and trained a protein Structural Sequence\nLanguage Model, SSLM. Preliminary experiments demonstrated that, compared to\nlarge-scale protein language models pre-trained on vast amounts of amino acid\nsequences, the \"structure-sequence\" enables the language model to learn more\ninformative protein features, enhancing and optimizing structural\nrepresentations. We have provided the code, model weights, and all related\nmaterials on https://github.com/GouWenrui/CPE-Pro-main.git.\n", "date": "2024-10-21", "categories": []}
{"id": "2410.16700", "title": "AskBeacon -- Performing genomic data exchange and analytics with natural\n  language", "abstract": "  Enabling clinicians and researchers to directly interact with global genomic\ndata resources by removing technological barriers is vital for medical\ngenomics. AskBeacon enables Large Language Models to be applied to securely\nshared cohorts via the GA4GH Beacon protocol. By simply \"asking\" Beacon,\nactionable insights can be gained, analyzed and made publication-ready.\n", "date": "2024-10-22", "categories": []}
{"id": "2312.06682", "title": "Learning to Denoise Biomedical Knowledge Graph for Robust Molecular\n  Interaction Prediction", "abstract": "  Molecular interaction prediction plays a crucial role in forecasting unknown\ninteractions between molecules, such as drug-target interaction (DTI) and\ndrug-drug interaction (DDI), which are essential in the field of drug discovery\nand therapeutics. Although previous prediction methods have yielded promising\nresults by leveraging the rich semantics and topological structure of\nbiomedical knowledge graphs (KGs), they have primarily focused on enhancing\npredictive performance without addressing the presence of inevitable noise and\ninconsistent semantics. This limitation has hindered the advancement of\nKG-based prediction methods. To address this limitation, we propose BioKDN\n(Biomedical Knowledge Graph Denoising Network) for robust molecular interaction\nprediction. BioKDN refines the reliable structure of local subgraphs by\ndenoising noisy links in a learnable manner, providing a general module for\nextracting task-relevant interactions. To enhance the reliability of the\nrefined structure, BioKDN maintains consistent and robust semantics by\nsmoothing relations around the target interaction. By maximizing the mutual\ninformation between reliable structure and smoothed relations, BioKDN\nemphasizes informative semantics to enable precise predictions. Experimental\nresults on real-world datasets show that BioKDN surpasses state-of-the-art\nmodels in DTI and DDI prediction tasks, confirming the effectiveness and\nrobustness of BioKDN in denoising unreliable interactions within contaminated\nKGs\n", "date": "2023-12-09", "categories": []}
{"id": "2410.16291", "title": "Accelerating Biological Spatial Cluster Analysis with the Parallel\n  Integral Image Technique", "abstract": "  Spatial cluster analysis (SCA) offers valuable insights into biological\nimages; a common SCA technique is sliding window analysis (SWA). Unfortunately,\nSWA's computational cost hinders its application to larger images, limiting its\nuse to small-scale images. With advancements in high-resolution microscopy,\nimages now exceed the capabilities of previous SWA approaches, reaching sizes\nup to 70,000 by 85,000 pixels. To overcome these limitations, this paper\nintroduces the parallel integral image approach to SWA, surpassing previous\nmethods. We achieve a remarkable speedup of 131,806x on small-scale images and\nconsistent speedups of over 10,000x on a variety of large-scale microscopy\nimages. We analyze the computational complexity advantages of the parallel\nintegral image approach and present experimental results that validate the\nsuperior performance of integral-image-based methods. Our approach is made\navailable as an open-source Python PIP package available at\nhttps://github.com/OckermanSethGVSU/BioPII.\n", "date": "2024-10-06", "categories": []}
{"id": "2410.16302", "title": "Computational design of target-specific linear peptide binders with\n  TransformerBeta", "abstract": "  The computational prediction and design of peptide binders targeting specific\nlinear epitopes is crucial in biological and biomedical research, yet it\nremains challenging due to their highly dynamic nature and the scarcity of\nexperimentally solved binding data. To address this problem, we built an\nunprecedentedly large-scale library of peptide pairs within stable secondary\nstructures (beta sheets), leveraging newly available AlphaFold predicted\nstructures. We then developed a machine learning method based on the\nTransformer architecture for the design of specific linear binders, in analogy\nto a language translation task. Our method, TransformerBeta, accurately\npredicts specific beta strand interactions and samples sequences with beta\nsheet-like molecular properties, while capturing interpretable physico-chemical\ninteraction patterns. As such, it can propose specific candidate binders\ntargeting linear epitope for experimental validation to inform protein design.\n", "date": "2024-10-07", "categories": []}
{"id": "2410.16474", "title": "QuickBind: A Light-Weight And Interpretable Molecular Docking Model", "abstract": "  Predicting a ligand's bound pose to a target protein is a key component of\nearly-stage computational drug discovery. Recent developments in machine\nlearning methods have focused on improving pose quality at the cost of model\nruntime. For high-throughput virtual screening applications, this exposes a\ncapability gap that can be filled by moderately accurate but fast pose\nprediction. To this end, we developed QuickBind, a light-weight pose prediction\nalgorithm. We assess QuickBind on widely used benchmarks and find that it\nprovides an attractive trade-off between model accuracy and runtime. To\nfacilitate virtual screening applications, we augment QuickBind with a binding\naffinity module and demonstrate its capabilities for multiple\nclinically-relevant drug targets. Finally, we investigate the mechanistic basis\nby which QuickBind makes predictions and find that it has learned key\nphysicochemical properties of molecular docking, providing new insights into\nhow machine learning models generate protein-ligand poses. By virtue of its\nsimplicity, QuickBind can serve as both an effective virtual screening tool and\na minimal test bed for exploring new model architectures and innovations. Model\ncode and weights are available at https://github.com/aqlaboratory/QuickBind .\n", "date": "2024-10-21", "categories": []}
{"id": "2410.16673", "title": "Efficient Antibody Structure Refinement Using Energy-Guided SE(3) Flow\n  Matching", "abstract": "  Antibodies are proteins produced by the immune system that recognize and bind\nto specific antigens, and their 3D structures are crucial for understanding\ntheir binding mechanism and designing therapeutic interventions. The\nspecificity of antibody-antigen binding predominantly depends on the\ncomplementarity-determining regions (CDR) within antibodies. Despite recent\nadvancements in antibody structure prediction, the quality of predicted CDRs\nremains suboptimal. In this paper, we develop a novel antibody structure\nrefinement method termed FlowAB based on energy-guided flow matching. FlowAB\nadopts the powerful deep generative method SE(3) flow matching and\nsimultaneously incorporates important physical prior knowledge into the flow\nmodel to guide the generation process. The extensive experiments demonstrate\nthat FlowAB can significantly improve the antibody CDR structures. It achieves\nnew state-of-the-art performance on the antibody structure prediction task when\nused in conjunction with an appropriate prior model while incurring only\nmarginal computational overhead. This advantage makes FlowAB a practical tool\nin antibody engineering.\n", "date": "2024-10-22", "categories": []}
{"id": "2410.17173", "title": "Reinforcement learning on structure-conditioned categorical diffusion\n  for protein inverse folding", "abstract": "  Protein inverse folding-that is, predicting an amino acid sequence that will\nfold into the desired 3D structure-is an important problem for structure-based\nprotein design. Machine learning based methods for inverse folding typically\nuse recovery of the original sequence as the optimization objective. However,\ninverse folding is a one-to-many problem where several sequences can fold to\nthe same structure. Moreover, for many practical applications, it is often\ndesirable to have multiple, diverse sequences that fold into the target\nstructure since it allows for more candidate sequences for downstream\noptimizations. Here, we demonstrate that although recent inverse folding\nmethods show increased sequence recovery, their \"foldable diversity\"-i.e. their\nability to generate multiple non-similar sequences that fold into the\nstructures consistent with the target-does not increase. To address this, we\npresent RL-DIF, a categorical diffusion model for inverse folding that is\npre-trained on sequence recovery and tuned via reinforcement learning on\nstructural consistency. We find that RL-DIF achieves comparable sequence\nrecovery and structural consistency to benchmark models but shows greater\nfoldable diversity: experiments show RL-DIF can achieve an foldable diversity\nof 29% on CATH 4.2, compared to 23% from models trained on the same dataset.\nThe PyTorch model weights and sampling code are available on GitHub.\n", "date": "2024-10-22", "categories": []}
{"id": "2306.06276", "title": "Contrastive Learning for Predicting Cancer Prognosis Using Gene\n  Expression Values", "abstract": "  Recent advancements in image classification have demonstrated that\ncontrastive learning (CL) can aid in further learning tasks by acquiring good\nfeature representation from a limited number of data samples. In this paper, we\napplied CL to tumor transcriptomes and clinical data to learn feature\nrepresentations in a low-dimensional space. We then utilized these learned\nfeatures to train a classifier to categorize tumors into a high- or low-risk\ngroup of recurrence. Using data from The Cancer Genome Atlas (TCGA), we\ndemonstrated that CL can significantly improve classification accuracy.\nSpecifically, our CL-based classifiers achieved an area under the receiver\noperating characteristic curve (AUC) greater than 0.8 for 14 types of cancer,\nand an AUC greater than 0.9 for 2 types of cancer. We also developed CL-based\nCox (CLCox) models for predicting cancer prognosis. Our CLCox models trained\nwith the TCGA data outperformed existing methods significantly in predicting\nthe prognosis of 19 types of cancer under consideration. The performance of\nCLCox models and CL-based classifiers trained with TCGA lung and prostate\ncancer data were validated using the data from two independent cohorts. We also\nshow that the CLCox model trained with the whole transcriptome significantly\noutperforms the Cox model trained with the 21 genes of Oncotype DX that is in\nclinical use for breast cancer patients. CL-based classifiers and CLCox models\nfor 19 types of cancer are publicly available and can be used to predict cancer\nprognosis using the RNA-seq transcriptome of an individual tumor. Python codes\nfor model training and testing are also publicly accessible, and can be applied\nto train new CL-based models using gene expression data of tumors.\n", "date": "2023-06-09", "categories": []}
{"id": "2310.03946", "title": "Improved prediction of ligand-protein binding affinities by\n  meta-modeling", "abstract": "  The accurate screening of candidate drug ligands against target proteins\nthrough computational approaches is of prime interest to drug development\nefforts. Such virtual screening depends in part on methods to predict the\nbinding affinity between ligands and proteins. Many computational models for\nbinding affinity prediction have been developed, but with varying results\nacross targets. Given that ensembling or meta-modeling approaches have shown\ngreat promise in reducing model-specific biases, we develop a framework to\nintegrate published force-field-based empirical docking and sequence-based deep\nlearning models. In building this framework, we evaluate many combinations of\nindividual base models, training databases, and several meta-modeling\napproaches. We show that many of our meta-models significantly improve affinity\npredictions over base models. Our best meta-models achieve comparable\nperformance to state-of-the-art deep learning tools exclusively based on 3D\nstructures, while allowing for improved database scalability and flexibility\nthrough the explicit inclusion of features such as physicochemical properties\nor molecular descriptors. We further demonstrate improved generalization\ncapability by our models using a large-scale benchmark of affinity prediction\nas well as a virtual screening application benchmark. Overall, we demonstrate\nthat diverse modeling approaches can be ensembled together to gain meaningful\nimprovement in binding affinity prediction.\n", "date": "2023-10-05", "categories": []}
{"id": "2410.09968", "title": "Deep-Ace: LSTM-based Prokaryotic Lysine Acetylation Site Predictor", "abstract": "  Acetylation of lysine residues (K-Ace) is a post-translation modification\noccurring in both prokaryotes and eukaryotes. It plays a crucial role in\ndisease pathology and cell biology hence it is important to identify these\nK-Ace sites. In the past, many machine learning-based models using hand-crafted\nfeatures and encodings have been used to find and analyze the characteristics\nof K-Ace sites however these methods ignore long term relationships within\nsequences and therefore observe performance degradation. In the current work we\npropose Deep-Ace, a deep learning-based framework using Long-Short-Term-Memory\n(LSTM) network which has the ability to understand and encode long-term\nrelationships within a sequence. Such relations are vital for learning\ndiscriminative and effective sequence representations. In the work reported\nhere, the use of LSTM to extract deep features as well as for prediction of\nK-Ace sites using fully connected layers for eight different species of\nprokaryotic models (including B. subtilis, C. glutamicum, E. coli, G.\nkaustophilus, S. eriocheiris, B. velezensis, S. typhimurium, and M.\ntuberculosis) has been explored. Our proposed method has outperformed existing\nstate of the art models achieving accuracy as 0.80, 0.79, 0.71, 0.75, 0.80,\n0.83, 0.756, and 0.82 respectively for eight bacterial species mentioned above.\nThe method with minor modifications can be used for eukaryotic systems and can\nserve as a tool for the prognosis and diagnosis of various diseases in humans.\n", "date": "2024-10-13", "categories": []}
{"id": "2410.16109", "title": "Interpreting Microbiome Relative Abundance Data Using Symbolic\n  Regression", "abstract": "  Understanding the complex interactions within the microbiome is crucial for\ndeveloping effective diagnostic and therapeutic strategies. Traditional machine\nlearning models often lack interpretability, which is essential for clinical\nand biological insights. This paper explores the application of symbolic\nregression (SR) to microbiome relative abundance data, with a focus on\ncolorectal cancer (CRC). SR, known for its high interpretability, is compared\nagainst traditional machine learning models, e.g., random forest, gradient\nboosting decision trees. These models are evaluated based on performance\nmetrics such as F1 score and accuracy. We utilize 71 studies encompassing, from\nvarious cohorts, over 10,000 samples across 749 species features. Our results\nindicate that SR not only competes reasonably well in terms of predictive\nperformance, but also excels in model interpretability. SR provides explicit\nmathematical expressions that offer insights into the biological relationships\nwithin the microbiome, a crucial advantage for clinical and biological\ninterpretation. Our experiments also show that SR can help understand complex\nmodels like XGBoost via knowledge distillation. To aid in reproducibility and\nfurther research, we have made the code openly available at\nhttps://github.com/swag2198/microbiome-symbolic-regression .\n", "date": "2024-10-18", "categories": []}
{"id": "2410.14426", "title": "Predicting time-varying flux and balance in metabolic systems using\n  structured neural-ODE processes", "abstract": "  We develop a novel data-driven framework as an alternative to dynamic flux\nbalance analysis, bypassing the demand for deep domain knowledge and manual\nefforts to formulate the optimization problem. The proposed framework is\nend-to-end, which trains a structured neural ODE process (SNODEP) model to\nestimate flux and balance samples using gene-expression time-series data.\nSNODEP is designed to circumvent the limitations of the standard neural ODE\nprocess model, including restricting the latent and decoder sampling\ndistributions to be normal and lacking structure between context points for\ncalculating the latent, thus more suitable for modeling the underlying dynamics\nof a metabolic system. Through comprehensive experiments ($156$ in total), we\ndemonstrate that SNODEP not only predicts the unseen time points of real-world\ngene-expression data and the flux and balance estimates well but can even\ngeneralize to more challenging unseen knockout configurations and irregular\ndata sampling scenarios, all essential for metabolic pathway analysis. We hope\nour work can serve as a catalyst for building more scalable and powerful models\nfor genome-scale metabolic analysis. Our code is available at:\n\\url{https://github.com/TrustMLRG/SNODEP}.\n", "date": "2024-10-18", "categories": []}
{"id": "2410.14433", "title": "A Bioinformatic Approach Validated Utilizing Machine Learning Algorithms\n  to Identify Relevant Biomarkers and Crucial Pathways in Gallbladder Cancer", "abstract": "  Gallbladder cancer (GBC) is the most frequent cause of disease among biliary\ntract neoplasms. Identifying the molecular mechanisms and biomarkers linked to\nGBC progression has been a significant challenge in scientific research. Few\nrecent studies have explored the roles of biomarkers in GBC. Our study aimed to\nidentify biomarkers in GBC using machine learning (ML) and bioinformatics\ntechniques. We compared GBC tumor samples with normal samples to identify\ndifferentially expressed genes (DEGs) from two microarray datasets (GSE100363,\nGSE139682) obtained from the NCBI GEO database. A total of 146 DEGs were found,\nwith 39 up-regulated and 107 down-regulated genes. Functional enrichment\nanalysis of these DEGs was performed using Gene Ontology (GO) terms and\nREACTOME pathways through DAVID. The protein-protein interaction network was\nconstructed using the STRING database. To identify hub genes, we applied three\nranking algorithms: Degree, MNC, and Closeness Centrality. The intersection of\nhub genes from these algorithms yielded 11 hub genes. Simultaneously, two\nfeature selection methods (Pearson correlation and recursive feature\nelimination) were used to identify significant gene subsets. We then developed\nML models using SVM and RF on the GSE100363 dataset, with validation on\nGSE139682, to determine the gene subset that best distinguishes GBC samples.\nThe hub genes outperformed the other gene subsets. Finally, NTRK2, COL14A1,\nSCN4B, ATP1A2, SLC17A7, SLIT3, COL7A1, CLDN4, CLEC3B, ADCYAP1R1, and MFAP4 were\nidentified as crucial genes, with SLIT3, COL7A1, and CLDN4 being strongly\nlinked to GBC development and prediction.\n", "date": "2024-10-18", "categories": []}
{"id": "2409.13057", "title": "Natural Language Processing Methods for the Study of Protein-Ligand\n  Interactions", "abstract": "  Recent advances in Natural Language Processing (NLP) have ignited interest in\ndeveloping effective methods for predicting protein-ligand interactions (PLIs)\ngiven their relevance to drug discovery and protein engineering efforts and the\never-growing volume of biochemical sequence and structural data available. The\nparallels between human languages and the \"languages\" used to represent\nproteins and ligands have enabled the use of NLP machine learning approaches to\nadvance PLI studies. In this review, we explain where and how such approaches\nhave been applied in the recent literature and discuss useful mechanisms such\nas long short-term memory, transformers, and attention. We conclude with a\ndiscussion of the current limitations of NLP methods for the study of PLIs as\nwell as key challenges that need to be addressed in future work.\n", "date": "2024-09-19", "categories": []}
{"id": "2410.13257", "title": "scFusionTTT: Single-cell transcriptomics and proteomics fusion with\n  Test-Time Training layers", "abstract": "  Single-cell multi-omics (scMulti-omics) refers to the paired multimodal data,\nsuch as Cellular Indexing of Transcriptomes and Epitopes by Sequencing\n(CITE-seq), where the regulation of each cell was measured from different\nmodalities, i.e. genes and proteins. scMulti-omics can reveal heterogeneity\ninside tumors and understand the distinct genetic properties of diverse cell\ntypes, which is crucial to targeted therapy. Currently, deep learning methods\nbased on attention structures in the bioinformatics area face two challenges.\nThe first challenge is the vast number of genes in a single cell. Traditional\nattention-based modules struggled to effectively leverage all gene information\ndue to their limited capacity for long-context learning and high-complexity\ncomputing. The second challenge is that genes in the human genome are ordered\nand influence each other's expression. Most of the methods ignored this\nsequential information. The recently introduced Test-Time Training (TTT) layer\nis a novel sequence modeling approach, particularly suitable for handling long\ncontexts like genomics data because TTT layer is a linear complexity sequence\nmodeling structure and is better suited to data with sequential relationships.\nIn this paper, we propose scFusionTTT, a novel method for Single-Cell\nmultimodal omics Fusion with TTT-based masked autoencoder. Of note, we combine\nthe order information of genes and proteins in the human genome with the TTT\nlayer, fuse multimodal omics, and enhance unimodal omics analysis. Finally, the\nmodel employs a three-stage training strategy, which yielded the best\nperformance across most metrics in four multimodal omics datasets and four\nunimodal omics datasets, demonstrating the superior performance of our model.\nThe dataset and code will be available on\nhttps://github.com/DM0815/scFusionTTT.\n", "date": "2024-10-17", "categories": []}
{"id": "2405.18749", "title": "A SARS-CoV-2 Interaction Dataset and VHH Sequence Corpus for Antibody\n  Language Models", "abstract": "  Antibodies are crucial proteins produced by the immune system to eliminate\nharmful foreign substances and have become pivotal therapeutic agents for\ntreating human diseases. To accelerate the discovery of antibody therapeutics,\nthere is growing interest in constructing language models using antibody\nsequences. However, the applicability of pre-trained language models for\nantibody discovery has not been thoroughly evaluated due to the scarcity of\nlabeled datasets. To overcome these limitations, we introduce AVIDa-SARS-CoV-2,\na dataset featuring the antigen-variable domain of heavy chain of heavy chain\nantibody (VHH) interactions obtained from two alpacas immunized with severe\nacute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike proteins.\nAVIDa-SARS-CoV-2 includes binary labels indicating the binding or non-binding\nof diverse VHH sequences to 12 SARS-CoV-2 mutants, such as the Delta and\nOmicron variants. Furthermore, we release VHHCorpus-2M, a pre-training dataset\nfor antibody language models, containing over two million VHH sequences. We\nreport benchmark results for predicting SARS-CoV-2-VHH binding using VHHBERT\npre-trained on VHHCorpus-2M and existing general protein and antibody-specific\npre-trained language models. These results confirm that AVIDa-SARS-CoV-2\nprovides valuable benchmarks for evaluating the representation capabilities of\nantibody language models for binding prediction, thereby facilitating the\ndevelopment of AI-driven antibody discovery. The datasets are available at\nhttps://datasets.cognanous.com.\n", "date": "2024-05-29", "categories": []}
{"id": "2410.11224", "title": "DeltaDock: A Unified Framework for Accurate, Efficient, and Physically\n  Reliable Molecular Docking", "abstract": "  Molecular docking, a technique for predicting ligand binding poses, is\ncrucial in structure-based drug design for understanding protein-ligand\ninteractions. Recent advancements in docking methods, particularly those\nleveraging geometric deep learning (GDL), have demonstrated significant\nefficiency and accuracy advantages over traditional sampling methods. Despite\nthese advancements, current methods are often tailored for specific docking\nsettings, and limitations such as the neglect of protein side-chain structures,\ndifficulties in handling large binding pockets, and challenges in predicting\nphysically valid structures exist. To accommodate various docking settings and\nachieve accurate, efficient, and physically reliable docking, we propose a\nnovel two-stage docking framework, DeltaDock, consisting of pocket prediction\nand site-specific docking. We innovatively reframe the pocket prediction task\nas a pocket-ligand alignment problem rather than direct prediction in the first\nstage. Then we follow a bi-level coarse-to-fine iterative refinement process to\nperform site-specific docking. Comprehensive experiments demonstrate the\nsuperior performance of DeltaDock. Notably, in the blind docking setting,\nDeltaDock achieves a 31\\% relative improvement over the docking success rate\ncompared with the previous state-of-the-art GDL model. With the consideration\nof physical validity, this improvement increases to about 300\\%.\n", "date": "2024-10-15", "categories": []}
{"id": "2410.11910", "title": "Explainable AI Methods for Multi-Omics Analysis: A Survey", "abstract": "  Advancements in high-throughput technologies have led to a shift from\ntraditional hypothesis-driven methodologies to data-driven approaches.\nMulti-omics refers to the integrative analysis of data derived from multiple\n'omes', such as genomics, proteomics, transcriptomics, metabolomics, and\nmicrobiomics. This approach enables a comprehensive understanding of biological\nsystems by capturing different layers of biological information. Deep learning\nmethods are increasingly utilized to integrate multi-omics data, offering\ninsights into molecular interactions and enhancing research into complex\ndiseases. However, these models, with their numerous interconnected layers and\nnonlinear relationships, often function as black boxes, lacking transparency in\ndecision-making processes. To overcome this challenge, explainable artificial\nintelligence (xAI) methods are crucial for creating transparent models that\nallow clinicians to interpret and work with complex data more effectively. This\nreview explores how xAI can improve the interpretability of deep learning\nmodels in multi-omics research, highlighting its potential to provide\nclinicians with clear insights, thereby facilitating the effective application\nof such models in clinical settings.\n", "date": "2024-10-15", "categories": []}
{"id": "2410.12249", "title": "Devil in the Tail: A Multi-Modal Framework for Drug-Drug Interaction\n  Prediction in Long Tail Distinction", "abstract": "  Drug-drug interaction (DDI) identification is a crucial aspect of\npharmacology research. There are many DDI types (hundreds), and they are not\nevenly distributed with equal chance to occur. Some of the rarely occurred DDI\ntypes are often high risk and could be life-critical if overlooked,\nexemplifying the long-tailed distribution problem. Existing models falter\nagainst this distribution challenge and overlook the multi-faceted nature of\ndrugs in DDI prediction. In this paper, a novel multi-modal deep learning-based\nframework, namely TFDM, is introduced to leverage multiple properties of a drug\nto achieve DDI classification. The proposed framework fuses multimodal features\nof drugs, including graph-based, molecular structure, Target and Enzyme, for\nDDI identification. To tackle the challenge posed by the distribution skewness\nacross categories, a novel loss function called Tailed Focal Loss is\nintroduced, aimed at further enhancing the model performance and address\ngradient vanishing problem of focal loss in extremely long-tailed dataset.\nIntensive experiments over 4 challenging long-tailed dataset demonstrate that\nthe TFMD outperforms the most recent SOTA methods in long-tailed DDI\nclassification tasks. The source code is released to reproduce our experiment\nresults: https://github.com/IcurasLW/TFMD_Longtailed_DDI.git\n", "date": "2024-10-16", "categories": []}
{"id": "2410.12655", "title": "Position Specific Scoring Is All You Need? Revisiting Protein Sequence\n  Classification Tasks", "abstract": "  Understanding the structural and functional characteristics of proteins are\ncrucial for developing preventative and curative strategies that impact fields\nfrom drug discovery to policy development. An important and popular technique\nfor examining how amino acids make up these characteristics of the protein\nsequences with position-specific scoring (PSS). While the string kernel is\ncrucial in natural language processing (NLP), it is unclear if string kernels\ncan extract biologically meaningful information from protein sequences, despite\nthe fact that they have been shown to be effective in the general sequence\nanalysis tasks. In this work, we propose a weighted PSS kernel matrix (or\nW-PSSKM), that combines a PSS representation of protein sequences, which\nencodes the frequency information of each amino acid in a sequence, with the\nnotion of the string kernel. This results in a novel kernel function that\noutperforms many other approaches for protein sequence classification. We\nperform extensive experimentation to evaluate the proposed method. Our findings\ndemonstrate that the W-PSSKM significantly outperforms existing baselines and\nstate-of-the-art methods and achieves up to 45.1\\% improvement in\nclassification accuracy.\n", "date": "2024-10-16", "categories": []}
{"id": "2410.11499", "title": "BSM: Small but Powerful Biological Sequence Model for Genes and Proteins", "abstract": "  Modeling biological sequences such as DNA, RNA, and proteins is crucial for\nunderstanding complex processes like gene regulation and protein synthesis.\nHowever, most current models either focus on a single type or treat multiple\ntypes of data separately, limiting their ability to capture cross-modal\nrelationships. We propose that by learning the relationships between these\nmodalities, the model can enhance its understanding of each type. To address\nthis, we introduce BSM, a small but powerful mixed-modal biological sequence\nfoundation model, trained on three types of data: RefSeq, Gene Related\nSequences, and interleaved biological sequences from the web. These datasets\ncapture the genetic flow, gene-protein relationships, and the natural\nco-occurrence of diverse biological data, respectively. By training on\nmixed-modal data, BSM significantly enhances learning efficiency and\ncross-modal representation, outperforming models trained solely on unimodal\ndata. With only 110M parameters, BSM achieves performance comparable to much\nlarger models across both single-modal and mixed-modal tasks, and uniquely\ndemonstrates in-context learning capability for mixed-modal tasks, which is\nabsent in existing models. Further scaling to 270M parameters demonstrates even\ngreater performance gains, highlighting the potential of BSM as a significant\nadvancement in multimodal biological sequence modeling.\n", "date": "2024-10-15", "categories": []}
{"id": "2311.16207", "title": "ALNSynergy: a graph convolutional network with multi-representation\n  alignment for drug synergy prediction", "abstract": "  Drug combination refers to the use of two or more drugs to treat a specific\ndisease at the same time. It is currently the mainstream way to treat complex\ndiseases. Compared with single drugs, drug combinations have better efficacy\nand can better inhibit toxicity and drug resistance. The computational model\nbased on deep learning concatenates the representation of multiple drugs and\nthe corresponding cell line feature as input, and the output is whether the\ndrug combination can have an inhibitory effect on the cell line. However, this\nstrategy of concatenating multiple representations has the following defects:\nthe alignment of drug representation and cell line representation is ignored,\nresulting in the synergistic relationship not being reflected positionally in\nthe embedding space. Moreover, the alignment measurement function in deep\nlearning cannot be suitable for drug synergy prediction tasks due to\ndifferences in input types. Therefore, in this work, we propose ALNSynergy, a\ngraph convolutional network with multi-representation alignment for predicting\ndrug synergy. In the ALNSynergy model, we designed a multi-representation\nalignment function suitable for the drug synergy prediction task so that the\npositional relationship between drug representations and cell line\nrepresentation is reflected in the embedding space. In addition, the vector\nmodulus of drug representations and cell line representation is considered to\nimprove the accuracy of calculation results and accelerate model convergence.\nFinally, many relevant experiments were run on multiple drug synergy datasets\nto verify the effectiveness of the above innovative elements and the excellence\nof the ALNSynergy model.\n", "date": "2023-11-27", "categories": []}
{"id": "2408.00040", "title": "Barlow Twins Deep Neural Network for Advanced 1D Drug-Target Interaction\n  Prediction", "abstract": "  Accurate prediction of drug-target interactions is critical for advancing\ndrug discovery. By reducing time and cost, machine learning and deep learning\ncan accelerate this laborious discovery process. In a novel approach,\nBarlowDTI, we utilise the powerful Barlow Twins architecture for\nfeature-extraction while considering the structure of the target protein. Our\nmethod achieves state-of-the-art predictive performance against multiple\nestablished benchmarks using only one-dimensional input. The use of gradient\nboosting machine as the underlying predictor ensures fast and efficient\npredictions without the need for substantial computational resources. We also\ninvestigate how the model reaches its decision based on individual training\nsamples. By comparing co-crystal structures, we find that BarlowDTI effectively\nexploits catalytically active and stabilising residues, highlighting the\nmodel's ability to generalise from one-dimensional input data. In addition, we\nfurther benchmark new baselines against existing methods. Together, these\ninnovations improve the efficiency and effectiveness of drug-target interaction\npredictions, providing robust tools for accelerating drug development and\ndeepening the understanding of molecular interactions. Therefore, we provide an\neasy-to-use web interface that can be freely accessed at\nhttps://www.bio.nat.tum.de/oc2/barlowdti .\n", "date": "2024-07-31", "categories": []}
{"id": "2409.11654", "title": "How to Build the Virtual Cell with Artificial Intelligence: Priorities\n  and Opportunities", "abstract": "  The cell is arguably the most fundamental unit of life and is central to\nunderstanding biology. Accurate modeling of cells is important for this\nunderstanding as well as for determining the root causes of disease. Recent\nadvances in artificial intelligence (AI), combined with the ability to generate\nlarge-scale experimental data, present novel opportunities to model cells. Here\nwe propose a vision of leveraging advances in AI to construct virtual cells,\nhigh-fidelity simulations of cells and cellular systems under different\nconditions that are directly learned from biological data across measurements\nand scales. We discuss desired capabilities of such AI Virtual Cells, including\ngenerating universal representations of biological entities across scales, and\nfacilitating interpretable in silico experiments to predict and understand\ntheir behavior using virtual instruments. We further address the challenges,\nopportunities and requirements to realize this vision including data needs,\nevaluation strategies, and community standards and engagement to ensure\nbiological accuracy and broad utility. We envision a future where AI Virtual\nCells help identify new drug targets, predict cellular responses to\nperturbations, as well as scale hypothesis exploration. With open science\ncollaborations across the biomedical ecosystem that includes academia,\nphilanthropy, and the biopharma and AI industries, a comprehensive predictive\nunderstanding of cell mechanisms and interactions has come into reach.\n", "date": "2024-09-18", "categories": []}
{"id": "2410.09543", "title": "Boltzmann-Aligned Inverse Folding Model as a Predictor of Mutational\n  Effects on Protein-Protein Interactions", "abstract": "  Predicting the change in binding free energy ($\\Delta \\Delta G$) is crucial\nfor understanding and modulating protein-protein interactions, which are\ncritical in drug design. Due to the scarcity of experimental $\\Delta \\Delta G$\ndata, existing methods focus on pre-training, while neglecting the importance\nof alignment. In this work, we propose the Boltzmann Alignment technique to\ntransfer knowledge from pre-trained inverse folding models to $\\Delta \\Delta G$\nprediction. We begin by analyzing the thermodynamic definition of $\\Delta\n\\Delta G$ and introducing the Boltzmann distribution to connect energy with\nprotein conformational distribution. However, the protein conformational\ndistribution is intractable; therefore, we employ Bayes' theorem to circumvent\ndirect estimation and instead utilize the log-likelihood provided by protein\ninverse folding models for $\\Delta \\Delta G$ estimation. Compared to previous\ninverse folding-based methods, our method explicitly accounts for the unbound\nstate of protein complex in the $\\Delta \\Delta G$ thermodynamic cycle,\nintroducing a physical inductive bias and achieving both supervised and\nunsupervised state-of-the-art (SoTA) performance. Experimental results on\nSKEMPI v2 indicate that our method achieves Spearman coefficients of 0.3201\n(unsupervised) and 0.5134 (supervised), significantly surpassing the previously\nreported SoTA values of 0.2632 and 0.4324, respectively. Futhermore, we\ndemonstrate the capability of our method on binding energy prediction,\nprotein-protein docking and antibody optimization tasks.\n", "date": "2024-10-12", "categories": []}
{"id": "2410.10144", "title": "Unified Representation of Genomic and Biomedical Concepts through\n  Multi-Task, Multi-Source Contrastive Learning", "abstract": "  We introduce GENomic Encoding REpresentation with Language Model (GENEREL), a\nframework designed to bridge genetic and biomedical knowledge bases. What sets\nGENEREL apart is its ability to fine-tune language models to infuse biological\nknowledge behind clinical concepts such as diseases and medications. This\nfine-tuning enables the model to capture complex biomedical relationships more\neffectively, enriching the understanding of how genomic data connects to\nclinical outcomes. By constructing a unified embedding space for biomedical\nconcepts and a wide range of common SNPs from sources such as patient-level\ndata, biomedical knowledge graphs, and GWAS summaries, GENEREL aligns the\nembeddings of SNPs and clinical concepts through multi-task contrastive\nlearning. This allows the model to adapt to diverse natural language\nrepresentations of biomedical concepts while bypassing the limitations of\ntraditional code mapping systems across different data sources. Our experiments\ndemonstrate GENEREL's ability to effectively capture the nuanced relationships\nbetween SNPs and clinical concepts. GENEREL also emerges to discern the degree\nof relatedness, potentially allowing for a more refined identification of\nconcepts. This pioneering approach in constructing a unified embedding system\nfor both SNPs and biomedical concepts enhances the potential for data\nintegration and discovery in biomedical research.\n", "date": "2024-10-14", "categories": []}
{"id": "2410.08867", "title": "Prediction by Machine Learning Analysis of Genomic Data Phenotypic Frost\n  Tolerance in Perccottus glenii", "abstract": "  Analysis of the genome sequence of Perccottus glenii, the only fish known to\npossess freeze tolerance, holds significant importance for understanding how\norganisms adapt to extreme environments, Traditional biological analysis\nmethods are time-consuming and have limited accuracy, To address these issues,\nwe will employ machine learning techniques to analyze the gene sequences of\nPerccottus glenii, with Neodontobutis hainanens as a comparative group,\nFirstly, we have proposed five gene sequence vectorization methods and a method\nfor handling ultra-long gene sequences, We conducted a comparative study on the\nthree vectorization methods: ordinal encoding, One-Hot encoding, and K-mer\nencoding, to identify the optimal encoding method, Secondly, we constructed\nfour classification models: Random Forest, LightGBM, XGBoost, and Decision\nTree, The dataset used by these classification models was extracted from the\nNational Center for Biotechnology Information database, and we vectorized the\nsequence matrices using the optimal encoding method, K-mer, The Random Forest\nmodel, which is the optimal model, achieved a classification accuracy of up to\n99, 98 , Lastly, we utilized SHAP values to conduct an interpretable analysis\nof the optimal classification model, Through ten-fold cross-validation and the\nAUC metric, we identified the top 10 features that contribute the most to the\nmodel's classification accuracy, This demonstrates that machine learning\nmethods can effectively replace traditional manual analysis in identifying\ngenes associated with the freeze tolerance phenotype in Perccottus glenii.\n", "date": "2024-10-11", "categories": []}
{"id": "2406.10840", "title": "CBGBench: Fill in the Blank of Protein-Molecule Complex Binding Graph", "abstract": "  Structure-based drug design (SBDD) aims to generate potential drugs that can\nbind to a target protein and is greatly expedited by the aid of AI techniques\nin generative models. However, a lack of systematic understanding persists due\nto the diverse settings, complex implementation, difficult reproducibility, and\ntask singularity. Firstly, the absence of standardization can lead to unfair\ncomparisons and inconclusive insights. To address this dilemma, we propose\nCBGBench, a comprehensive benchmark for SBDD, that unifies the task as a\ngenerative heterogeneous graph completion, analogous to fill-in-the-blank of\nthe 3D complex binding graph. By categorizing existing methods based on their\nattributes, CBGBench facilitates a modular and extensible framework that\nimplements various cutting-edge methods. Secondly, a single task on \\textit{de\nnovo} molecule generation can hardly reflect their capabilities. To broaden the\nscope, we have adapted these models to a range of tasks essential in drug\ndesign, which are considered sub-tasks within the graph fill-in-the-blank\ntasks. These tasks include the generative designation of \\textit{de novo}\nmolecules, linkers, fragments, scaffolds, and sidechains, all conditioned on\nthe structures of protein pockets. Our evaluations are conducted with fairness,\nencompassing comprehensive perspectives on interaction, chemical properties,\ngeometry authenticity, and substructure validity. We further provide the\npre-trained versions of the state-of-the-art models and deep insights with\nanalysis from empirical studies. The codebase for CBGBench is publicly\naccessible at \\url{https://github.com/Edapinenut/CBGBench}.\n", "date": "2024-06-16", "categories": []}
{"id": "2410.07260", "title": "Precision Cancer Classification and Biomarker Identification from mRNA\n  Gene Expression via Dimensionality Reduction and Explainable AI", "abstract": "  Gene expression analysis is a critical method for cancer classification,\nenabling precise diagnoses through the identification of unique molecular\nsignatures associated with various tumors. Identifying cancer-specific genes\nfrom gene expression values enables a more tailored and personalized treatment\napproach. However, the high dimensionality of mRNA gene expression data poses\nchallenges for analysis and data extraction. This research presents a\ncomprehensive pipeline designed to accurately identify 33 distinct cancer types\nand their corresponding gene sets. It incorporates a combination of\nnormalization and feature selection techniques to reduce dataset dimensionality\neffectively while ensuring high performance. Notably, our pipeline successfully\nidentifies a substantial number of cancer-specific genes using a reduced\nfeature set of just 500, in contrast to using the full dataset comprising\n19,238 features. By employing an ensemble approach that combines three\ntop-performing classifiers, a classification accuracy of 96.61% was achieved.\nFurthermore, we leverage Explainable AI to elucidate the biological\nsignificance of the identified cancer-specific genes, employing Differential\nGene Expression (DGE) analysis.\n", "date": "2024-10-08", "categories": []}
{"id": "2408.12817", "title": "Data-Driven Parametrization of Molecular Mechanics Force Fields for\n  Expansive Chemical Space Coverage", "abstract": "  A force field is a critical component in molecular dynamics simulations for\ncomputational drug discovery. It must achieve high accuracy within the\nconstraints of molecular mechanics' (MM) limited functional forms, which offers\nhigh computational efficiency. With the rapid expansion of synthetically\naccessible chemical space, traditional look-up table approaches face\nsignificant challenges. In this study, we address this issue using a modern\ndata-driven approach, developing ByteFF, an Amber-compatible force field for\ndrug-like molecules. To create ByteFF, we generated an expansive and highly\ndiverse molecular dataset at the B3LYP-D3(BJ)/DZVP level of theory. This\ndataset includes 2.4 million optimized molecular fragment geometries with\nanalytical Hessian matrices, along with 3.2 million torsion profiles. We then\ntrained an edge-augmented, symmetry-preserving molecular graph neural network\n(GNN) on this dataset, employing a carefully optimized training strategy. Our\nmodel predicts all bonded and non-bonded MM force field parameters for\ndrug-like molecules simultaneously across a broad chemical space. ByteFF\ndemonstrates state-of-the-art performance on various benchmark datasets,\nexcelling in predicting relaxed geometries, torsional energy profiles, and\nconformational energies and forces. Its exceptional accuracy and expansive\nchemical space coverage make ByteFF a valuable tool for multiple stages of\ncomputational drug discovery.\n", "date": "2024-08-23", "categories": []}
{"id": "2410.06211", "title": "A mechanistically interpretable neural network for regulatory genomics", "abstract": "  Deep neural networks excel in mapping genomic DNA sequences to associated\nreadouts (e.g., protein-DNA binding). Beyond prediction, the goal of these\nnetworks is to reveal to scientists the underlying motifs (and their syntax)\nwhich drive genome regulation. Traditional methods that extract motifs from\nconvolutional filters suffer from the uninterpretable dispersion of information\nacross filters and layers. Other methods which rely on importance scores can be\nunstable and unreliable. Instead, we designed a novel mechanistically\ninterpretable architecture for regulatory genomics, where motifs and their\nsyntax are directly encoded and readable from the learned weights and\nactivations. We provide theoretical and empirical evidence of our\narchitecture's full expressivity, while still being highly interpretable.\nThrough several experiments, we show that our architecture excels in de novo\nmotif discovery and motif instance calling, is robust to variable sequence\ncontexts, and enables fully interpretable generation of novel functional\nsequences.\n", "date": "2024-10-08", "categories": []}
{"id": "2409.06744", "title": "ProteinBench: A Holistic Evaluation of Protein Foundation Models", "abstract": "  Recent years have witnessed a surge in the development of protein foundation\nmodels, significantly improving performance in protein prediction and\ngenerative tasks ranging from 3D structure prediction and protein design to\nconformational dynamics. However, the capabilities and limitations associated\nwith these models remain poorly understood due to the absence of a unified\nevaluation framework. To fill this gap, we introduce ProteinBench, a holistic\nevaluation framework designed to enhance the transparency of protein foundation\nmodels. Our approach consists of three key components: (i) A taxonomic\nclassification of tasks that broadly encompass the main challenges in the\nprotein domain, based on the relationships between different protein\nmodalities; (ii) A multi-metric evaluation approach that assesses performance\nacross four key dimensions: quality, novelty, diversity, and robustness; and\n(iii) In-depth analyses from various user objectives, providing a holistic view\nof model performance. Our comprehensive evaluation of protein foundation models\nreveals several key findings that shed light on their current capabilities and\nlimitations. To promote transparency and facilitate further research, we\nrelease the evaluation dataset, code, and a public leaderboard publicly for\nfurther analysis and a general modular toolkit. We intend for ProteinBench to\nbe a living benchmark for establishing a standardized, in-depth evaluation\nframework for protein foundation models, driving their development and\napplication while fostering collaboration within the field.\n", "date": "2024-09-10", "categories": []}
{"id": "2410.03927", "title": "End-to-End Reaction Field Energy Modeling via Deep Learning based\n  Voxel-to-voxel Transform", "abstract": "  In computational biochemistry and biophysics, understanding the role of\nelectrostatic interactions is crucial for elucidating the structure, dynamics,\nand function of biomolecules. The Poisson-Boltzmann (PB) equation is a\nfoundational tool for modeling these interactions by describing the\nelectrostatic potential in and around charged molecules. However, solving the\nPB equation presents significant computational challenges due to the complexity\nof biomolecular surfaces and the need to account for mobile ions. While\ntraditional numerical methods for solving the PB equation are accurate, they\nare computationally expensive and scale poorly with increasing system size. To\naddress these challenges, we introduce PBNeF, a novel machine learning approach\ninspired by recent advancements in neural network-based partial differential\nequation solvers. Our method formulates the input and boundary electrostatic\nconditions of the PB equation into a learnable voxel representation, enabling\nthe use of a neural field transformer to predict the PB solution and,\nsubsequently, the reaction field potential energy. Extensive experiments\ndemonstrate that PBNeF achieves over a 100-fold speedup compared to traditional\nPB solvers, while maintaining accuracy comparable to the Generalized Born (GB)\nmodel.\n", "date": "2024-10-04", "categories": []}
{"id": "2406.12056", "title": "Learning Molecular Representation in a Cell", "abstract": "  Predicting drug efficacy and safety in vivo requires information on\nbiological responses (e.g., cell morphology and gene expression) to small\nmolecule perturbations. However, current molecular representation learning\nmethods do not provide a comprehensive view of cell states under these\nperturbations and struggle to remove noise, hindering model generalization. We\nintroduce the Information Alignment (InfoAlign) approach to learn molecular\nrepresentations through the information bottleneck method in cells. We\nintegrate molecules and cellular response data as nodes into a context graph,\nconnecting them with weighted edges based on chemical, biological, and\ncomputational criteria. For each molecule in a training batch, InfoAlign\noptimizes the encoder's latent representation with a minimality objective to\ndiscard redundant structural information. A sufficiency objective decodes the\nrepresentation to align with different feature spaces from the molecule's\nneighborhood in the context graph. We demonstrate that the proposed sufficiency\nobjective for alignment is tighter than existing encoder-based contrastive\nmethods. Empirically, we validate representations from InfoAlign in two\ndownstream applications: molecular property prediction against up to 27\nbaseline methods across four datasets, plus zero-shot molecule-morphology\nmatching.\n", "date": "2024-06-17", "categories": []}
{"id": "2410.01853", "title": "Recovering Time-Varying Networks From Single-Cell Data", "abstract": "  Gene regulation is a dynamic process that underlies all aspects of human\ndevelopment, disease response, and other key biological processes. The\nreconstruction of temporal gene regulatory networks has conventionally relied\non regression analysis, graphical models, or other types of relevance networks.\nWith the large increase in time series single-cell data, new approaches are\nneeded to address the unique scale and nature of this data for reconstructing\nsuch networks. Here, we develop a deep neural network, Marlene, to infer\ndynamic graphs from time series single-cell gene expression data. Marlene\nconstructs directed gene networks using a self-attention mechanism where the\nweights evolve over time using recurrent units. By employing meta learning, the\nmodel is able to recover accurate temporal networks even for rare cell types.\nIn addition, Marlene can identify gene interactions relevant to specific\nbiological responses, including COVID-19 immune response, fibrosis, and aging.\n", "date": "2024-10-01", "categories": []}
{"id": "2410.01858", "title": "Long-range gene expression prediction with token alignment of large\n  language model", "abstract": "  Gene expression is a cellular process that plays a fundamental role in human\nphenotypical variations and diseases. Despite advances of deep learning models\nfor gene expression prediction, recent benchmarks have revealed their inability\nto learn distal regulatory grammar. Here, we address this challenge by\nleveraging a pretrained large language model to enhance gene expression\nprediction. We introduce Genetic sequence Token Alignment (GTA), which aligns\ngenetic sequence features with natural language tokens, allowing for symbolic\nreasoning of genomic sequence features via the frozen language model. This\ncross-modal adaptation learns the regulatory grammar and allows us to further\nincorporate gene-specific human annotations as prompts, enabling in-context\nlearning that is not possible with existing models. Trained on lymphoblastoid\ncells, GTA was evaluated on cells from the Geuvadis consortium and outperforms\nstate-of-the-art models such as Enformer, achieving a Spearman correlation of\n0.65, a 10\\% improvement. Additionally, GTA offers improved interpretation of\nlong-range interactions through the identification of the most meaningful\nsections of the input genetic context. GTA represents a powerful and novel\ncross-modal approach to gene expression prediction by utilizing a pretrained\nlanguage model, in a paradigm shift from conventional gene expression models\ntrained only on sequence data.\n", "date": "2024-10-02", "categories": []}
{"id": "2410.02085", "title": "Multi-Omic and Quantum Machine Learning Integration for Lung Subtypes\n  Classification", "abstract": "  Quantum Machine Learning (QML) is a red-hot field that brings novel\ndiscoveries and exciting opportunities to resolve, speed up, or refine the\nanalysis of a wide range of computational problems. In the realm of biomedical\nresearch and personalized medicine, the significance of multi-omics integration\nlies in its ability to provide a thorough and holistic comprehension of complex\nbiological systems. This technology links fundamental research to clinical\npractice. The insights gained from integrated omics data can be translated into\nclinical tools for diagnosis, prognosis, and treatment planning. The fusion of\nquantum computing and machine learning holds promise for unraveling complex\npatterns within multi-omics datasets, providing unprecedented insights into the\nmolecular landscape of lung cancer. Due to the heterogeneity, complexity, and\nhigh dimensionality of multi-omic cancer data, characterized by the vast number\nof features (such as gene expression, micro-RNA, and DNA methylation) relative\nto the limited number of lung cancer patient samples, our prime motivation for\nthis paper is the integration of multi-omic data, unique feature selection, and\ndiagnostic classification of lung subtypes: lung squamous cell carcinoma\n(LUSC-I) and lung adenocarcinoma (LUAD-II) using quantum machine learning. We\ndeveloped a method for finding the best differentiating features between LUAD\nand LUSC datasets, which has the potential for biomarker discovery.\n", "date": "2024-10-02", "categories": []}
{"id": "2409.09828", "title": "Latent Diffusion Models for Controllable RNA Sequence Generation", "abstract": "  This work presents RNAdiffusion, a latent diffusion model for generating and\noptimizing discrete RNA sequences of variable lengths. RNA is a key\nintermediary between DNA and protein, exhibiting high sequence diversity and\ncomplex three-dimensional structures to support a wide range of functions. We\nutilize pretrained BERT-type models to encode raw RNA sequences into\ntoken-level, biologically meaningful representations. A Query Transformer is\nemployed to compress such representations into a set of fixed-length latent\nvectors, with an autoregressive decoder trained to reconstruct RNA sequences\nfrom these latent variables. We then develop a continuous diffusion model\nwithin this latent space. To enable optimization, we integrate the gradients of\nreward models--surrogates for RNA functional properties--into the backward\ndiffusion process, thereby generating RNAs with high reward scores. Empirical\nresults confirm that RNAdiffusion generates non-coding RNAs that align with\nnatural distributions across various biological metrics. Further, we fine-tune\nthe diffusion model on mRNA 5' untranslated regions (5'-UTRs) and optimize\nsequences for high translation efficiencies. Our guided diffusion model\neffectively generates diverse 5'-UTRs with high Mean Ribosome Loading (MRL) and\nTranslation Efficiency (TE), outperforming baselines in balancing rewards and\nstructural stability trade-off. Our findings hold potential for advancing RNA\nsequence-function research and therapeutic RNA design.\n", "date": "2024-09-15", "categories": []}
{"id": "2410.01784", "title": "OmniGenBench: Automating Large-scale in-silico Benchmarking for Genomic\n  Foundation Models", "abstract": "  The advancements in artificial intelligence in recent years, such as Large\nLanguage Models (LLMs), have fueled expectations for breakthroughs in genomic\nfoundation models (GFMs). The code of nature, hidden in diverse genomes since\nthe very beginning of life's evolution, holds immense potential for impacting\nhumans and ecosystems through genome modeling. Recent breakthroughs in GFMs,\nsuch as Evo, have attracted significant investment and attention to genomic\nmodeling, as they address long-standing challenges and transform in-silico\ngenomic studies into automated, reliable, and efficient paradigms. In the\ncontext of this flourishing era of consecutive technological revolutions in\ngenomics, GFM studies face two major challenges: the lack of GFM benchmarking\ntools and the absence of open-source software for diverse genomics. These\nchallenges hinder the rapid evolution of GFMs and their wide application in\ntasks such as understanding and synthesizing genomes, problems that have\npersisted for decades. To address these challenges, we introduce GFMBench, a\nframework dedicated to GFM-oriented benchmarking. GFMBench standardizes\nbenchmark suites and automates benchmarking for a wide range of open-source\nGFMs. It integrates millions of genomic sequences across hundreds of genomic\ntasks from four large-scale benchmarks, democratizing GFMs for a wide range of\nin-silico genomic applications. Additionally, GFMBench is released as\nopen-source software, offering user-friendly interfaces and diverse tutorials,\napplicable for AutoBench and complex tasks like RNA design and structure\nprediction. To facilitate further advancements in genome modeling, we have\nlaunched a public leaderboard showcasing the benchmark performance derived from\nAutoBench. GFMBench represents a step toward standardizing GFM benchmarking and\ndemocratizing GFM applications.\n", "date": "2024-10-02", "categories": []}
{"id": "2405.16206", "title": "GlycanML: A Multi-Task and Multi-Structure Benchmark for Glycan Machine\n  Learning", "abstract": "  Glycans are basic biomolecules and perform essential functions within living\norganisms. The rapid increase of functional glycan data provides a good\nopportunity for machine learning solutions to glycan understanding. However,\nthere still lacks a standard machine learning benchmark for glycan property and\nfunction prediction. In this work, we fill this blank by building a\ncomprehensive benchmark for Glycan Machine Learning (GlycanML). The GlycanML\nbenchmark consists of diverse types of tasks including glycan taxonomy\nprediction, glycan immunogenicity prediction, glycosylation type prediction,\nand protein-glycan interaction prediction. Glycans can be represented by both\nsequences and graphs in GlycanML, which enables us to extensively evaluate\nsequence-based models and graph neural networks (GNNs) on benchmark tasks.\nFurthermore, by concurrently performing eight glycan taxonomy prediction tasks,\nwe introduce the GlycanML-MTL testbed for multi-task learning (MTL) algorithms.\nAlso, we evaluate how taxonomy prediction can boost other three function\nprediction tasks by MTL. Experimental results show the superiority of modeling\nglycans with multi-relational GNNs, and suitable MTL methods can further boost\nmodel performance. We provide all datasets and source codes at\nhttps://github.com/GlycanML/GlycanML and maintain a leaderboard at\nhttps://GlycanML.github.io/project\n", "date": "2024-05-25", "categories": []}
{"id": "2406.06841", "title": "CompassDock: Comprehensive Accurate Assessment Approach for Deep\n  Learning-Based Molecular Docking in Inference and Fine-Tuning", "abstract": "  Datasets used for molecular docking, such as PDBBind, contain technical\nvariability - they are noisy. Although the origins of the noise have been\ndiscussed, a comprehensive analysis of the physical, chemical, and bioactivity\ncharacteristics of the datasets is still lacking. To address this gap, we\nintroduce the Comprehensive Accurate Assessment (Compass). Compass integrates\ntwo key components: PoseCheck, which examines ligand strain energy,\nprotein-ligand steric clashes, and interactions, and AA-Score, a new empirical\nscoring function for calculating binding affinity energy. Together, these form\na unified workflow that assesses both the physical/chemical properties and\nbioactivity favorability of ligands and protein-ligand interactions. Our\nanalysis of the PDBBind dataset using Compass reveals substantial noise in the\nground truth data. Additionally, we propose CompassDock, which incorporates the\nCompass module with DiffDock, the state-of-the-art deep learning-based\nmolecular docking method, to enable accurate assessment of docked ligands\nduring inference. Finally, we present a new paradigm for enhancing molecular\ndocking model performance by fine-tuning with Compass Scores, which encompass\nbinding affinity energy, strain energy, and the number of steric clashes\nidentified by Compass. Our results show that, while fine-tuning without Compass\nimproves the percentage of docked poses with RMSD < 2{\\AA}, it leads to a\ndecrease in physical/chemical and bioactivity favorability. In contrast,\nfine-tuning with Compass shows a limited improvement in RMSD < 2{\\AA} but\nenhances the physical/chemical and bioactivity favorability of the ligand\nconformation. The source code is available publicly at\nhttps://github.com/BIMSBbioinfo/CompassDock.\n", "date": "2024-06-10", "categories": []}
{"id": "2408.13659", "title": "ReactZyme: A Benchmark for Enzyme-Reaction Prediction", "abstract": "  Enzymes, with their specific catalyzed reactions, are necessary for all\naspects of life, enabling diverse biological processes and adaptations.\nPredicting enzyme functions is essential for understanding biological pathways,\nguiding drug development, enhancing bioproduct yields, and facilitating\nevolutionary studies. Addressing the inherent complexities, we introduce a new\napproach to annotating enzymes based on their catalyzed reactions. This method\nprovides detailed insights into specific reactions and is adaptable to newly\ndiscovered reactions, diverging from traditional classifications by protein\nfamily or expert-derived reaction classes. We employ machine learning\nalgorithms to analyze enzyme reaction datasets, delivering a much more refined\nview on the functionality of enzymes. Our evaluation leverages the largest\nenzyme-reaction dataset to date, derived from the SwissProt and Rhea databases\nwith entries up to January 8, 2024. We frame the enzyme-reaction prediction as\na retrieval problem, aiming to rank enzymes by their catalytic ability for\nspecific reactions. With our model, we can recruit proteins for novel reactions\nand predict reactions in novel proteins, facilitating enzyme discovery and\nfunction annotation (https://github.com/WillHua127/ReactZyme).\n", "date": "2024-08-24", "categories": []}
{"id": "2410.00327", "title": "EnzymeFlow: Generating Reaction-specific Enzyme Catalytic Pockets\n  through Flow Matching and Co-Evolutionary Dynamics", "abstract": "  Enzyme design is a critical area in biotechnology, with applications ranging\nfrom drug development to synthetic biology. Traditional methods for enzyme\nfunction prediction or protein binding pocket design often fall short in\ncapturing the dynamic and complex nature of enzyme-substrate interactions,\nparticularly in catalytic processes. To address the challenges, we introduce\nEnzymeFlow, a generative model that employs flow matching with hierarchical\npre-training and enzyme-reaction co-evolution to generate catalytic pockets for\nspecific substrates and catalytic reactions. Additionally, we introduce a\nlarge-scale, curated, and validated dataset of enzyme-reaction pairs,\nspecifically designed for the catalytic pocket generation task, comprising a\ntotal of $328,192$ pairs. By incorporating evolutionary dynamics and\nreaction-specific adaptations, EnzymeFlow becomes a powerful model for\ndesigning enzyme pockets, which is capable of catalyzing a wide range of\nbiochemical reactions. Experiments on the new dataset demonstrate the model's\neffectiveness in designing high-quality, functional enzyme catalytic pockets,\npaving the way for advancements in enzyme engineering and synthetic biology. We\nprovide EnzymeFlow code at https://github.com/WillHua127/EnzymeFlow with\nnotebook demonstration at\nhttps://github.com/WillHua127/EnzymeFlow/blob/main/enzymeflow_demo.ipynb.\n", "date": "2024-10-01", "categories": []}
{"id": "2409.20227", "title": "Assessing interaction recovery of predicted protein-ligand poses", "abstract": "  The field of protein-ligand pose prediction has seen significant advances in\nrecent years, with machine learning-based methods now being commonly used in\nlieu of classical docking methods or even to predict all-atom protein-ligand\ncomplex structures. Most contemporary studies focus on the accuracy and\nphysical plausibility of ligand placement to determine pose quality, often\nneglecting a direct assessment of the interactions observed with the protein.\nIn this work, we demonstrate that ignoring protein-ligand interaction\nfingerprints can lead to overestimation of model performance, most notably in\nrecent protein-ligand cofolding models which often fail to recapitulate key\ninteractions.\n", "date": "2024-09-30", "categories": []}
{"id": "2407.13027", "title": "SpaRED benchmark: Enhancing Gene Expression Prediction from Histology\n  Images with Spatial Transcriptomics Completion", "abstract": "  Spatial Transcriptomics is a novel technology that aligns histology images\nwith spatially resolved gene expression profiles. Although groundbreaking, it\nstruggles with gene capture yielding high corruption in acquired data. Given\npotential applications, recent efforts have focused on predicting\ntranscriptomic profiles solely from histology images. However, differences in\ndatabases, preprocessing techniques, and training hyperparameters hinder a fair\ncomparison between methods. To address these challenges, we present a\nsystematically curated and processed database collected from 26 public sources,\nrepresenting an 8.6-fold increase compared to previous works. Additionally, we\npropose a state-of-the-art transformer based completion technique for inferring\nmissing gene expression, which significantly boosts the performance of\ntranscriptomic profile predictions across all datasets. Altogether, our\ncontributions constitute the most comprehensive benchmark of gene expression\nprediction from histology images to date and a stepping stone for future\nresearch on spatial transcriptomics.\n", "date": "2024-07-17", "categories": []}
{"id": "2409.18201", "title": "Loop-Diffusion: an equivariant diffusion model for designing and scoring\n  protein loops", "abstract": "  Predicting protein functional characteristics from structure remains a\ncentral problem in protein science, with broad implications from understanding\nthe mechanisms of disease to designing novel therapeutics. Unfortunately,\ncurrent machine learning methods are limited by scarce and biased experimental\ndata, and physics-based methods are either too slow to be useful, or too\nsimplified to be accurate. In this work, we present Loop-Diffusion, an energy\nbased diffusion model which leverages a dataset of general protein loops from\nthe entire protein universe to learn an energy function that generalizes to\nfunctional prediction tasks. We evaluate Loop-Diffusion's performance on\nscoring TCR-pMHC interfaces and demonstrate state-of-the-art results in\nrecognizing binding-enhancing mutations.\n", "date": "2024-09-26", "categories": []}
{"id": "2409.17265", "title": "CodonMPNN for Organism Specific and Codon Optimal Inverse Folding", "abstract": "  Generating protein sequences conditioned on protein structures is an\nimpactful technique for protein engineering. When synthesizing engineered\nproteins, they are commonly translated into DNA and expressed in an organism\nsuch as yeast. One difficulty in this process is that the expression rates can\nbe low due to suboptimal codon sequences for expressing a protein in a host\norganism. We propose CodonMPNN, which generates a codon sequence conditioned on\na protein backbone structure and an organism label. If naturally occurring DNA\nsequences are close to codon optimality, CodonMPNN could learn to generate\ncodon sequences with higher expression yields than heuristic codon choices for\ngenerated amino acid sequences. Experiments show that CodonMPNN retains the\nperformance of previous inverse folding approaches and recovers wild-type\ncodons more frequently than baselines. Furthermore, CodonMPNN has a higher\nlikelihood of generating high-fitness codon sequences than low-fitness codon\nsequences for the same protein sequence. Code is available at\nhttps://github.com/HannesStark/CodonMPNN.\n", "date": "2024-09-25", "categories": []}
{"id": "2405.02628", "title": "Contrastive Dual-Interaction Graph Neural Network for Molecular Property\n  Prediction", "abstract": "  Molecular property prediction is a key component of AI-driven drug discovery\nand molecular characterization learning. Despite recent advances, existing\nmethods still face challenges such as limited ability to generalize, and\ninadequate representation of learning from unlabeled data, especially for tasks\nspecific to molecular structures. To address these limitations, we introduce\nDIG-Mol, a novel self-supervised graph neural network framework for molecular\nproperty prediction. This architecture leverages the power of contrast learning\nwith dual interaction mechanisms and unique molecular graph enhancement\nstrategies. DIG-Mol integrates a momentum distillation network with two\ninterconnected networks to efficiently improve molecular characterization. The\nframework's ability to extract key information about molecular structure and\nhigher-order semantics is supported by minimizing loss of contrast. We have\nestablished DIG-Mol's state-of-the-art performance through extensive\nexperimental evaluation in a variety of molecular property prediction tasks. In\naddition to demonstrating superior transferability in a small number of\nlearning scenarios, our visualizations highlight DIG-Mol's enhanced\ninterpretability and representation capabilities. These findings confirm the\neffectiveness of our approach in overcoming challenges faced by traditional\nmethods and mark a significant advance in molecular property prediction.\n", "date": "2024-05-04", "categories": []}
{"id": "2409.16298", "title": "BetterBodies: Reinforcement Learning guided Diffusion for Antibody\n  Sequence Design", "abstract": "  Antibodies offer great potential for the treatment of various diseases.\nHowever, the discovery of therapeutic antibodies through traditional wet lab\nmethods is expensive and time-consuming. The use of generative models in\ndesigning antibodies therefore holds great promise, as it can reduce the time\nand resources required. Recently, the class of diffusion models has gained\nconsiderable traction for their ability to synthesize diverse and high-quality\nsamples. In their basic form, however, they lack mechanisms to optimize for\nspecific properties, such as binding affinity to an antigen. In contrast, the\nclass of offline Reinforcement Learning (RL) methods has demonstrated strong\nperformance in navigating large search spaces, including scenarios where\nfrequent real-world interaction, such as interaction with a wet lab, is\nimpractical. Our novel method, BetterBodies, which combines Variational\nAutoencoders (VAEs) with RL guided latent diffusion, is able to generate novel\nsets of antibody CDRH3 sequences from different data distributions. Using the\nAbsolut! simulator, we demonstrate the improved affinity of our novel sequences\nto the SARS-CoV spike receptor-binding domain. Furthermore, we reflect\nbiophysical properties in the VAE latent space using a contrastive loss and add\na novel Q-function based filtering to enhance the affinity of generated\nsequences. In conclusion, methods such as ours have the potential to have great\nimplications for real-world biological sequence design, where the generation of\nnovel high-affinity binders is a cost-intensive endeavor.\n", "date": "2024-09-09", "categories": []}
{"id": "2409.16333", "title": "Predicting Distance matrix with large language models", "abstract": "  Structural prediction has long been considered critical in RNA research,\nespecially following the success of AlphaFold2 in protein studies, which has\ndrawn significant attention to the field. While recent advances in machine\nlearning and data accumulation have effectively addressed many biological\ntasks, particularly in protein related research. RNA structure prediction\nremains a significant challenge due to data limitations. Obtaining RNA\nstructural data is difficult because traditional methods such as nuclear\nmagnetic resonance spectroscopy, Xray crystallography, and electron microscopy\nare expensive and time consuming. Although several RNA 3D structure prediction\nmethods have been proposed, their accuracy is still limited. Predicting RNA\nstructural information at another level, such as distance maps, remains highly\nvaluable. Distance maps provide a simplified representation of spatial\nconstraints between nucleotides, capturing essential relationships without\nrequiring a full 3D model. This intermediate level of structural information\ncan guide more accurate 3D modeling and is computationally less intensive,\nmaking it a useful tool for improving structural predictions. In this work, we\ndemonstrate that using only primary sequence information, we can accurately\ninfer the distances between RNA bases by utilizing a large pretrained RNA\nlanguage model coupled with a well trained downstream transformer.\n", "date": "2024-09-24", "categories": []}
{"id": "2401.17612", "title": "IGCN: Integrative Graph Convolution Networks for patient level insights\n  and biomarker discovery in multi-omics integration", "abstract": "  Developing computational tools for integrative analysis across multiple types\nof omics data has been of immense importance in cancer molecular biology and\nprecision medicine research. While recent advancements have yielded integrative\nprediction solutions for multi-omics data, these methods lack a comprehensive\nand cohesive understanding of the rationale behind their specific predictions.\nTo shed light on personalized medicine and unravel previously unknown\ncharacteristics within integrative analysis of multi-omics data, we introduce a\nnovel integrative neural network approach for cancer molecular subtype and\nbiomedical classification applications, named Integrative Graph Convolutional\nNetworks (IGCN). IGCN can identify which types of omics receive more emphasis\nfor each patient to predict a certain class. Additionally, IGCN has the\ncapability to pinpoint significant biomarkers from a range of omics data types.\nTo demonstrate the superiority of IGCN, we compare its performance with other\nstate-of-the-art approaches across different cancer subtype and biomedical\nclassification tasks.\n", "date": "2024-01-31", "categories": []}
{"id": "2403.14088", "title": "Protein Conformation Generation via Force-Guided SE(3) Diffusion Models", "abstract": "  The conformational landscape of proteins is crucial to understanding their\nfunctionality in complex biological processes. Traditional physics-based\ncomputational methods, such as molecular dynamics (MD) simulations, suffer from\nrare event sampling and long equilibration time problems, hindering their\napplications in general protein systems. Recently, deep generative modeling\ntechniques, especially diffusion models, have been employed to generate novel\nprotein conformations. However, existing score-based diffusion methods cannot\nproperly incorporate important physical prior knowledge to guide the generation\nprocess, causing large deviations in the sampled protein conformations from the\nequilibrium distribution. In this paper, to overcome these limitations, we\npropose a force-guided SE(3) diffusion model, ConfDiff, for protein\nconformation generation. By incorporating a force-guided network with a mixture\nof data-based score models, ConfDiff can generate protein conformations with\nrich diversity while preserving high fidelity. Experiments on a variety of\nprotein conformation prediction tasks, including 12 fast-folding proteins and\nthe Bovine Pancreatic Trypsin Inhibitor (BPTI), demonstrate that our method\nsurpasses the state-of-the-art method.\n", "date": "2024-03-21", "categories": []}
{"id": "2409.15612", "title": "Revolutionizing Biomarker Discovery: Leveraging Generative AI for\n  Bio-Knowledge-Embedded Continuous Space Exploration", "abstract": "  Biomarker discovery is vital in advancing personalized medicine, offering\ninsights into disease diagnosis, prognosis, and therapeutic efficacy.\nTraditionally, the identification and validation of biomarkers heavily depend\non extensive experiments and statistical analyses. These approaches are\ntime-consuming, demand extensive domain expertise, and are constrained by the\ncomplexity of biological systems. These limitations motivate us to ask: Can we\nautomatically identify the effective biomarker subset without substantial human\nefforts? Inspired by the success of generative AI, we think that the intricate\nknowledge of biomarker identification can be compressed into a continuous\nembedding space, thus enhancing the search for better biomarkers. Thus, we\npropose a new biomarker identification framework with two important modules:1)\ntraining data preparation and 2) embedding-optimization-generation. The first\nmodule uses a multi-agent system to automatically collect pairs of biomarker\nsubsets and their corresponding prediction accuracy as training data. These\ndata establish a strong knowledge base for biomarker identification. The second\nmodule employs an encoder-evaluator-decoder learning paradigm to compress the\nknowledge of the collected data into a continuous space. Then, it utilizes\ngradient-based search techniques and autoregressive-based reconstruction to\nefficiently identify the optimal subset of biomarkers. Finally, we conduct\nextensive experiments on three real-world datasets to show the efficiency,\nrobustness, and effectiveness of our method.\n", "date": "2024-09-23", "categories": []}
{"id": "2409.14040", "title": "PepINVENT: Generative peptide design beyond the natural amino acids", "abstract": "  Peptides play a crucial role in the drug design and discovery whether as a\ntherapeutic modality or a delivery agent. Non-natural amino acids (NNAAs) have\nbeen used to enhance the peptide properties from binding affinity, plasma\nstability to permeability. Incorporating novel NNAAs facilitates the design of\nmore effective peptides with improved properties. The generative models used in\nthe field, have focused on navigating the peptide sequence space. The sequence\nspace is formed by combinations of a predefined set of amino acids. However,\nthere is still a need for a tool to explore the peptide landscape beyond this\nenumerated space to unlock and effectively incorporate de novo design of new\namino acids. To thoroughly explore the theoretical chemical space of the\npeptides, we present PepINVENT, a novel generative AI-based tool as an\nextension to the small molecule molecular design platform, REINVENT. PepINVENT\nnavigates the vast space of natural and non-natural amino acids to propose\nvalid, novel, and diverse peptide designs. The generative model can serve as a\ncentral tool for peptide-related tasks, as it was not trained on peptides with\nspecific properties or topologies. The prior was trained to understand the\ngranularity of peptides and to design amino acids for filling the masked\npositions within a peptide. PepINVENT coupled with reinforcement learning\nenables the goal-oriented design of peptides using its chemistry-informed\ngenerative capabilities. This study demonstrates PepINVENT's ability to explore\nthe peptide space with unique and novel designs, and its capacity for property\noptimization in the context of therapeutically relevant peptides. Our tool can\nbe employed for multi-parameter learning objectives, peptidomimetics, lead\noptimization, and variety of other tasks within the peptide domain.\n", "date": "2024-09-21", "categories": []}
{"id": "2409.13259", "title": "A generalizable framework for unlocking missing reactions in\n  genome-scale metabolic networks using deep learning", "abstract": "  Incomplete knowledge of metabolic processes hinders the accuracy of\nGEnome-scale Metabolic models (GEMs), which in turn impedes advancements in\nsystems biology and metabolic engineering. Existing gap-filling methods\ntypically rely on phenotypic data to minimize the disparity between\ncomputational predictions and experimental results. However, there is still a\nlack of an automatic and precise gap-filling method for initial state GEMs\nbefore experimental data and annotated genomes become available. In this study,\nwe introduce CLOSEgaps, a deep learning-driven tool that addresses the\ngap-filling issue by modeling it as a hyperedge prediction problem within GEMs.\nSpecifically, CLOSEgaps maps metabolic networks as hypergraphs and learns their\nhyper-topology features to identify missing reactions and gaps by leveraging\nhypothetical reactions. This innovative approach allows for the\ncharacterization and curation of both known and hypothetical reactions within\nmetabolic networks. Extensive results demonstrate that CLOSEgaps accurately\ngap-filling over 96% of artificially introduced gaps for various GEMs.\nFurthermore, CLOSEgaps enhances phenotypic predictions for 24 GEMs and also\nfinds a notable improvement in producing four crucial metabolites (Lactate,\nEthanol, Propionate, and Succinate) in two organisms. As a broadly applicable\nsolution for any GEM, CLOSEgaps represents a promising model to automate the\ngap-filling process and uncover missing connections between reactions and\nobserved metabolic phenotypes.\n", "date": "2024-09-20", "categories": []}
{"id": "2409.12926", "title": "MaskMol: Knowledge-guided Molecular Image Pre-Training Framework for\n  Activity Cliffs", "abstract": "  Activity cliffs, which refer to pairs of molecules that are structurally\nsimilar but show significant differences in their potency, can lead to model\nrepresentation collapse and make the model challenging to distinguish them. Our\nresearch indicates that as molecular similarity increases, graph-based methods\nstruggle to capture these nuances, whereas image-based approaches effectively\nretain the distinctions. Thus, we developed MaskMol, a knowledge-guided\nmolecular image self-supervised learning framework. MaskMol accurately learns\nthe representation of molecular images by considering multiple levels of\nmolecular knowledge, such as atoms, bonds, and substructures. By utilizing\npixel masking tasks, MaskMol extracts fine-grained information from molecular\nimages, overcoming the limitations of existing deep learning models in\nidentifying subtle structural changes. Experimental results demonstrate\nMaskMol's high accuracy and transferability in activity cliff estimation and\ncompound potency prediction across 20 different macromolecular targets,\noutperforming 25 state-of-the-art deep learning and machine learning\napproaches. Visualization analyses reveal MaskMol's high biological\ninterpretability in identifying activity cliff-relevant molecular\nsubstructures. Notably, through MaskMol, we identified candidate EP4 inhibitors\nthat could be used to treat tumors. This study not only raises awareness about\nactivity cliffs but also introduces a novel method for molecular image\nrepresentation learning and virtual screening, advancing drug discovery and\nproviding new insights into structure-activity relationships (SAR).\n", "date": "2024-09-02", "categories": []}
{"id": "2408.12413", "title": "Dynamic PDB: A New Dataset and a SE(3) Model Extension by Integrating\n  Dynamic Behaviors and Physical Properties in Protein Structures", "abstract": "  Despite significant progress in static protein structure collection and\nprediction, the dynamic behavior of proteins, one of their most vital\ncharacteristics, has been largely overlooked in prior research. This oversight\ncan be attributed to the limited availability, diversity, and heterogeneity of\ndynamic protein datasets. To address this gap, we propose to enhance existing\nprestigious static 3D protein structural databases, such as the Protein Data\nBank (PDB), by integrating dynamic data and additional physical properties.\nSpecifically, we introduce a large-scale dataset, Dynamic PDB, encompassing\napproximately 12.6K proteins, each subjected to all-atom molecular dynamics\n(MD) simulations lasting 1 microsecond to capture conformational changes.\nFurthermore, we provide a comprehensive suite of physical properties, including\natomic velocities and forces, potential and kinetic energies of proteins, and\nthe temperature of the simulation environment, recorded at 1 picosecond\nintervals throughout the simulations. For benchmarking purposes, we evaluate\nstate-of-the-art methods on the proposed dataset for the task of trajectory\nprediction. To demonstrate the value of integrating richer physical properties\nin the study of protein dynamics and related model design, we base our approach\non the SE(3) diffusion model and incorporate these physical properties into the\ntrajectory prediction process. Preliminary results indicate that this\nstraightforward extension of the SE(3) model yields improved accuracy, as\nmeasured by MAE and RMSD, when the proposed physical properties are taken into\nconsideration. https://fudan-generative-vision.github.io/dynamicPDB/ .\n", "date": "2024-08-22", "categories": []}
{"id": "2409.10964", "title": "Active learning for energy-based antibody optimization and enhanced\n  screening", "abstract": "  Accurate prediction and optimization of protein-protein binding affinity is\ncrucial for therapeutic antibody development. Although machine learning-based\nprediction methods $\\Delta\\Delta G$ are suitable for large-scale mutant\nscreening, they struggle to predict the effects of multiple mutations for\ntargets without existing binders. Energy function-based methods, though more\naccurate, are time consuming and not ideal for large-scale screening. To\naddress this, we propose an active learning workflow that efficiently trains a\ndeep learning model to learn energy functions for specific targets, combining\nthe advantages of both approaches. Our method integrates the RDE-Network deep\nlearning model with Rosetta's energy function-based Flex ddG to efficiently\nexplore mutants. In a case study targeting HER2-binding Trastuzumab mutants,\nour approach significantly improved the screening performance over random\nselection and demonstrated the ability to identify mutants with better binding\nproperties without experimental $\\Delta\\Delta G$ data. This workflow advances\ncomputational antibody design by combining machine learning, physics-based\ncomputations, and active learning to achieve more efficient antibody\ndevelopment.\n", "date": "2024-09-17", "categories": []}
{"id": "2409.10579", "title": "Recent advances in deep learning and language models for studying the\n  microbiome", "abstract": "  Recent advancements in deep learning, particularly large language models\n(LLMs), made a significant impact on how researchers study microbiome and\nmetagenomics data. Microbial protein and genomic sequences, like natural\nlanguages, form a language of life, enabling the adoption of LLMs to extract\nuseful insights from complex microbial ecologies. In this paper, we review\napplications of deep learning and language models in analyzing microbiome and\nmetagenomics data. We focus on problem formulations, necessary datasets, and\nthe integration of language modeling techniques. We provide an extensive\noverview of protein/genomic language modeling and their contributions to\nmicrobiome studies. We also discuss applications such as novel viromics\nlanguage modeling, biosynthetic gene cluster prediction, and knowledge\nintegration for metagenomics studies.\n", "date": "2024-09-15", "categories": []}
{"id": "2402.14482", "title": "SpanSeq: Similarity-based sequence data splitting method for improved\n  development and assessment of deep learning projects", "abstract": "  The use of deep learning models in computational biology has increased\nmassively in recent years, and it is expected to continue with the current\nadvances in the fields such as Natural Language Processing. These models,\nalthough able to draw complex relations between input and target, are also\ninclined to learn noisy deviations from the pool of data used during their\ndevelopment. In order to assess their performance on unseen data (their\ncapacity to generalize), it is common to split the available data randomly into\ndevelopment (train/validation) and test sets. This procedure, although\nstandard, has been shown to produce dubious assessments of generalization due\nto the existing similarity between samples in the databases used. In this work,\nwe present SpanSeq, a database partition method for machine learning that can\nscale to most biological sequences (genes, proteins and genomes) in order to\navoid data leakage between sets. We also explore the effect of not restraining\nsimilarity between sets by reproducing the development of two state-of-the-art\nmodels on bioinformatics, not only confirming the consequences of randomly\nsplitting databases on the model assessment, but expanding those repercussions\nto the model development. SpanSeq is available at\nhttps://github.com/genomicepidemiology/SpanSeq.\n", "date": "2024-02-22", "categories": []}
{"id": "2409.08302", "title": "How Molecules Impact Cells: Unlocking Contrastive PhenoMolecular\n  Retrieval", "abstract": "  Predicting molecular impact on cellular function is a core challenge in\ntherapeutic design. Phenomic experiments, designed to capture cellular\nmorphology, utilize microscopy based techniques and demonstrate a high\nthroughput solution for uncovering molecular impact on the cell. In this work,\nwe learn a joint latent space between molecular structures and microscopy\nphenomic experiments, aligning paired samples with contrastive learning.\nSpecifically, we study the problem ofContrastive PhenoMolecular Retrieval,\nwhich consists of zero-shot molecular structure identification conditioned on\nphenomic experiments. We assess challenges in multi-modal learning of phenomics\nand molecular modalities such as experimental batch effect, inactive molecule\nperturbations, and encoding perturbation concentration. We demonstrate improved\nmulti-modal learner retrieval through (1) a uni-modal pre-trained phenomics\nmodel, (2) a novel inter sample similarity aware loss, and (3) models\nconditioned on a representation of molecular concentration. Following this\nrecipe, we propose MolPhenix, a molecular phenomics model. MolPhenix leverages\na pre-trained phenomics model to demonstrate significant performance gains\nacross perturbation concentrations, molecular scaffolds, and activity\nthresholds. In particular, we demonstrate an 8.1x improvement in zero shot\nmolecular retrieval of active molecules over the previous state-of-the-art,\nreaching 77.33% in top-1% accuracy. These results open the door for machine\nlearning to be applied in virtual phenomics screening, which can significantly\nbenefit drug discovery applications.\n", "date": "2024-09-10", "categories": []}
{"id": "2405.12356", "title": "Coarse-graining conformational dynamics with multi-dimensional\n  generalized Langevin equation: how, when, and why", "abstract": "  A data-driven ab initio generalized Langevin equation (AIGLE) approach is\ndeveloped to learn and simulate high-dimensional, heterogeneous, coarse-grained\nconformational dynamics. Constrained by the fluctuation-dissipation theorem,\nthe approach can build coarse-grained models in dynamical consistency with\nall-atom molecular dynamics. We also propose practical criteria for AIGLE to\nenforce long-term dynamical consistency. Case studies of a toy polymer, with 20\ncoarse-grained sites, and the alanine dipeptide, with two dihedral angles,\nelucidate why one should adopt AIGLE or its Markovian limit for modeling\ncoarse-grained conformational dynamics in practice.\n", "date": "2024-05-20", "categories": []}
{"id": "2409.05484", "title": "CRADLE-VAE: Enhancing Single-Cell Gene Perturbation Modeling with\n  Counterfactual Reasoning-based Artifact Disentanglement", "abstract": "  Predicting cellular responses to various perturbations is a critical focus in\ndrug discovery and personalized therapeutics, with deep learning models playing\na significant role in this endeavor. Single-cell datasets contain technical\nartifacts that may hinder the predictability of such models, which poses\nquality control issues highly regarded in this area. To address this, we\npropose CRADLE-VAE, a causal generative framework tailored for single-cell gene\nperturbation modeling, enhanced with counterfactual reasoning-based artifact\ndisentanglement. Throughout training, CRADLE-VAE models the underlying latent\ndistribution of technical artifacts and perturbation effects present in\nsingle-cell datasets. It employs counterfactual reasoning to effectively\ndisentangle such artifacts by modulating the latent basal spaces and learns\nrobust features for generating cellular response data with improved quality.\nExperimental results demonstrate that this approach improves not only treatment\neffect estimation performance but also generative quality as well. The\nCRADLE-VAE codebase is publicly available at\nhttps://github.com/dmis-lab/CRADLE-VAE.\n", "date": "2024-09-09", "categories": []}
{"id": "2409.05938", "title": "DeepFM-Crispr: Prediction of CRISPR On-Target Effects via Deep Learning", "abstract": "  Since the advent of CRISPR-Cas9, a groundbreaking gene-editing technology\nthat enables precise genomic modifications via a short RNA guide sequence,\nthere has been a marked increase in the accessibility and application of this\ntechnology across various fields. The success of CRISPR-Cas9 has spurred\nfurther investment and led to the discovery of additional CRISPR systems,\nincluding CRISPR-Cas13. Distinct from Cas9, which targets DNA, Cas13 targets\nRNA, offering unique advantages for gene modulation. We focus on Cas13d, a\nvariant known for its collateral activity where it non-specifically cleaves\nadjacent RNA molecules upon activation, a feature critical to its function. We\nintroduce DeepFM-Crispr, a novel deep learning model developed to predict the\non-target efficiency and evaluate the off-target effects of Cas13d. This model\nharnesses a large language model to generate comprehensive representations rich\nin evolutionary and structural data, thereby enhancing predictions of RNA\nsecondary structures and overall sgRNA efficacy. A transformer-based\narchitecture processes these inputs to produce a predictive efficacy score.\nComparative experiments show that DeepFM-Crispr not only surpasses traditional\nmodels but also outperforms recent state-of-the-art deep learning methods in\nterms of prediction accuracy and reliability.\n", "date": "2024-09-09", "categories": []}
{"id": "2406.05832", "title": "Improving Antibody Design with Force-Guided Sampling in Diffusion Models", "abstract": "  Antibodies, crucial for immune defense, primarily rely on\ncomplementarity-determining regions (CDRs) to bind and neutralize antigens,\nsuch as viruses. The design of these CDRs determines the antibody's affinity\nand specificity towards its target. Generative models, particularly denoising\ndiffusion probabilistic models (DDPMs), have shown potential to advance the\nstructure-based design of CDR regions. However, only a limited dataset of bound\nantibody-antigen structures is available, and generalization to\nout-of-distribution interfaces remains a challenge. Physics based force-fields,\nwhich approximate atomic interactions, offer a coarse but universal source of\ninformation to better mold designs to target interfaces. Integrating this\nfoundational information into diffusion models is, therefore, highly desirable.\nHere, we propose a novel approach to enhance the sampling process of diffusion\nmodels by integrating force field energy-based feedback. Our model, DiffForce,\nemploys forces to guide the diffusion sampling process, effectively blending\nthe two distributions. Through extensive experiments, we demonstrate that our\nmethod guides the model to sample CDRs with lower energy, enhancing both the\nstructure and sequence of the generated antibodies.\n", "date": "2024-06-09", "categories": []}
{"id": "2409.05047", "title": "Machine Learning-Based Prediction of Key Genes Correlated to the\n  Subretinal Lesion Severity in a Mouse Model of Age-Related Macular\n  Degeneration", "abstract": "  Age-related macular degeneration (AMD) is a major cause of blindness in older\nadults, severely affecting vision and quality of life. Despite advances in\nunderstanding AMD, the molecular factors driving the severity of subretinal\nscarring (fibrosis) remain elusive, hampering the development of effective\ntherapies. This study introduces a machine learning-based framework to predict\nkey genes that are strongly correlated with lesion severity and to identify\npotential therapeutic targets to prevent subretinal fibrosis in AMD. Using an\noriginal RNA sequencing (RNA-seq) dataset from the diseased retinas of JR5558\nmice, we developed a novel and specific feature engineering technique,\nincluding pathway-based dimensionality reduction and gene-based feature\nexpansion, to enhance prediction accuracy. Two iterative experiments were\nconducted by leveraging Ridge and ElasticNet regression models to assess\nbiological relevance and gene impact. The results highlight the biological\nsignificance of several key genes and demonstrate the framework's effectiveness\nin identifying novel therapeutic targets. The key findings provide valuable\ninsights for advancing drug discovery efforts and improving treatment\nstrategies for AMD, with the potential to enhance patient outcomes by targeting\nthe underlying genetic mechanisms of subretinal lesion development.\n", "date": "2024-09-08", "categories": []}
{"id": "2409.05804", "title": "Celcomen: spatial causal disentanglement for single-cell and tissue\n  perturbation modeling", "abstract": "  Celcomen leverages a mathematical causality framework to disentangle intra-\nand inter- cellular gene regulation programs in spatial transcriptomics and\nsingle-cell data through a generative graph neural network. It can learn\ngene-gene interactions, as well as generate post-perturbation counterfactual\nspatial transcriptomics, thereby offering access to experimentally inaccessible\nsamples. We validated its disentanglement, identifiability, and counterfactual\nprediction capabilities through simulations and in clinically relevant human\nglioblastoma, human fetal spleen, and mouse lung cancer samples. Celcomen\nprovides the means to model disease and therapy induced changes allowing for\nnew insights into single-cell spatially resolved tissue responses relevant to\nhuman health.\n", "date": "2024-09-09", "categories": []}
{"id": "2310.03223", "title": "TacoGFN: Target-conditioned GFlowNet for Structure-based Drug Design", "abstract": "  Searching the vast chemical space for drug-like molecules that bind with a\nprotein pocket is a challenging task in drug discovery. Recently,\nstructure-based generative models have been introduced which promise to be more\nefficient by learning to generate molecules for any given protein structure.\nHowever, since they learn the distribution of a limited protein-ligand complex\ndataset, structure-based methods do not yet outperform optimization-based\nmethods that generate binding molecules for just one pocket. To overcome\nlimitations on data while leveraging learning across protein targets, we choose\nto model the reward distribution conditioned on pocket structure, instead of\nthe training data distribution. We design TacoGFN, a novel GFlowNet-based\napproach for structure-based drug design, which can generate molecules\nconditioned on any protein pocket structure with probabilities proportional to\nits affinity and property rewards. In the generative setting for\nCrossDocked2020 benchmark, TacoGFN attains a state-of-the-art success rate of\n$56.0\\%$ and $-8.44$ kcal/mol in median Vina Dock score while improving the\ngeneration time by multiple orders of magnitude. Fine-tuning TacoGFN further\nimproves the median Vina Dock score to $-10.93$ kcal/mol and the success rate\nto $88.8\\%$, outperforming all optimization-based methods.\n", "date": "2023-10-05", "categories": []}
{"id": "2408.10689", "title": "Genesis: Towards the Automation of Systems Biology Research", "abstract": "  The cutting edge of applying AI to science is the closed-loop automation of\nscientific research: robot scientists. We have previously developed two robot\nscientists: `Adam' (for yeast functional biology), and `Eve' (for early-stage\ndrug design)). We are now developing a next generation robot scientist Genesis.\nWith Genesis we aim to demonstrate that an area of science can be investigated\nusing robot scientists unambiguously faster, and at lower cost, than with human\nscientists. Here we report progress on the Genesis project. Genesis is designed\nto automatically improve system biology models with thousands of interacting\ncausal components. When complete Genesis will be able to initiate and execute\nin parallel one thousand hypothesis-led closed-loop cycles of experiment\nper-day. Here we describe the core Genesis hardware: the one thousand\ncomputer-controlled $\\mu$-bioreactors. For the integrated Mass Spectrometry\nplatform we have developed AutonoMS, a system to automatically run, process,\nand analyse high-throughput experiments. We have also developed Genesis-DB, a\ndatabase system designed to enable software agents access to large quantities\nof structured domain information. We have developed RIMBO (Revisions for\nImprovements of Models in Biology Ontology) to describe the planned hundreds of\nthousands of changes to the models. We have demonstrated the utility of this\ninfrastructure by developed two relational learning bioinformatic projects.\nFinally, we describe LGEM+ a relational learning system for the automated\nabductive improvement of genome-scale metabolic models.\n", "date": "2024-08-20", "categories": []}
{"id": "2309.03919", "title": "A hybrid quantum-classical fusion neural network to improve\n  protein-ligand binding affinity predictions for drug discovery", "abstract": "  The field of drug discovery hinges on the accurate prediction of binding\naffinity between prospective drug molecules and target proteins, especially\nwhen such proteins directly influence disease progression. However, estimating\nbinding affinity demands significant financial and computational resources.\nWhile state-of-the-art methodologies employ classical machine learning (ML)\ntechniques, emerging hybrid quantum machine learning (QML) models have shown\npromise for enhanced performance, owing to their inherent parallelism and\ncapacity to manage exponential increases in data dimensionality. Despite these\nadvances, existing models encounter issues related to convergence stability and\nprediction accuracy. This paper introduces a novel hybrid quantum-classical\ndeep learning model tailored for binding affinity prediction in drug discovery.\nSpecifically, the proposed model synergistically integrates 3D and spatial\ngraph convolutional neural networks within an optimized quantum architecture.\nSimulation results demonstrate a 6% improvement in prediction accuracy relative\nto existing classical models, as well as a significantly more stable\nconvergence performance compared to previous classical approaches.\n", "date": "2023-09-06", "categories": []}
{"id": "2401.04478", "title": "TwinBooster: Synergising Large Language Models with Barlow Twins and\n  Gradient Boosting for Enhanced Molecular Property Prediction", "abstract": "  The success of drug discovery and development relies on the precise\nprediction of molecular activities and properties. While in silico molecular\nproperty prediction has shown remarkable potential, its use has been limited so\nfar to assays for which large amounts of data are available. In this study, we\nuse a fine-tuned large language model to integrate biological assays based on\ntheir textual information, coupled with Barlow Twins, a Siamese neural network\nusing a novel self-supervised learning approach. This architecture uses both\nassay information and molecular fingerprints to extract the true molecular\ninformation. TwinBooster enables the prediction of properties of unseen\nbioassays and molecules by providing state-of-the-art zero-shot learning tasks.\nRemarkably, our artificial intelligence pipeline shows excellent performance on\nthe FS-Mol benchmark. This breakthrough demonstrates the application of deep\nlearning to critical property prediction tasks where data is typically scarce.\nBy accelerating the early identification of active molecules in drug discovery\nand development, this method has the potential to help streamline the\nidentification of novel therapeutics.\n", "date": "2024-01-09", "categories": []}
{"id": "2408.16068", "title": "Identification of Prognostic Biomarkers for Stage III Non-Small Cell\n  Lung Carcinoma in Female Nonsmokers Using Machine Learning", "abstract": "  Lung cancer remains a leading cause of cancer-related deaths globally, with\nnon-small cell lung cancer (NSCLC) being the most common subtype. This study\naimed to identify key biomarkers associated with stage III NSCLC in non-smoking\nfemales using gene expression profiling from the GDS3837 dataset. Utilizing\nXGBoost, a machine learning algorithm, the analysis achieved a strong\npredictive performance with an AUC score of 0.835. The top biomarkers\nidentified - CCAAT enhancer binding protein alpha (C/EBP-alpha), lactate\ndehydrogenase A4 (LDHA), UNC-45 myosin chaperone B (UNC-45B), checkpoint kinase\n1 (CHK1), and hypoxia-inducible factor 1 subunit alpha (HIF-1-alpha) - have\nbeen validated in the literature as being significantly linked to lung cancer.\nThese findings highlight the potential of these biomarkers for early diagnosis\nand personalized therapy, emphasizing the value of integrating machine learning\nwith molecular profiling in cancer research.\n", "date": "2024-08-28", "categories": []}
{"id": "2408.15299", "title": "TourSynbio: A Multi-Modal Large Model and Agent Framework to Bridge Text\n  and Protein Sequences for Protein Engineering", "abstract": "  The structural similarities between protein sequences and natural languages\nhave led to parallel advancements in deep learning across both domains. While\nlarge language models (LLMs) have achieved much progress in the domain of\nnatural language processing, their potential in protein engineering remains\nlargely unexplored. Previous approaches have equipped LLMs with protein\nunderstanding capabilities by incorporating external protein encoders, but this\nfails to fully leverage the inherent similarities between protein sequences and\nnatural languages, resulting in sub-optimal performance and increased model\ncomplexity. To address this gap, we present TourSynbio-7B, the first\nmulti-modal large model specifically designed for protein engineering tasks\nwithout external protein encoders. TourSynbio-7B demonstrates that LLMs can\ninherently learn to understand proteins as language. The model is post-trained\nand instruction fine-tuned on InternLM2-7B using ProteinLMDataset, a dataset\ncomprising 17.46 billion tokens of text and protein sequence for\nself-supervised pretraining and 893K instructions for supervised fine-tuning.\nTourSynbio-7B outperforms GPT-4 on the ProteinLMBench, a benchmark of 944\nmanually verified multiple-choice questions, with 62.18% accuracy. Leveraging\nTourSynbio-7B's enhanced protein sequence understanding capability, we\nintroduce TourSynbio-Agent, an innovative framework capable of performing\nvarious protein engineering tasks, including mutation analysis, inverse\nfolding, protein folding, and visualization. TourSynbio-Agent integrates\npreviously disconnected deep learning models in the protein engineering domain,\noffering a unified conversational user interface for improved usability.\nFinally, we demonstrate the efficacy of TourSynbio-7B and TourSynbio-Agent\nthrough two wet lab case studies on vanilla key enzyme modification and steroid\ncompound catalysis.\n", "date": "2024-08-27", "categories": []}
{"id": "2408.15956", "title": "Generating Binary Species Range Maps", "abstract": "  Accurately predicting the geographic ranges of species is crucial for\nassisting conservation efforts. Traditionally, range maps were manually created\nby experts. However, species distribution models (SDMs) and, more recently,\ndeep learning-based variants offer a potential automated alternative. Deep\nlearning-based SDMs generate a continuous probability representing the\npredicted presence of a species at a given location, which must be binarized by\nsetting per-species thresholds to obtain binary range maps. However, selecting\nappropriate per-species thresholds to binarize these predictions is non-trivial\nas different species can require distinct thresholds. In this work, we evaluate\ndifferent approaches for automatically identifying the best thresholds for\nbinarizing range maps using presence-only data. This includes approaches that\nrequire the generation of additional pseudo-absence data, along with ones that\nonly require presence data. We also propose an extension of an existing\npresence-only technique that is more robust to outliers. We perform a detailed\nevaluation of different thresholding techniques on the tasks of binary range\nestimation and large-scale fine-grained visual classification, and we\ndemonstrate improved performance over existing pseudo-absence free approaches\nusing our method.\n", "date": "2024-08-28", "categories": []}
{"id": "2408.15032", "title": "Mamba2MIL: State Space Duality Based Multiple Instance Learning for\n  Computational Pathology", "abstract": "  Computational pathology (CPath) has significantly advanced the clinical\npractice of pathology. Despite the progress made, Multiple Instance Learning\n(MIL), a promising paradigm within CPath, continues to face challenges,\nparticularly related to incomplete information utilization. Existing\nframeworks, such as those based on Convolutional Neural Networks (CNNs),\nattention, and selective scan space state sequential model (SSM), lack\nsufficient flexibility and scalability in fusing diverse features, and cannot\neffectively fuse diverse features. Additionally, current approaches do not\nadequately exploit order-related and order-independent features, resulting in\nsuboptimal utilization of sequence information. To address these limitations,\nwe propose a novel MIL framework called Mamba2MIL. Our framework utilizes the\nstate space duality model (SSD) to model long sequences of patches of whole\nslide images (WSIs), which, combined with weighted feature selection, supports\nthe fusion processing of more branching features and can be extended according\nto specific application needs. Moreover, we introduce a sequence transformation\nmethod tailored to varying WSI sizes, which enhances sequence-independent\nfeatures while preserving local sequence information, thereby improving\nsequence information utilization. Extensive experiments demonstrate that\nMamba2MIL surpasses state-of-the-art MIL methods. We conducted extensive\nexperiments across multiple datasets, achieving improvements in nearly all\nperformance metrics. Specifically, on the NSCLC dataset, Mamba2MIL achieves a\nbinary tumor classification AUC of 0.9533 and an accuracy of 0.8794. On the\nBRACS dataset, it achieves a multiclass classification AUC of 0.7986 and an\naccuracy of 0.4981. The code is available at\nhttps://github.com/YuqiZhang-Buaa/Mamba2MIL.\n", "date": "2024-08-27", "categories": []}
{"id": "2311.02333", "title": "Understanding the Natural Language of DNA using Encoder-Decoder\n  Foundation Models with Byte-level Precision", "abstract": "  This paper presents the Ensemble Nucleotide Byte-level Encoder-Decoder\n(ENBED) foundation model, analyzing DNA sequences at byte-level precision with\nan encoder-decoder Transformer architecture. ENBED uses a sub-quadratic\nimplementation of attention to develop an efficient model capable of\nsequence-to-sequence transformations, generalizing previous genomic models with\nencoder-only or decoder-only architectures. We use Masked Language Modeling to\npre-train the foundation model using reference genome sequences and apply it in\nthe following downstream tasks: (1) identification of enhancers, promotors and\nsplice sites, (2) recognition of sequences containing base call mismatches and\ninsertion/deletion errors, an advantage over tokenization schemes involving\nmultiple base pairs, which lose the ability to analyze with byte-level\nprecision, (3) identification of biological function annotations of genomic\nsequences, and (4) generating mutations of the Influenza virus using the\nencoder-decoder architecture and validating them against real-world\nobservations. In each of these tasks, we demonstrate significant improvement as\ncompared to the existing state-of-the-art results.\n", "date": "2023-11-04", "categories": []}
{"id": "2408.06720", "title": "Multimodal Analysis of White Blood Cell Differentiation in Acute Myeloid\n  Leukemia Patients using a \\beta-Variational Autoencoder", "abstract": "  Biomedical imaging and RNA sequencing with single-cell resolution improves\nour understanding of white blood cell diseases like leukemia. By combining\nmorphological and transcriptomic data, we can gain insights into cellular\nfunctions and trajectoriess involved in blood cell differentiation. However,\nexisting methodologies struggle with integrating morphological and\ntranscriptomic data, leaving a significant research gap in comprehensively\nunderstanding the dynamics of cell differentiation. Here, we introduce an\nunsupervised method that explores and reconstructs these two modalities and\nuncovers the relationship between different subtypes of white blood cells from\nhuman peripheral blood smears in terms of morphology and their corresponding\ntranscriptome. Our method is based on a beta-variational autoencoder\n({\\ss}-VAE) with a customized loss function, incorporating a R-CNN architecture\nto distinguish single-cell from background and to minimize any interference\nfrom artifacts. This implementation of {\\ss}-VAE shows good reconstruction\ncapability along with continuous latent embeddings, while maintaining clear\ndifferentiation between single-cell classes. Our novel approach is especially\nhelpful to uncover the correlation of two latent features in complex biological\nprocesses such as formation of granules in the cell (granulopoiesis) with gene\nexpression patterns. It thus provides a unique tool to improve the\nunderstanding of white blood cell maturation for biomedicine and diagnostics.\n", "date": "2024-08-13", "categories": []}
{"id": "2408.12625", "title": "Machine Learning Potentials: A Roadmap Toward Next-Generation\n  Biomolecular Simulations", "abstract": "  Machine learning potentials offer a revolutionary, unifying framework for\nmolecular simulations across scales, from quantum chemistry to coarse-grained\nmodels. Here, I explore their potential to dramatically improve accuracy and\nscalability in simulating complex molecular systems. I discuss key challenges\nthat must be addressed to fully realize their transformative potential in\nchemical biology and related fields.\n", "date": "2024-08-17", "categories": []}
{"id": "2408.12636", "title": "Wave-LSTM: Multi-scale analysis of somatic whole genome copy number\n  profiles", "abstract": "  Changes in the number of copies of certain parts of the genome, known as copy\nnumber alterations (CNAs), due to somatic mutation processes are a hallmark of\nmany cancers. This genomic complexity is known to be associated with poorer\noutcomes for patients but describing its contribution in detail has been\ndifficult. Copy number alterations can affect large regions spanning whole\nchromosomes or the entire genome itself but can also be localised to only small\nsegments of the genome and no methods exist that allow this multi-scale nature\nto be quantified. In this paper, we address this using Wave-LSTM, a signal\ndecomposition approach designed to capture the multi-scale structure of complex\nwhole genome copy number profiles. Using wavelet-based source separation in\ncombination with deep learning-based attention mechanisms. We show that\nWave-LSTM can be used to derive multi-scale representations from copy number\nprofiles which can be used to decipher sub-clonal structures from single-cell\ncopy number data and to improve survival prediction performance from patient\ntumour profiles.\n", "date": "2024-08-22", "categories": []}
{"id": "2408.13012", "title": "Personalised Medicine: Establishing predictive machine learning models\n  for drug responses in patient derived cell culture", "abstract": "  The concept of personalised medicine in cancer therapy is becoming\nincreasingly important. There already exist drugs administered specifically for\npatients with tumours presenting well-defined mutations. However, the field is\nstill in its infancy, and personalised treatments are far from being standard\nof care. Personalised medicine is often associated with the utilisation of\nomics data. Yet, implementation of multi-omics data has proven difficult, due\nto the variety and scale of the information within the data, as well as the\ncomplexity behind the myriad of interactions taking place within the cell. An\nalternative approach to precision medicine is to employ a function-based\nprofile of the cell. This involves screening a range of drugs against patient\nderived cells. Here we demonstrate a proof-of-concept, where a collection of\ndrug screens against a highly diverse set of patient-derived cell lines, are\nleveraged to identify putative treatment options for a 'new patient'. We show\nthat this methodology is highly efficient in ranking the drugs according to\ntheir activity towards the target cells. We argue that this approach offers\ngreat potential, as activities can be efficiently imputed from various subsets\nof the drug treated cell lines that do not necessarily originate from the same\ntissue type.\n", "date": "2024-08-23", "categories": []}
{"id": "2408.11356", "title": "One-step Structure Prediction and Screening for Protein-Ligand Complexes\n  using Multi-Task Geometric Deep Learning", "abstract": "  Understanding the structure of the protein-ligand complex is crucial to drug\ndevelopment. Existing virtual structure measurement and screening methods are\ndominated by docking and its derived methods combined with deep learning.\nHowever, the sampling and scoring methodology have largely restricted the\naccuracy and efficiency. Here, we show that these two fundamental tasks can be\naccurately tackled with a single model, namely LigPose, based on multi-task\ngeometric deep learning. By representing the ligand and the protein pair as a\ngraph, LigPose directly optimizes the three-dimensional structure of the\ncomplex, with the learning of binding strength and atomic interactions as\nauxiliary tasks, enabling its one-step prediction ability without docking\ntools. Extensive experiments show LigPose achieved state-of-the-art performance\non major tasks in drug research. Its considerable improvements indicate a\npromising paradigm of AI-based pipeline for drug development.\n", "date": "2024-08-21", "categories": []}
{"id": "2408.10247", "title": "MetaEnzyme: Meta Pan-Enzyme Learning for Task-Adaptive Redesign", "abstract": "  Enzyme design plays a crucial role in both industrial production and biology.\nHowever, this field faces challenges due to the lack of comprehensive\nbenchmarks and the complexity of enzyme design tasks, leading to a dearth of\nsystematic research. Consequently, computational enzyme design is relatively\noverlooked within the broader protein domain and remains in its early stages.\nIn this work, we address these challenges by introducing MetaEnzyme, a staged\nand unified enzyme design framework. We begin by employing a cross-modal\nstructure-to-sequence transformation architecture, as the feature-driven\nstarting point to obtain initial robust protein representation. Subsequently,\nwe leverage domain adaptive techniques to generalize specific enzyme design\ntasks under low-resource conditions. MetaEnzyme focuses on three fundamental\nlow-resource enzyme redesign tasks: functional design (FuncDesign), mutation\ndesign (MutDesign), and sequence generation design (SeqDesign). Through novel\nunified paradigm and enhanced representation capabilities, MetaEnzyme\ndemonstrates adaptability to diverse enzyme design tasks, yielding outstanding\nresults. Wet lab experiments further validate these findings, reinforcing the\nefficacy of the redesign process.\n", "date": "2024-08-05", "categories": []}
{"id": "2209.10545", "title": "SGC: A semi-supervised pipeline for gene clustering using self-training\n  approach in gene co-expression networks", "abstract": "  A widely used approach for extracting information from gene expression data\nemploy the construction of a gene co-expression network and the subsequent\napplication of algorithms that discover network structure. In particular, a\ncommon goal is the computational discovery of gene clusters, commonly called\nmodules. When applied on a novel gene expression dataset, the quality of the\ncomputed modules can be evaluated automatically, using Gene Ontology\nenrichment, a method that measures the frequencies of Gene Ontology terms in\nthe computed modules and evaluates their statistical likelihood. In this work\nwe propose SGC a novel pipeline for gene clustering based on relatively recent\nseminal work in the mathematics of spectral network theory. SGC consists of\nmultiple novel steps that enable the computation of highly enriched modules in\nan unsupervised manner. But unlike all existing frameworks, it further\nincorporates a novel step that leverages Gene Ontology information in a\nsemi-supervised clustering method that further improves the quality of the\ncomputed modules. Comparing with already well-known existing frameworks, we\nshow that SGC results in higher enrichment in real data. In particular, in 12\nreal gene expression datasets, SGC outperforms in all except one.\n", "date": "2022-09-21", "categories": []}
{"id": "2407.07930", "title": "Token-Mol 1.0: Tokenized drug design with large language model", "abstract": "  Significant interests have recently risen in leveraging sequence-based large\nlanguage models (LLMs) for drug design. However, most current applications of\nLLMs in drug discovery lack the ability to comprehend three-dimensional (3D)\nstructures, thereby limiting their effectiveness in tasks that explicitly\ninvolve molecular conformations. In this study, we introduced Token-Mol, a\ntoken-only 3D drug design model. This model encodes all molecular information,\nincluding 2D and 3D structures, as well as molecular property data, into\ntokens, which transforms classification and regression tasks in drug discovery\ninto probabilistic prediction problems, thereby enabling learning through a\nunified paradigm. Token-Mol is built on the transformer decoder architecture\nand trained using random causal masking techniques. Additionally, we proposed\nthe Gaussian cross-entropy (GCE) loss function to overcome the challenges in\nregression tasks, significantly enhancing the capacity of LLMs to learn\ncontinuous numerical values. Through a combination of fine-tuning and\nreinforcement learning (RL), Token-Mol achieves performance comparable to or\nsurpassing existing task-specific methods across various downstream tasks,\nincluding pocket-based molecular generation, conformation generation, and\nmolecular property prediction. Compared to existing molecular pre-trained\nmodels, Token-Mol exhibits superior proficiency in handling a wider range of\ndownstream tasks essential for drug design. Notably, our approach improves\nregression task accuracy by approximately 30% compared to similar token-only\nmethods. Token-Mol overcomes the precision limitations of token-only models and\nhas the potential to integrate seamlessly with general models such as ChatGPT,\npaving the way for the development of a universal artificial intelligence drug\ndesign model that facilitates rapid and high-quality drug design by experts.\n", "date": "2024-07-10", "categories": []}
{"id": "2408.06402", "title": "PhaGO: Protein function annotation for bacteriophages by integrating the\n  genomic context", "abstract": "  Bacteriophages are viruses that target bacteria, playing a crucial role in\nmicrobial ecology. Phage proteins are important in understanding phage biology,\nsuch as virus infection, replication, and evolution. Although a large number of\nnew phages have been identified via metagenomic sequencing, many of them have\nlimited protein function annotation. Accurate function annotation of phage\nproteins presents several challenges, including their inherent diversity and\nthe scarcity of annotated ones. Existing tools have yet to fully leverage the\nunique properties of phages in annotating protein functions. In this work, we\npropose a new protein function annotation tool for phages by leveraging the\nmodular genomic structure of phage genomes. By employing embeddings from the\nlatest protein foundation models and Transformer to capture contextual\ninformation between proteins in phage genomes, PhaGO surpasses state-of-the-art\nmethods in annotating diverged proteins and proteins with uncommon functions by\n6.78% and 13.05% improvement, respectively. PhaGO can annotate proteins lacking\nhomology search results, which is critical for characterizing the rapidly\naccumulating phage genomes. We demonstrate the utility of PhaGO by identifying\n688 potential holins in phages, which exhibit high structural conservation with\nknown holins. The results show the potential of PhaGO to extend our\nunderstanding of newly discovered phages.\n", "date": "2024-08-12", "categories": []}
{"id": "2408.08341", "title": "Exploring Latent Space for Generating Peptide Analogs Using Protein\n  Language Models", "abstract": "  Generating peptides with desired properties is crucial for drug discovery and\nbiotechnology. Traditional sequence-based and structure-based methods often\nrequire extensive datasets, which limits their effectiveness. In this study, we\nproposed a novel method that utilized autoencoder shaped models to explore the\nprotein embedding space, and generate novel peptide analogs by leveraging\nprotein language models. The proposed method requires only a single sequence of\ninterest, avoiding the need for large datasets. Our results show significant\nimprovements over baseline models in similarity indicators of peptide\nstructures, descriptors and bioactivities. The proposed method validated\nthrough Molecular Dynamics simulations on TIGIT inhibitors, demonstrates that\nour method produces peptide analogs with similar yet distinct properties,\nhighlighting its potential to enhance peptide screening processes.\n", "date": "2024-08-15", "categories": []}
{"id": "2408.07709", "title": "Pretrained-Guided Conditional Diffusion Models for Microbiome Data\n  Analysis", "abstract": "  Emerging evidence indicates that human cancers are intricately linked to\nhuman microbiomes, forming an inseparable connection. However, due to limited\nsample sizes and significant data loss during collection for various reasons,\nsome machine learning methods have been proposed to address the issue of\nmissing data. These methods have not fully utilized the known clinical\ninformation of patients to enhance the accuracy of data imputation. Therefore,\nwe introduce mbVDiT, a novel pre-trained conditional diffusion model for\nmicrobiome data imputation and denoising, which uses the unmasked data and\npatient metadata as conditional guidance for imputating missing values. It is\nalso uses VAE to integrate the the other public microbiome datasets to enhance\nmodel performance. The results on the microbiome datasets from three different\ncancer types demonstrate the performance of our methods in comparison with\nexisting methods.\n", "date": "2024-08-10", "categories": []}
{"id": "2305.19800", "title": "Accurate and Efficient Structural Ensemble Generation of Macrocyclic\n  Peptides using Internal Coordinate Diffusion", "abstract": "  Macrocyclic peptides are an emerging therapeutic modality, yet computational\napproaches for accurately sampling their diverse 3D ensembles remain\nchallenging due to their conformational diversity and geometric constraints.\nHere, we introduce RINGER, a diffusion-based transformer model using a\nredundant internal coordinate representation that generates three-dimensional\nconformational ensembles of macrocyclic peptides from their 2D representations.\nRINGER provides fast backbone and side-chain sampling while respecting key\nstructural invariances of cyclic peptides. Through extensive benchmarking and\nanalysis against gold-standard conformer ensembles of cyclic peptides generated\nwith metadynamics, we demonstrate how RINGER generates both high-quality and\ndiverse geometries at a fraction of the computational cost. Our work lays the\nfoundation for improved sampling of cyclic geometries and the development of\ngeometric learning methods for peptides.\n", "date": "2023-05-30", "categories": []}
{"id": "2408.07233", "title": "Pan-cancer gene set discovery via scRNA-seq for optimal deep learning\n  based downstream tasks", "abstract": "  The application of machine learning to transcriptomics data has led to\nsignificant advances in cancer research. However, the high dimensionality and\ncomplexity of RNA sequencing (RNA-seq) data pose significant challenges in\npan-cancer studies. This study hypothesizes that gene sets derived from\nsingle-cell RNA sequencing (scRNA-seq) data will outperform those selected\nusing bulk RNA-seq in pan-cancer downstream tasks. We analyzed scRNA-seq data\nfrom 181 tumor biopsies across 13 cancer types. High-dimensional weighted gene\nco-expression network analysis (hdWGCNA) was performed to identify relevant\ngene sets, which were further refined using XGBoost for feature selection.\nThese gene sets were applied to downstream tasks using TCGA pan-cancer RNA-seq\ndata and compared to six reference gene sets and oncogenes from OncoKB\nevaluated with deep learning models, including multilayer perceptrons (MLPs)\nand graph neural networks (GNNs). The XGBoost-refined hdWGCNA gene set\ndemonstrated higher performance in most tasks, including tumor mutation burden\nassessment, microsatellite instability classification, mutation prediction,\ncancer subtyping, and grading. In particular, genes such as DPM1, BAD, and\nFKBP4 emerged as important pan-cancer biomarkers, with DPM1 consistently\nsignificant across tasks. This study presents a robust approach for feature\nselection in cancer genomics by integrating scRNA-seq data and advanced\nanalysis techniques, offering a promising avenue for improving predictive\naccuracy in cancer research.\n", "date": "2024-08-13", "categories": []}
{"id": "2408.06391", "title": "Autoregressive Enzyme Function Prediction with Multi-scale\n  Multi-modality Fusion", "abstract": "  Accurate prediction of enzyme function is crucial for elucidating biological\nmechanisms and driving innovation across various sectors. Existing deep\nlearning methods tend to rely solely on either sequence data or structural data\nand predict the EC number as a whole, neglecting the intrinsic hierarchical\nstructure of EC numbers. To address these limitations, we introduce MAPred, a\nnovel multi-modality and multi-scale model designed to autoregressively predict\nthe EC number of proteins. MAPred integrates both the primary amino acid\nsequence and the 3D tokens of proteins, employing a dual-pathway approach to\ncapture comprehensive protein characteristics and essential local functional\nsites. Additionally, MAPred utilizes an autoregressive prediction network to\nsequentially predict the digits of the EC number, leveraging the hierarchical\norganization of EC classifications. Evaluations on benchmark datasets,\nincluding New-392, Price, and New-815, demonstrate that our method outperforms\nexisting models, marking a significant advance in the reliability and\ngranularity of protein function prediction within bioinformatics.\n", "date": "2024-08-11", "categories": []}
{"id": "2401.08738", "title": "Machine Learning-Based Analysis of Ebola Virus' Impact on Gene\n  Expression in Nonhuman Primates", "abstract": "  This study introduces the Supervised Magnitude-Altitude Scoring (SMAS)\nmethodology, a machine learning-based approach, for analyzing gene expression\ndata obtained from nonhuman primates (NHPs) infected with Ebola virus (EBOV).\nWe utilize a comprehensive dataset of NanoString gene expression profiles from\nEbola-infected NHPs, deploying the SMAS system for nuanced host-pathogen\ninteraction analysis. SMAS effectively combines gene selection based on\nstatistical significance and expression changes, employing linear classifiers\nsuch as logistic regression to accurately differentiate between RT-qPCR\npositive and negative NHP samples. A key finding of our research is the\nidentification of IFI6 and IFI27 as critical biomarkers, demonstrating\nexceptional predictive performance with 100% accuracy and Area Under the Curve\n(AUC) metrics in classifying various stages of Ebola infection. Alongside IFI6\nand IFI27, genes, including MX1, OAS1, and ISG15, were significantly\nupregulated, highlighting their essential roles in the immune response to EBOV.\nOur results underscore the efficacy of the SMAS method in revealing complex\ngenetic interactions and response mechanisms during EBOV infection. This\nresearch provides valuable insights into EBOV pathogenesis and aids in\ndeveloping more precise diagnostic tools and therapeutic strategies to address\nEBOV infection in particular and viral infection in general.\n", "date": "2024-01-16", "categories": []}
{"id": "2405.18724", "title": "Adapting Differential Molecular Representation with Hierarchical Prompts\n  for Multi-label Property Prediction", "abstract": "  Accurate prediction of molecular properties is crucial in drug discovery.\nTraditional methods often overlook that real-world molecules typically exhibit\nmultiple property labels with complex correlations. To this end, we propose a\nnovel framework, HiPM, which stands for hierarchical prompted molecular\nrepresentation learning framework. HiPM leverages task-aware prompts to enhance\nthe differential expression of tasks in molecular representations and mitigate\nnegative transfer caused by conflicts in individual task information. Our\nframework comprises two core components: the Molecular Representation Encoder\n(MRE) and the Task-Aware Prompter (TAP). MRE employs a hierarchical\nmessage-passing network architecture to capture molecular features at both the\natom and motif levels. Meanwhile, TAP utilizes agglomerative hierarchical\nclustering algorithm to construct a prompt tree that reflects task affinity and\ndistinctiveness, enabling the model to consider multi-granular correlation\ninformation among tasks, thereby effectively handling the complexity of\nmulti-label property prediction. Extensive experiments demonstrate that HiPM\nachieves state-of-the-art performance across various multi-label datasets,\noffering a novel perspective on multi-label molecular representation learning.\n", "date": "2024-05-29", "categories": []}
{"id": "2408.05258", "title": "scASDC: Attention Enhanced Structural Deep Clustering for Single-cell\n  RNA-seq Data", "abstract": "  Single-cell RNA sequencing (scRNA-seq) data analysis is pivotal for\nunderstanding cellular heterogeneity. However, the high sparsity and complex\nnoise patterns inherent in scRNA-seq data present significant challenges for\ntraditional clustering methods. To address these issues, we propose a deep\nclustering method, Attention-Enhanced Structural Deep Embedding Graph\nClustering (scASDC), which integrates multiple advanced modules to improve\nclustering accuracy and robustness.Our approach employs a multi-layer graph\nconvolutional network (GCN) to capture high-order structural relationships\nbetween cells, termed as the graph autoencoder module. To mitigate the\noversmoothing issue in GCNs, we introduce a ZINB-based autoencoder module that\nextracts content information from the data and learns latent representations of\ngene expression. These modules are further integrated through an attention\nfusion mechanism, ensuring effective combination of gene expression and\nstructural information at each layer of the GCN. Additionally, a\nself-supervised learning module is incorporated to enhance the robustness of\nthe learned embeddings. Extensive experiments demonstrate that scASDC\noutperforms existing state-of-the-art methods, providing a robust and effective\nsolution for single-cell clustering tasks. Our method paves the way for more\naccurate and meaningful analysis of single-cell RNA sequencing data,\ncontributing to better understanding of cellular heterogeneity and biological\nprocesses. All code and public datasets used in this paper are available at\n\\url{https://github.com/wenwenmin/scASDC} and\n\\url{https://zenodo.org/records/12814320}.\n", "date": "2024-08-09", "categories": []}
{"id": "2408.06244", "title": "3D Reconstruction of Protein Structures from Multi-view AFM Images using\n  Neural Radiance Fields (NeRFs)", "abstract": "  Recent advancements in deep learning for predicting 3D protein structures\nhave shown promise, particularly when leveraging inputs like protein sequences\nand Cryo-Electron microscopy (Cryo-EM) images. However, these techniques often\nfall short when predicting the structures of protein complexes (PCs), which\ninvolve multiple proteins. In our study, we investigate using atomic force\nmicroscopy (AFM) combined with deep learning to predict the 3D structures of\nPCs. AFM generates height maps that depict the PCs in various random\norientations, providing a rich information for training a neural network to\npredict the 3D structures. We then employ the pre-trained UpFusion model (which\nutilizes a conditional diffusion model for synthesizing novel views) to train\nan instance-specific NeRF model for 3D reconstruction. The performance of\nUpFusion is evaluated through zero-shot predictions of 3D protein structures\nusing AFM images. The challenge, however, lies in the time-intensive and\nimpractical nature of collecting actual AFM images. To address this, we use a\nvirtual AFM imaging process that transforms a `PDB' protein file into\nmulti-view 2D virtual AFM images via volume rendering techniques. We\nextensively validate the UpFusion architecture using both virtual and actual\nmulti-view AFM images. Our results include a comparison of structures predicted\nwith varying numbers of views and different sets of views. This novel approach\nholds significant potential for enhancing the accuracy of protein complex\nstructure predictions with further fine-tuning of the UpFusion network.\n", "date": "2024-08-12", "categories": []}
{"id": "2408.04847", "title": "A Pipeline for Data-Driven Learning of Topological Features with\n  Applications to Protein Stability Prediction", "abstract": "  In this paper, we propose a data-driven method to learn interpretable\ntopological features of biomolecular data and demonstrate the efficacy of\nparsimonious models trained on topological features in predicting the stability\nof synthetic mini proteins. We compare models that leverage\nautomatically-learned structural features against models trained on a large set\nof biophysical features determined by subject-matter experts (SME). Our models,\nbased only on topological features of the protein structures, achieved 92%-99%\nof the performance of SME-based models in terms of the average precision score.\nBy interrogating model performance and feature importance metrics, we extract\nnumerous insights that uncover high correlations between topological features\nand SME features. We further showcase how combining topological features and\nSME features can lead to improved model performance over either feature set\nused in isolation, suggesting that, in some settings, topological features may\nprovide new discriminating information not captured in existing SME features\nthat are useful for protein stability prediction.\n", "date": "2024-08-09", "categories": []}
{"id": "2408.05065", "title": "Masked adversarial neural network for cell type deconvolution in spatial\n  transcriptomics", "abstract": "  Accurately determining cell type composition in disease-relevant tissues is\ncrucial for identifying disease targets. Most existing spatial transcriptomics\n(ST) technologies cannot achieve single-cell resolution, making it challenging\nto accurately determine cell types. To address this issue, various\ndeconvolution methods have been developed. Most of these methods use\nsingle-cell RNA sequencing (scRNA-seq) data from the same tissue as a reference\nto infer cell types in ST data spots. However, they often overlook the\ndifferences between scRNA-seq and ST data. To overcome this limitation, we\npropose a Masked Adversarial Neural Network (MACD). MACD employs adversarial\nlearning to align real ST data with simulated ST data generated from scRNA-seq\ndata. By mapping them into a unified latent space, it can minimize the\ndifferences between the two types of data. Additionally, MACD uses masking\ntechniques to effectively learn the features of real ST data and mitigate\nnoise. We evaluated MACD on 32 simulated datasets and 2 real datasets,\ndemonstrating its accuracy in performing cell type deconvolution. All code and\npublic datasets used in this paper are available at\nhttps://github.com/wenwenmin/MACD and https://zenodo.org/records/12804822.\n", "date": "2024-08-09", "categories": []}
{"id": "2209.12635", "title": "ImmunoLingo: Linguistics-based formalization of the antibody language", "abstract": "  Apparent parallels between natural language and biological sequence have led\nto a recent surge in the application of deep language models (LMs) to the\nanalysis of antibody and other biological sequences. However, a lack of a\nrigorous linguistic formalization of biological sequence languages, which would\ndefine basic components, such as lexicon (i.e., the discrete units of the\nlanguage) and grammar (i.e., the rules that link sequence well-formedness,\nstructure, and meaning) has led to largely domain-unspecific applications of\nLMs, which do not take into account the underlying structure of the biological\nsequences studied. A linguistic formalization, on the other hand, establishes\nlinguistically-informed and thus domain-adapted components for LM applications.\nIt would facilitate a better understanding of how differences and similarities\nbetween natural language and biological sequences influence the quality of LMs,\nwhich is crucial for the design of interpretable models with extractable\nsequence-functions relationship rules, such as the ones underlying the antibody\nspecificity prediction problem. Deciphering the rules of antibody specificity\nis crucial to accelerating rational and in silico biotherapeutic drug design.\nHere, we formalize the properties of the antibody language and thereby\nestablish not only a foundation for the application of linguistic tools in\nadaptive immune receptor analysis but also for the systematic immunolinguistic\nstudies of immune receptor specificity in general.\n", "date": "2022-09-26", "categories": []}
{"id": "2407.20879", "title": "A Scalable Tool For Analyzing Genomic Variants Of Humans Using Knowledge\n  Graphs and Machine Learning", "abstract": "  The integration of knowledge graphs and graph machine learning (GML) in\ngenomic data analysis offers several opportunities for understanding complex\ngenetic relationships, especially at the RNA level. We present a comprehensive\napproach for leveraging these technologies to analyze genomic variants,\nspecifically in the context of RNA sequencing (RNA-seq) data from COVID-19\npatient samples. The proposed method involves extracting variant-level genetic\ninformation, annotating the data with additional metadata using SnpEff, and\nconverting the enriched Variant Call Format (VCF) files into Resource\nDescription Framework (RDF) triples. The resulting knowledge graph is further\nenhanced with patient metadata and stored in a graph database, facilitating\nefficient querying and indexing. We utilize the Deep Graph Library (DGL) to\nperform graph machine learning tasks, including node classification with\nGraphSAGE and Graph Convolutional Networks (GCNs). Our approach demonstrates\nsignificant utility using our proposed tool, VariantKG, in three key scenarios:\nenriching graphs with new VCF data, creating subgraphs based on user-defined\nfeatures, and conducting graph machine learning for node classification.\n", "date": "2024-07-30", "categories": []}
{"id": "2408.01156", "title": "TCR-GPT: Integrating Autoregressive Model and Reinforcement Learning for\n  T-Cell Receptor Repertoires Generation", "abstract": "  T-cell receptors (TCRs) play a crucial role in the immune system by\nrecognizing and binding to specific antigens presented by infected or cancerous\ncells. Understanding the sequence patterns of TCRs is essential for developing\ntargeted immune therapies and designing effective vaccines. Language models,\nsuch as auto-regressive transformers, offer a powerful solution to this problem\nby learning the probability distributions of TCR repertoires, enabling the\ngeneration of new TCR sequences that inherit the underlying patterns of the\nrepertoire. We introduce TCR-GPT, a probabilistic model built on a decoder-only\ntransformer architecture, designed to uncover and replicate sequence patterns\nin TCR repertoires. TCR-GPT demonstrates an accuracy of 0.953 in inferring\nsequence probability distributions measured by Pearson correlation coefficient.\nFurthermore, by leveraging Reinforcement Learning(RL), we adapted the\ndistribution of TCR sequences to generate TCRs capable of recognizing specific\npeptides, offering significant potential for advancing targeted immune\ntherapies and vaccine development. With the efficacy of RL, fine-tuned\npretrained TCR-GPT models demonstrated the ability to produce TCR repertoires\nlikely to bind specific peptides, illustrating RL's efficiency in enhancing the\nmodel's adaptability to the probability distributions of biologically relevant\nTCR sequences.\n", "date": "2024-08-02", "categories": []}
{"id": "2407.20660", "title": "What makes for good morphology representations for spatial omics?", "abstract": "  Spatial omics has transformed our understanding of tissue architecture by\npreserving spatial context of gene expression patterns. Simultaneously,\nadvances in imaging AI have enabled extraction of morphological features\ndescribing the tissue. The intersection of spatial omics and imaging AI\npresents opportunities for a more holistic understanding. In this review we\nintroduce a framework for categorizing spatial omics-morphology combination\nmethods, focusing on how morphological features can be translated or integrated\ninto spatial omics analyses. By translation we mean finding morphological\nfeatures that spatially correlate with gene expression patterns with the\npurpose of predicting gene expression. Such features can be used to generate\nsuper-resolution gene expression maps or infer genetic information from\nclinical H&E-stained samples. By integration we mean finding morphological\nfeatures that spatially complement gene expression patterns with the purpose of\nenriching information. Such features can be used to define spatial domains,\nespecially where gene expression has preceded morphological changes and where\nmorphology remains after gene expression. We discuss learning strategies and\ndirections for further development of the field.\n", "date": "2024-07-30", "categories": []}
{"id": "2407.21028", "title": "Antibody DomainBed: Out-of-Distribution Generalization in Therapeutic\n  Protein Design", "abstract": "  Machine learning (ML) has demonstrated significant promise in accelerating\ndrug design. Active ML-guided optimization of therapeutic molecules typically\nrelies on a surrogate model predicting the target property of interest. The\nmodel predictions are used to determine which designs to evaluate in the lab,\nand the model is updated on the new measurements to inform the next cycle of\ndecisions. A key challenge is that the experimental feedback from each cycle\ninspires changes in the candidate proposal or experimental protocol for the\nnext cycle, which lead to distribution shifts. To promote robustness to these\nshifts, we must account for them explicitly in the model training. We apply\ndomain generalization (DG) methods to classify the stability of interactions\nbetween an antibody and antigen across five domains defined by design cycles.\nOur results suggest that foundational models and ensembling improve predictive\nperformance on out-of-distribution domains. We publicly release our codebase\nextending the DG benchmark ``DomainBed,'' and the associated dataset of\nantibody sequences and structures emulating distribution shifts across design\ncycles.\n", "date": "2024-07-15", "categories": []}
{"id": "2407.21298", "title": "A Vectorization Method Induced By Maximal Margin Classification For\n  Persistent Diagrams", "abstract": "  Persistent homology is an effective method for extracting topological\ninformation, represented as persistent diagrams, of spatial structure data.\nHence it is well-suited for the study of protein structures. Attempts to\nincorporate Persistent homology in machine learning methods of protein function\nprediction have resulted in several techniques for vectorizing persistent\ndiagrams. However, current vectorization methods are excessively artificial and\ncannot ensure the effective utilization of information or the rationality of\nthe methods. To address this problem, we propose a more geometrical\nvectorization method of persistent diagrams based on maximal margin\nclassification for Banach space, and additionaly propose a framework that\nutilizes topological data analysis to identify proteins with specific\nfunctions. We evaluated our vectorization method using a binary classification\ntask on proteins and compared it with the statistical methods that exhibit the\nbest performance among thirteen commonly used vectorization methods. The\nexperimental results indicate that our approach surpasses the statistical\nmethods in both robustness and precision.\n", "date": "2024-07-31", "categories": []}
{"id": "2407.21714", "title": "UMMAN: Unsupervised Multi-graph Merge Adversarial Network for Disease\n  Prediction Based on Intestinal Flora", "abstract": "  The abundance of intestinal flora is closely related to human diseases, but\ndiseases are not caused by a single gut microbe. Instead, they result from the\ncomplex interplay of numerous microbial entities. This intricate and implicit\nconnection among gut microbes poses a significant challenge for disease\nprediction using abundance information from OTU data. Recently, several methods\nhave shown potential in predicting corresponding diseases. However, these\nmethods fail to learn the inner association among gut microbes from different\nhosts, leading to unsatisfactory performance. In this paper, we present a novel\narchitecture, Unsupervised Multi-graph Merge Adversarial Network (UMMAN). UMMAN\ncan obtain the embeddings of nodes in the Multi-Graph in an unsupervised\nscenario, so that it helps learn the multiplex association. Our method is the\nfirst to combine Graph Neural Network with the task of intestinal flora disease\nprediction. We employ complex relation-types to construct the Original-Graph\nand disrupt the relationships among nodes to generate corresponding\nShuffled-Graph. We introduce the Node Feature Global Integration (NFGI) module\nto represent the global features of the graph. Furthermore, we design a joint\nloss comprising adversarial loss and hybrid attention loss to ensure that the\nreal graph embedding aligns closely with the Original-Graph and diverges from\nthe Shuffled-Graph. Comprehensive experiments on five classical OTU gut\nmicrobiome datasets demonstrate the effectiveness and stability of our method.\n(We will release our code soon.)\n", "date": "2024-07-31", "categories": []}
{"id": "2407.20518", "title": "High-Resolution Spatial Transcriptomics from Histology Images using\n  HisToSGE", "abstract": "  Spatial transcriptomics (ST) is a groundbreaking genomic technology that\nenables spatial localization analysis of gene expression within tissue\nsections. However, it is significantly limited by high costs and sparse spatial\nresolution. An alternative, more cost-effective strategy is to use deep\nlearning methods to predict high-density gene expression profiles from\nhistological images. However, existing methods struggle to capture rich image\nfeatures effectively or rely on low-dimensional positional coordinates, making\nit difficult to accurately predict high-resolution gene expression profiles. To\naddress these limitations, we developed HisToSGE, a method that employs a\nPathology Image Large Model (PILM) to extract rich image features from\nhistological images and utilizes a feature learning module to robustly generate\nhigh-resolution gene expression profiles. We evaluated HisToSGE on four ST\ndatasets, comparing its performance with five state-of-the-art baseline\nmethods. The results demonstrate that HisToSGE excels in generating\nhigh-resolution gene expression profiles and performing downstream tasks such\nas spatial domain identification. All code and public datasets used in this\npaper are available at https://github.com/wenwenmin/HisToSGE and\nhttps://zenodo.org/records/12792163.\n", "date": "2024-07-30", "categories": []}
{"id": "2405.18507", "title": "Injecting Hierarchical Biological Priors into Graph Neural Networks for\n  Flow Cytometry Prediction", "abstract": "  In the complex landscape of hematologic samples such as peripheral blood or\nbone marrow derived from flow cytometry (FC) data, cell-level prediction\npresents profound challenges. This work explores injecting hierarchical prior\nknowledge into graph neural networks (GNNs) for single-cell multi-class\nclassification of tabular cellular data. By representing the data as graphs and\nencoding hierarchical relationships between classes, we propose our\nhierarchical plug-in method to be applied to several GNN models, namely,\nFCHC-GNN, and effectively designed to capture neighborhood information crucial\nfor single-cell FC domain. Extensive experiments on our cohort of 19 distinct\npatients, demonstrate that incorporating hierarchical biological constraints\nboosts performance significantly across multiple metrics compared to baseline\nGNNs without such priors. The proposed approach highlights the importance of\nstructured inductive biases for gaining improved generalization in complex\nbiological prediction tasks.\n", "date": "2024-05-28", "categories": []}
{"id": "2407.19039", "title": "GraphBPE: Molecular Graphs Meet Byte-Pair Encoding", "abstract": "  With the increasing attention to molecular machine learning, various\ninnovations have been made in designing better models or proposing more\ncomprehensive benchmarks. However, less is studied on the data preprocessing\nschedule for molecular graphs, where a different view of the molecular graph\ncould potentially boost the model's performance. Inspired by the Byte-Pair\nEncoding (BPE) algorithm, a subword tokenization method popularly adopted in\nNatural Language Processing, we propose GraphBPE, which tokenizes a molecular\ngraph into different substructures and acts as a preprocessing schedule\nindependent of the model architectures. Our experiments on 3 graph-level\nclassification and 3 graph-level regression datasets show that data\npreprocessing could boost the performance of models for molecular graphs, and\nGraphBPE is effective for small classification datasets and it performs on par\nwith other tokenization methods across different model architectures.\n", "date": "2024-07-26", "categories": []}
{"id": "2407.19073", "title": "On Machine Learning Approaches for Protein-Ligand Binding Affinity\n  Prediction", "abstract": "  Binding affinity optimization is crucial in early-stage drug discovery. While\nnumerous machine learning methods exist for predicting ligand potency, their\ncomparative efficacy remains unclear. This study evaluates the performance of\nclassical tree-based models and advanced neural networks in protein-ligand\nbinding affinity prediction. Our comprehensive benchmarking encompasses 2D\nmodels utilizing ligand-only RDKit embeddings and Large Language Model (LLM)\nligand representations, as well as 3D neural networks incorporating bound\nprotein-ligand conformations. We assess these models across multiple standard\ndatasets, examining various predictive scenarios including classification,\nranking, regression, and active learning. Results indicate that simpler models\ncan surpass more complex ones in specific tasks, while 3D models leveraging\nstructural information become increasingly competitive with larger training\ndatasets containing compounds with labelled affinity data against multiple\ntargets. Pre-trained 3D models, by incorporating protein pocket environments,\ndemonstrate significant advantages in data-scarce scenarios for specific\nbinding pockets. Additionally, LLM pretraining on 2D ligand data enhances\ncomplex model performance, providing versatile embeddings that outperform\ntraditional RDKit features in computational efficiency. Finally, we show that\ncombining 2D and 3D model strengths improves active learning outcomes beyond\ncurrent state-of-the-art approaches. These findings offer valuable insights for\noptimizing machine learning strategies in drug discovery pipelines.\n", "date": "2024-07-15", "categories": []}
{"id": "2407.19349", "title": "Predicting T-Cell Receptor Specificity", "abstract": "  Researching the specificity of TCR contributes to the development of\nimmunotherapy and provides new opportunities and strategies for personalized\ncancer immunotherapy. Therefore, we established a TCR generative specificity\ndetection framework consisting of an antigen selector and a TCR classifier\nbased on the Random Forest algorithm, aiming to efficiently screen out TCRs and\ntarget antigens and achieve TCR specificity prediction. Furthermore, we used\nthe k-fold validation method to compare the performance of our model with\nordinary deep learning methods. The result proves that adding a classifier to\nthe model based on the random forest algorithm is very effective, and our model\ngenerally outperforms ordinary deep learning methods. Moreover, we put forward\nfeasible optimization suggestions for the shortcomings and challenges of our\nmodel found during model implementation.\n", "date": "2024-07-27", "categories": []}
{"id": "2407.18317", "title": "CavDetect: A DBSCAN Algorithm based Novel Cavity Detection Model on\n  Protein Structure", "abstract": "  Cavities on the structures of proteins are formed due to interaction between\nproteins and some small molecules, known as ligands. These are basically the\nlocations where ligands bind with proteins. Actual detection of such locations\nis all-important to succeed in the entire drug design process. This study\nproposes a Voronoi Tessellation based novel cavity detection model that is used\nto detect cavities on the structure of proteins. As the atom space of protein\nstructure is dense and of large volumes and the DBSCAN (Density Based Spatial\nClustering of Applications with Noise) algorithm can handle such type of data\nvery well as well as it is not mandatory to have knowledge about the numbers of\nclusters (cavities) in data as priori in this algorithm, this study proposes to\nimplement the proposed algorithm with the DBSCAN algorithm.\n", "date": "2024-07-25", "categories": []}
{"id": "2309.03227", "title": "Learning a Patent-Informed Biomedical Knowledge Graph Reveals\n  Technological Potential of Drug Repositioning Candidates", "abstract": "  Drug repositioning-a promising strategy for discovering new therapeutic uses\nfor existing drugs-has been increasingly explored in the computational science\nliterature using biomedical databases. However, the technological potential of\ndrug repositioning candidates has often been overlooked. This study presents a\nnovel protocol to comprehensively analyse various sources such as\npharmaceutical patents and biomedical databases, and identify drug\nrepositioning candidates with both technological potential and scientific\nevidence. To this end, first, we constructed a scientific biomedical knowledge\ngraph (s-BKG) comprising relationships between drugs, diseases, and genes\nderived from biomedical databases. Our protocol involves identifying drugs that\nexhibit limited association with the target disease but are closely located in\nthe s-BKG, as potential drug candidates. We constructed a patent-informed\nbiomedical knowledge graph (p-BKG) by adding pharmaceutical patent information.\nFinally, we developed a graph embedding protocol to ascertain the structure of\nthe p-BKG, thereby calculating the relevance scores of those candidates with\ntarget disease-related patents to evaluate their technological potential. Our\ncase study on Alzheimer's disease demonstrates its efficacy and feasibility,\nwhile the quantitative outcomes and systematic methods are expected to bridge\nthe gap between computational discoveries and successful market applications in\ndrug repositioning research.\n", "date": "2023-09-04", "categories": []}
{"id": "2405.15700", "title": "Trackastra: Transformer-based cell tracking for live-cell microscopy", "abstract": "  Cell tracking is a ubiquitous image analysis task in live-cell microscopy.\nUnlike multiple object tracking (MOT) for natural images, cell tracking\ntypically involves hundreds of similar-looking objects that can divide in each\nframe, making it a particularly challenging problem. Current state-of-the-art\napproaches follow the tracking-by-detection paradigm, i.e. first all cells are\ndetected per frame and successively linked in a second step to form\nbiologically consistent cell tracks. Linking is commonly solved via discrete\noptimization methods, which require manual tuning of hyperparameters for each\ndataset and are therefore cumbersome to use in practice. Here we propose\nTrackastra, a general purpose cell tracking approach that uses a simple\ntransformer architecture to directly learn pairwise associations of cells\nwithin a temporal window from annotated data. Importantly, unlike existing\ntransformer-based MOT pipelines, our learning architecture also accounts for\ndividing objects such as cells and allows for accurate tracking even with\nsimple greedy linking, thus making strides towards removing the requirement for\na complex linking step. The proposed architecture operates on the full\nspatio-temporal context of detections within a time window by avoiding the\ncomputational burden of processing dense images. We show that our tracking\napproach performs on par with or better than highly tuned state-of-the-art cell\ntracking algorithms for various biological datasets, such as bacteria, cell\ncultures and fluorescent particles. We provide code at\nhttps://github.com/weigertlab/trackastra.\n", "date": "2024-05-24", "categories": []}
{"id": "2407.16709", "title": "LSTM Autoencoder-based Deep Neural Networks for Barley\n  Genotype-to-Phenotype Prediction", "abstract": "  Artificial Intelligence (AI) has emerged as a key driver of precision\nagriculture, facilitating enhanced crop productivity, optimized resource use,\nfarm sustainability, and informed decision-making. Also, the expansion of\ngenome sequencing technology has greatly increased crop genomic resources,\ndeepening our understanding of genetic variation and enhancing desirable crop\ntraits to optimize performance in various environments. There is increasing\ninterest in using machine learning (ML) and deep learning (DL) algorithms for\ngenotype-to-phenotype prediction due to their excellence in capturing complex\ninteractions within large, high-dimensional datasets. In this work, we propose\na new LSTM autoencoder-based model for barley genotype-to-phenotype prediction,\nspecifically for flowering time and grain yield estimation, which could\npotentially help optimize yields and management practices. Our model\noutperformed the other baseline methods, demonstrating its potential in\nhandling complex high-dimensional agricultural datasets and enhancing crop\nphenotype prediction performance.\n", "date": "2024-07-21", "categories": []}
{"id": "2407.16984", "title": "scGHSOM: Hierarchical clustering and visualization of single-cell and\n  CRISPR data using growing hierarchical SOM", "abstract": "  High-dimensional single-cell data poses significant challenges in identifying\nunderlying biological patterns due to the complexity and heterogeneity of\ncellular states. We propose a comprehensive gene-cell dependency visualization\nvia unsupervised clustering, Growing Hierarchical Self-Organizing Map (GHSOM),\nspecifically designed for analyzing high-dimensional single-cell data like\nsingle-cell sequencing and CRISPR screens. GHSOM is applied to cluster samples\nin a hierarchical structure such that the self-growth structure of clusters\nsatisfies the required variations between and within. We propose a novel\nSignificant Attributes Identification Algorithm to identify features that\ndistinguish clusters. This algorithm pinpoints attributes with minimal\nvariation within a cluster but substantial variation between clusters. These\nkey attributes can then be used for targeted data retrieval and downstream\nanalysis. Furthermore, we present two innovative visualization tools: Cluster\nFeature Map and Cluster Distribution Map. The Cluster Feature Map highlights\nthe distribution of specific features across the hierarchical structure of\nGHSOM clusters. This allows for rapid visual assessment of cluster uniqueness\nbased on chosen features. The Cluster Distribution Map depicts leaf clusters as\ncircles on the GHSOM grid, with circle size reflecting cluster data size and\ncolor customizable to visualize features like cell type or other attributes. We\napply our analysis to three single-cell datasets and one CRISPR dataset\n(cell-gene database) and evaluate clustering methods with internal and external\nCH and ARI scores. GHSOM performs well, being the best performer in internal\nevaluation (CH=4.2). In external evaluation, GHSOM has the third-best\nperformance of all methods.\n", "date": "2024-07-24", "categories": []}
{"id": "2407.15880", "title": "Diff4VS: HIV-inhibiting Molecules Generation with Classifier Guidance\n  Diffusion for Virtual Screening", "abstract": "  The AIDS epidemic has killed 40 million people and caused serious global\nproblems. The identification of new HIV-inhibiting molecules is of great\nimportance for combating the AIDS epidemic. Here, the Classifier Guidance\nDiffusion model and ligand-based virtual screening strategy are combined to\ndiscover potential HIV-inhibiting molecules for the first time. We call it\nDiff4VS. An extra classifier is trained using the HIV molecule dataset, and the\ngradient of the classifier is used to guide the Diffusion to generate\nHIV-inhibiting molecules. Experiments show that Diff4VS can generate more\ncandidate HIV-inhibiting molecules than other methods. Inspired by ligand-based\nvirtual screening, a new metric DrugIndex is proposed. The DrugIndex is the\nratio of the proportion of candidate drug molecules in the generated molecule\nto the proportion of candidate drug molecules in the training set. DrugIndex\nprovides a new evaluation method for evolving molecular generative models from\na pharmaceutical perspective. Besides, we report a new phenomenon observed when\nusing molecule generation models for virtual screening. Compared to real\nmolecules, the generated molecules have a lower proportion that is highly\nsimilar to known drug molecules. We call it Degradation in molecule generation.\nBased on the data analysis, the Degradation may result from the difficulty of\ngenerating molecules with a specific structure in the generative model. Our\nresearch contributes to the application of generative models in drug design\nfrom method, metric, and phenomenon analysis.\n", "date": "2024-07-20", "categories": []}
{"id": "2407.07055", "title": "Multicell-Fold: geometric learning in folding multicellular life", "abstract": "  During developmental processes such as embryogenesis, how a group of cells\nfold into specific structures, is a central question in biology that defines\nhow living organisms form. Establishing tissue-level morphology critically\nrelies on how every single cell decides to position itself relative to its\nneighboring cells. Despite its importance, it remains a major challenge to\nunderstand and predict the behavior of every cell within the living tissue over\ntime during such intricate processes. To tackle this question, we propose a\ngeometric deep learning model that can predict multicellular folding and\nembryogenesis, accurately capturing the highly convoluted spatial interactions\namong cells. We demonstrate that multicellular data can be represented with\nboth granular and foam-like physical pictures through a unified graph data\nstructure, considering both cellular interactions and cell junction networks.\nWe successfully use our model to achieve two important tasks, interpretable 4-D\nmorphological sequence alignment, and predicting local cell rearrangements\nbefore they occur at single-cell resolution. Furthermore, using an activation\nmap and ablation studies, we demonstrate that cell geometries and cell junction\nnetworks together regulate local cell rearrangement which is critical for\nembryo morphogenesis. This approach provides a novel paradigm to study\nmorphogenesis, highlighting a unified data structure and harnessing the power\nof geometric deep learning to accurately model the mechanisms and behaviors of\ncells during development. It offers a pathway toward creating a unified dynamic\nmorphological atlas for a variety of developmental processes such as\nembryogenesis.\n", "date": "2024-07-09", "categories": []}
{"id": "2407.14968", "title": "Technical report: Improving the properties of molecules generated by\n  LIMO", "abstract": "  This technical report investigates variants of the Latent Inceptionism on\nMolecules (LIMO) framework to improve the properties of generated molecules. We\nconduct ablative studies of molecular representation, decoder model, and\nsurrogate model training scheme. The experiments suggest that an autogressive\nTransformer decoder with GroupSELFIES achieves the best average properties for\nthe random generation task.\n", "date": "2024-07-20", "categories": []}
{"id": "2401.04082", "title": "Improved motif-scaffolding with SE(3) flow matching", "abstract": "  Protein design often begins with the knowledge of a desired function from a\nmotif which motif-scaffolding aims to construct a functional protein around.\nRecently, generative models have achieved breakthrough success in designing\nscaffolds for a range of motifs. However, generated scaffolds tend to lack\nstructural diversity, which can hinder success in wet-lab validation. In this\nwork, we extend FrameFlow, an SE(3) flow matching model for protein backbone\ngeneration, to perform motif-scaffolding with two complementary approaches. The\nfirst is motif amortization, in which FrameFlow is trained with the motif as\ninput using a data augmentation strategy. The second is motif guidance, which\nperforms scaffolding using an estimate of the conditional score from FrameFlow\nwithout additional training. On a benchmark of 24 biologically meaningful\nmotifs, we show our method achieves 2.5 times more designable and unique\nmotif-scaffolds compared to state-of-the-art. Code:\nhttps://github.com/microsoft/protein-frame-flow\n", "date": "2024-01-08", "categories": []}
{"id": "2402.01542", "title": "Learning Collective Variables with Synthetic Data Augmentation through\n  Physics-Inspired Geodesic Interpolation", "abstract": "  In molecular dynamics simulations, rare events, such as protein folding, are\ntypically studied using enhanced sampling techniques, most of which are based\non the definition of a collective variable (CV) along which acceleration\noccurs. Obtaining an expressive CV is crucial, but often hindered by the lack\nof information about the particular event, e.g., the transition from unfolded\nto folded conformation. We propose a simulation-free data augmentation strategy\nusing physics-inspired metrics to generate geodesic interpolations resembling\nprotein folding transitions, thereby improving sampling efficiency without true\ntransition state samples. This new data can be used to improve the accuracy of\nclassifier-based methods. Alternatively, a regression-based learning scheme for\nCV models can be adopted by leveraging the interpolation progress parameter.\n", "date": "2024-02-02", "categories": []}
{"id": "2407.13780", "title": "Generative Model for Small Molecules with Latent Space RL Fine-Tuning to\n  Protein Targets", "abstract": "  A specific challenge with deep learning approaches for molecule generation is\ngenerating both syntactically valid and chemically plausible molecular string\nrepresentations. To address this, we propose a novel generative latent-variable\ntransformer model for small molecules that leverages a recently proposed\nmolecular string representation called SAFE. We introduce a modification to\nSAFE to reduce the number of invalid fragmented molecules generated during\ntraining and use this to train our model. Our experiments show that our model\ncan generate novel molecules with a validity rate > 90% and a fragmentation\nrate < 1% by sampling from a latent space. By fine-tuning the model using\nreinforcement learning to improve molecular docking, we significantly increase\nthe number of hit candidates for five specific protein targets compared to the\npre-trained model, nearly doubling this number for certain targets.\nAdditionally, our top 5% mean docking scores are comparable to the current\nstate-of-the-art (SOTA), and we marginally outperform SOTA on three of the five\ntargets.\n", "date": "2024-07-02", "categories": []}
{"id": "2407.14185", "title": "Achieving Well-Informed Decision-Making in Drug Discovery: A\n  Comprehensive Calibration Study using Neural Network-Based Structure-Activity\n  Models", "abstract": "  In the drug discovery process, where experiments can be costly and\ntime-consuming, computational models that predict drug-target interactions are\nvaluable tools to accelerate the development of new therapeutic agents.\nEstimating the uncertainty inherent in these neural network predictions\nprovides valuable information that facilitates optimal decision-making when\nrisk assessment is crucial. However, such models can be poorly calibrated,\nwhich results in unreliable uncertainty estimates that do not reflect the true\npredictive uncertainty. In this study, we compare different metrics, including\naccuracy and calibration scores, used for model hyperparameter tuning to\ninvestigate which model selection strategy achieves well-calibrated models.\nFurthermore, we propose to use a computationally efficient Bayesian uncertainty\nestimation method named Bayesian Linear Probing (BLP), which generates\nHamiltonian Monte Carlo (HMC) trajectories to obtain samples for the parameters\nof a Bayesian Logistic Regression fitted to the hidden layer of the baseline\nneural network. We report that BLP improves model calibration and achieves the\nperformance of common uncertainty quantification methods by combining the\nbenefits of uncertainty estimation and probability calibration methods.\nFinally, we show that combining post hoc calibration method with\nwell-performing uncertainty quantification approaches can boost model accuracy\nand calibration.\n", "date": "2024-07-19", "categories": []}
{"id": "2407.03897", "title": "gFlora: a topology-aware method to discover functional co-response\n  groups in soil microbial communities", "abstract": "  We aim to learn the functional co-response group: a group of taxa whose\nco-response effect (the representative characteristic of the group showing the\ntotal topological abundance of taxa) co-responds (associates well\nstatistically) to a functional variable. Different from the state-of-the-art\nmethod, we model the soil microbial community as an ecological co-occurrence\nnetwork with the taxa as nodes (weighted by their abundance) and their\nrelationships (a combination from both spatial and functional ecological\naspects) as edges (weighted by the strength of the relationships). Then, we\ndesign a method called gFlora which notably uses graph convolution over this\nco-occurrence network to get the co-response effect of the group, such that the\nnetwork topology is also considered in the discovery process. We evaluate\ngFlora on two real-world soil microbiome datasets (bacteria and nematodes) and\ncompare it with the state-of-the-art method. gFlora outperforms this on all\nevaluation metrics, and discovers new functional evidence for taxa which were\nso far under-studied. We show that the graph convolution step is crucial to\ntaxa with relatively low abundance (thus removing the bias towards taxa with\nhigher abundance), and the discovered bacteria of different genera are\ndistributed in the co-occurrence network but still tightly connected among\nthemselves, demonstrating that topologically they fill different but\ncollaborative functional roles in the ecological community.\n", "date": "2024-07-04", "categories": []}
{"id": "2407.13023", "title": "A Three-Stage Algorithm for the Closest String Problem on Artificial and\n  Real Gene Sequences", "abstract": "  The Closest String Problem is an NP-hard problem that aims to find a string\nthat has the minimum distance from all sequences that belong to the given set\nof strings. Its applications can be found in coding theory, computational\nbiology, and designing degenerated primers, among others. There are efficient\nexact algorithms that have reached high-quality solutions for binary sequences.\nHowever, there is still room for improvement concerning the quality of\nsolutions over DNA and protein sequences. In this paper, we introduce a\nthree-stage algorithm that comprises the following process: first, we apply a\nnovel alphabet pruning method to reduce the search space for effectively\nfinding promising search regions. Second, a variant of beam search to find a\nheuristic solution is employed. This method utilizes a newly developed guiding\nfunction based on an expected distance heuristic score of partial solutions.\nLast, we introduce a local search to improve the quality of the solution\nobtained from the beam search. Furthermore, due to the lack of real-world\nbenchmarks, two real-world datasets are introduced to verify the robustness of\nthe method. The extensive experimental results show that the proposed method\noutperforms the previous approaches from the literature.\n", "date": "2024-07-17", "categories": []}
{"id": "2407.13182", "title": "SpaDiT: Diffusion Transformer for Spatial Gene Expression Prediction\n  using scRNA-seq", "abstract": "  The rapid development of spatial transcriptomics (ST) technologies is\nrevolutionizing our understanding of the spatial organization of biological\ntissues. Current ST methods, categorized into next-generation sequencing-based\n(seq-based) and fluorescence in situ hybridization-based (image-based) methods,\noffer innovative insights into the functional dynamics of biological tissues.\nHowever, these methods are limited by their cellular resolution and the\nquantity of genes they can detect. To address these limitations, we propose\nSpaDiT, a deep learning method that utilizes a diffusion generative model to\nintegrate scRNA-seq and ST data for the prediction of undetected genes. By\nemploying a Transformer-based diffusion model, SpaDiT not only accurately\npredicts unknown genes but also effectively generates the spatial structure of\nST genes. We have demonstrated the effectiveness of SpaDiT through extensive\nexperiments on both seq-based and image-based ST data. SpaDiT significantly\ncontributes to ST gene prediction methods with its innovative approach.\nCompared to eight leading baseline methods, SpaDiT achieved state-of-the-art\nperformance across multiple metrics, highlighting its substantial\nbioinformatics contribution.\n", "date": "2024-07-18", "categories": []}
{"id": "2407.12053", "title": "Improving AlphaFlow for Efficient Protein Ensembles Generation", "abstract": "  Investigating conformational landscapes of proteins is a crucial way to\nunderstand their biological functions and properties. AlphaFlow stands out as a\nsequence-conditioned generative model that introduces flexibility into\nstructure prediction models by fine-tuning AlphaFold under the flow-matching\nframework. Despite the advantages of efficient sampling afforded by\nflow-matching, AlphaFlow still requires multiple runs of AlphaFold to finally\ngenerate one single conformation. Due to the heavy consumption of AlphaFold,\nits applicability is limited in sampling larger set of protein ensembles or the\nlonger chains within a constrained timeframe. In this work, we propose a\nfeature-conditioned generative model called AlphaFlow-Lit to realize efficient\nprotein ensembles generation. In contrast to the full fine-tuning on the entire\nstructure, we focus solely on the light-weight structure module to reconstruct\nthe conformation. AlphaFlow-Lit performs on-par with AlphaFlow and surpasses\nits distilled version without pretraining, all while achieving a significant\nsampling acceleration of around 47 times. The advancement in efficiency\nshowcases the potential of AlphaFlow-Lit in enabling faster and more scalable\ngeneration of protein ensembles.\n", "date": "2024-07-08", "categories": []}
{"id": "2407.11057", "title": "SPIN: SE(3)-Invariant Physics Informed Network for Binding Affinity\n  Prediction", "abstract": "  Accurate prediction of protein-ligand binding affinity is crucial for rapid\nand efficient drug development. Recently, the importance of predicting binding\naffinity has led to increased attention on research that models the\nthree-dimensional structure of protein-ligand complexes using graph neural\nnetworks to predict binding affinity. However, traditional methods often fail\nto accurately model the complex's spatial information or rely solely on\ngeometric features, neglecting the principles of protein-ligand binding. This\ncan lead to overfitting, resulting in models that perform poorly on independent\ndatasets and ultimately reducing their usefulness in real drug development. To\naddress this issue, we propose SPIN, a model designed to achieve superior\ngeneralization by incorporating various inductive biases applicable to this\ntask, beyond merely training on empirical data from datasets. For prediction,\nwe defined two types of inductive biases: a geometric perspective that\nmaintains consistent binding affinity predictions regardless of the complexs\nrotations and translations, and a physicochemical perspective that necessitates\nminimal binding free energy along their reaction coordinate for effective\nprotein-ligand binding. These prior knowledge inputs enable the SPIN to\noutperform comparative models in benchmark sets such as CASF-2016 and CSAR HiQ.\nFurthermore, we demonstrated the practicality of our model through virtual\nscreening experiments and validated the reliability and potential of our\nproposed model based on experiments assessing its interpretability.\n", "date": "2024-07-10", "categories": []}
{"id": "2407.09811", "title": "CellAgent: An LLM-driven Multi-Agent Framework for Automated Single-cell\n  Data Analysis", "abstract": "  Single-cell RNA sequencing (scRNA-seq) data analysis is crucial for\nbiological research, as it enables the precise characterization of cellular\nheterogeneity. However, manual manipulation of various tools to achieve desired\noutcomes can be labor-intensive for researchers. To address this, we introduce\nCellAgent (http://cell.agent4science.cn/), an LLM-driven multi-agent framework,\nspecifically designed for the automatic processing and execution of scRNA-seq\ndata analysis tasks, providing high-quality results with no human intervention.\nFirstly, to adapt general LLMs to the biological field, CellAgent constructs\nLLM-driven biological expert roles - planner, executor, and evaluator - each\nwith specific responsibilities. Then, CellAgent introduces a hierarchical\ndecision-making mechanism to coordinate these biological experts, effectively\ndriving the planning and step-by-step execution of complex data analysis tasks.\nFurthermore, we propose a self-iterative optimization mechanism, enabling\nCellAgent to autonomously evaluate and optimize solutions, thereby guaranteeing\noutput quality. We evaluate CellAgent on a comprehensive benchmark dataset\nencompassing dozens of tissues and hundreds of distinct cell types. Evaluation\nresults consistently show that CellAgent effectively identifies the most\nsuitable tools and hyperparameters for single-cell analysis tasks, achieving\noptimal performance. This automated framework dramatically reduces the workload\nfor science data analyses, bringing us into the \"Agent for Science\" era.\n", "date": "2024-07-13", "categories": []}
{"id": "2407.10452", "title": "GraphPrint: Extracting Features from 3D Protein Structure for Drug\n  Target Affinity Prediction", "abstract": "  Accurate drug target affinity prediction can improve drug candidate\nselection, accelerate the drug discovery process, and reduce drug production\ncosts. Previous work focused on traditional fingerprints or used features\nextracted based on the amino acid sequence in the protein, ignoring its 3D\nstructure which affects its binding affinity. In this work, we propose\nGraphPrint: a framework for incorporating 3D protein structure features for\ndrug target affinity prediction. We generate graph representations for protein\n3D structures using amino acid residue location coordinates and combine them\nwith drug graph representation and traditional features to jointly learn drug\ntarget affinity. Our model achieves a mean square error of 0.1378 and a\nconcordance index of 0.8929 on the KIBA dataset and improves over using\ntraditional protein features alone. Our ablation study shows that the 3D\nprotein structure-based features provide information complementary to\ntraditional features.\n", "date": "2024-07-15", "categories": []}
{"id": "2407.09274", "title": "Unifying Sequences, Structures, and Descriptions for Any-to-Any Protein\n  Generation with the Large Multimodal Model HelixProtX", "abstract": "  Proteins are fundamental components of biological systems and can be\nrepresented through various modalities, including sequences, structures, and\ntextual descriptions. Despite the advances in deep learning and scientific\nlarge language models (LLMs) for protein research, current methodologies\npredominantly focus on limited specialized tasks -- often predicting one\nprotein modality from another. These approaches restrict the understanding and\ngeneration of multimodal protein data. In contrast, large multimodal models\nhave demonstrated potential capabilities in generating any-to-any content like\ntext, images, and videos, thus enriching user interactions across various\ndomains. Integrating these multimodal model technologies into protein research\noffers significant promise by potentially transforming how proteins are\nstudied. To this end, we introduce HelixProtX, a system built upon the large\nmultimodal model, aiming to offer a comprehensive solution to protein research\nby supporting any-to-any protein modality generation. Unlike existing methods,\nit allows for the transformation of any input protein modality into any desired\nprotein modality. The experimental results affirm the advanced capabilities of\nHelixProtX, not only in generating functional descriptions from amino acid\nsequences but also in executing critical tasks such as designing protein\nsequences and structures from textual descriptions. Preliminary findings\nindicate that HelixProtX consistently achieves superior accuracy across a range\nof protein-related tasks, outperforming existing state-of-the-art models. By\nintegrating multimodal large models into protein research, HelixProtX opens new\navenues for understanding protein biology, thereby promising to accelerate\nscientific discovery.\n", "date": "2024-07-12", "categories": []}
{"id": "2407.09355", "title": "FastImpute: A Baseline for Open-source, Reference-Free Genotype\n  Imputation Methods -- A Case Study in PRS313", "abstract": "  Genotype imputation enhances genetic data by predicting missing SNPs using\nreference haplotype information. Traditional methods leverage linkage\ndisequilibrium (LD) to infer untyped SNP genotypes, relying on the similarity\nof LD structures between genotyped target sets and fully sequenced reference\npanels. Recently, reference-free deep learning-based methods have emerged,\noffering a promising alternative by predicting missing genotypes without\nexternal databases, thereby enhancing privacy and accessibility. However, these\nmethods often produce models with tens of millions of parameters, leading to\nchallenges such as the need for substantial computational resources to train\nand inefficiency for client-sided deployment. Our study addresses these\nlimitations by introducing a baseline for a novel genotype imputation pipeline\nthat supports client-sided imputation models generalizable across any\ngenotyping chip and genomic region. This approach enhances patient privacy by\nperforming imputation directly on edge devices. As a case study, we focus on\nPRS313, a polygenic risk score comprising 313 SNPs used for breast cancer risk\nprediction. Utilizing consumer genetic panels such as 23andMe, our model\ndemocratizes access to personalized genetic insights by allowing 23andMe users\nto obtain their PRS313 score. We demonstrate that simple linear regression can\nsignificantly improve the accuracy of PRS313 scores when calculated using SNPs\nimputed from consumer gene panels, such as 23andMe. Our linear regression model\nachieved an R^2 of 0.86, compared to 0.33 without imputation and 0.28 with\nsimple imputation (substituting missing SNPs with the minor allele frequency).\nThese findings suggest that popular SNP analysis libraries could benefit from\nintegrating linear regression models for genotype imputation, providing a\nviable and light-weight alternative to reference based imputation.\n", "date": "2024-07-12", "categories": []}
{"id": "2407.08216", "title": "Multimodal contrastive learning for spatial gene expression prediction\n  using histology images", "abstract": "  In recent years, the advent of spatial transcriptomics (ST) technology has\nunlocked unprecedented opportunities for delving into the complexities of gene\nexpression patterns within intricate biological systems. Despite its\ntransformative potential, the prohibitive cost of ST technology remains a\nsignificant barrier to its widespread adoption in large-scale studies. An\nalternative, more cost-effective strategy involves employing artificial\nintelligence to predict gene expression levels using readily accessible\nwhole-slide images (WSIs) stained with Hematoxylin and Eosin (H\\&E). However,\nexisting methods have yet to fully capitalize on multimodal information\nprovided by H&E images and ST data with spatial location. In this paper, we\npropose \\textbf{mclSTExp}, a multimodal contrastive learning with Transformer\nand Densenet-121 encoder for Spatial Transcriptomics Expression prediction. We\nconceptualize each spot as a \"word\", integrating its intrinsic features with\nspatial context through the self-attention mechanism of a Transformer encoder.\nThis integration is further enriched by incorporating image features via\ncontrastive learning, thereby enhancing the predictive capability of our model.\nOur extensive evaluation of \\textbf{mclSTExp} on two breast cancer datasets and\na skin squamous cell carcinoma dataset demonstrates its superior performance in\npredicting spatial gene expression. Moreover, mclSTExp has shown promise in\ninterpreting cancer-specific overexpressed genes, elucidating immune-related\ngenes, and identifying specialized spatial domains annotated by pathologists.\nOur source code is available at https://github.com/shizhiceng/mclSTExp.\n", "date": "2024-07-11", "categories": []}
{"id": "2311.12917", "title": "Orchard: building large cancer phylogenies using stochastic\n  combinatorial search", "abstract": "  Phylogenies depicting the evolutionary history of genetically heterogeneous\nsubpopulations of cells from the same cancer, i.e., cancer phylogenies, offer\nvaluable insights about cancer development and guide treatment strategies. Many\nmethods exist that reconstruct cancer phylogenies using point mutations\ndetected with bulk DNA sequencing. However, these methods become inaccurate\nwhen reconstructing phylogenies with more than 30 mutations, or, in some cases,\nfail to recover a phylogeny altogether. Here, we introduce Orchard, a cancer\nphylogeny reconstruction algorithm that is fast and accurate using up to 1000\nmutations. Orchard samples without replacement from a factorized approximation\nof the posterior distribution over phylogenies, a novel result derived in this\npaper. Each factor in this approximate posterior corresponds to a conditional\ndistribution for adding a new mutation to a partially built phylogeny. Orchard\noptimizes each factor sequentially, generating a sequence of incrementally\nlarger phylogenies that ultimately culminate in a complete tree containing all\nmutations. Our evaluations demonstrate that Orchard outperforms\nstate-of-the-art cancer phylogeny reconstruction methods in reconstructing more\nplausible phylogenies across 90 simulated cancers and 14 B-progenitor acute\nlymphoblastic leukemias (B-ALLs). Remarkably, Orchard accurately reconstructs\ncancer phylogenies using up to 1,000 mutations. Additionally, we demonstrate\nthat the large and accurate phylogenies reconstructed by Orchard are useful for\nidentifying patterns of somatic mutations and genetic variations among distinct\ncancer cell subpopulations.\n", "date": "2023-11-21", "categories": []}
{"id": "2406.19300", "title": "scTree: Discovering Cellular Hierarchies in the Presence of Batch\n  Effects in scRNA-seq Data", "abstract": "  We propose a novel method, scTree, for single-cell Tree Variational\nAutoencoders, extending a hierarchical clustering approach to single-cell RNA\nsequencing data. scTree corrects for batch effects while simultaneously\nlearning a tree-structured data representation. This VAE-based method allows\nfor a more in-depth understanding of complex cellular landscapes independently\nof the biasing effects of batches. We show empirically on seven datasets that\nscTree discovers the underlying clusters of the data and the hierarchical\nrelations between them, as well as outperforms established baseline methods\nacross these datasets. Additionally, we analyze the learned hierarchy to\nunderstand its biological relevance, thus underpinning the importance of\nintegrating batch correction directly into the clustering procedure.\n", "date": "2024-06-27", "categories": []}
{"id": "2407.05736", "title": "TransMA: an explainable multi-modal deep learning model for predicting\n  properties of ionizable lipid nanoparticles in mRNA delivery", "abstract": "  As the primary mRNA delivery vehicles, ionizable lipid nanoparticles (LNPs)\nexhibit excellent safety, high transfection efficiency, and strong immune\nresponse induction. However, the screening process for LNPs is time-consuming\nand costly. To expedite the identification of high-transfection-efficiency mRNA\ndrug delivery systems, we propose an explainable LNPs transfection efficiency\nprediction model, called TransMA. TransMA employs a multi-modal molecular\nstructure fusion architecture, wherein the fine-grained atomic spatial\nrelationship extractor named molecule 3D Transformer captures three-dimensional\nspatial features of the molecule, and the coarse-grained atomic sequence\nextractor named molecule Mamba captures one-dimensional molecular features. We\ndesign the mol-attention mechanism block, enabling it to align coarse and\nfine-grained atomic features and captures relationships between atomic spatial\nand sequential structures. TransMA achieves state-of-the-art performance in\npredicting transfection efficiency using the scaffold and cliff data splitting\nmethods on the current largest LNPs dataset, including Hela and RAW cell lines.\nMoreover, we find that TransMA captures the relationship between subtle\nstructural changes and significant transfection efficiency variations,\nproviding valuable insights for LNPs design. Additionally, TransMA's\npredictions on external transfection efficiency data maintain a consistent\norder with actual transfection efficiencies, demonstrating its robust\ngeneralization capability. The code, model and data are made publicly available\nat https://github.com/wklix/TransMA/tree/master. We hope that high-accuracy\ntransfection prediction models in the future can aid in LNPs design and initial\nscreening, thereby assisting in accelerating the mRNA design process.\n", "date": "2024-07-08", "categories": []}
{"id": "2407.03380", "title": "Multi-Peptide: Multimodality Leveraged Language-Graph Learning of\n  Peptide Properties", "abstract": "  Peptides are essential in biological processes and therapeutics. In this\nstudy, we introduce Multi-Peptide, an innovative approach that combines\ntransformer-based language models with Graph Neural Networks (GNNs) to predict\npeptide properties. We combine PeptideBERT, a transformer model tailored for\npeptide property prediction, with a GNN encoder to capture both sequence-based\nand structural features. By employing Contrastive Language-Image Pre-training\n(CLIP), Multi-Peptide aligns embeddings from both modalities into a shared\nlatent space, thereby enhancing the model's predictive accuracy. Evaluations on\nhemolysis and nonfouling datasets demonstrate Multi-Peptide's robustness,\nachieving state-of-the-art 86.185% accuracy in hemolysis prediction. This study\nhighlights the potential of multimodal learning in bioinformatics, paving the\nway for accurate and reliable predictions in peptide-based research and\napplications.\n", "date": "2024-07-02", "categories": []}
{"id": "2407.03381", "title": "SeqMate: A Novel Large Language Model Pipeline for Automating RNA\n  Sequencing", "abstract": "  RNA sequencing techniques, like bulk RNA-seq and Single Cell (sc) RNA-seq,\nare critical tools for the biologist looking to analyze the genetic\nactivity/transcriptome of a tissue or cell during an experimental procedure.\nPlatforms like Illumina's next-generation sequencing (NGS) are used to produce\nthe raw data for this experimental procedure. This raw FASTQ data must then be\nprepared via a complex series of data manipulations by bioinformaticians. This\nprocess currently takes place on an unwieldy textual user interface like a\nterminal/command line that requires the user to install and import multiple\nprogram packages, preventing the untrained biologist from initiating data\nanalysis. Open-source platforms like Galaxy have produced a more user-friendly\npipeline, yet the visual interface remains cluttered and highly technical,\nremaining uninviting for the natural scientist. To address this, SeqMate is a\nuser-friendly tool that allows for one-click analytics by utilizing the power\nof a large language model (LLM) to automate both data preparation and analysis\n(differential expression, trajectory analysis, etc). Furthermore, by utilizing\nthe power of generative AI, SeqMate is also capable of analyzing such findings\nand producing written reports of upregulated/downregulated/user-prompted genes\nwith sources cited from known repositories like PubMed, PDB, and Uniprot.\n", "date": "2024-07-02", "categories": []}
{"id": "2407.03392", "title": "M5: A Whole Genome Bacterial Encoder at Single Nucleotide Resolution", "abstract": "  A linear attention mechanism is described to extend the context length of an\nencoder only transformer, called M5 in this report, to a multi-million single\nnucleotide resolution foundation model pretrained on bacterial whole genomes.\nThe linear attention mechanism used approximates a full quadratic attention\nmechanism tightly and has a simple and lightweight implementation for the use\ncase when the key-query embedding dimensionality is low. The M5-small model is\nentirely trained and tested on one A100 GPU with 40gb of memory up to 196K\nnucleotides during training and 2M nucleotides during testing. We test the\nperformance of the M5-small model and record notable improvements in\nperformance as whole genome bacterial sequence lengths are increased as well as\ndemonstrating the stability of the full multi-head attention approximation used\nas sequence length is increased.\n", "date": "2024-07-03", "categories": []}
{"id": "2407.03428", "title": "NEBULA: Neural Empirical Bayes Under Latent Representations for\n  Efficient and Controllable Design of Molecular Libraries", "abstract": "  We present NEBULA, the first latent 3D generative model for scalable\ngeneration of large molecular libraries around a seed compound of interest.\nSuch libraries are crucial for scientific discovery, but it remains challenging\nto generate large numbers of high quality samples efficiently. 3D-voxel-based\nmethods have recently shown great promise for generating high quality samples\nde novo from random noise (Pinheiro et al., 2023). However, sampling in\n3D-voxel space is computationally expensive and use in library generation is\nprohibitively slow. Here, we instead perform neural empirical Bayes sampling\n(Saremi & Hyvarinen, 2019) in the learned latent space of a vector-quantized\nvariational autoencoder. NEBULA generates large molecular libraries nearly an\norder of magnitude faster than existing methods without sacrificing sample\nquality. Moreover, NEBULA generalizes better to unseen drug-like molecules, as\ndemonstrated on two public datasets and multiple recently released drugs. We\nexpect the approach herein to be highly enabling for machine learning-based\ndrug discovery. The code is available at\nhttps://github.com/prescient-design/nebula\n", "date": "2024-07-03", "categories": []}
{"id": "2407.04486", "title": "Variational and Explanatory Neural Networks for Encoding Cancer Profiles\n  and Predicting Drug Responses", "abstract": "  Human cancers present a significant public health challenge and require the\ndiscovery of novel drugs through translational research. Transcriptomics\nprofiling data that describes molecular activities in tumors and cancer cell\nlines are widely utilized for predicting anti-cancer drug responses. However,\nexisting AI models face challenges due to noise in transcriptomics data and\nlack of biological interpretability. To overcome these limitations, we\nintroduce VETE (Variational and Explanatory Transcriptomics Encoder), a novel\nneural network framework that incorporates a variational component to mitigate\nnoise effects and integrates traceable gene ontology into the neural network\narchitecture for encoding cancer transcriptomics data. Key innovations include\na local interpretability-guided method for identifying ontology paths, a\nvisualization tool to elucidate biological mechanisms of drug responses, and\nthe application of centralized large scale hyperparameter optimization. VETE\ndemonstrated robust accuracy in cancer cell line classification and drug\nresponse prediction. Additionally, it provided traceable biological\nexplanations for both tasks and offers insights into the mechanisms underlying\nits predictions. VETE bridges the gap between AI-driven predictions and\nbiologically meaningful insights in cancer research, which represents a\npromising advancement in the field.\n", "date": "2024-07-05", "categories": []}
{"id": "2308.07954", "title": "APACE: AlphaFold2 and advanced computing as a service for accelerated\n  discovery in biophysics", "abstract": "  The prediction of protein 3D structure from amino acid sequence is a\ncomputational grand challenge in biophysics, and plays a key role in robust\nprotein structure prediction algorithms, from drug discovery to genome\ninterpretation. The advent of AI models, such as AlphaFold, is revolutionizing\napplications that depend on robust protein structure prediction algorithms. To\nmaximize the impact, and ease the usability, of these novel AI tools we\nintroduce APACE, AlphaFold2 and advanced computing as a service, a novel\ncomputational framework that effectively handles this AI model and its TB-size\ndatabase to conduct accelerated protein structure prediction analyses in modern\nsupercomputing environments. We deployed APACE in the Delta and Polaris\nsupercomputers, and quantified its performance for accurate protein structure\npredictions using four exemplar proteins: 6AWO, 6OAN, 7MEZ, and 6D6U. Using up\nto 300 ensembles, distributed across 200 NVIDIA A100 GPUs, we found that APACE\nis up to two orders of magnitude faster than off-the-self AlphaFold2\nimplementations, reducing time-to-solution from weeks to minutes. This\ncomputational approach may be readily linked with robotics laboratories to\nautomate and accelerate scientific discovery.\n", "date": "2023-08-15", "categories": []}
{"id": "2405.03961", "title": "Structure-based drug design by denoising voxel grids", "abstract": "  We present VoxBind, a new score-based generative model for 3D molecules\nconditioned on protein structures. Our approach represents molecules as 3D\natomic density grids and leverages a 3D voxel-denoising network for learning\nand generation. We extend the neural empirical Bayes formalism (Saremi &\nHyvarinen, 2019) to the conditional setting and generate structure-conditioned\nmolecules with a two-step procedure: (i) sample noisy molecules from the\nGaussian-smoothed conditional distribution with underdamped Langevin MCMC using\nthe learned score function and (ii) estimate clean molecules from the noisy\nsamples with single-step denoising. Compared to the current state of the art,\nour model is simpler to train, significantly faster to sample from, and\nachieves better results on extensive in silico benchmarks -- the generated\nmolecules are more diverse, exhibit fewer steric clashes, and bind with higher\naffinity to protein pockets. The code is available at\nhttps://github.com/genentech/voxbind/.\n", "date": "2024-05-07", "categories": []}
{"id": "2406.14307", "title": "QuST-LLM: Integrating Large Language Models for Comprehensive Spatial\n  Transcriptomics Analysis", "abstract": "  In this paper, we introduce QuST-LLM, an innovative extension of QuPath that\nutilizes the capabilities of large language models (LLMs) to analyze and\ninterpret spatial transcriptomics (ST) data. In addition to simplifying the\nintricate and high-dimensional nature of ST data by offering a comprehensive\nworkflow that includes data loading, region selection, gene expression\nanalysis, and functional annotation, QuST-LLM employs LLMs to transform complex\nST data into understandable and detailed biological narratives based on gene\nontology annotations, thereby significantly improving the interpretability of\nST data. Consequently, users can interact with their own ST data using natural\nlanguage. Hence, QuST-LLM provides researchers with a potent functionality to\nunravel the spatial and functional complexities of tissues, fostering novel\ninsights and advancements in biomedical research. QuST-LLM is a part of QuST\nproject. The source code is hosted on GitHub and documentation is available at\n(https://github.com/huangch/qust).\n", "date": "2024-06-20", "categories": []}
{"id": "2406.14969", "title": "Uni-Mol2: Exploring Molecular Pretraining Model at Scale", "abstract": "  In recent years, pretraining models have made significant advancements in the\nfields of natural language processing (NLP), computer vision (CV), and life\nsciences. The significant advancements in NLP and CV are predominantly driven\nby the expansion of model parameters and data size, a phenomenon now recognized\nas the scaling laws. However, research exploring scaling law in molecular\npretraining models remains unexplored. In this work, we present Uni-Mol2 , an\ninnovative molecular pretraining model that leverages a two-track transformer\nto effectively integrate features at the atomic level, graph level, and\ngeometry structure level. Along with this, we systematically investigate the\nscaling law within molecular pretraining models, characterizing the power-law\ncorrelations between validation loss and model size, dataset size, and\ncomputational resources. Consequently, we successfully scale Uni-Mol2 to 1.1\nbillion parameters through pretraining on 800 million conformations, making it\nthe largest molecular pretraining model to date. Extensive experiments show\nconsistent improvement in the downstream tasks as the model size grows. The\nUni-Mol2 with 1.1B parameters also outperforms existing methods, achieving an\naverage 27% improvement on the QM9 and 14% on COMPAS-1D dataset.\n", "date": "2024-06-21", "categories": []}
{"id": "2407.00111", "title": "Accurate Prediction of Ligand-Protein Interaction Affinities with\n  Fine-Tuned Small Language Models", "abstract": "  We describe the accurate prediction of ligand-protein interaction (LPI)\naffinities, also known as drug-target interactions (DTI), with instruction\nfine-tuned pretrained generative small language models (SLMs). We achieved\naccurate predictions for a range of affinity values associated with\nligand-protein interactions on out-of-sample data in a zero-shot setting. Only\nthe SMILES string of the ligand and the amino acid sequence of the protein were\nused as the model inputs. Our results demonstrate a clear improvement over\nmachine learning (ML) and free-energy perturbation (FEP+) based methods in\naccurately predicting a range of ligand-protein interaction affinities, which\ncan be leveraged to further accelerate drug discovery campaigns against\nchallenging therapeutic targets.\n", "date": "2024-06-27", "categories": []}
{"id": "2407.00175", "title": "Permutation invariant multi-output Gaussian Processes for drug\n  combination prediction in cancer", "abstract": "  Dose-response prediction in cancer is an active application field in machine\nlearning. Using large libraries of \\textit{in-vitro} drug sensitivity screens,\nthe goal is to develop accurate predictive models that can be used to guide\nexperimental design or inform treatment decisions. Building on previous work\nthat makes use of permutation invariant multi-output Gaussian Processes in the\ncontext of dose-response prediction for drug combinations, we develop a\nvariational approximation to these models. The variational approximation\nenables a more scalable model that provides uncertainty quantification and\nnaturally handles missing data. Furthermore, we propose using a deep generative\nmodel to encode the chemical space in a continuous manner, enabling prediction\nfor new drugs and new combinations. We demonstrate the performance of our model\nin a simple setting using a high-throughput dataset and show that the model is\nable to efficiently borrow information across outputs.\n", "date": "2024-06-28", "categories": []}
{"id": "2307.07085", "title": "Machine-learned molecular mechanics force field for the simulation of\n  protein-ligand systems and beyond", "abstract": "  The development of reliable and extensible molecular mechanics (MM) force\nfields -- fast, empirical models characterizing the potential energy surface of\nmolecular systems -- is indispensable for biomolecular simulation and\ncomputer-aided drug design. Here, we introduce a generalized and extensible\nmachine-learned MM force field, \\texttt{espaloma-0.3}, and an end-to-end\ndifferentiable framework using graph neural networks to overcome the\nlimitations of traditional rule-based methods. Trained in a single GPU-day to\nfit a large and diverse quantum chemical dataset of over 1.1M energy and force\ncalculations, \\texttt{espaloma-0.3} reproduces quantum chemical energetic\nproperties of chemical domains highly relevant to drug discovery, including\nsmall molecules, peptides, and nucleic acids. Moreover, this force field\nmaintains the quantum chemical energy-minimized geometries of small molecules\nand preserves the condensed phase properties of peptides, self-consistently\nparametrizing proteins and ligands to produce stable simulations leading to\nhighly accurate predictions of binding free energies. This methodology\ndemonstrates significant promise as a path forward for systematically building\nmore accurate force fields that are easily extensible to new chemical domains\nof interest.\n", "date": "2023-07-13", "categories": []}
{"id": "2406.19156", "title": "Heterogeneous Causal Metapath Graph Neural Network for\n  Gene-Microbe-Disease Association Prediction", "abstract": "  The recent focus on microbes in human medicine highlights their potential\nrole in the genetic framework of diseases. To decode the complex interactions\namong genes, microbes, and diseases, computational predictions of\ngene-microbe-disease (GMD) associations are crucial. Existing methods primarily\naddress gene-disease and microbe-disease associations, but the more intricate\ntriple-wise GMD associations remain less explored. In this paper, we propose a\nHeterogeneous Causal Metapath Graph Neural Network (HCMGNN) to predict GMD\nassociations. HCMGNN constructs a heterogeneous graph linking genes, microbes,\nand diseases through their pairwise associations, and utilizes six predefined\ncausal metapaths to extract directed causal subgraphs, which facilitate the\nmulti-view analysis of causal relations among three entity types. Within each\nsubgraph, we employ a causal semantic sharing message passing network for node\nrepresentation learning, coupled with an attentive fusion method to integrate\nthese representations for predicting GMD associations. Our extensive\nexperiments show that HCMGNN effectively predicts GMD associations and\naddresses association sparsity issue by enhancing the graph's semantics and\nstructure.\n", "date": "2024-06-27", "categories": []}
{"id": "2402.08703", "title": "A Survey of Generative AI for de novo Drug Design: New Frontiers in\n  Molecule and Protein Generation", "abstract": "  Artificial intelligence (AI)-driven methods can vastly improve the\nhistorically costly drug design process, with various generative models already\nin widespread use. Generative models for de novo drug design, in particular,\nfocus on the creation of novel biological compounds entirely from scratch,\nrepresenting a promising future direction. Rapid development in the field,\ncombined with the inherent complexity of the drug design process, creates a\ndifficult landscape for new researchers to enter. In this survey, we organize\nde novo drug design into two overarching themes: small molecule and protein\ngeneration. Within each theme, we identify a variety of subtasks and\napplications, highlighting important datasets, benchmarks, and model\narchitectures and comparing the performance of top models. We take a broad\napproach to AI-driven drug design, allowing for both micro-level comparisons of\nvarious methods within each subtask and macro-level observations across\ndifferent fields. We discuss parallel challenges and approaches between the two\napplications and highlight future directions for AI-driven de novo drug design\nas a whole. An organized repository of all covered sources is available at\nhttps://github.com/gersteinlab/GenAI4Drug.\n", "date": "2024-02-13", "categories": []}
{"id": "2402.11363", "title": "Transformer-based de novo peptide sequencing for data-independent\n  acquisition mass spectrometry", "abstract": "  Tandem mass spectrometry (MS/MS) stands as the predominant high-throughput\ntechnique for comprehensively analyzing protein content within biological\nsamples. This methodology is a cornerstone driving the advancement of\nproteomics. In recent years, substantial strides have been made in\nData-Independent Acquisition (DIA) strategies, facilitating impartial and\nnon-targeted fragmentation of precursor ions. The DIA-generated MS/MS spectra\npresent a formidable obstacle due to their inherent high multiplexing nature.\nEach spectrum encapsulates fragmented product ions originating from multiple\nprecursor peptides. This intricacy poses a particularly acute challenge in de\nnovo peptide/protein sequencing, where current methods are ill-equipped to\naddress the multiplexing conundrum. In this paper, we introduce DiaTrans, a\ndeep-learning model based on transformer architecture. It deciphers peptide\nsequences from DIA mass spectrometry data. Our results show significant\nimprovements over existing STOA methods, including DeepNovo-DIA and PepNet.\nCasanovo-DIA enhances precision by 15.14% to 34.8%, recall by 11.62% to 31.94%\nat the amino acid level, and boosts precision by 59% to 81.36% at the peptide\nlevel. Integrating DIA data and our DiaTrans model holds considerable promise\nto uncover novel peptides and more comprehensive profiling of biological\nsamples. Casanovo-DIA is freely available under the GNU GPL license at\nhttps://github.com/Biocomputing-Research-Group/DiaTrans.\n", "date": "2024-02-17", "categories": []}
{"id": "2404.08722", "title": "VADA: a Data-Driven Simulator for Nanopore Sequencing", "abstract": "  Nanopore sequencing offers the ability for real-time analysis of long DNA\nsequences at a low cost, enabling new applications such as early detection of\ncancer. Due to the complex nature of nanopore measurements and the high cost of\nobtaining ground truth datasets, there is a need for nanopore simulators.\nExisting simulators rely on handcrafted rules and parameters and do not learn\nan internal representation that would allow for analysing underlying biological\nfactors of interest. Instead, we propose VADA, a purely data-driven method for\nsimulating nanopores based on an autoregressive latent variable model. We embed\nsubsequences of DNA and introduce a conditional prior to address the challenge\nof a collapsing conditioning. We introduce an auxiliary regressor on the latent\nvariable to encourage our model to learn an informative latent representation.\nWe empirically demonstrate that our model achieves competitive simulation\nperformance on experimental nanopore data. Moreover, we show we have learned an\ninformative latent representation that is predictive of the DNA labels. We\nhypothesize that other biological factors of interest, beyond the DNA labels,\ncan potentially be extracted from such a learned latent representation.\n", "date": "2024-04-12", "categories": []}
{"id": "2311.04846", "title": "Incorporating temporal dynamics of mutations to enhance the prediction\n  capability of antiretroviral therapy's outcome for HIV-1", "abstract": "  Motivation: In predicting HIV therapy outcomes, a critical clinical question\nis whether using historical information can enhance predictive capabilities\ncompared with current or latest available data analysis. This study analyses\nwhether historical knowledge, which includes viral mutations detected in all\ngenotypic tests before therapy, their temporal occurrence, and concomitant\nviral load measurements, can bring improvements. We introduce a method to weigh\nmutations, considering the previously enumerated factors and the reference\nmutation-drug Stanford resistance tables. We compare a model encompassing\nhistory (H) with one not using it (NH). Results: The H-model demonstrates\nsuperior discriminative ability, with a higher ROC-AUC score (76.34%) than the\nNH-model (74.98%). Significant Wilcoxon test results confirm that incorporating\nhistorical information improves consistently predictive accuracy for treatment\noutcomes. The better performance of the H-model might be attributed to its\nconsideration of latent HIV reservoirs, probably obtained when leveraging\nhistorical information. The findings emphasize the importance of temporal\ndynamics in mutations, offering insights into HIV infection complexities.\nHowever, our result also shows that prediction accuracy remains relatively high\neven when no historical information is available. Supplementary information:\nSupplementary material is available.\n", "date": "2023-11-08", "categories": []}
{"id": "2405.09585", "title": "An Embarrassingly Simple Approach to Enhance Transformer Performance in\n  Genomic Selection for Crop Breeding", "abstract": "  Genomic selection (GS), as a critical crop breeding strategy, plays a key\nrole in enhancing food production and addressing the global hunger crisis. The\npredominant approaches in GS currently revolve around employing statistical\nmethods for prediction. However, statistical methods often come with two main\nlimitations: strong statistical priors and linear assumptions. A recent trend\nis to capture the non-linear relationships between markers by deep learning.\nHowever, as crop datasets are commonly long sequences with limited samples, the\nrobustness of deep learning models, especially Transformers, remains a\nchallenge. In this work, to unleash the unexplored potential of attention\nmechanism for the task of interest, we propose a simple yet effective\nTransformer-based framework that enables end-to-end training of the whole\nsequence. Via experiments on rice3k and wheat3k datasets, we show that, with\nsimple tricks such as k-mer tokenization and random masking, Transformer can\nachieve overall superior performance against seminal methods on GS tasks of\ninterest.\n", "date": "2024-05-15", "categories": []}
{"id": "2212.14041", "title": "Deciphering RNA Secondary Structure Prediction: A Probabilistic K-Rook\n  Matching Perspective", "abstract": "  The secondary structure of ribonucleic acid (RNA) is more stable and\naccessible in the cell than its tertiary structure, making it essential for\nfunctional prediction. Although deep learning has shown promising results in\nthis field, current methods suffer from poor generalization and high\ncomplexity. In this work, we reformulate the RNA secondary structure prediction\nas a K-Rook problem, thereby simplifying the prediction process into\nprobabilistic matching within a finite solution space. Building on this\ninnovative perspective, we introduce RFold, a simple yet effective method that\nlearns to predict the most matching K-Rook solution from the given sequence.\nRFold employs a bi-dimensional optimization strategy that decomposes the\nprobabilistic matching problem into row-wise and column-wise components to\nreduce the matching complexity, simplifying the solving process while\nguaranteeing the validity of the output. Extensive experiments demonstrate that\nRFold achieves competitive performance and about eight times faster inference\nefficiency than the state-of-the-art approaches. The code and Colab demo are\navailable in (http://github.com/A4Bio/RFold).\n", "date": "2022-12-02", "categories": []}
{"id": "2406.12108", "title": "Computing in the Life Sciences: From Early Algorithms to Modern AI", "abstract": "  Computing in the life sciences has undergone a transformative evolution, from\nearly computational models in the 1950s to the applications of artificial\nintelligence (AI) and machine learning (ML) seen today. This paper highlights\nkey milestones and technological advancements through the historical\ndevelopment of computing in the life sciences. The discussion includes the\ninception of computational models for biological processes, the advent of\nbioinformatics tools, and the integration of AI/ML in modern life sciences\nresearch. Attention is given to AI-enabled tools used in the life sciences,\nsuch as scientific large language models and bio-AI tools, examining their\ncapabilities, limitations, and impact to biological risk. This paper seeks to\nclarify and establish essential terminology and concepts to ensure informed\ndecision-making and effective communication across disciplines.\n", "date": "2024-06-17", "categories": []}
{"id": "2406.13133", "title": "PathoLM: Identifying pathogenicity from the DNA sequence through the\n  Genome Foundation Model", "abstract": "  Pathogen identification is pivotal in diagnosing, treating, and preventing\ndiseases, crucial for controlling infections and safeguarding public health.\nTraditional alignment-based methods, though widely used, are computationally\nintense and reliant on extensive reference databases, often failing to detect\nnovel pathogens due to their low sensitivity and specificity. Similarly,\nconventional machine learning techniques, while promising, require large\nannotated datasets and extensive feature engineering and are prone to\noverfitting. Addressing these challenges, we introduce PathoLM, a cutting-edge\npathogen language model optimized for the identification of pathogenicity in\nbacterial and viral sequences. Leveraging the strengths of pre-trained DNA\nmodels such as the Nucleotide Transformer, PathoLM requires minimal data for\nfine-tuning, thereby enhancing pathogen detection capabilities. It effectively\ncaptures a broader genomic context, significantly improving the identification\nof novel and divergent pathogens. We developed a comprehensive data set\ncomprising approximately 30 species of viruses and bacteria, including ESKAPEE\npathogens, seven notably virulent bacterial strains resistant to antibiotics.\nAdditionally, we curated a species classification dataset centered specifically\non the ESKAPEE group. In comparative assessments, PathoLM dramatically\noutperforms existing models like DciPatho, demonstrating robust zero-shot and\nfew-shot capabilities. Furthermore, we expanded PathoLM-Sp for ESKAPEE species\nclassification, where it showed superior performance compared to other advanced\ndeep learning methods, despite the complexities of the task.\n", "date": "2024-06-19", "categories": []}
{"id": "2406.13162", "title": "AntibodyFlow: Normalizing Flow Model for Designing Antibody\n  Complementarity-Determining Regions", "abstract": "  Therapeutic antibodies have been extensively studied in drug discovery and\ndevelopment in the past decades. Antibodies are specialized protective proteins\nthat bind to antigens in a lock-to-key manner. The binding strength/affinity\nbetween an antibody and a specific antigen is heavily determined by the\ncomplementarity-determining regions (CDRs) on the antibodies. Existing machine\nlearning methods cast in silico development of CDRs as either sequence or 3D\ngraph (with a single chain) generation tasks and have achieved initial success.\nHowever, with CDR loops having specific geometry shapes, learning the 3D\ngeometric structures of CDRs remains a challenge. To address this issue, we\npropose AntibodyFlow, a 3D flow model to design antibody CDR loops.\nSpecifically, AntibodyFlow first constructs the distance matrix, then predicts\namino acids conditioned on the distance matrix. Also, AntibodyFlow conducts\nconstraint learning and constrained generation to ensure valid 3D structures.\nExperimental results indicate that AntibodyFlow outperforms the best baseline\nconsistently with up to 16.0% relative improvement in validity rate and 24.3%\nrelative reduction in geometric graph level error (root mean square deviation,\nRMSD).\n", "date": "2024-06-19", "categories": []}
{"id": "2406.14150", "title": "Multi-modal Transfer Learning between Biological Foundation Models", "abstract": "  Biological sequences encode fundamental instructions for the building blocks\nof life, in the form of DNA, RNA, and proteins. Modeling these sequences is key\nto understand disease mechanisms and is an active research area in\ncomputational biology. Recently, Large Language Models have shown great promise\nin solving certain biological tasks but current approaches are limited to a\nsingle sequence modality (DNA, RNA, or protein). Key problems in genomics\nintrinsically involve multiple modalities, but it remains unclear how to adapt\ngeneral-purpose sequence models to those cases. In this work we propose a\nmulti-modal model that connects DNA, RNA, and proteins by leveraging\ninformation from different pre-trained modality-specific encoders. We\ndemonstrate its capabilities by applying it to the largely unsolved problem of\npredicting how multiple RNA transcript isoforms originate from the same gene\n(i.e. same DNA sequence) and map to different transcription expression levels\nacross various human tissues. We show that our model, dubbed IsoFormer, is able\nto accurately predict differential transcript expression, outperforming\nexisting methods and leveraging the use of multiple modalities. Our framework\nalso achieves efficient transfer knowledge from the encoders pre-training as\nwell as in between modalities. We open-source our model, paving the way for new\nmulti-modal gene expression approaches.\n", "date": "2024-06-20", "categories": []}
{"id": "2405.06693", "title": "SurfPro: Functional Protein Design Based on Continuous Surface", "abstract": "  How can we design proteins with desired functions? We are motivated by a\nchemical intuition that both geometric structure and biochemical properties are\ncritical to a protein's function. In this paper, we propose SurfPro, a new\nmethod to generate functional proteins given a desired surface and its\nassociated biochemical properties. SurfPro comprises a hierarchical encoder\nthat progressively models the geometric shape and biochemical features of a\nprotein surface, and an autoregressive decoder to produce an amino acid\nsequence. We evaluate SurfPro on a standard inverse folding benchmark CATH 4.2\nand two functional protein design tasks: protein binder design and enzyme\ndesign. Our SurfPro consistently surpasses previous state-of-the-art inverse\nfolding methods, achieving a recovery rate of 57.78% on CATH 4.2 and higher\nsuccess rates in terms of protein-protein binding and enzyme-substrate\ninteraction scores.\n", "date": "2024-05-07", "categories": []}
{"id": "2406.11900", "title": "Horizon-wise Learning Paradigm Promotes Gene Splicing Identification", "abstract": "  Identifying gene splicing is a core and significant task confronted in modern\ncollaboration between artificial intelligence and bioinformatics. Past decades\nhave witnessed great efforts on this concern, such as the bio-plausible\nsplicing pattern AT-CG and the famous SpliceAI. In this paper, we propose a\nnovel framework for the task of gene splicing identification, named\nHorizon-wise Gene Splicing Identification (H-GSI). The proposed H-GSI follows\nthe horizon-wise identification paradigm and comprises four components: the\npre-processing procedure transforming string data into tensors, the sliding\nwindow technique handling long sequences, the SeqLab model, and the predictor.\nIn contrast to existing studies that process gene information with a truncated\nfixed-length sequence, H-GSI employs a horizon-wise identification paradigm in\nwhich all positions in a sequence are predicted with only one forward\ncomputation, improving accuracy and efficiency. The experiments conducted on\nthe real-world Human dataset show that our proposed H-GSI outperforms SpliceAI\nand achieves the best accuracy of 97.20\\%. The source code is available from\nthis link.\n", "date": "2024-06-15", "categories": []}
{"id": "2305.01475", "title": "Cancer-inspired Genomics Mapper Model for the Generation of Synthetic\n  DNA Sequences with Desired Genomics Signatures", "abstract": "  Genome data are crucial in modern medicine, offering significant potential\nfor diagnosis and treatment. Thanks to technological advancements, many\nmillions of healthy and diseased genomes have already been sequenced; however,\nobtaining the most suitable data for a specific study, and specifically for\nvalidation studies, remains challenging with respect to scale and access.\nTherefore, in silico genomics sequence generators have been proposed as a\npossible solution. However, the current generators produce inferior data using\nmostly shallow (stochastic) connections, detected with limited computational\ncomplexity in the training data. This means they do not take the appropriate\nbiological relations and constraints, that originally caused the observed\nconnections, into consideration. To address this issue, we propose\ncancer-inspired genomics mapper model (CGMM), that combines genetic algorithm\n(GA) and deep learning (DL) methods to tackle this challenge. CGMM mimics\nprocesses that generate genetic variations and mutations to transform readily\navailable control genomes into genomes with the desired phenotypes. We\ndemonstrate that CGMM can generate synthetic genomes of selected phenotypes\nsuch as ancestry and cancer that are indistinguishable from real genomes of\nsuch phenotypes, based on unsupervised clustering. Our results show that CGMM\noutperforms four current state-of-the-art genomics generators on two different\ntasks, suggesting that CGMM will be suitable for a wide range of purposes in\ngenomic medicine, especially for much-needed validation studies.\n", "date": "2023-05-01", "categories": []}
{"id": "2406.10867", "title": "Geometric-informed GFlowNets for Structure-Based Drug Design", "abstract": "  The rise of cost involved with drug discovery and current speed of which they\nare discover, underscore the need for more efficient structure-based drug\ndesign (SBDD) methods. We employ Generative Flow Networks (GFlowNets), to\neffectively explore the vast combinatorial space of drug-like molecules, which\ntraditional virtual screening methods fail to cover. We introduce a novel\nmodification to the GFlowNet framework by incorporating trigonometrically\nconsistent embeddings, previously utilized in tasks involving protein\nconformation and protein-ligand interactions, to enhance the model's ability to\ngenerate molecules tailored to specific protein pockets. We have modified the\nexisting protein conditioning used by GFlowNets, blending geometric information\nfrom both protein and ligand embeddings to achieve more geometrically\nconsistent embeddings. Experiments conducted using CrossDocked2020 demonstrated\nan improvement in the binding affinity between generated molecules and protein\npockets for both single and multi-objective tasks, compared to previous work.\nAdditionally, we propose future work aimed at further increasing the geometric\ninformation captured in protein-ligand interactions.\n", "date": "2024-06-16", "categories": []}
{"id": "2403.12995", "title": "ESM All-Atom: Multi-scale Protein Language Model for Unified Molecular\n  Modeling", "abstract": "  Protein language models have demonstrated significant potential in the field\nof protein engineering. However, current protein language models primarily\noperate at the residue scale, which limits their ability to provide information\nat the atom level. This limitation prevents us from fully exploiting the\ncapabilities of protein language models for applications involving both\nproteins and small molecules. In this paper, we propose ESM-AA (ESM All-Atom),\na novel approach that enables atom-scale and residue-scale unified molecular\nmodeling. ESM-AA achieves this by pre-training on multi-scale code-switch\nprotein sequences and utilizing a multi-scale position encoding to capture\nrelationships among residues and atoms. Experimental results indicate that\nESM-AA surpasses previous methods in protein-molecule tasks, demonstrating the\nfull utilization of protein language models. Further investigations reveal that\nthrough unified molecular modeling, ESM-AA not only gains molecular knowledge\nbut also retains its understanding of proteins. The source codes of ESM-AA are\npublicly released at https://github.com/zhengkangjie/ESM-AA.\n", "date": "2024-03-05", "categories": []}
{"id": "2406.08686", "title": "Opportunities in deep learning methods development for computational\n  biology", "abstract": "  Advances in molecular technologies underlie an enormous growth in the size of\ndata sets pertaining to biology and biomedicine. These advances parallel those\nin the deep learning subfield of machine learning. Components in the\ndifferentiable programming toolbox that makes deep learning possible are\nallowing computer scientists to address an increasingly large array of problems\nwith flexible and effective tools. However many of these tools have not fully\nproliferated into the computational biology and bioinformatics fields. In this\nperspective we survey some of these advances and highlight exemplary examples\nof their utilization in the biosciences, with the goal of increasing awareness\namong practitioners of emerging opportunities to blend expert knowledge with\nnewly emerging deep learning architectural tools.\n", "date": "2024-06-12", "categories": []}
{"id": "2403.15274", "title": "Bioinformatics and Biomedical Informatics with ChatGPT: Year One Review", "abstract": "  The year 2023 marked a significant surge in the exploration of applying large\nlanguage model (LLM) chatbots, notably ChatGPT, across various disciplines. We\nsurveyed the applications of ChatGPT in bioinformatics and biomedical\ninformatics throughout the year, covering omics, genetics, biomedical text\nmining, drug discovery, biomedical image understanding, bioinformatics\nprogramming, and bioinformatics education. Our survey delineates the current\nstrengths and limitations of this chatbot in bioinformatics and offers insights\ninto potential avenues for future developments.\n", "date": "2024-03-22", "categories": []}
{"id": "2406.07567", "title": "A GRASP-based memetic algorithm with path relinking for the far from\n  most string problem", "abstract": "  The FAR FROM MOST STRING PROBLEM (FFMSP) is a string selection problem. The\nobjective is to find a string whose distance to other strings in a certain\ninput set is above a given threshold for as many of those strings as possible.\nThis problem has links with some tasks in computational biology and its\nresolution has been shown to be very hard. We propose a memetic algorithm (MA)\nto tackle the FFMSP. This MA exploits a heuristic objective function for the\nproblem and features initialization of the population via a Greedy Randomized\nAdaptive Search Procedure (GRASP) metaheuristic, intensive recombination via\npath relinking and local improvement via hill climbing. An extensive empirical\nevaluation using problem instances of both random and biological origin is done\nto assess parameter sensitivity and draw performance comparisons with other\nstate-of-the-art techniques. The MA is shown to perform better than these\nlatter techniques with statistical significance.\n", "date": "2024-05-27", "categories": []}
{"id": "2406.07763", "title": "Gene-Level Representation Learning via Interventional Style Transfer in\n  Optical Pooled Screening", "abstract": "  Optical pooled screening (OPS) combines automated microscopy and genetic\nperturbations to systematically study gene function in a scalable and\ncost-effective way. Leveraging the resulting data requires extracting\nbiologically informative representations of cellular perturbation phenotypes\nfrom images. We employ a style-transfer approach to learn gene-level feature\nrepresentations from images of genetically perturbed cells obtained via OPS.\nOur method outperforms widely used engineered features in clustering gene\nrepresentations according to gene function, demonstrating its utility for\nuncovering latent biological relationships. This approach offers a promising\nalternative to investigate the role of genes in health and disease.\n", "date": "2024-06-11", "categories": []}
{"id": "2406.07770", "title": "DualBind: A Dual-Loss Framework for Protein-Ligand Binding Affinity\n  Prediction", "abstract": "  Accurate prediction of protein-ligand binding affinities is crucial for drug\ndevelopment. Recent advances in machine learning show promising results on this\ntask. However, these methods typically rely heavily on labeled data, which can\nbe scarce or unreliable, or they rely on assumptions like Boltzmann-distributed\ndata that may not hold true in practice. Here, we present DualBind, a novel\nframework that integrates supervised mean squared error (MSE) with unsupervised\ndenoising score matching (DSM) to accurately learn the binding energy function.\nDualBind not only addresses the limitations of DSM-only models by providing\nmore accurate absolute affinity predictions but also improves generalizability\nand reduces reliance on labeled data compared to MSE-only models. Our\nexperimental results demonstrate that DualBind excels in predicting binding\naffinities and can effectively utilize both labeled and unlabeled data to\nenhance performance.\n", "date": "2024-06-11", "categories": []}
{"id": "2403.08797", "title": "Evolutionary Algorithms Simulating Molecular Evolution: A New Field\n  Proposal", "abstract": "  The genetic blueprint for the essential functions of life is encoded in DNA,\nwhich is translated into proteins -- the engines driving most of our metabolic\nprocesses. Recent advancements in genome sequencing have unveiled a vast\ndiversity of protein families, but compared to the massive search space of all\npossible amino acid sequences, the set of known functional families is minimal.\nOne could say nature has a limited protein \"vocabulary.\" The major question for\ncomputational biologists, therefore, is whether this vocabulary can be expanded\nto include useful proteins that went extinct long ago, or maybe never evolved\nin the first place. We outline a computational approach to solving this\nproblem. By merging evolutionary algorithms, machine learning (ML), and\nbioinformatics, we can facilitate the development of completely novel proteins\nwhich have never existed before. We envision this work forming a new sub-field\nof computational evolution we dub evolutionary algorithms simulating molecular\nevolution (EASME).\n", "date": "2024-02-01", "categories": []}
{"id": "2405.06708", "title": "LangCell: Language-Cell Pre-training for Cell Identity Understanding", "abstract": "  Cell identity encompasses various semantic aspects of a cell, including cell\ntype, pathway information, disease information, and more, which are essential\nfor biologists to gain insights into its biological characteristics.\nUnderstanding cell identity from the transcriptomic data, such as annotating\ncell types, has become an important task in bioinformatics. As these semantic\naspects are determined by human experts, it is impossible for AI models to\neffectively carry out cell identity understanding tasks without the supervision\nsignals provided by single-cell and label pairs. The single-cell pre-trained\nlanguage models (PLMs) currently used for this task are trained only on a\nsingle modality, transcriptomics data, lack an understanding of cell identity\nknowledge. As a result, they have to be fine-tuned for downstream tasks and\nstruggle when lacking labeled data with the desired semantic labels. To address\nthis issue, we propose an innovative solution by constructing a unified\nrepresentation of single-cell data and natural language during the pre-training\nphase, allowing the model to directly incorporate insights related to cell\nidentity. More specifically, we introduce $\\textbf{LangCell}$, the first\n$\\textbf{Lang}$uage-$\\textbf{Cell}$ pre-training framework. LangCell utilizes\ntexts enriched with cell identity information to gain a profound comprehension\nof cross-modal knowledge. Results from experiments conducted on different\nbenchmarks show that LangCell is the only single-cell PLM that can work\neffectively in zero-shot cell identity understanding scenarios, and also\nsignificantly outperforms existing models in few-shot and fine-tuning cell\nidentity understanding scenarios.\n", "date": "2024-05-09", "categories": []}
{"id": "2406.06553", "title": "Ensemble Model With Bert,Roberta and Xlnet For Molecular property\n  prediction", "abstract": "  This paper presents a novel approach for predicting molecular properties with\nhigh accuracy without the need for extensive pre-training. Employing ensemble\nlearning and supervised fine-tuning of BERT, RoBERTa, and XLNet, our method\ndemonstrates significant effectiveness compared to existing advanced models.\nCrucially, it addresses the issue of limited computational resources faced by\nexperimental groups, enabling them to accurately predict molecular properties.\nThis innovation provides a cost-effective and resource-efficient solution,\npotentially advancing further research in the molecular domain.\n", "date": "2024-05-30", "categories": []}
{"id": "2406.07263", "title": "Active learning for affinity prediction of antibodies", "abstract": "  The primary objective of most lead optimization campaigns is to enhance the\nbinding affinity of ligands. For large molecules such as antibodies,\nidentifying mutations that enhance antibody affinity is particularly\nchallenging due to the combinatorial explosion of potential mutations. When the\nstructure of the antibody-antigen complex is available, relative binding free\nenergy (RBFE) methods can offer valuable insights into how different mutations\nwill impact the potency and selectivity of a drug candidate, thereby reducing\nthe reliance on costly and time-consuming wet-lab experiments. However,\naccurately simulating the physics of large molecules is computationally\nintensive. We present an active learning framework that iteratively proposes\npromising sequences for simulators to evaluate, thereby accelerating the search\nfor improved binders. We explore different modeling approaches to identify the\nmost effective surrogate model for this task, and evaluate our framework both\nusing pre-computed pools of data and in a realistic full-loop setting.\n", "date": "2024-06-11", "categories": []}
{"id": "2406.07418", "title": "Enhanced Gene Selection in Single-Cell Genomics: Pre-Filtering Synergy\n  and Reinforced Optimization", "abstract": "  Recent advancements in single-cell genomics necessitate precision in gene\npanel selection to interpret complex biological data effectively. Those methods\naim to streamline the analysis of scRNA-seq data by focusing on the most\ninformative genes that contribute significantly to the specific analysis task.\nTraditional selection methods, which often rely on expert domain knowledge,\nembedded machine learning models, or heuristic-based iterative optimization,\nare prone to biases and inefficiencies that may obscure critical genomic\nsignals. Recognizing the limitations of traditional methods, we aim to\ntranscend these constraints with a refined strategy. In this study, we\nintroduce an iterative gene panel selection strategy that is applicable to\nclustering tasks in single-cell genomics. Our method uniquely integrates\nresults from other gene selection algorithms, providing valuable preliminary\nboundaries or prior knowledge as initial guides in the search space to enhance\nthe efficiency of our framework. Furthermore, we incorporate the stochastic\nnature of the exploration process in reinforcement learning (RL) and its\ncapability for continuous optimization through reward-based feedback. This\ncombination mitigates the biases inherent in the initial boundaries and\nharnesses RL's adaptability to refine and target gene panel selection\ndynamically. To illustrate the effectiveness of our method, we conducted\ndetailed comparative experiments, case studies, and visualization analysis.\n", "date": "2024-06-11", "categories": []}
{"id": "2401.14442", "title": "Improving Antibody Humanness Prediction using Patent Data", "abstract": "  We investigate the potential of patent data for improving the antibody\nhumanness prediction using a multi-stage, multi-loss training process.\nHumanness serves as a proxy for the immunogenic response to antibody\ntherapeutics, one of the major causes of attrition in drug discovery and a\nchallenging obstacle for their use in clinical settings. We pose the initial\nlearning stage as a weakly-supervised contrastive-learning problem, where each\nantibody sequence is associated with possibly multiple identifiers of function\nand the objective is to learn an encoder that groups them according to their\npatented properties. We then freeze a part of the contrastive encoder and\ncontinue training it on the patent data using the cross-entropy loss to predict\nthe humanness score of a given antibody sequence. We illustrate the utility of\nthe patent data and our approach by performing inference on three different\nimmunogenicity datasets, unseen during training. Our empirical results\ndemonstrate that the learned model consistently outperforms the alternative\nbaselines and establishes new state-of-the-art on five out of six inference\ntasks, irrespective of the used metric.\n", "date": "2024-01-25", "categories": []}
{"id": "2402.03781", "title": "MolTC: Towards Molecular Relational Modeling In Language Models", "abstract": "  Molecular Relational Learning (MRL), aiming to understand interactions\nbetween molecular pairs, plays a pivotal role in advancing biochemical\nresearch. Recently, the adoption of large language models (LLMs), known for\ntheir vast knowledge repositories and advanced logical inference capabilities,\nhas emerged as a promising way for efficient and effective MRL. Despite their\npotential, these methods predominantly rely on the textual data, thus not fully\nharnessing the wealth of structural information inherent in molecular graphs.\nMoreover, the absence of a unified framework exacerbates the issue of\ninformation underutilization, as it hinders the sharing of interaction\nmechanism learned across diverse datasets. To address these challenges, this\nwork proposes a novel LLM-based multi-modal framework for Molecular inTeraction\nprediction following Chain-of-Thought (CoT) theory, termed MolTC, which\neffectively integrate graphical information of two molecules in pair. To train\nMolTC efficiently, we introduce a Multi-hierarchical CoT concept to refine its\ntraining paradigm, and conduct a comprehensive Molecular Interactive\nInstructions dataset for the development of biochemical LLMs involving MRL. Our\nexperiments, conducted across various datasets involving over 4,000,000\nmolecular pairs, exhibit the superiority of our method over current GNN and\nLLM-based baselines. Code is available at https://github.com/MangoKiller/MolTC.\n", "date": "2024-02-06", "categories": []}
{"id": "2405.18768", "title": "RNAFlow: RNA Structure & Sequence Design via Inverse Folding-Based Flow\n  Matching", "abstract": "  The growing significance of RNA engineering in diverse biological\napplications has spurred interest in developing AI methods for structure-based\nRNA design. While diffusion models have excelled in protein design, adapting\nthem for RNA presents new challenges due to RNA's conformational flexibility\nand the computational cost of fine-tuning large structure prediction models. To\nthis end, we propose RNAFlow, a flow matching model for protein-conditioned RNA\nsequence-structure design. Its denoising network integrates an RNA inverse\nfolding model and a pre-trained RosettaFold2NA network for generation of RNA\nsequences and structures. The integration of inverse folding in the structure\ndenoising process allows us to simplify training by fixing the structure\nprediction network. We further enhance the inverse folding model by\nconditioning it on inferred conformational ensembles to model dynamic RNA\nconformations. Evaluation on protein-conditioned RNA structure and sequence\ngeneration tasks demonstrates RNAFlow's advantage over existing RNA design\nmethods.\n", "date": "2024-05-29", "categories": []}
{"id": "2406.05738", "title": "Smiles2Dock: an open large-scale multi-task dataset for ML-based\n  molecular docking", "abstract": "  Docking is a crucial component in drug discovery aimed at predicting the\nbinding conformation and affinity between small molecules and target proteins.\nML-based docking has recently emerged as a prominent approach, outpacing\ntraditional methods like DOCK and AutoDock Vina in handling the growing scale\nand complexity of molecular libraries. However, the availability of\ncomprehensive and user-friendly datasets for training and benchmarking ML-based\ndocking algorithms remains limited. We introduce Smiles2Dock, an open\nlarge-scale multi-task dataset for molecular docking. We created a framework\ncombining P2Rank and AutoDock Vina to dock 1.7 million ligands from the ChEMBL\ndatabase against 15 AlphaFold proteins, giving us more than 25 million\nprotein-ligand binding scores. The dataset leverages a wide range of\nhigh-accuracy AlphaFold protein models, encompasses a diverse set of\nbiologically relevant compounds and enables researchers to benchmark all major\napproaches for ML-based docking such as Graph, Transformer and CNN-based\nmethods. We also introduce a novel Transformer-based architecture for docking\nscores prediction and set it as an initial benchmark for our dataset. Our\ndataset and code are publicly available to support the development of novel\nML-based methods for molecular docking to advance scientific research in this\nfield.\n", "date": "2024-06-09", "categories": []}
{"id": "2402.04997", "title": "Generative Flows on Discrete State-Spaces: Enabling Multimodal Flows\n  with Applications to Protein Co-Design", "abstract": "  Combining discrete and continuous data is an important capability for\ngenerative models. We present Discrete Flow Models (DFMs), a new flow-based\nmodel of discrete data that provides the missing link in enabling flow-based\ngenerative models to be applied to multimodal continuous and discrete data\nproblems. Our key insight is that the discrete equivalent of continuous space\nflow matching can be realized using Continuous Time Markov Chains. DFMs benefit\nfrom a simple derivation that includes discrete diffusion models as a specific\ninstance while allowing improved performance over existing diffusion-based\napproaches. We utilize our DFMs method to build a multimodal flow-based\nmodeling framework. We apply this capability to the task of protein co-design,\nwherein we learn a model for jointly generating protein structure and sequence.\nOur approach achieves state-of-the-art co-design performance while allowing the\nsame multimodal model to be used for flexible generation of the sequence or\nstructure.\n", "date": "2024-02-07", "categories": []}
{"id": "2403.03234", "title": "Caduceus: Bi-Directional Equivariant Long-Range DNA Sequence Modeling", "abstract": "  Large-scale sequence modeling has sparked rapid advances that now extend into\nbiology and genomics. However, modeling genomic sequences introduces challenges\nsuch as the need to model long-range token interactions, the effects of\nupstream and downstream regions of the genome, and the reverse complementarity\n(RC) of DNA. Here, we propose an architecture motivated by these challenges\nthat builds off the long-range Mamba block, and extends it to a BiMamba\ncomponent that supports bi-directionality, and to a MambaDNA block that\nadditionally supports RC equivariance. We use MambaDNA as the basis of\nCaduceus, the first family of RC equivariant bi-directional long-range DNA\nlanguage models, and we introduce pre-training and fine-tuning strategies that\nyield Caduceus DNA foundation models. Caduceus outperforms previous long-range\nmodels on downstream benchmarks; on a challenging long-range variant effect\nprediction task, Caduceus exceeds the performance of 10x larger models that do\nnot leverage bi-directionality or equivariance.\n", "date": "2024-03-05", "categories": []}
{"id": "2406.02381", "title": "Kirigami: large convolutional kernels improve deep learning-based RNA\n  secondary structure prediction", "abstract": "  We introduce a novel fully convolutional neural network (FCN) architecture\nfor predicting the secondary structure of ribonucleic acid (RNA) molecules.\nInterpreting RNA structures as weighted graphs, we employ deep learning to\nestimate the probability of base pairing between nucleotide residues. Unique to\nour model are its massive 11-pixel kernels, which we argue provide a distinct\nadvantage for FCNs on the specialized domain of RNA secondary structures. On a\nwidely adopted, standardized test set comprised of 1,305 molecules, the\naccuracy of our method exceeds that of current state-of-the-art (SOTA)\nsecondary structure prediction software, achieving a Matthews Correlation\nCoefficient (MCC) over 11-40% higher than that of other leading methods on\noverall structures and 58-400% higher on pseudoknots specifically.\n", "date": "2024-06-04", "categories": []}
{"id": "2406.03686", "title": "BindGPT: A Scalable Framework for 3D Molecular Design via Language\n  Modeling and Reinforcement Learning", "abstract": "  Generating novel active molecules for a given protein is an extremely\nchallenging task for generative models that requires an understanding of the\ncomplex physical interactions between the molecule and its environment. In this\npaper, we present a novel generative model, BindGPT which uses a conceptually\nsimple but powerful approach to create 3D molecules within the protein's\nbinding site. Our model produces molecular graphs and conformations jointly,\neliminating the need for an extra graph reconstruction step. We pretrain\nBindGPT on a large-scale dataset and fine-tune it with reinforcement learning\nusing scores from external simulation software. We demonstrate how a single\npretrained language model can serve at the same time as a 3D molecular\ngenerative model, conformer generator conditioned on the molecular graph, and a\npocket-conditioned 3D molecule generator. Notably, the model does not make any\nrepresentational equivariance assumptions about the domain of generation. We\nshow how such simple conceptual approach combined with pretraining and scaling\ncan perform on par or better than the current best specialized diffusion\nmodels, language models, and graph neural networks while being two orders of\nmagnitude cheaper to sample.\n", "date": "2024-06-06", "categories": []}
{"id": "2402.00054", "title": "Predicting loss-of-function impact of genetic mutations: a machine\n  learning approach", "abstract": "  The innovation of next-generation sequencing (NGS) techniques has\nsignificantly reduced the price of genome sequencing, lowering barriers to\nfuture medical research; it is now feasible to apply genome sequencing to\nstudies where it would have previously been cost-inefficient. Identifying\ndamaging or pathogenic mutations in vast amounts of complex, high-dimensional\ngenome sequencing data may be of particular interest to researchers. Thus, this\npaper's aims were to train machine learning models on the attributes of a\ngenetic mutation to predict LoFtool scores (which measure a gene's intolerance\nto loss-of-function mutations). These attributes included, but were not limited\nto, the position of a mutation on a chromosome, changes in amino acids, and\nchanges in codons caused by the mutation. Models were built using the\nunivariate feature selection technique f-regression combined with K-nearest\nneighbors (KNN), Support Vector Machine (SVM), Random Sample Consensus\n(RANSAC), Decision Trees, Random Forest, and Extreme Gradient Boosting\n(XGBoost). These models were evaluated using five-fold cross-validated averages\nof r-squared, mean squared error, root mean squared error, mean absolute error,\nand explained variance. The findings of this study include the training of\nmultiple models with testing set r-squared values of 0.97.\n", "date": "2024-01-26", "categories": []}
{"id": "2406.01627", "title": "GenBench: A Benchmarking Suite for Systematic Evaluation of Genomic\n  Foundation Models", "abstract": "  The Genomic Foundation Model (GFM) paradigm is expected to facilitate the\nextraction of generalizable representations from massive genomic data, thereby\nenabling their application across a spectrum of downstream applications.\nDespite advancements, a lack of evaluation framework makes it difficult to\nensure equitable assessment due to experimental settings, model intricacy,\nbenchmark datasets, and reproducibility challenges. In the absence of\nstandardization, comparative analyses risk becoming biased and unreliable. To\nsurmount this impasse, we introduce GenBench, a comprehensive benchmarking\nsuite specifically tailored for evaluating the efficacy of Genomic Foundation\nModels. GenBench offers a modular and expandable framework that encapsulates a\nvariety of state-of-the-art methodologies. Through systematic evaluations of\ndatasets spanning diverse biological domains with a particular emphasis on both\nshort-range and long-range genomic tasks, firstly including the three most\nimportant DNA tasks covering Coding Region, Non-Coding Region, Genome\nStructure, etc. Moreover, We provide a nuanced analysis of the interplay\nbetween model architecture and dataset characteristics on task-specific\nperformance. Our findings reveal an interesting observation: independent of the\nnumber of parameters, the discernible difference in preference between the\nattention-based and convolution-based models on short- and long-range tasks may\nprovide insights into the future design of GFM.\n", "date": "2024-06-01", "categories": []}
{"id": "2406.02610", "title": "MoFormer: Multi-objective Antimicrobial Peptide Generation Based on\n  Conditional Transformer Joint Multi-modal Fusion Descriptor", "abstract": "  Deep learning holds a big promise for optimizing existing peptides with more\ndesirable properties, a critical step towards accelerating new drug discovery.\nDespite the recent emergence of several optimized Antimicrobial peptides(AMP)\ngeneration methods, multi-objective optimizations remain still quite\nchallenging for the idealism-realism tradeoff. Here, we establish a\nmulti-objective AMP synthesis pipeline (MoFormer) for the simultaneous\noptimization of multi-attributes of AMPs. MoFormer improves the desired\nattributes of AMP sequences in a highly structured latent space, guided by\nconditional constraints and fine-grained multi-descriptor.We show that MoFormer\noutperforms existing methods in the generation task of enhanced antimicrobial\nactivity and minimal hemolysis. We also utilize a Pareto-based non-dominated\nsorting algorithm and proxies based on large model fine-tuning to\nhierarchically rank the candidates. We demonstrate substantial property\nimprovement using MoFormer from two perspectives: (1) employing molecular\nsimulations and scoring interactions among amino acids to decipher the\nstructure and functionality of AMPs; (2) visualizing latent space to examine\nthe qualities and distribution features, verifying an effective means to\nfacilitate multi-objective optimization AMPs with design constraints\n", "date": "2024-06-03", "categories": []}
{"id": "2406.02990", "title": "Predicting Genetic Mutation from Whole Slide Images via\n  Biomedical-Linguistic Knowledge Enhanced Multi-label Classification", "abstract": "  Predicting genetic mutations from whole slide images is indispensable for\ncancer diagnosis. However, existing work training multiple binary\nclassification models faces two challenges: (a) Training multiple binary\nclassifiers is inefficient and would inevitably lead to a class imbalance\nproblem. (b) The biological relationships among genes are overlooked, which\nlimits the prediction performance. To tackle these challenges, we innovatively\ndesign a Biological-knowledge enhanced PathGenomic multi-label Transformer to\nimprove genetic mutation prediction performances. BPGT first establishes a\nnovel gene encoder that constructs gene priors by two carefully designed\nmodules: (a) A gene graph whose node features are the genes' linguistic\ndescriptions and the cancer phenotype, with edges modeled by genes' pathway\nassociations and mutation consistencies. (b) A knowledge association module\nthat fuses linguistic and biomedical knowledge into gene priors by\ntransformer-based graph representation learning, capturing the intrinsic\nrelationships between different genes' mutations. BPGT then designs a label\ndecoder that finally performs genetic mutation prediction by two tailored\nmodules: (a) A modality fusion module that firstly fuses the gene priors with\ncritical regions in WSIs and obtains gene-wise mutation logits. (b) A\ncomparative multi-label loss that emphasizes the inherent comparisons among\nmutation status to enhance the discrimination capabilities. Sufficient\nexperiments on The Cancer Genome Atlas benchmark demonstrate that BPGT\noutperforms the state-of-the-art.\n", "date": "2024-06-05", "categories": []}
{"id": "2406.03141", "title": "Floating Anchor Diffusion Model for Multi-motif Scaffolding", "abstract": "  Motif scaffolding seeks to design scaffold structures for constructing\nproteins with functions derived from the desired motif, which is crucial for\nthe design of vaccines and enzymes. Previous works approach the problem by\ninpainting or conditional generation. Both of them can only scaffold motifs\nwith fixed positions, and the conditional generation cannot guarantee the\npresence of motifs. However, prior knowledge of the relative motif positions in\na protein is not readily available, and constructing a protein with multiple\nfunctions in one protein is more general and significant because of the\nsynergies between functions. We propose a Floating Anchor Diffusion (FADiff)\nmodel. FADiff allows motifs to float rigidly and independently in the process\nof diffusion, which guarantees the presence of motifs and automates the motif\nposition design. Our experiments demonstrate the efficacy of FADiff with high\nsuccess rates and designable novel scaffolds. To the best of our knowledge,\nFADiff is the first work to tackle the challenge of scaffolding multiple motifs\nwithout relying on the expertise of relative motif positions in the protein.\nCode is available at https://github.com/aim-uofa/FADiff.\n", "date": "2024-06-05", "categories": []}
{"id": "2406.03403", "title": "Structure-based Drug Design Benchmark: Do 3D Methods Really Dominate?", "abstract": "  Currently, the field of structure-based drug design is dominated by three\nmain types of algorithms: search-based algorithms, deep generative models, and\nreinforcement learning. While existing works have typically focused on\ncomparing models within a single algorithmic category, cross-algorithm\ncomparisons remain scarce. In this paper, to fill the gap, we establish a\nbenchmark to evaluate the performance of sixteen models across these different\nalgorithmic foundations by assessing the pharmaceutical properties of the\ngenerated molecules and their docking affinities with specified target\nproteins. We highlight the unique advantages of each algorithmic approach and\noffer recommendations for the design of future SBDD models. We emphasize that\n1D/2D ligand-centric drug design methods can be used in SBDD by treating the\ndocking function as a black-box oracle, which is typically neglected. The\nempirical results show that 1D/2D methods achieve competitive performance\ncompared with 3D-based methods that use the 3D structure of the target protein\nexplicitly. Also, AutoGrow4, a 2D molecular graph-based genetic algorithm,\ndominates SBDD in terms of optimization ability. The relevant code is available\nin https://github.com/zkysfls/2024-sbdd-benchmark.\n", "date": "2024-06-04", "categories": []}
{"id": "2406.03406", "title": "LncRNA-disease association prediction method based on heterogeneous\n  information completion and convolutional neural network", "abstract": "  The emerging research shows that lncRNA has crucial research value in a\nseries of complex human diseases. Therefore, the accurate identification of\nlncRNA-disease associations (LDAs) is very important for the warning and\ntreatment of diseases. However, most of the existing methods have limitations\nin identifying nonlinear LDAs, and it remains a huge challenge to predict new\nLDAs. In this paper, a deep learning model based on a heterogeneous network and\nconvolutional neural network (CNN) is proposed for lncRNA-disease association\nprediction, named HCNNLDA. The heterogeneous network containing the lncRNA,\ndisease, and miRNA nodes, is constructed firstly. The embedding matrix of a\nlncRNA-disease node pair is constructed according to various biological\npremises about lncRNAs, diseases, and miRNAs. Then, the low-dimensional feature\nrepresentation is fully learned by the convolutional neural network. In the\nend, the XGBoot classifier model is trained to predict the potential LDAs.\nHCNNLDA obtains a high AUC value of 0.9752 and AUPR of 0.9740 under the 5-fold\ncross-validation. The experimental results show that the proposed model has\nbetter performance than that of several latest prediction models. Meanwhile,\nthe effectiveness of HCNNLDA in identifying novel LDAs is further demonstrated\nby case studies of three diseases. To sum up, HCNNLDA is a feasible calculation\nmodel to predict LDAs.\n", "date": "2024-06-02", "categories": []}
{"id": "2406.01617", "title": "LightCPPgen: An Explainable Machine Learning Pipeline for Rational\n  Design of Cell Penetrating Peptides", "abstract": "  Cell-penetrating peptides (CPPs) are powerful vectors for the intracellular\ndelivery of a diverse array of therapeutic molecules. Despite their potential,\nthe rational design of CPPs remains a challenging task that often requires\nextensive experimental efforts and iterations. In this study, we introduce an\ninnovative approach for the de novo design of CPPs, leveraging the strengths of\nmachine learning (ML) and optimization algorithms. Our strategy, named\nLightCPPgen, integrates a LightGBM-based predictive model with a genetic\nalgorithm (GA), enabling the systematic generation and optimization of CPP\nsequences. At the core of our methodology is the development of an accurate,\nefficient, and interpretable predictive model, which utilizes 20 explainable\nfeatures to shed light on the critical factors influencing CPP translocation\ncapacity. The CPP predictive model works synergistically with an optimization\nalgorithm, which is tuned to enhance computational efficiency while maintaining\noptimization performance. The GA solutions specifically target the candidate\nsequences' penetrability score, while trying to maximize similarity with the\noriginal non-penetrating peptide in order to retain its original biological and\nphysicochemical properties. By prioritizing the synthesis of only the most\npromising CPP candidates, LightCPPgen can drastically reduce the time and cost\nassociated with wet lab experiments. In summary, our research makes a\nsubstantial contribution to the field of CPP design, offering a robust\nframework that combines ML and optimization techniques to facilitate the\nrational design of penetrating peptides, by enhancing the explainability and\ninterpretability of the design process.\n", "date": "2024-05-31", "categories": []}
{"id": "2406.01622", "title": "Sifting through the Noise: A Survey of Diffusion Probabilistic Models\n  and Their Applications to Biomolecules", "abstract": "  Diffusion probabilistic models have made their way into a number of\nhigh-profile applications since their inception. In particular, there has been\na wave of research into using diffusion models in the prediction and design of\nbiomolecular structures and sequences. Their growing ubiquity makes it\nimperative for researchers in these fields to understand them. This paper\nserves as a general overview for the theory behind these models and the current\nstate of research. We first introduce diffusion models and discuss common\nmotifs used when applying them to biomolecules. We then present the significant\noutcomes achieved through the application of these models in generative and\npredictive tasks. This survey aims to provide readers with a comprehensive\nunderstanding of the increasingly critical role of diffusion models.\n", "date": "2024-05-31", "categories": []}
{"id": "2406.01650", "title": "TAGMol: Target-Aware Gradient-guided Molecule Generation", "abstract": "  3D generative models have shown significant promise in structure-based drug\ndesign (SBDD), particularly in discovering ligands tailored to specific target\nbinding sites. Existing algorithms often focus primarily on ligand-target\nbinding, characterized by binding affinity. Moreover, models trained solely on\ntarget-ligand distribution may fall short in addressing the broader objectives\nof drug discovery, such as the development of novel ligands with desired\nproperties like drug-likeness, and synthesizability, underscoring the\nmultifaceted nature of the drug design process. To overcome these challenges,\nwe decouple the problem into molecular generation and property prediction. The\nlatter synergistically guides the diffusion sampling process, facilitating\nguided diffusion and resulting in the creation of meaningful molecules with the\ndesired properties. We call this guided molecular generation process as TAGMol.\nThrough experiments on benchmark datasets, TAGMol demonstrates superior\nperformance compared to state-of-the-art baselines, achieving a 22% improvement\nin average Vina Score and yielding favorable outcomes in essential auxiliary\nproperties. This establishes TAGMol as a comprehensive framework for drug\ngeneration.\n", "date": "2024-06-03", "categories": []}
{"id": "2405.10812", "title": "VQDNA: Unleashing the Power of Vector Quantization for Multi-Species\n  Genomic Sequence Modeling", "abstract": "  Similar to natural language models, pre-trained genome language models are\nproposed to capture the underlying intricacies within genomes with unsupervised\nsequence modeling. They have become essential tools for researchers and\npractitioners in biology. However, the hand-crafted tokenization policies used\nin these models may not encode the most discriminative patterns from the\nlimited vocabulary of genomic data. In this paper, we introduce VQDNA, a\ngeneral-purpose framework that renovates genome tokenization from the\nperspective of genome vocabulary learning. By leveraging vector-quantized\ncodebooks as learnable vocabulary, VQDNA can adaptively tokenize genomes into\npattern-aware embeddings in an end-to-end manner. To further push its limits,\nwe propose Hierarchical Residual Quantization (HRQ), where varying scales of\ncodebooks are designed in a hierarchy to enrich the genome vocabulary in a\ncoarse-to-fine manner. Extensive experiments on 32 genome datasets demonstrate\nVQDNA's superiority and favorable parameter efficiency compared to existing\ngenome language models. Notably, empirical analysis of SARS-CoV-2 mutations\nreveals the fine-grained pattern awareness and biological significance of\nlearned HRQ vocabulary, highlighting its untapped potential for broader\napplications in genomics.\n", "date": "2024-05-13", "categories": []}
{"id": "2406.00164", "title": "DYNA: Disease-Specific Language Model for Variant Pathogenicity", "abstract": "  Clinical variant classification of pathogenic versus benign genetic variants\nremains a challenge in clinical genetics. Recently, the proposition of genomic\nfoundation models has improved the generic variant effect prediction (VEP)\naccuracy via weakly-supervised or unsupervised training. However, these VEPs\nare not disease-specific, limiting their adaptation at the point of care. To\naddress this problem, we propose DYNA: Disease-specificity fine-tuning via a\nSiamese neural network broadly applicable to all genomic foundation models for\nmore effective variant effect predictions in disease-specific contexts. We\nevaluate DYNA in two distinct disease-relevant tasks. For coding VEPs, we focus\non various cardiovascular diseases, where gene-disease relationships of\nloss-of-function vs. gain-of-function dictate disease-specific VEP. For\nnon-coding VEPs, we apply DYNA to an essential post-transcriptional regulatory\naxis of RNA splicing, the most common non-coding pathogenic mechanism in\nestablished clinical VEP guidelines. In both cases, DYNA fine-tunes various\npre-trained genomic foundation models on small, rare variant sets. The DYNA\nfine-tuned models show superior performance in the held-out rare variant\ntesting set and are further replicated in large, clinically-relevant variant\nannotations in ClinVAR. Thus, DYNA offers a potent disease-specific variant\neffect prediction method, excelling in intra-gene generalization and\ngeneralization to unseen genetic variants, making it particularly valuable for\ndisease associations and clinical applicability.\n", "date": "2024-05-31", "categories": []}
{"id": "2406.00735", "title": "Full-Atom Peptide Design based on Multi-modal Flow Matching", "abstract": "  Peptides, short chains of amino acid residues, play a vital role in numerous\nbiological processes by interacting with other target molecules, offering\nsubstantial potential in drug discovery. In this work, we present PepFlow, the\nfirst multi-modal deep generative model grounded in the flow-matching framework\nfor the design of full-atom peptides that target specific protein receptors.\nDrawing inspiration from the crucial roles of residue backbone orientations and\nside-chain dynamics in protein-peptide interactions, we characterize the\npeptide structure using rigid backbone frames within the $\\mathrm{SE}(3)$\nmanifold and side-chain angles on high-dimensional tori. Furthermore, we\nrepresent discrete residue types in the peptide sequence as categorical\ndistributions on the probability simplex. By learning the joint distributions\nof each modality using derived flows and vector fields on corresponding\nmanifolds, our method excels in the fine-grained design of full-atom peptides.\nHarnessing the multi-modal paradigm, our approach adeptly tackles various tasks\nsuch as fix-backbone sequence design and side-chain packing through partial\nsampling. Through meticulously crafted experiments, we demonstrate that PepFlow\nexhibits superior performance in comprehensive benchmarks, highlighting its\nsignificant potential in computational peptide design and analysis.\n", "date": "2024-06-02", "categories": []}
{"id": "2310.05764", "title": "Harmonic Self-Conditioned Flow Matching for Multi-Ligand Docking and\n  Binding Site Design", "abstract": "  A significant amount of protein function requires binding small molecules,\nincluding enzymatic catalysis. As such, designing binding pockets for small\nmolecules has several impactful applications ranging from drug synthesis to\nenergy storage. Towards this goal, we first develop HarmonicFlow, an improved\ngenerative process over 3D protein-ligand binding structures based on our\nself-conditioned flow matching objective. FlowSite extends this flow model to\njointly generate a protein pocket's discrete residue types and the molecule's\nbinding 3D structure. We show that HarmonicFlow improves upon state-of-the-art\ngenerative processes for docking in simplicity, generality, and average sample\nquality in pocket-level docking. Enabled by this structure modeling, FlowSite\ndesigns binding sites substantially better than baseline approaches.\n", "date": "2023-10-09", "categories": []}
{"id": "2402.05841", "title": "Dirichlet Flow Matching with Applications to DNA Sequence Design", "abstract": "  Discrete diffusion or flow models could enable faster and more controllable\nsequence generation than autoregressive models. We show that na\\\"ive linear\nflow matching on the simplex is insufficient toward this goal since it suffers\nfrom discontinuities in the training target and further pathologies. To\novercome this, we develop Dirichlet flow matching on the simplex based on\nmixtures of Dirichlet distributions as probability paths. In this framework, we\nderive a connection between the mixtures' scores and the flow's vector field\nthat allows for classifier and classifier-free guidance. Further, we provide\ndistilled Dirichlet flow matching, which enables one-step sequence generation\nwith minimal performance hits, resulting in $O(L)$ speedups compared to\nautoregressive models. On complex DNA sequence generation tasks, we demonstrate\nsuperior performance compared to all baselines in distributional metrics and in\nachieving desired design targets for generated sequences. Finally, we show that\nour classifier-free guidance approach improves unconditional generation and is\neffective for generating DNA that satisfies design targets. Code is available\nat https://github.com/HannesStark/dirichlet-flow-matching.\n", "date": "2024-02-08", "categories": []}
{"id": "2402.17810", "title": "BioT5+: Towards Generalized Biological Understanding with IUPAC\n  Integration and Multi-task Tuning", "abstract": "  Recent research trends in computational biology have increasingly focused on\nintegrating text and bio-entity modeling, especially in the context of\nmolecules and proteins. However, previous efforts like BioT5 faced challenges\nin generalizing across diverse tasks and lacked a nuanced understanding of\nmolecular structures, particularly in their textual representations (e.g.,\nIUPAC). This paper introduces BioT5+, an extension of the BioT5 framework,\ntailored to enhance biological research and drug discovery. BioT5+ incorporates\nseveral novel features: integration of IUPAC names for molecular understanding,\ninclusion of extensive bio-text and molecule data from sources like bioRxiv and\nPubChem, the multi-task instruction tuning for generality across tasks, and a\nnumerical tokenization technique for improved processing of numerical data.\nThese enhancements allow BioT5+ to bridge the gap between molecular\nrepresentations and their textual descriptions, providing a more holistic\nunderstanding of biological entities, and largely improving the grounded\nreasoning of bio-text and bio-sequences. The model is pre-trained and\nfine-tuned with a large number of experiments, including \\emph{3 types of\nproblems (classification, regression, generation), 15 kinds of tasks, and 21\ntotal benchmark datasets}, demonstrating the remarkable performance and\nstate-of-the-art results in most cases. BioT5+ stands out for its ability to\ncapture intricate relationships in biological data, thereby contributing\nsignificantly to bioinformatics and computational biology. Our code is\navailable at \\url{https://github.com/QizhiPei/BioT5}.\n", "date": "2024-02-27", "categories": []}
{"id": "2405.20863", "title": "ABodyBuilder3: Improved and scalable antibody structure predictions", "abstract": "  Accurate prediction of antibody structure is a central task in the design and\ndevelopment of monoclonal antibodies, notably to understand both their\ndevelopability and their binding properties. In this article, we introduce\nABodyBuilder3, an improved and scalable antibody structure prediction model\nbased on ImmuneBuilder. We achieve a new state-of-the-art accuracy in the\nmodelling of CDR loops by leveraging language model embeddings, and show how\npredicted structures can be further improved through careful relaxation\nstrategies. Finally, we incorporate a predicted Local Distance Difference Test\ninto the model output to allow for a more accurate estimation of uncertainties.\n", "date": "2024-05-31", "categories": []}
{"id": "2405.18655", "title": "CAVACHON: a hierarchical variational autoencoder to integrate\n  multi-modal single-cell data", "abstract": "  Paired single-cell sequencing technologies enable the simultaneous\nmeasurement of complementary modalities of molecular data at single-cell\nresolution. Along with the advances in these technologies, many methods based\non variational autoencoders have been developed to integrate these data.\nHowever, these methods do not explicitly incorporate prior biological\nrelationships between the data modalities, which could significantly enhance\nmodeling and interpretation. We propose a novel probabilistic learning\nframework that explicitly incorporates conditional independence relationships\nbetween multi-modal data as a directed acyclic graph using a generalized\nhierarchical variational autoencoder. We demonstrate the versatility of our\nframework across various applications pertinent to single-cell multi-omics data\nintegration. These include the isolation of common and distinct information\nfrom different modalities, modality-specific differential analysis, and\nintegrated cell clustering. We anticipate that the proposed framework can\nfacilitate the construction of highly flexible graphical models that can\ncapture the complexities of biological hypotheses and unravel the connections\nbetween different biological data types, such as different modalities of paired\nsingle-cell multi-omics data. The implementation of the proposed framework can\nbe found in the repository https://github.com/kuijjerlab/CAVACHON.\n", "date": "2024-05-28", "categories": []}
{"id": "2405.18968", "title": "UniIF: Unified Molecule Inverse Folding", "abstract": "  Molecule inverse folding has been a long-standing challenge in chemistry and\nbiology, with the potential to revolutionize drug discovery and material\nscience. Despite specified models have been proposed for different small- or\nmacro-molecules, few have attempted to unify the learning process, resulting in\nredundant efforts. Complementary to recent advancements in molecular structure\nprediction, such as RoseTTAFold All-Atom and AlphaFold3, we propose the unified\nmodel UniIF for the inverse folding of all molecules. We do such unification in\ntwo levels: 1) Data-Level: We propose a unified block graph data form for all\nmolecules, including the local frame building and geometric feature\ninitialization. 2) Model-Level: We introduce a geometric block attention\nnetwork, comprising a geometric interaction, interactive attention and virtual\nlong-term dependency modules, to capture the 3D interactions of all molecules.\nThrough comprehensive evaluations across various tasks such as protein design,\nRNA design, and material design, we demonstrate that our proposed method\nsurpasses state-of-the-art methods on all tasks. UniIF offers a versatile and\neffective solution for general molecule inverse folding.\n", "date": "2024-05-29", "categories": []}
{"id": "2405.18986", "title": "Robust Optimization in Protein Fitness Landscapes Using Reinforcement\n  Learning in Latent Space", "abstract": "  Proteins are complex molecules responsible for different functions in nature.\nEnhancing the functionality of proteins and cellular fitness can significantly\nimpact various industries. However, protein optimization using computational\nmethods remains challenging, especially when starting from low-fitness\nsequences. We propose LatProtRL, an optimization method to efficiently traverse\na latent space learned by an encoder-decoder leveraging a large protein\nlanguage model. To escape local optima, our optimization is modeled as a Markov\ndecision process using reinforcement learning acting directly in latent space.\nWe evaluate our approach on two important fitness optimization tasks,\ndemonstrating its ability to achieve comparable or superior fitness over\nbaseline methods. Our findings and in vitro evaluation show that the generated\nsequences can reach high-fitness regions, suggesting a substantial potential of\nLatProtRL in lab-in-the-loop scenarios.\n", "date": "2024-05-29", "categories": []}
{"id": "2312.05340", "title": "Transition Path Sampling with Boltzmann Generator-based MCMC Moves", "abstract": "  Sampling all possible transition paths between two 3D states of a molecular\nsystem has various applications ranging from catalyst design to drug discovery.\nCurrent approaches to sample transition paths use Markov chain Monte Carlo and\nrely on time-intensive molecular dynamics simulations to find new paths. Our\napproach operates in the latent space of a normalizing flow that maps from the\nmolecule's Boltzmann distribution to a Gaussian, where we propose new paths\nwithout requiring molecular simulations. Using alanine dipeptide, we explore\nMetropolis-Hastings acceptance criteria in the latent space for exact sampling\nand investigate different latent proposal mechanisms.\n", "date": "2023-12-08", "categories": []}
{"id": "2405.17802", "title": "Multi-level Interaction Modeling for Protein Mutational Effect\n  Prediction", "abstract": "  Protein-protein interactions are central mediators in many biological\nprocesses. Accurately predicting the effects of mutations on interactions is\ncrucial for guiding the modulation of these interactions, thereby playing a\nsignificant role in therapeutic development and drug discovery. Mutations\ngenerally affect interactions hierarchically across three levels: mutated\nresidues exhibit different sidechain conformations, which lead to changes in\nthe backbone conformation, eventually affecting the binding affinity between\nproteins. However, existing methods typically focus only on sidechain-level\ninteraction modeling, resulting in suboptimal predictions. In this work, we\npropose a self-supervised multi-level pre-training framework, ProMIM, to fully\ncapture all three levels of interactions with well-designed pretraining\nobjectives. Experiments show ProMIM outperforms all the baselines on the\nstandard benchmark, especially on mutations where significant changes in\nbackbone conformations may occur. In addition, leading results from zero-shot\nevaluations for SARS-CoV-2 mutational effect prediction and antibody\noptimization underscore the potential of ProMIM as a powerful next-generation\ntool for developing novel therapeutic approaches and new drugs.\n", "date": "2024-05-28", "categories": []}
{"id": "2405.15928", "title": "PatchProt: Hydrophobic patch prediction using protein foundation models", "abstract": "  Hydrophobic patches on protein surfaces play important functional roles in\nprotein-protein and protein-ligand interactions. Large hydrophobic surfaces are\nalso involved in the progression of aggregation diseases. Predicting exposed\nhydrophobic patches from a protein sequence has been shown to be a difficult\ntask. Fine-tuning foundation models allows for adapting a model to the specific\nnuances of a new task using a much smaller dataset. Additionally, multi-task\ndeep learning offers a promising solution for addressing data gaps,\nsimultaneously outperforming single-task methods. In this study, we harnessed a\nrecently released leading large language model ESM-2. Efficient fine-tuning of\nESM-2 was achieved by leveraging a recently developed parameter-efficient\nfine-tuning method. This approach enabled comprehensive training of model\nlayers without excessive parameters and without the need to include a\ncomputationally expensive multiple sequence analysis. We explored several\nrelated tasks, at local (residue) and global (protein) levels, to improve the\nrepresentation of the model. As a result, our fine-tuned ESM-2 model,\nPatchProt, cannot only predict hydrophobic patch areas but also outperforms\nexisting methods at predicting primary tasks, including secondary structure and\nsurface accessibility predictions. Importantly, our analysis shows that\nincluding related local tasks can improve predictions on more difficult global\ntasks. This research sets a new standard for sequence-based protein property\nprediction and highlights the remarkable potential of fine-tuning foundation\nmodels enriching the model representation by training over related tasks.\n", "date": "2024-05-24", "categories": []}
{"id": "2405.16205", "title": "GeneAgent: Self-verification Language Agent for Gene Set Knowledge\n  Discovery using Domain Databases", "abstract": "  Gene set knowledge discovery is essential for advancing human functional\ngenomics. Recent studies have shown promising performance by harnessing the\npower of Large Language Models (LLMs) on this task. Nonetheless, their results\nare subject to several limitations common in LLMs such as hallucinations. In\nresponse, we present GeneAgent, a first-of-its-kind language agent featuring\nself-verification capability. It autonomously interacts with various biological\ndatabases and leverages relevant domain knowledge to improve accuracy and\nreduce hallucination occurrences. Benchmarking on 1,106 gene sets from\ndifferent sources, GeneAgent consistently outperforms standard GPT-4 by a\nsignificant margin. Moreover, a detailed manual review confirms the\neffectiveness of the self-verification module in minimizing hallucinations and\ngenerating more reliable analytical narratives. To demonstrate its practical\nutility, we apply GeneAgent to seven novel gene sets derived from mouse B2905\nmelanoma cell lines, with expert evaluations showing that GeneAgent offers\nnovel insights into gene functions and subsequently expedites knowledge\ndiscovery.\n", "date": "2024-05-25", "categories": []}
{"id": "2405.17066", "title": "Saturn: Sample-efficient Generative Molecular Design using Memory\n  Manipulation", "abstract": "  Generative molecular design for drug discovery has very recently achieved a\nwave of experimental validation, with language-based backbones being the most\ncommon architectures employed. The most important factor for downstream success\nis whether an in silico oracle is well correlated with the desired end-point.\nTo this end, current methods use cheaper proxy oracles with higher throughput\nbefore evaluating the most promising subset with high-fidelity oracles. The\nability to directly optimize high-fidelity oracles would greatly enhance\ngenerative design and be expected to improve hit rates. However, current models\nare not efficient enough to consider such a prospect, exemplifying the sample\nefficiency problem. In this work, we introduce Saturn, which leverages the\nAugmented Memory algorithm and demonstrates the first application of the Mamba\narchitecture for generative molecular design. We elucidate how experience\nreplay with data augmentation improves sample efficiency and how Mamba\nsynergistically exploits this mechanism. Saturn outperforms 22 models on\nmulti-parameter optimization tasks relevant to drug discovery and may possess\nsufficient sample efficiency to consider the prospect of directly optimizing\nhigh-fidelity oracles.\n", "date": "2024-05-27", "categories": []}
{"id": "2306.03218", "title": "Optimal transport for automatic alignment of untargeted metabolomic data", "abstract": "  Untargeted metabolomic profiling through liquid chromatography-mass\nspectrometry (LC-MS) measures a vast array of metabolites within biospecimens,\nadvancing drug development, disease diagnosis, and risk prediction. However,\nthe low throughput of LC-MS poses a major challenge for biomarker discovery,\nannotation, and experimental comparison, necessitating the merging of multiple\ndatasets. Current data pooling methods encounter practical limitations due to\ntheir vulnerability to data variations and hyperparameter dependence. Here we\nintroduce GromovMatcher, a flexible and user-friendly algorithm that\nautomatically combines LC-MS datasets using optimal transport. By capitalizing\non feature intensity correlation structures, GromovMatcher delivers superior\nalignment accuracy and robustness compared to existing approaches. This\nalgorithm scales to thousands of features requiring minimal hyperparameter\ntuning. Manually curated datasets for validating alignment algorithms are\nlimited in the field of untargeted metabolomics, and hence we develop a dataset\nsplit procedure to generate pairs of validation datasets to test the alignments\nproduced by GromovMatcher and other methods. Applying our method to\nexperimental patient studies of liver and pancreatic cancer, we discover shared\nmetabolic features related to patient alcohol intake, demonstrating how\nGromovMatcher facilitates the search for biomarkers associated with lifestyle\nrisk factors linked to several cancer types.\n", "date": "2023-06-05", "categories": []}
{"id": "2405.14925", "title": "PILOT: Equivariant diffusion for pocket conditioned de novo ligand\n  generation with multi-objective guidance via importance sampling", "abstract": "  The generation of ligands that both are tailored to a given protein pocket\nand exhibit a range of desired chemical properties is a major challenge in\nstructure-based drug design. Here, we propose an in-silico approach for the\n$\\textit{de novo}$ generation of 3D ligand structures using the equivariant\ndiffusion model PILOT, combining pocket conditioning with a large-scale\npre-training and property guidance. Its multi-objective trajectory-based\nimportance sampling strategy is designed to direct the model towards molecules\nthat not only exhibit desired characteristics such as increased binding\naffinity for a given protein pocket but also maintains high synthetic\naccessibility. This ensures the practicality of sampled molecules, thus\nmaximizing their potential for the drug discovery pipeline. PILOT significantly\noutperforms existing methods across various metrics on the common benchmark\ndataset CrossDocked2020. Moreover, we employ PILOT to generate novel ligands\nfor unseen protein pockets from the Kinodata-3D dataset, which encompasses a\nsubstantial portion of the human kinome. The generated structures exhibit\npredicted $IC_{50}$ values indicative of potent biological activity, which\nhighlights the potential of PILOT as a powerful tool for structure-based drug\ndesign.\n", "date": "2024-05-23", "categories": []}
{"id": "2405.15489", "title": "Out of Many, One: Designing and Scaffolding Proteins at the Scale of the\n  Structural Universe with Genie 2", "abstract": "  Protein diffusion models have emerged as a promising approach for protein\ndesign. One such pioneering model is Genie, a method that asymmetrically\nrepresents protein structures during the forward and backward processes, using\nsimple Gaussian noising for the former and expressive SE(3)-equivariant\nattention for the latter. In this work we introduce Genie 2, extending Genie to\ncapture a larger and more diverse protein structure space through architectural\ninnovations and massive data augmentation. Genie 2 adds motif scaffolding\ncapabilities via a novel multi-motif framework that designs co-occurring motifs\nwith unspecified inter-motif positions and orientations. This makes possible\ncomplex protein designs that engage multiple interaction partners and perform\nmultiple functions. On both unconditional and conditional generation, Genie 2\nachieves state-of-the-art performance, outperforming all known methods on key\ndesign metrics including designability, diversity, and novelty. Genie 2 also\nsolves more motif scaffolding problems than other methods and does so with more\nunique and varied solutions. Taken together, these advances set a new standard\nfor structure-based protein design. Genie 2 inference and training code, as\nwell as model weights, are freely available at:\nhttps://github.com/aqlaboratory/genie2.\n", "date": "2024-05-24", "categories": []}
{"id": "2207.13842", "title": "Dive into Machine Learning Algorithms for Influenza Virus Host\n  Prediction with Hemagglutinin Sequences", "abstract": "  Influenza viruses mutate rapidly and can pose a threat to public health,\nespecially to those in vulnerable groups. Throughout history, influenza A\nviruses have caused pandemics between different species. It is important to\nidentify the origin of a virus in order to prevent the spread of an outbreak.\nRecently, there has been increasing interest in using machine learning\nalgorithms to provide fast and accurate predictions for viral sequences. In\nthis study, real testing data sets and a variety of evaluation metrics were\nused to evaluate machine learning algorithms at different taxonomic levels. As\nhemagglutinin is the major protein in the immune response, only hemagglutinin\nsequences were used and represented by position-specific scoring matrix and\nword embedding. The results suggest that the 5-grams-transformer neural network\nis the most effective algorithm for predicting viral sequence origins, with\napproximately 99.54% AUCPR, 98.01% F1 score and 96.60% MCC at a higher\nclassification level, and approximately 94.74% AUCPR, 87.41% F1 score and\n80.79% MCC at a lower classification level.\n", "date": "2022-07-28", "categories": []}
{"id": "2310.13913", "title": "Pre-Training on Large-Scale Generated Docking Conformations with\n  HelixDock to Unlock the Potential of Protein-ligand Structure Prediction\n  Models", "abstract": "  Protein-ligand structure prediction is an essential task in drug discovery,\npredicting the binding interactions between small molecules (ligands) and\ntarget proteins (receptors). Recent advances have incorporated deep learning\ntechniques to improve the accuracy of protein-ligand structure prediction.\nNevertheless, the experimental validation of docking conformations remains\ncostly, it raises concerns regarding the generalizability of these deep\nlearning-based methods due to the limited training data. In this work, we show\nthat by pre-training on a large-scale docking conformation generated by\ntraditional physics-based docking tools and then fine-tuning with a limited set\nof experimentally validated receptor-ligand complexes, we can obtain a\nprotein-ligand structure prediction model with outstanding performance.\nSpecifically, this process involved the generation of 100 million docking\nconformations for protein-ligand pairings, an endeavor consuming roughly 1\nmillion CPU core days. The proposed model, HelixDock, aims to acquire the\nphysical knowledge encapsulated by the physics-based docking tools during the\npre-training phase. HelixDock has been rigorously benchmarked against both\nphysics-based and deep learning-based baselines, demonstrating its exceptional\nprecision and robust transferability in predicting binding confirmation. In\naddition, our investigation reveals the scaling laws governing pre-trained\nprotein-ligand structure prediction models, indicating a consistent enhancement\nin performance with increases in model parameters and the volume of\npre-training data. Moreover, we applied HelixDock to several drug\ndiscovery-related tasks to validate its practical utility. HelixDock\ndemonstrates outstanding capabilities on both cross-docking and structure-based\nvirtual screening benchmarks.\n", "date": "2023-10-21", "categories": []}
{"id": "2405.14536", "title": "Regressor-free Molecule Generation to Support Drug Response Prediction", "abstract": "  Drug response prediction (DRP) is a crucial phase in drug discovery, and the\nmost important metric for its evaluation is the IC50 score. DRP results are\nheavily dependent on the quality of the generated molecules. Existing molecule\ngeneration methods typically employ classifier-based guidance, enabling\nsampling within the IC50 classification range. However, these methods fail to\nensure the sampling space range's effectiveness, generating numerous\nineffective molecules. Through experimental and theoretical study, we\nhypothesize that conditional generation based on the target IC50 score can\nobtain a more effective sampling space. As a result, we introduce\nregressor-free guidance molecule generation to ensure sampling within a more\neffective space and support DRP. Regressor-free guidance combines a diffusion\nmodel's score estimation with a regression controller model's gradient based on\nnumber labels. To effectively map regression labels between drugs and cell\nlines, we design a common-sense numerical knowledge graph that constrains the\norder of text representations. Experimental results on the real-world dataset\nfor the DRP task demonstrate our method's effectiveness in drug discovery. The\ncode is available at:https://anonymous.4open.science/r/RMCD-DBD1.\n", "date": "2024-05-23", "categories": []}
{"id": "2405.14545", "title": "A Cross-Field Fusion Strategy for Drug-Target Interaction Prediction", "abstract": "  Drug-target interaction (DTI) prediction is a critical component of the drug\ndiscovery process. In the drug development engineering field, predicting novel\ndrug-target interactions is extremely crucial.However, although existing\nmethods have achieved high accuracy levels in predicting known drugs and drug\ntargets, they fail to utilize global protein information during DTI prediction.\nThis leads to an inability to effectively predict interaction the interactions\nbetween novel drugs and their targets. As a result, the cross-field information\nfusion strategy is employed to acquire local and global protein information.\nThus, we propose the siamese drug-target interaction SiamDTI prediction method,\nwhich utilizes a double channel network structure for cross-field supervised\nlearning.Experimental results on three benchmark datasets demonstrate that\nSiamDTI achieves higher accuracy levels than other state-of-the-art (SOTA)\nmethods on novel drugs and targets.Additionally, SiamDTI's performance with\nknown drugs and targets is comparable to that of SOTA approachs. The code is\navailable at https://anonymous.4open.science/r/DDDTI-434D.\n", "date": "2024-05-23", "categories": []}
{"id": "2306.09391", "title": "Multi-omics Prediction from High-content Cellular Imaging with Deep\n  Learning", "abstract": "  High-content cellular imaging, transcriptomics, and proteomics data provide\nrich and complementary views on the molecular layers of biology that influence\ncellular states and function. However, the biological determinants through\nwhich changes in multi-omics measurements influence cellular morphology have\nnot yet been systematically explored, and the degree to which cell imaging\ncould potentially enable the prediction of multi-omics directly from cell\nimaging data is therefore currently unclear. Here, we address the question of\nwhether it is possible to predict bulk multi-omics measurements directly from\ncell images using Image2Omics - a deep learning approach that predicts\nmulti-omics in a cell population directly from high-content images of cells\nstained with multiplexed fluorescent dyes. We perform an experimental\nevaluation in gene-edited macrophages derived from human induced pluripotent\nstem cells (hiPSC) under multiple stimulation conditions and demonstrate that\nImage2Omics achieves significantly better performance in predicting\ntranscriptomics and proteomics measurements directly from cell images than\npredictions based on the mean observed training set abundance. We observed\nsignificant predictability of abundances for 4927 (18.72%; 95% CI: 6.52%,\n35.52%) and 3521 (13.38%; 95% CI: 4.10%, 32.21%) transcripts out of 26137 in M1\nand M2-stimulated macrophages respectively and for 422 (8.46%; 95% CI: 0.58%,\n25.83%) and 697 (13.98%; 95% CI: 2.41%, 32.83%) proteins out of 4986 in M1 and\nM2-stimulated macrophages respectively. Our results show that some transcript\nand protein abundances are predictable from cell imaging and that cell imaging\nmay potentially, in some settings and depending on the mechanisms of interest\nand desired performance threshold, even be a scalable and resource-efficient\nsubstitute for multi-omics measurements.\n", "date": "2023-06-15", "categories": []}
{"id": "2307.14361", "title": "A Hybrid Machine Learning Model for Classifying Gene Mutations in Cancer\n  using LSTM, BiLSTM, CNN, GRU, and GloVe", "abstract": "  In our study, we introduce a novel hybrid ensemble model that synergistically\ncombines LSTM, BiLSTM, CNN, GRU, and GloVe embeddings for the classification of\ngene mutations in cancer. This model was rigorously tested using Kaggle's\nPersonalized Medicine: Redefining Cancer Treatment dataset, demonstrating\nexceptional performance across all evaluation metrics. Notably, our approach\nachieved a training accuracy of 80.6%, precision of 81.6%, recall of 80.6%, and\nan F1 score of 83.1%, alongside a significantly reduced Mean Squared Error\n(MSE) of 2.596. These results surpass those of advanced transformer models and\ntheir ensembles, showcasing our model's superior capability in handling the\ncomplexities of gene mutation classification. The accuracy and efficiency of\ngene mutation classification are paramount in the era of precision medicine,\nwhere tailored treatment plans based on individual genetic profiles can\ndramatically improve patient outcomes and save lives. Our model's remarkable\nperformance highlights its potential in enhancing the precision of cancer\ndiagnoses and treatments, thereby contributing significantly to the advancement\nof personalized healthcare.\n", "date": "2023-07-24", "categories": []}
{"id": "2405.07452", "title": "PLA-SGCN: Protein-Ligand Binding Affinity Prediction by Integrating\n  Similar Pairs and Semi-supervised Graph Convolutional Network", "abstract": "  The protein-ligand binding affinity (PLA) prediction goal is to predict\nwhether or not the ligand could bind to a protein sequence. Recently, in PLA\nprediction, deep learning has received much attention. Two steps are involved\nin deep learning-based approaches: feature extraction and task prediction step.\nMany deep learning-based approaches concentrate on introducing new feature\nextraction networks or integrating auxiliary knowledge like protein-protein\ninteraction networks or gene ontology knowledge. Then, a task prediction\nnetwork is designed simply using some fully connected layers. This paper aims\nto integrate retrieved similar hard protein-ligand pairs in PLA prediction\n(i.e., task prediction step) using a semi-supervised graph convolutional\nnetwork (GCN). Hard protein-ligand pairs are retrieved for each input query\nsample based on the manifold smoothness constraint. Then, a graph is learned\nautomatically in which each node is a protein-ligand pair, and each edge\nrepresents the similarity between pairs. In other words, an end-to-end\nframework is proposed that simultaneously retrieves hard similar samples,\nlearns protein-ligand descriptor, learns the graph topology of the input sample\nwith retrieved similar hard samples (learn adjacency matrix), and learns a\nsemi-supervised GCN to predict the binding affinity (as task predictor). The\ntraining step adjusts the parameter values, and in the inference step, the\nlearned model is fine-tuned for each input sample. To evaluate the proposed\napproach, it is applied to the four well-known PDBbind, Davis, KIBA, and\nBindingDB datasets. The results show that the proposed method significantly\nperforms better than the comparable approaches.\n", "date": "2024-05-13", "categories": []}
{"id": "2405.11280", "title": "Joint Analysis of Single-Cell Data across Cohorts with Missing\n  Modalities", "abstract": "  Joint analysis of multi-omic single-cell data across cohorts has\nsignificantly enhanced the comprehensive analysis of cellular processes.\nHowever, most of the existing approaches for this purpose require access to\nsamples with complete modality availability, which is impractical in many\nreal-world scenarios. In this paper, we propose (Single-Cell Cross-Cohort\nCross-Category) integration, a novel framework that learns unified cell\nrepresentations under domain shift without requiring full-modality reference\nsamples. Our generative approach learns rich cross-modal and cross-domain\nrelationships that enable imputation of these missing modalities. Through\nexperiments on real-world multi-omic datasets, we demonstrate that offers a\nrobust solution to single-cell tasks such as cell type clustering, cell type\nclassification, and feature imputation.\n", "date": "2024-05-18", "categories": []}
{"id": "2404.10260", "title": "HelixFold-Multimer: Elevating Protein Complex Structure Prediction to\n  New Heights", "abstract": "  While monomer protein structure prediction tools boast impressive accuracy,\nthe prediction of protein complex structures remains a daunting challenge in\nthe field. This challenge is particularly pronounced in scenarios involving\ncomplexes with protein chains from different species, such as antigen-antibody\ninteractions, where accuracy often falls short. Limited by the accuracy of\ncomplex prediction, tasks based on precise protein-protein interaction analysis\nalso face obstacles. In this report, we highlight the ongoing advancements of\nour protein complex structure prediction model, HelixFold-Multimer,\nunderscoring its enhanced performance. HelixFold-Multimer provides precise\npredictions for diverse protein complex structures, especially in therapeutic\nprotein interactions. Notably, HelixFold-Multimer achieves remarkable success\nin antigen-antibody and peptide-protein structure prediction, greatly\nsurpassing AlphaFold 3. HelixFold-Multimer is now available for public use on\nthe PaddleHelix platform, offering both a general version and an\nantigen-antibody version. Researchers can conveniently access and utilize this\nservice for their development needs.\n", "date": "2024-04-16", "categories": []}
{"id": "2311.18803", "title": "BioCLIP: A Vision Foundation Model for the Tree of Life", "abstract": "  Images of the natural world, collected by a variety of cameras, from drones\nto individual phones, are increasingly abundant sources of biological\ninformation. There is an explosion of computational methods and tools,\nparticularly computer vision, for extracting biologically relevant information\nfrom images for science and conservation. Yet most of these are bespoke\napproaches designed for a specific task and are not easily adaptable or\nextendable to new questions, contexts, and datasets. A vision model for general\norganismal biology questions on images is of timely need. To approach this, we\ncurate and release TreeOfLife-10M, the largest and most diverse ML-ready\ndataset of biology images. We then develop BioCLIP, a foundation model for the\ntree of life, leveraging the unique properties of biology captured by\nTreeOfLife-10M, namely the abundance and variety of images of plants, animals,\nand fungi, together with the availability of rich structured biological\nknowledge. We rigorously benchmark our approach on diverse fine-grained biology\nclassification tasks and find that BioCLIP consistently and substantially\noutperforms existing baselines (by 16% to 17% absolute). Intrinsic evaluation\nreveals that BioCLIP has learned a hierarchical representation conforming to\nthe tree of life, shedding light on its strong generalizability.\nhttps://imageomics.github.io/bioclip has models, data and code.\n", "date": "2023-11-30", "categories": []}
{"id": "2405.06645", "title": "On Recovering Higher-order Interactions from Protein Language Models", "abstract": "  Protein language models leverage evolutionary information to perform\nstate-of-the-art 3D structure and zero-shot variant prediction. Yet, extracting\nand explaining all the mutational interactions that govern model predictions\nremains difficult as it requires querying the entire amino acid space for $n$\nsites using $20^n$ sequences, which is computationally expensive even for\nmoderate values of $n$ (e.g., $n\\sim10$). Although approaches to lower the\nsample complexity exist, they often limit the interpretability of the model to\njust single and pairwise interactions. Recently, computationally scalable\nalgorithms relying on the assumption of sparsity in the Fourier domain have\nemerged to learn interactions from experimental data. However, extracting\ninteractions from language models poses unique challenges: it's unclear if\nsparsity is always present or if it is the only metric needed to assess the\nutility of Fourier algorithms. Herein, we develop a framework to do a\nsystematic Fourier analysis of the protein language model ESM2 applied on three\nproteins-green fluorescent protein (GFP), tumor protein P53 (TP53), and G\ndomain B1 (GB1)-across various sites for 228 experiments. We demonstrate that\nESM2 is dominated by three regions in the sparsity-ruggedness plane, two of\nwhich are better suited for sparse Fourier transforms. Validations on two\nsample proteins demonstrate recovery of all interactions with $R^2=0.72$ in the\nmore sparse region and $R^2=0.66$ in the more dense region, using only 7\nmillion out of $20^{10}\\sim10^{13}$ ESM2 samples, reducing the computational\ntime by a staggering factor of 15,000. All codes and data are available on our\nGitHub repository https://github.com/amirgroup-codes/InteractionRecovery.\n", "date": "2024-03-15", "categories": []}
{"id": "2405.06654", "title": "PROflow: An iterative refinement model for PROTAC-induced structure\n  prediction", "abstract": "  Proteolysis targeting chimeras (PROTACs) are small molecules that trigger the\nbreakdown of traditionally ``undruggable'' proteins by binding simultaneously\nto their targets and degradation-associated proteins. A key challenge in their\nrational design is understanding their structural basis of activity. Due to the\nlack of crystal structures (18 in the PDB), existing PROTAC docking methods\nhave been forced to simplify the problem into a distance-constrained\nprotein-protein docking task. To address the data issue, we develop a novel\npseudo-data generation scheme that requires only binary protein-protein\ncomplexes. This new dataset enables PROflow, an iterative refinement model for\nPROTAC-induced structure prediction that models the full PROTAC flexibility\nduring constrained protein-protein docking. PROflow outperforms the\nstate-of-the-art across docking metrics and runtime. Its inference speed\nenables the large-scale screening of PROTAC designs, and computed properties of\npredicted structures achieve statistically significant correlations with\npublished degradation activities.\n", "date": "2024-04-10", "categories": []}
{"id": "2405.06992", "title": "ResSurv: Cancer Survival Analysis Prediction Model Based on Residual\n  Networks", "abstract": "  Survival prediction is an important branch of cancer prognosis analysis. The\nmodel that predicts survival risk through TCGA genomics data can discover genes\nrelated to cancer and provide diagnosis and treatment recommendations based on\npatient characteristics. We found that deep learning models based on Cox\nproportional hazards often suffer from overfitting when dealing with\nhigh-throughput data. Moreover, we found that as the number of network layers\nincreases, the experimental results will not get better, and network\ndegradation will occur. Based on this problem, we propose a new framework based\non Deep Residual Learning. Combine the ideas of Cox proportional hazards and\nResidual. And name it ResSurv. First, ResSurv is a feed-forward deep learning\nnetwork stacked by multiple basic ResNet Blocks. In each ResNet Block, we add a\nNormalization Layer to prevent gradient disappearance and gradient explosion.\nSecondly, for the loss function of the neural network, we inherited the Cox\nproportional hazards methods, applied the semi-parametric of the CPH model to\nthe neural network, combined with the partial likelihood model, established the\nloss function, and performed backpropagation and gradient update. Finally, we\ncompared ResSurv networks of different depths and found that we can effectively\nextract high-dimensional features. Ablation experiments and comparative\nexperiments prove that our model has reached SOTA(state of the art) in the\nfield of deep learning, and our network can effectively extract deep\ninformation.\n", "date": "2024-05-11", "categories": []}
{"id": "2405.07622", "title": "De novo antibody design with SE(3) diffusion", "abstract": "  We introduce IgDiff, an antibody variable domain diffusion model based on a\ngeneral protein backbone diffusion framework which was extended to handle\nmultiple chains. Assessing the designability and novelty of the structures\ngenerated with our model, we find that IgDiff produces highly designable\nantibodies that can contain novel binding regions. The backbone dihedral angles\nof sampled structures show good agreement with a reference antibody\ndistribution. We verify these designed antibodies experimentally and find that\nall express with high yield. Finally, we compare our model with a\nstate-of-the-art generative backbone diffusion model on a range of antibody\ndesign tasks, such as the design of the complementarity determining regions or\nthe pairing of a light chain to an existing heavy chain, and show improved\nproperties and designability.\n", "date": "2024-05-13", "categories": []}
{"id": "2405.07879", "title": "On the Relation Between Autoencoders and Non-negative Matrix\n  Factorization, and Their Application for Mutational Signature Extraction", "abstract": "  The aim of this study is to provide a foundation to understand the\nrelationship between non-negative matrix factorization (NMF) and non-negative\nautoencoders enabling proper interpretation and understanding of\nautoencoder-based alternatives to NMF. Since its introduction, NMF has been a\npopular tool for extracting interpretable, low-dimensional representations of\nhigh-dimensional data. However, recently, several studies have proposed to\nreplace NMF with autoencoders. This increasing popularity of autoencoders\nwarrants an investigation on whether this replacement is in general valid and\nreasonable. Moreover, the exact relationship between non-negative autoencoders\nand NMF has not been thoroughly explored. Thus, a main aim of this study is to\ninvestigate in detail the relationship between non-negative autoencoders and\nNMF. We find that the connection between the two models can be established\nthrough convex NMF, which is a restricted case of NMF. In particular, convex\nNMF is a special case of an autoencoder. The performance of NMF and\nautoencoders is compared within the context of extraction of mutational\nsignatures from cancer genomics data. We find that the reconstructions based on\nNMF are more accurate compared to autoencoders, while the signatures extracted\nusing both methods show comparable consistencies and values when externally\nvalidated. These findings suggest that the non-negative autoencoders\ninvestigated in this article do not provide an improvement of NMF in the field\nof mutational signature extraction.\n", "date": "2024-05-13", "categories": []}
{"id": "2405.06511", "title": "Towards Less Biased Data-driven Scoring with Deep Learning-Based\n  End-to-end Database Search in Tandem Mass Spectrometry", "abstract": "  Peptide identification in mass spectrometry-based proteomics is crucial for\nunderstanding protein function and dynamics. Traditional database search\nmethods, though widely used, rely on heuristic scoring functions and\nstatistical estimations have to be introduced for a higher identification rate.\nHere, we introduce DeepSearch, the first deep learning-based end-to-end\ndatabase search method for tandem mass spectrometry. DeepSearch leverages a\nmodified transformer-based encoder-decoder architecture under the contrastive\nlearning framework. Unlike conventional methods that rely on ion-to-ion\nmatching, DeepSearch adopts a data-driven approach to score peptide spectrum\nmatches. DeepSearch is also the first deep learning-based method that can\nprofile variable post-translational modifications in a zero-shot manner. We\nshowed that DeepSearch's scoring scheme expressed less bias and did not require\nany statistical estimation. We validated DeepSearch's accuracy and robustness\nacross various datasets, including those from species with diverse protein\ncompositions and a modification-enriched dataset. DeepSearch sheds new light on\ndatabase search methods in tandem mass spectrometry.\n", "date": "2024-05-08", "categories": []}
{"id": "2311.13466", "title": "Accelerating Inference in Molecular Diffusion Models with Latent\n  Representations of Protein Structure", "abstract": "  Diffusion generative models have emerged as a powerful framework for\naddressing problems in structural biology and structure-based drug design.\nThese models operate directly on 3D molecular structures. Due to the\nunfavorable scaling of graph neural networks (GNNs) with graph size as well as\nthe relatively slow inference speeds inherent to diffusion models, many\nexisting molecular diffusion models rely on coarse-grained representations of\nprotein structure to make training and inference feasible. However, such\ncoarse-grained representations discard essential information for modeling\nmolecular interactions and impair the quality of generated structures. In this\nwork, we present a novel GNN-based architecture for learning latent\nrepresentations of molecular structure. When trained end-to-end with a\ndiffusion model for de novo ligand design, our model achieves comparable\nperformance to one with an all-atom protein representation while exhibiting a\n3-fold reduction in inference time.\n", "date": "2023-11-22", "categories": []}
{"id": "2308.01323", "title": "Evaluation of network-guided random forest for disease gene discovery", "abstract": "  Gene network information is believed to be beneficial for disease module and\npathway identification, but has not been explicitly utilized in the standard\nrandom forest (RF) algorithm for gene expression data analysis. We investigate\nthe performance of a network-guided RF where the network information is\nsummarized into a sampling probability of predictor variables which is further\nused in the construction of the RF. Our results suggest that network-guided RF\ndoes not provide better disease prediction than the standard RF. In terms of\ndisease gene discovery, if disease genes form module(s), network-guided RF\nidentifies them more accurately. In addition, when disease status is\nindependent from genes in the given network, spurious gene selection results\ncan occur when using network information, especially on hub genes. Our\nempirical analysis on two balanced microarray and RNA-Seq breast cancer\ndatasets from The Cancer Genome Atlas (TCGA) for classification of progesterone\nreceptor (PR) status also demonstrates that network-guided RF can identify\ngenes from PGR-related pathways, which leads to a better connected module of\nidentified genes.\n", "date": "2023-08-02", "categories": []}
{"id": "2404.17626", "title": "Using Pre-training and Interaction Modeling for ancestry-specific\n  disease prediction in UK Biobank", "abstract": "  Recent genome-wide association studies (GWAS) have uncovered the genetic\nbasis of complex traits, but show an under-representation of non-European\ndescent individuals, underscoring a critical gap in genetic research. Here, we\nassess whether we can improve disease prediction across diverse ancestries\nusing multiomic data. We evaluate the performance of Group-LASSO\nINTERaction-NET (glinternet) and pretrained lasso in disease prediction\nfocusing on diverse ancestries in the UK Biobank. Models were trained on data\nfrom White British and other ancestries and validated across a cohort of over\n96,000 individuals for 8 diseases. Out of 96 models trained, we report 16 with\nstatistically significant incremental predictive performance in terms of\nROC-AUC scores (p-value < 0.05), found for diabetes, arthritis, gall stones,\ncystitis, asthma and osteoarthritis. For the interaction and pretrained models\nthat outperformed the baseline, the PRS score was the primary driver behind\nprediction. Our findings indicate that both interaction terms and pre-training\ncan enhance prediction accuracy but for a limited set of diseases and moderate\nimprovements in accuracy\n", "date": "2024-04-26", "categories": []}
{"id": "2405.03726", "title": "sc-OTGM: Single-Cell Perturbation Modeling by Solving Optimal Mass\n  Transport on the Manifold of Gaussian Mixtures", "abstract": "  Influenced by breakthroughs in LLMs, single-cell foundation models are\nemerging. While these models show successful performance in cell type\nclustering, phenotype classification, and gene perturbation response\nprediction, it remains to be seen if a simpler model could achieve comparable\nor better results, especially with limited data. This is important, as the\nquantity and quality of single-cell data typically fall short of the standards\nin textual data used for training LLMs. Single-cell sequencing often suffers\nfrom technical artifacts, dropout events, and batch effects. These challenges\nare compounded in a weakly supervised setting, where the labels of cell states\ncan be noisy, further complicating the analysis. To tackle these challenges, we\npresent sc-OTGM, streamlined with less than 500K parameters, making it\napproximately 100x more compact than the foundation models, offering an\nefficient alternative. sc-OTGM is an unsupervised model grounded in the\ninductive bias that the scRNAseq data can be generated from a combination of\nthe finite multivariate Gaussian distributions. The core function of sc-OTGM is\nto create a probabilistic latent space utilizing a GMM as its prior\ndistribution and distinguish between distinct cell populations by learning\ntheir respective marginal PDFs. It uses a Hit-and-Run Markov chain sampler to\ndetermine the OT plan across these PDFs within the GMM framework. We evaluated\nour model against a CRISPR-mediated perturbation dataset, called CROP-seq,\nconsisting of 57 one-gene perturbations. Our results demonstrate that sc-OTGM\nis effective in cell state classification, aids in the analysis of differential\ngene expression, and ranks genes for target identification through a\nrecommender system. It also predicts the effects of single-gene perturbations\non downstream gene regulation and generates synthetic scRNA-seq data\nconditioned on specific cell states.\n", "date": "2024-05-06", "categories": []}
{"id": "2405.03879", "title": "Scalable Amortized GPLVMs for Single Cell Transcriptomics Data", "abstract": "  Dimensionality reduction is crucial for analyzing large-scale single-cell\nRNA-seq data. Gaussian Process Latent Variable Models (GPLVMs) offer an\ninterpretable dimensionality reduction method, but current scalable models lack\neffectiveness in clustering cell types. We introduce an improved model, the\namortized stochastic variational Bayesian GPLVM (BGPLVM), tailored for\nsingle-cell RNA-seq with specialized encoder, kernel, and likelihood designs.\nThis model matches the performance of the leading single-cell variational\ninference (scVI) approach on synthetic and real-world COVID datasets and\neffectively incorporates cell-cycle and batch information to reveal more\ninterpretable latent structures as we demonstrate on an innate immunity\ndataset.\n", "date": "2024-05-06", "categories": []}
{"id": "2405.04078", "title": "WISER: Weak supervISion and supErvised Representation learning to\n  improve drug response prediction in cancer", "abstract": "  Cancer, a leading cause of death globally, occurs due to genomic changes and\nmanifests heterogeneously across patients. To advance research on personalized\ntreatment strategies, the effectiveness of various drugs on cells derived from\ncancers (`cell lines') is experimentally determined in laboratory settings.\nNevertheless, variations in the distribution of genomic data and drug responses\nbetween cell lines and humans arise due to biological and environmental\ndifferences. Moreover, while genomic profiles of many cancer patients are\nreadily available, the scarcity of corresponding drug response data limits the\nability to train machine learning models that can predict drug response in\npatients effectively. Recent cancer drug response prediction methods have\nlargely followed the paradigm of unsupervised domain-invariant representation\nlearning followed by a downstream drug response classification step.\nIntroducing supervision in both stages is challenging due to heterogeneous\npatient response to drugs and limited drug response data. This paper addresses\nthese challenges through a novel representation learning method in the first\nphase and weak supervision in the second. Experimental results on real patient\ndata demonstrate the efficacy of our method (WISER) over state-of-the-art\nalternatives on predicting personalized drug response.\n", "date": "2024-05-07", "categories": []}
{"id": "2405.02354", "title": "Heterogeneous network and graph attention auto-encoder for\n  LncRNA-disease association prediction", "abstract": "  The emerging research shows that lncRNAs are associated with a series of\ncomplex human diseases. However, most of the existing methods have limitations\nin identifying nonlinear lncRNA-disease associations (LDAs), and it remains a\nhuge challenge to predict new LDAs. Therefore, the accurate identification of\nLDAs is very important for the warning and treatment of diseases. In this work,\nmultiple sources of biomedical data are fully utilized to construct\ncharacteristics of lncRNAs and diseases, and linear and nonlinear\ncharacteristics are effectively integrated. Furthermore, a novel deep learning\nmodel based on graph attention automatic encoder is proposed, called HGATELDA.\nTo begin with, the linear characteristics of lncRNAs and diseases are created\nby the miRNA-lncRNA interaction matrix and miRNA-disease interaction matrix.\nFollowing this, the nonlinear features of diseases and lncRNAs are extracted\nusing a graph attention auto-encoder, which largely retains the critical\ninformation and effectively aggregates the neighborhood information of nodes.\nIn the end, LDAs can be predicted by fusing the linear and nonlinear\ncharacteristics of diseases and lncRNA. The HGATELDA model achieves an\nimpressive AUC value of 0.9692 when evaluated using a 5-fold cross-validation\nindicating its superior performance in comparison to several recent prediction\nmodels. Meanwhile, the effectiveness of HGATELDA in identifying novel LDAs is\nfurther demonstrated by case studies. the HGATELDA model appears to be a viable\ncomputational model for predicting LDAs.\n", "date": "2024-05-03", "categories": []}
{"id": "2405.02534", "title": "A Multi-Domain Multi-Task Approach for Feature Selection from Bulk RNA\n  Datasets", "abstract": "  In this paper a multi-domain multi-task algorithm for feature selection in\nbulk RNAseq data is proposed. Two datasets are investigated arising from mouse\nhost immune response to Salmonella infection. Data is collected from several\nstrains of collaborative cross mice. Samples from the spleen and liver serve as\nthe two domains. Several machine learning experiments are conducted and the\nsmall subset of discriminative across domains features have been extracted in\neach case. The algorithm proves viable and underlines the benefits of across\ndomain feature selection by extracting new subset of discriminative features\nwhich couldn't be extracted only by one-domain approach.\n", "date": "2024-05-04", "categories": []}
{"id": "2402.09330", "title": "3D-based RNA function prediction tools in rnaglib", "abstract": "  Understanding the connection between complex structural features of RNA and\nbiological function is a fundamental challenge in evolutionary studies and in\nRNA design. However, building datasets of RNA 3D structures and making\nappropriate modeling choices remains time-consuming and lacks standardization.\nIn this chapter, we describe the use of rnaglib, to train supervised and\nunsupervised machine learning-based function prediction models on datasets of\nRNA 3D structures.\n", "date": "2024-02-14", "categories": []}
{"id": "2405.01616", "title": "Generative Active Learning for the Search of Small-molecule Protein\n  Binders", "abstract": "  Despite substantial progress in machine learning for scientific discovery in\nrecent years, truly de novo design of small molecules which exhibit a property\nof interest remains a significant challenge. We introduce LambdaZero, a\ngenerative active learning approach to search for synthesizable molecules.\nPowered by deep reinforcement learning, LambdaZero learns to search over the\nvast space of molecules to discover candidates with a desired property. We\napply LambdaZero with molecular docking to design novel small molecules that\ninhibit the enzyme soluble Epoxide Hydrolase 2 (sEH), while enforcing\nconstraints on synthesizability and drug-likeliness. LambdaZero provides an\nexponential speedup in terms of the number of calls to the expensive molecular\ndocking oracle, and LambdaZero de novo designed molecules reach docking scores\nthat would otherwise require the virtual screening of a hundred billion\nmolecules. Importantly, LambdaZero discovers novel scaffolds of synthesizable,\ndrug-like inhibitors for sEH. In in vitro experimental validation, a series of\nligands from a generated quinazoline-based scaffold were synthesized, and the\nlead inhibitor N-(4,6-di(pyrrolidin-1-yl)quinazolin-2-yl)-N-methylbenzamide\n(UM0152893) displayed sub-micromolar enzyme inhibition of sEH.\n", "date": "2024-05-02", "categories": []}
{"id": "2405.01983", "title": "Model-based reinforcement learning for protein backbone design", "abstract": "  Designing protein nanomaterials of predefined shape and characteristics has\nthe potential to dramatically impact the medical industry. Machine learning\n(ML) has proven successful in protein design, reducing the need for expensive\nwet lab experiment rounds. However, challenges persist in efficiently exploring\nthe protein fitness landscapes to identify optimal protein designs. In\nresponse, we propose the use of AlphaZero to generate protein backbones,\nmeeting shape and structural scoring requirements. We extend an existing Monte\nCarlo tree search (MCTS) framework by incorporating a novel threshold-based\nreward and secondary objectives to improve design precision. This innovation\nconsiderably outperforms existing approaches, leading to protein backbones that\nbetter respect structural scores. The application of AlphaZero is novel in the\ncontext of protein backbone design and demonstrates promising performance.\nAlphaZero consistently surpasses baseline MCTS by more than 100% in top-down\nprotein design tasks. Additionally, our application of AlphaZero with secondary\nobjectives uncovers further promising outcomes, indicating the potential of\nmodel-based reinforcement learning (RL) in navigating the intricate and nuanced\naspects of protein design\n", "date": "2024-05-03", "categories": []}
{"id": "2405.00751", "title": "F$^3$low: Frame-to-Frame Coarse-grained Molecular Dynamics with SE(3)\n  Guided Flow Matching", "abstract": "  Molecular dynamics (MD) is a crucial technique for simulating biological\nsystems, enabling the exploration of their dynamic nature and fostering an\nunderstanding of their functions and properties. To address exploration\ninefficiency, emerging enhanced sampling approaches like coarse-graining (CG)\nand generative models have been employed. In this work, we propose a\n\\underline{Frame-to-Frame} generative model with guided\n\\underline{Flow}-matching (F$3$low) for enhanced sampling, which (a) extends\nthe domain of CG modeling to the SE(3) Riemannian manifold; (b) retreating CGMD\nsimulations as autoregressively sampling guided by the former frame via\nflow-matching models; (c) targets the protein backbone, offering improved\ninsights into secondary structure formation and intricate folding pathways.\nCompared to previous methods, F$3$low allows for broader exploration of\nconformational space. The ability to rapidly generate diverse conformations via\nforce-free generative paradigm on SE(3) paves the way toward efficient enhanced\nsampling methods.\n", "date": "2024-05-01", "categories": []}
{"id": "2405.00753", "title": "HMAMP: Hypervolume-Driven Multi-Objective Antimicrobial Peptides Design", "abstract": "  Antimicrobial peptides (AMPs) have exhibited unprecedented potential as\nbiomaterials in combating multidrug-resistant bacteria. Despite the increasing\nadoption of artificial intelligence for novel AMP design, challenges pertaining\nto conflicting attributes such as activity, hemolysis, and toxicity have\nsignificantly impeded the progress of researchers. This paper introduces a\nparadigm shift by considering multiple attributes in AMP design.\n  Presented herein is a novel approach termed Hypervolume-driven\nMulti-objective Antimicrobial Peptide Design (HMAMP), which prioritizes the\nsimultaneous optimization of multiple attributes of AMPs. By synergizing\nreinforcement learning and a gradient descent algorithm rooted in the\nhypervolume maximization concept, HMAMP effectively expands exploration space\nand mitigates the issue of pattern collapse. This method generates a wide array\nof prospective AMP candidates that strike a balance among diverse attributes.\nFurthermore, we pinpoint knee points along the Pareto front of these candidate\nAMPs. Empirical results across five benchmark models substantiate that\nHMAMP-designed AMPs exhibit competitive performance and heightened diversity. A\ndetailed analysis of the helical structures and molecular dynamics simulations\nfor ten potential candidate AMPs validates the superiority of HMAMP in the\nrealm of multi-objective AMP design. The ability of HMAMP to systematically\ncraft AMPs considering multiple attributes marks a pioneering milestone,\nestablishing a universal computational framework for the multi-objective design\nof AMPs.\n", "date": "2024-05-01", "categories": []}
{"id": "2405.00070", "title": "Bayesian-Guided Generation of Synthetic Microbiomes with Minimized\n  Pathogenicity", "abstract": "  Synthetic microbiomes offer new possibilities for modulating microbiota, to\naddress the barriers in multidtug resistance (MDR) research. We present a\nBayesian optimization approach to enable efficient searching over the space of\nsynthetic microbiome variants to identify candidates predictive of reduced MDR.\nMicrobiome datasets were encoded into a low-dimensional latent space using\nautoencoders. Sampling from this space allowed generation of synthetic\nmicrobiome signatures. Bayesian optimization was then implemented to select\nvariants for biological screening to maximize identification of designs with\nrestricted MDR pathogens based on minimal samples. Four acquisition functions\nwere evaluated: expected improvement, upper confidence bound, Thompson\nsampling, and probability of improvement. Based on each strategy, synthetic\nsamples were prioritized according to their MDR detection. Expected\nimprovement, upper confidence bound, and probability of improvement\nconsistently produced synthetic microbiome candidates with significantly fewer\nsearches than Thompson sampling. By combining deep latent space mapping and\nBayesian learning for efficient guided screening, this study demonstrated the\nfeasibility of creating bespoke synthetic microbiomes with customized MDR\nprofiles.\n", "date": "2024-04-29", "categories": []}
{"id": "2404.18960", "title": "Leak Proof CMap; a framework for training and evaluation of cell line\n  agnostic L1000 similarity methods", "abstract": "  The Connectivity Map (CMap) is a large publicly available database of\ncellular transcriptomic responses to chemical and genetic perturbations built\nusing a standardized acquisition protocol known as the L1000 technique.\nDatabases such as CMap provide an exciting opportunity to enrich drug discovery\nefforts, providing a 'known' phenotypic landscape to explore and enabling the\ndevelopment of state of the art techniques for enhanced information extraction\nand better informed decisions. Whilst multiple methods for measuring phenotypic\nsimilarity and interrogating profiles have been developed, the field is\nseverely lacking standardized benchmarks using appropriate data splitting for\ntraining and unbiased evaluation of machine learning methods. To address this,\nwe have developed 'Leak Proof CMap' and exemplified its application to a set of\ncommon transcriptomic and generic phenotypic similarity methods along with an\nexemplar triplet loss-based method. Benchmarking in three critical performance\nareas (compactness, distinctness, and uniqueness) is conducted using carefully\ncrafted data splits ensuring no similar cell lines or treatments with shared or\nclosely matching responses or mechanisms of action are present in training,\nvalidation, or test sets. This enables testing of models with unseen samples\nakin to exploring treatments with novel modes of action in novel patient\nderived cell lines. With a carefully crafted benchmark and data splitting\nregime in place, the tooling now exists to create performant phenotypic\nsimilarity methods for use in personalized medicine (novel cell lines) and to\nbetter augment high throughput phenotypic screening technologies with the L1000\ntranscriptomic technology.\n", "date": "2024-04-29", "categories": []}
{"id": "2404.17601", "title": "Nested Inheritance Dynamics", "abstract": "  The idea of the inheritance of biological processes, such as the\ndevelopmental process or the life cycle of an organism, has been discussed in\nthe biology literature, but formal mathematical descriptions and plausible data\nanalysis frameworks are lacking. We introduce an extension of the nested\nDirichlet Process (nDP) to a multiscale model to aid in understanding the\nmechanisms by which biological processes are inherited, remain stable, and are\nmodified across generations. To address these issues, we introduce Nested\nInheritance Dynamics Algorithm (NIDA). At its primary level, NIDA encompasses\nall processes unfolding within an individual organism's lifespan. The secondary\nlevel delineates the dynamics through which these processes evolve or remain\nstable over time. This framework allows for the specification of a physical\nsystem model at either scale, thus promoting seamless integration with\nestablished models of development and heredity.\n", "date": "2024-04-23", "categories": []}
{"id": "2404.16907", "title": "Season combinatorial intervention predictions with Salt & Peper", "abstract": "  Interventions play a pivotal role in the study of complex biological systems.\nIn drug discovery, genetic interventions (such as CRISPR base editing) have\nbecome central to both identifying potential therapeutic targets and\nunderstanding a drug's mechanism of action. With the advancement of CRISPR and\nthe proliferation of genome-scale analyses such as transcriptomics, a new\nchallenge is to navigate the vast combinatorial space of concurrent genetic\ninterventions. Addressing this, our work concentrates on estimating the effects\nof pairwise genetic combinations on the cellular transcriptome. We introduce\ntwo novel contributions: Salt, a biologically-inspired baseline that posits the\nmostly additive nature of combination effects, and Peper, a deep learning model\nthat extends Salt's additive assumption to achieve unprecedented accuracy. Our\ncomprehensive comparison against existing state-of-the-art methods, grounded in\ndiverse metrics, and our out-of-distribution analysis highlight the limitations\nof current models in realistic settings. This analysis underscores the\nnecessity for improved modelling techniques and data acquisition strategies,\npaving the way for more effective exploration of genetic intervention effects.\n", "date": "2024-04-25", "categories": []}
{"id": "2403.00024", "title": "FlowCyt: A Comparative Study of Deep Learning Approaches for Multi-Class\n  Classification in Flow Cytometry Benchmarking", "abstract": "  This paper presents FlowCyt, the first comprehensive benchmark for\nmulti-class single-cell classification in flow cytometry data. The dataset\ncomprises bone marrow samples from 30 patients, with each cell characterized by\ntwelve markers. Ground truth labels identify five hematological cell types: T\nlymphocytes, B lymphocytes, Monocytes, Mast cells, and Hematopoietic\nStem/Progenitor Cells (HSPCs). Experiments utilize supervised inductive\nlearning and semi-supervised transductive learning on up to 1 million cells per\npatient. Baseline methods include Gaussian Mixture Models, XGBoost, Random\nForests, Deep Neural Networks, and Graph Neural Networks (GNNs). GNNs\ndemonstrate superior performance by exploiting spatial relationships in\ngraph-encoded data. The benchmark allows standardized evaluation of clinically\nrelevant classification tasks, along with exploratory analyses to gain insights\ninto hematological cell phenotypes. This represents the first public flow\ncytometry benchmark with a richly annotated, heterogeneous dataset. It will\nempower the development and rigorous assessment of novel methodologies for\nsingle-cell analysis.\n", "date": "2024-02-28", "categories": []}
{"id": "2403.07592", "title": "Accurate Spatial Gene Expression Prediction by integrating\n  Multi-resolution features", "abstract": "  Recent advancements in Spatial Transcriptomics (ST) technology have\nfacilitated detailed gene expression analysis within tissue contexts. However,\nthe high costs and methodological limitations of ST necessitate a more robust\npredictive model. In response, this paper introduces TRIPLEX, a novel deep\nlearning framework designed to predict spatial gene expression from Whole Slide\nImages (WSIs). TRIPLEX uniquely harnesses multi-resolution features, capturing\ncellular morphology at individual spots, the local context around these spots,\nand the global tissue organization. By integrating these features through an\neffective fusion strategy, TRIPLEX achieves accurate gene expression\nprediction. Our comprehensive benchmark study, conducted on three public ST\ndatasets and supplemented with Visium data from 10X Genomics, demonstrates that\nTRIPLEX outperforms current state-of-the-art models in Mean Squared Error\n(MSE), Mean Absolute Error (MAE), and Pearson Correlation Coefficient (PCC).\nThe model's predictions align closely with ground truth gene expression\nprofiles and tumor annotations, underscoring TRIPLEX's potential in advancing\ncancer diagnosis and treatment.\n", "date": "2024-03-12", "categories": []}
{"id": "2404.14986", "title": "$\\texttt{MiniMol}$: A Parameter-Efficient Foundation Model for Molecular\n  Learning", "abstract": "  In biological tasks, data is rarely plentiful as it is generated from\nhard-to-gather measurements. Therefore, pre-training foundation models on large\nquantities of available data and then transfer to low-data downstream tasks is\na promising direction. However, how to design effective foundation models for\nmolecular learning remains an open question, with existing approaches typically\nfocusing on models with large parameter capacities. In this work, we propose\n$\\texttt{MiniMol}$, a foundational model for molecular learning with 10 million\nparameters. $\\texttt{MiniMol}$ is pre-trained on a mix of roughly 3300 sparsely\ndefined graph- and node-level tasks of both quantum and biological nature. The\npre-training dataset includes approximately 6 million molecules and 500 million\nlabels. To demonstrate the generalizability of $\\texttt{MiniMol}$ across tasks,\nwe evaluate it on downstream tasks from the Therapeutic Data Commons (TDC)\nADMET group showing significant improvements over the prior state-of-the-art\nfoundation model across 17 tasks. $\\texttt{MiniMol}$ will be a public and\nopen-sourced model for future research.\n", "date": "2024-04-23", "categories": []}
{"id": "2307.14367", "title": "Prot2Text: Multimodal Protein's Function Generation with GNNs and\n  Transformers", "abstract": "  In recent years, significant progress has been made in the field of protein\nfunction prediction with the development of various machine-learning\napproaches. However, most existing methods formulate the task as a\nmulti-classification problem, i.e. assigning predefined labels to proteins. In\nthis work, we propose a novel approach, Prot2Text, which predicts a protein's\nfunction in a free text style, moving beyond the conventional binary or\ncategorical classifications. By combining Graph Neural Networks(GNNs) and Large\nLanguage Models(LLMs), in an encoder-decoder framework, our model effectively\nintegrates diverse data types including protein sequence, structure, and\ntextual annotation and description. This multimodal approach allows for a\nholistic representation of proteins' functions, enabling the generation of\ndetailed and accurate functional descriptions. To evaluate our model, we\nextracted a multimodal protein dataset from SwissProt, and demonstrate\nempirically the effectiveness of Prot2Text. These results highlight the\ntransformative impact of multimodal models, specifically the fusion of GNNs and\nLLMs, empowering researchers with powerful tools for more accurate function\nprediction of existing as well as first-to-see proteins.\n", "date": "2023-07-25", "categories": []}
{"id": "2404.13265", "title": "F5C-finder: An Explainable and Ensemble Biological Language Model for\n  Predicting 5-Formylcytidine Modifications on mRNA", "abstract": "  As a prevalent and dynamically regulated epigenetic modification,\n5-formylcytidine (f5C) is crucial in various biological processes. However,\ntraditional experimental methods for f5C detection are often laborious and\ntime-consuming, limiting their ability to map f5C sites across the\ntranscriptome comprehensively. While computational approaches offer a\ncost-effective and high-throughput alternative, no recognition model for f5C\nhas been developed to date. Drawing inspiration from language models in natural\nlanguage processing, this study presents f5C-finder, an ensemble neural\nnetwork-based model utilizing multi-head attention for the identification of\nf5C. Five distinct feature extraction methods were employed to construct five\nindividual artificial neural networks, and these networks were subsequently\nintegrated through ensemble learning to create f5C-finder. 10-fold\ncross-validation and independent tests demonstrate that f5C-finder achieves\nstate-of-the-art (SOTA) performance with AUC of 0.807 and 0.827, respectively.\nThe result highlights the effectiveness of biological language model in\ncapturing both the order (sequential) and functional meaning (semantics) within\ngenomes. Furthermore, the built-in interpretability allows us to understand\nwhat the model is learning, creating a bridge between identifying key\nsequential elements and a deeper exploration of their biological functions.\n", "date": "2024-04-20", "categories": []}
{"id": "2404.10450", "title": "Graph Neural Networks for Protein-Protein Interactions -- A Short Survey", "abstract": "  Protein-protein interactions (PPIs) play key roles in a broad range of\nbiological processes. Numerous strategies have been proposed for predicting\nPPIs, and among them, graph-based methods have demonstrated promising outcomes\nowing to the inherent graph structure of PPI networks. This paper reviews\nvarious graph-based methodologies, and discusses their applications in PPI\nprediction. We classify these approaches into two primary groups based on their\nmodel structures. The first category employs Graph Neural Networks (GNN) or\nGraph Convolutional Networks (GCN), while the second category utilizes Graph\nAttention Networks (GAT), Graph Auto-Encoders and Graph-BERT. We highlight the\ndistinctive methodologies of each approach in managing the graph-structured\ndata inherent in PPI networks and anticipate future research directions in this\ndomain.\n", "date": "2024-04-16", "categories": []}
{"id": "2404.10573", "title": "AAVDiff: Experimental Validation of Enhanced Viability and Diversity in\n  Recombinant Adeno-Associated Virus (AAV) Capsids through Diffusion Generation", "abstract": "  Recombinant adeno-associated virus (rAAV) vectors have revolutionized gene\ntherapy, but their broad tropism and suboptimal transduction efficiency limit\ntheir clinical applications. To overcome these limitations, researchers have\nfocused on designing and screening capsid libraries to identify improved\nvectors. However, the large sequence space and limited resources present\nchallenges in identifying viable capsid variants. In this study, we propose an\nend-to-end diffusion model to generate capsid sequences with enhanced\nviability. Using publicly available AAV2 data, we generated 38,000 diverse AAV2\nviral protein (VP) sequences, and evaluated 8,000 for viral selection. The\nresults attested the superiority of our model compared to traditional methods.\nAdditionally, in the absence of AAV9 capsid data, apart from one wild-type\nsequence, we used the same model to directly generate a number of viable\nsequences with up to 9 mutations. we transferred the remaining 30,000 samples\nto the AAV9 domain. Furthermore, we conducted mutagenesis on AAV9 VP\nhypervariable regions VI and V, contributing to the continuous improvement of\nthe AAV9 VP sequence. This research represents a significant advancement in the\ndesign and functional validation of rAAV vectors, offering innovative solutions\nto enhance specificity and transduction efficiency in gene therapy\napplications.\n", "date": "2024-04-16", "categories": []}
{"id": "2402.06079", "title": "DiscDiff: Latent Diffusion Model for DNA Sequence Generation", "abstract": "  This paper introduces a novel framework for DNA sequence generation,\ncomprising two key components: DiscDiff, a Latent Diffusion Model (LDM)\ntailored for generating discrete DNA sequences, and Absorb-Escape, a\npost-training algorithm designed to refine these sequences. Absorb-Escape\nenhances the realism of the generated sequences by correcting `round errors'\ninherent in the conversion process between latent and input spaces. Our\napproach not only sets new standards in DNA sequence generation but also\ndemonstrates superior performance over existing diffusion models, in generating\nboth short and long DNA sequences. Additionally, we introduce EPD-GenDNA, the\nfirst comprehensive, multi-species dataset for DNA generation, encompassing\n160,000 unique sequences from 15 species. We hope this study will advance the\ngenerative modelling of DNA, with potential implications for gene therapy and\nprotein production.\n", "date": "2024-02-08", "categories": []}
{"id": "2404.10094", "title": "Towards DNA-Encoded Library Generation with GFlowNets", "abstract": "  DNA-encoded libraries (DELs) are a powerful approach for rapidly screening\nlarge numbers of diverse compounds. One of the key challenges in using DELs is\nlibrary design, which involves choosing the building blocks that will be\ncombinatorially combined to produce the final library. In this paper we\nconsider the task of protein-protein interaction (PPI) biased DEL design. To\nthis end, we evaluate several machine learning algorithms on the PPI modulation\ntask and use them as a reward for the proposed GFlowNet-based generative\napproach. We additionally investigate the possibility of using structural\ninformation about building blocks to design a hierarchical action space for the\nGFlowNet. The observed results indicate that GFlowNets are a promising approach\nfor generating diverse combinatorial library candidates.\n", "date": "2024-04-15", "categories": []}
{"id": "2404.10561", "title": "HiGraphDTI: Hierarchical Graph Representation Learning for Drug-Target\n  Interaction Prediction", "abstract": "  The discovery of drug-target interactions (DTIs) plays a crucial role in\npharmaceutical development. The deep learning model achieves more accurate\nresults in DTI prediction due to its ability to extract robust and expressive\nfeatures from drug and target chemical structures. However, existing deep\nlearning methods typically generate drug features via aggregating molecular\natom representations, ignoring the chemical properties carried by motifs, i.e.,\nsubstructures of the molecular graph. The atom-drug double-level molecular\nrepresentation learning can not fully exploit structure information and fails\nto interpret the DTI mechanism from the motif perspective. In addition,\nsequential model-based target feature extraction either fuses limited\ncontextual information or requires expensive computational resources. To tackle\nthe above issues, we propose a hierarchical graph representation learning-based\nDTI prediction method (HiGraphDTI). Specifically, HiGraphDTI learns\nhierarchical drug representations from triple-level molecular graphs to\nthoroughly exploit chemical information embedded in atoms, motifs, and\nmolecules. Then, an attentional feature fusion module incorporates information\nfrom different receptive fields to extract expressive target features.Last, the\nhierarchical attention mechanism identifies crucial molecular segments, which\noffers complementary views for interpreting interaction mechanisms. The\nexperiment results not only demonstrate the superiority of HiGraphDTI to the\nstate-of-the-art methods, but also confirm the practical ability of our model\nin interaction interpretation and new DTI discovery.\n", "date": "2024-04-16", "categories": []}
{"id": "2306.01745", "title": "Biomarker Discovery with Quantum Neural Networks: A Case-study in\n  CTLA4-Activation Pathways", "abstract": "  Biomarker discovery is a challenging task due to the massive search space.\nQuantum computing and quantum Artificial Intelligence (quantum AI) can be used\nto address the computational problem of biomarker discovery tasks. We propose a\nQuantum Neural Networks (QNNs) architecture to discover biomarkers for input\nactivation pathways. The Maximum Relevance, Minimum Redundancy (mRMR) criteria\nis used to score biomarker candidate sets. Our proposed model is economical\nsince the neural solution can be delivered on constrained hardware. We\ndemonstrate the proof of concept on four activation pathways associated with\nCTLA4, including (1) CTLA4-activation stand-alone, (2) CTLA4-CD8A-CD8B\nco-activation, (3) CTLA4-CD2 co-activation, and (4)\nCTLA4-CD2-CD48-CD53-CD58-CD84 co-activation. The model indicates new biomarkers\nassociated with the mutational activation of CLTA4-associated pathways,\nincluding 20 genes: CLIC4, CPE, ETS2, FAM107A, GPR116, HYOU1, LCN2, MACF1,\nMT1G, NAPA, NDUFS5, PAK1, PFN1, PGAP3, PPM1G, PSMD8, RNF213, SLC25A3, UBA1, and\nWLS. We open source the implementation at:\nhttps://github.com/namnguyen0510/Biomarker-Discovery-with-Quantum-Neural-Networks.\n", "date": "2023-05-15", "categories": []}
{"id": "2307.08509", "title": "Kernel-Based Testing for Single-Cell Differential Analysis", "abstract": "  Single-cell technologies offer insights into molecular feature distributions,\nbut comparing them poses challenges. We propose a kernel-testing framework for\nnon-linear cell-wise distribution comparison, analyzing gene expression and\nepigenomic modifications. Our method allows feature-wise and global\ntranscriptome/epigenome comparisons, revealing cell population heterogeneities.\nUsing a classifier based on embedding variability, we identify transitions in\ncell states, overcoming limitations of traditional single-cell analysis.\nApplied to single-cell ChIP-Seq data, our approach identifies untreated breast\ncancer cells with an epigenomic profile resembling persister cells. This\ndemonstrates the effectiveness of kernel testing in uncovering subtle\npopulation variations that might be missed by other methods.\n", "date": "2023-07-17", "categories": []}
{"id": "2308.04526", "title": "Large-Scale Multi-Hypotheses Cell Tracking Using Ultrametric Contours\n  Maps", "abstract": "  In this work, we describe a method for large-scale 3D cell-tracking through a\nsegmentation selection approach. The proposed method is effective at tracking\ncells across large microscopy datasets on two fronts: (i) It can solve problems\ncontaining millions of segmentation instances in terabyte-scale 3D+t datasets;\n(ii) It achieves competitive results with or without deep learning, which\nrequires 3D annotated data, that is scarce in the fluorescence microscopy\nfield. The proposed method computes cell tracks and segments using a hierarchy\nof segmentation hypotheses and selects disjoint segments by maximizing the\noverlap between adjacent frames. We show that this method achieves\nstate-of-the-art results in 3D images from the cell tracking challenge and has\na faster integer linear programming formulation. Moreover, our framework is\nflexible and supports segmentations from off-the-shelf cell segmentation models\nand can combine them into an ensemble that improves tracking. The code is\navailable https://github.com/royerlab/ultrack.\n", "date": "2023-08-08", "categories": []}
{"id": "2310.02391", "title": "SE(3)-Stochastic Flow Matching for Protein Backbone Generation", "abstract": "  The computational design of novel protein structures has the potential to\nimpact numerous scientific disciplines greatly. Toward this goal, we introduce\nFoldFlow, a series of novel generative models of increasing modeling power\nbased on the flow-matching paradigm over $3\\mathrm{D}$ rigid motions -- i.e.\nthe group $\\text{SE}(3)$ -- enabling accurate modeling of protein backbones. We\nfirst introduce FoldFlow-Base, a simulation-free approach to learning\ndeterministic continuous-time dynamics and matching invariant target\ndistributions on $\\text{SE}(3)$. We next accelerate training by incorporating\nRiemannian optimal transport to create FoldFlow-OT, leading to the construction\nof both more simple and stable flows. Finally, we design FoldFlow-SFM, coupling\nboth Riemannian OT and simulation-free training to learn stochastic\ncontinuous-time dynamics over $\\text{SE}(3)$. Our family of FoldFlow,\ngenerative models offers several key advantages over previous approaches to the\ngenerative modeling of proteins: they are more stable and faster to train than\ndiffusion-based approaches, and our models enjoy the ability to map any\ninvariant source distribution to any invariant target distribution over\n$\\text{SE}(3)$. Empirically, we validate FoldFlow, on protein backbone\ngeneration of up to $300$ amino acids leading to high-quality designable,\ndiverse, and novel samples.\n", "date": "2023-10-03", "categories": []}
{"id": "2404.06481", "title": "GeoDirDock: Guiding Docking Along Geodesic Paths", "abstract": "  This work introduces GeoDirDock (GDD), a novel approach to molecular docking\nthat enhances the accuracy and physical plausibility of ligand docking\npredictions. GDD guides the denoising process of a diffusion model along\ngeodesic paths within multiple spaces representing translational, rotational,\nand torsional degrees of freedom. Our method leverages expert knowledge to\ndirect the generative modeling process, specifically targeting desired\nprotein-ligand interaction regions. We demonstrate that GDD significantly\noutperforms existing blind docking methods in terms of RMSD accuracy and\nphysicochemical pose realism. Our results indicate that incorporating domain\nexpertise into the diffusion process leads to more biologically relevant\ndocking predictions. Additionally, we explore the potential of GDD for lead\noptimization in drug discovery through angle transfer in maximal common\nsubstructure (MCS) docking, showcasing its capability to predict ligand\norientations for chemically similar compounds accurately.\n", "date": "2024-04-09", "categories": []}
{"id": "2404.04299", "title": "GENEVIC: GENetic data Exploration and Visualization via Intelligent\n  interactive Console", "abstract": "  Summary: The vast generation of genetic data poses a significant challenge in\nefficiently uncovering valuable knowledge. Introducing GENEVIC, an AI-driven\nchat framework that tackles this challenge by bridging the gap between genetic\ndata generation and biomedical knowledge discovery. Leveraging generative AI,\nnotably ChatGPT, it serves as a biologist's 'copilot'. It automates the\nanalysis, retrieval, and visualization of customized domain-specific genetic\ninformation, and integrates functionalities to generate protein interaction\nnetworks, enrich gene sets, and search scientific literature from PubMed,\nGoogle Scholar, and arXiv, making it a comprehensive tool for biomedical\nresearch. In its pilot phase, GENEVIC is assessed using a curated database that\nranks genetic variants associated with Alzheimer's disease, schizophrenia, and\ncognition, based on their effect weights from the Polygenic Score Catalog, thus\nenabling researchers to prioritize genetic variants in complex diseases.\nGENEVIC's operation is user-friendly, accessible without any specialized\ntraining, secured by Azure OpenAI's HIPAA-compliant infrastructure, and\nevaluated for its efficacy through real-time query testing. As a prototype,\nGENEVIC is set to advance genetic research, enabling informed biomedical\ndecisions.\n  Availability and implementation: GENEVIC is publicly accessible at\nhttps://genevic-anath2024.streamlit.app. The underlying code is open-source and\navailable via GitHub at https://github.com/anath2110/GENEVIC.git.\n", "date": "2024-04-04", "categories": []}
{"id": "2404.04547", "title": "Exhaustive Exploitation of Nature-inspired Computation for Cancer\n  Screening in an Ensemble Manner", "abstract": "  Accurate screening of cancer types is crucial for effective cancer detection\nand precise treatment selection. However, the association between gene\nexpression profiles and tumors is often limited to a small number of biomarker\ngenes. While computational methods using nature-inspired algorithms have shown\npromise in selecting predictive genes, existing techniques are limited by\ninefficient search and poor generalization across diverse datasets. This study\npresents a framework termed Evolutionary Optimized Diverse Ensemble Learning\n(EODE) to improve ensemble learning for cancer classification from gene\nexpression data. The EODE methodology combines an intelligent grey wolf\noptimization algorithm for selective feature space reduction, guided random\ninjection modeling for ensemble diversity enhancement, and subset model\noptimization for synergistic classifier combinations. Extensive experiments\nwere conducted across 35 gene expression benchmark datasets encompassing varied\ncancer types. Results demonstrated that EODE obtained significantly improved\nscreening accuracy over individual and conventionally aggregated models. The\nintegrated optimization of advanced feature selection, directed specialized\nmodeling, and cooperative classifier ensembles helps address key challenges in\ncurrent nature-inspired approaches. This provides an effective framework for\nrobust and generalized ensemble learning with gene expression biomarkers.\nSpecifically, we have opened EODE source code on Github at\nhttps://github.com/wangxb96/EODE.\n", "date": "2024-04-06", "categories": []}
{"id": "2402.11950", "title": "A novel molecule generative model of VAE combined with Transformer for\n  unseen structure generation", "abstract": "  Recently, molecule generation using deep learning has been actively\ninvestigated in drug discovery. In this field, Transformer and VAE are widely\nused as powerful models, but they are rarely used in combination due to\nstructural and performance mismatch of them. This study proposes a model that\ncombines these two models through structural and parameter optimization in\nhandling diverse molecules. The proposed model shows comparable performance to\nexisting models in generating molecules, and showed by far superior performance\nin generating molecules with unseen structures. Another advantage of this VAE\nmodel is that it generates molecules from latent representation, and therefore\nproperties of molecules can be easily predicted or conditioned with it, and\nindeed, we show that the latent representation of the model successfully\npredicts molecular properties. Ablation study suggested the advantage of VAE\nover other generative models like language model in generating novel molecules.\nIt also indicated that the latent representation can be shortened to ~32\ndimensional variables without loss of reconstruction, suggesting the\npossibility of a much smaller molecular descriptor or model than existing ones.\nThis study is expected to provide a virtual chemical library containing a wide\nvariety of compounds for virtual screening and to enable efficient screening.\n", "date": "2024-02-19", "categories": []}
{"id": "2404.01693", "title": "HeMeNet: Heterogeneous Multichannel Equivariant Network for Protein\n  Multitask Learning", "abstract": "  Understanding and leveraging the 3D structures of proteins is central to a\nvariety of biological and drug discovery tasks. While deep learning has been\napplied successfully for structure-based protein function prediction tasks,\ncurrent methods usually employ distinct training for each task. However, each\nof the tasks is of small size, and such a single-task strategy hinders the\nmodels' performance and generalization ability. As some labeled 3D protein\ndatasets are biologically related, combining multi-source datasets for\nlarger-scale multi-task learning is one way to overcome this problem. In this\npaper, we propose a neural network model to address multiple tasks jointly upon\nthe input of 3D protein structures. In particular, we first construct a\nstandard structure-based multi-task benchmark called Protein-MT, consisting of\n6 biologically relevant tasks, including affinity prediction and property\nprediction, integrated from 4 public datasets. Then, we develop a novel graph\nneural network for multi-task learning, dubbed Heterogeneous Multichannel\nEquivariant Network (HeMeNet), which is E(3) equivariant and able to capture\nheterogeneous relationships between different atoms. Besides, HeMeNet can\nachieve task-specific learning via the task-aware readout mechanism. Extensive\nevaluations on our benchmark verify the effectiveness of multi-task learning,\nand our model generally surpasses state-of-the-art models.\n", "date": "2024-04-02", "categories": []}
{"id": "2403.20109", "title": "Mol-AIR: Molecular Reinforcement Learning with Adaptive Intrinsic\n  Rewards for Goal-directed Molecular Generation", "abstract": "  Optimizing techniques for discovering molecular structures with desired\nproperties is crucial in artificial intelligence(AI)-based drug discovery.\nCombining deep generative models with reinforcement learning has emerged as an\neffective strategy for generating molecules with specific properties. Despite\nits potential, this approach is ineffective in exploring the vast chemical\nspace and optimizing particular chemical properties. To overcome these\nlimitations, we present Mol-AIR, a reinforcement learning-based framework using\nadaptive intrinsic rewards for effective goal-directed molecular generation.\nMol-AIR leverages the strengths of both history-based and learning-based\nintrinsic rewards by exploiting random distillation network and counting-based\nstrategies. In benchmark tests, Mol-AIR demonstrates superior performance over\nexisting approaches in generating molecules with desired properties without any\nprior knowledge, including penalized LogP, QED, and celecoxib similarity. We\nbelieve that Mol-AIR represents a significant advancement in drug discovery,\noffering a more efficient path to discovering novel therapeutics.\n", "date": "2024-03-29", "categories": []}
{"id": "2310.00926", "title": "Integration of Graph Neural Network and Neural-ODEs for Tumor Dynamic\n  Prediction", "abstract": "  In anti-cancer drug development, a major scientific challenge is\ndisentangling the complex relationships between high-dimensional genomics data\nfrom patient tumor samples, the corresponding tumor's organ of origin, the drug\ntargets associated with given treatments and the resulting treatment response.\nFurthermore, to realize the aspirations of precision medicine in identifying\nand adjusting treatments for patients depending on the therapeutic response,\nthere is a need for building tumor dynamic models that can integrate both\nlongitudinal tumor size as well as multimodal, high-content data. In this work,\nwe take a step towards enhancing personalized tumor dynamic predictions by\nproposing a heterogeneous graph encoder that utilizes a bipartite Graph\nConvolutional Neural network (GCN) combined with Neural Ordinary Differential\nEquations (Neural-ODEs). We applied the methodology to a large collection of\npatient-derived xenograft (PDX) data, spanning a wide variety of treatments (as\nwell as their combinations) on tumors that originated from a number of\ndifferent organs. We first show that the methodology is able to discover a\ntumor dynamic model that significantly improves upon an empirical model which\nis in current use. Additionally, we show that the graph encoder is able to\neffectively utilize multimodal data to enhance tumor predictions. Our findings\nindicate that the methodology holds significant promise and offers potential\napplications in pre-clinical settings.\n", "date": "2023-10-02", "categories": []}
{"id": "2402.12190", "title": "Towards AI-Based Precision Oncology: A Machine Learning Framework for\n  Personalized Counterfactual Treatment Suggestions based on Multi-Omics Data", "abstract": "  AI-driven precision oncology has the transformative potential to reshape\ncancer treatment by leveraging the power of AI models to analyze the\ninteraction between complex patient characteristics and their corresponding\ntreatment outcomes. New technological platforms have facilitated the timely\nacquisition of multimodal data on tumor biology at an unprecedented resolution,\nsuch as single-cell multi-omics data, making this quality and quantity of data\navailable for data-driven improved clinical decision-making. In this work, we\npropose a modular machine learning framework designed for personalized\ncounterfactual cancer treatment suggestions based on an ensemble of machine\nlearning experts trained on diverse multi-omics technologies. These specialized\ncounterfactual experts per technology are consistently aggregated into a more\npowerful expert with superior performance and can provide both confidence and\nan explanation of its decision. The framework is tailored to address critical\nchallenges inherent in data-driven cancer research, including the\nhigh-dimensional nature of the data, and the presence of treatment assignment\nbias in the retrospective observational data. The framework is showcased\nthrough comprehensive demonstrations using data from in-vitro and in-vivo\ntreatment responses from a cohort of patients with ovarian cancer. Our method\naims to empower clinicians with a reality-centric decision-support tool\nincluding probabilistic treatment suggestions with calibrated confidence and\npersonalized explanations for tailoring treatment strategies to multi-omics\ncharacteristics of individual cancer patients.\n", "date": "2024-02-19", "categories": []}
{"id": "2308.05115", "title": "PTransIPs: Identification of phosphorylation sites enhanced by protein\n  PLM embeddings", "abstract": "  Phosphorylation is pivotal in numerous fundamental cellular processes and\nplays a significant role in the onset and progression of various diseases. The\naccurate identification of these phosphorylation sites is crucial for\nunraveling the molecular mechanisms within cells and during viral infections,\npotentially leading to the discovery of novel therapeutic targets. In this\nstudy, we develop PTransIPs, a new deep learning framework for the\nidentification of phosphorylation sites. Independent testing results\ndemonstrate that PTransIPs outperforms existing state-of-the-art (SOTA)\nmethods, achieving AUCs of 0.9232 and 0.9660 for the identification of\nphosphorylated S/T and Y sites, respectively. PTransIPs contributes from three\naspects. 1) PTransIPs is the first to apply protein pre-trained language model\n(PLM) embeddings to this task. It utilizes ProtTrans and EMBER2 to extract\nsequence and structure embeddings, respectively, as additional inputs into the\nmodel, effectively addressing issues of dataset size and overfitting, thus\nenhancing model performance; 2) PTransIPs is based on Transformer architecture,\noptimized through the integration of convolutional neural networks and TIM loss\nfunction, providing practical insights for model design and training; 3) The\nencoding of amino acids in PTransIPs enables it to serve as a universal\nframework for other peptide bioactivity tasks, with its excellent performance\nshown in extended experiments of this paper. Our code, data and models are\npublicly available at https://github.com/StatXzy7/PTransIPs.\n", "date": "2023-08-08", "categories": []}
{"id": "2403.14736", "title": "NaNa and MiGu: Semantic Data Augmentation Techniques to Enhance Protein\n  Classification in Graph Neural Networks", "abstract": "  Protein classification tasks are essential in drug discovery. Real-world\nprotein structures are dynamic, which will determine the properties of\nproteins. However, the existing machine learning methods, like ProNet (Wang et\nal., 2022a), only access limited conformational characteristics and protein\nside-chain features, leading to impractical protein structure and inaccuracy of\nprotein classes in their predictions. In this paper, we propose novel semantic\ndata augmentation methods, Novel Augmentation of New Node Attributes (NaNa),\nand Molecular Interactions and Geometric Upgrading (MiGu) to incorporate\nbackbone chemical and side-chain biophysical information into protein\nclassification tasks and a co-embedding residual learning framework.\nSpecifically, we leverage molecular biophysical, secondary structure, chemical\nbonds, and ionic features of proteins to facilitate protein classification\ntasks. Furthermore, our semantic augmentation methods and the co-embedding\nresidual learning framework can improve the performance of GIN (Xu et al.,\n2019) on EC and Fold datasets (Bairoch, 2000; Andreeva et al., 2007) by 16.41%\nand 11.33% respectively. Our code is available at\nhttps://github.com/r08b46009/Code_for_MIGU_NANA/tree/main.\n", "date": "2024-03-21", "categories": []}
{"id": "2403.17293", "title": "Tracing and segmentation of molecular patterns in 3-dimensional\n  cryo-et/em density maps through algorithmic image processing and deep\n  learning-based techniques", "abstract": "  Understanding the structures of biological macromolecules is highly important\nas they are closely associated with cellular functionalities. Comprehending the\nprecise organization actin filaments is crucial because they form the dynamic\ncytoskeleton, which offers structural support to cells and connects the cell's\ninterior with its surroundings. However, determining the precise organization\nof actin filaments is challenging due to the poor quality of cryo-electron\ntomography (cryo-ET) images, which suffer from low signal-to-noise (SNR) ratios\nand the presence of missing wedge, as well as diverse shape characteristics of\nactin filaments. To address these formidable challenges, the primary component\nof this dissertation focuses on developing sophisticated computational\ntechniques for tracing actin filaments. In particular, three novel\nmethodologies have been developed: i) BundleTrac, for tracing bundle-like actin\nfilaments found in Stereocilium, ii) Spaghetti Tracer, for tracing filaments\nthat move individually with loosely cohesive movements, and iii) Struwwel\nTracer, for tracing randomly orientated actin filaments in the actin network.\nThe second component of the dissertation introduces a convolutional neural\nnetwork (CNN) based segmentation model to determine the location of protein\nsecondary structures, such as helices and beta-sheets, in medium-resolution\n(5-10 Angstrom) 3-dimensional cryo-electron microscopy (cryo-EM) images. This\nmethodology later evolved into a tool named DeepSSETracer. The final component\nof the dissertation presents a novel algorithm, cylindrical fit measure, to\nestimate image structure match at helix regions in medium-resolution cryo-EM\nimages. Overall, my dissertation has made significant contributions to\naddressing critical research challenges in structural biology by introducing\nvarious computational methods and tools.\n", "date": "2024-03-26", "categories": []}
{"id": "2403.17889", "title": "Large scale paired antibody language models", "abstract": "  Antibodies are proteins produced by the immune system that can identify and\nneutralise a wide variety of antigens with high specificity and affinity, and\nconstitute the most successful class of biotherapeutics. With the advent of\nnext-generation sequencing, billions of antibody sequences have been collected\nin recent years, though their application in the design of better therapeutics\nhas been constrained by the sheer volume and complexity of the data. To address\nthis challenge, we present IgBert and IgT5, the best performing\nantibody-specific language models developed to date which can consistently\nhandle both paired and unpaired variable region sequences as input. These\nmodels are trained comprehensively using the more than two billion unpaired\nsequences and two million paired sequences of light and heavy chains present in\nthe Observed Antibody Space dataset. We show that our models outperform\nexisting antibody and protein language models on a diverse range of design and\nregression tasks relevant to antibody engineering. This advancement marks a\nsignificant leap forward in leveraging machine learning, large scale data sets\nand high-performance computing for enhancing antibody design for therapeutic\ndevelopment.\n", "date": "2024-03-26", "categories": []}
{"id": "2306.03111", "title": "Bootstrapped Training of Score-Conditioned Generator for Offline Design\n  of Biological Sequences", "abstract": "  We study the problem of optimizing biological sequences, e.g., proteins, DNA,\nand RNA, to maximize a black-box score function that is only evaluated in an\noffline dataset. We propose a novel solution, bootstrapped training of\nscore-conditioned generator (BootGen) algorithm. Our algorithm repeats a\ntwo-stage process. In the first stage, our algorithm trains the biological\nsequence generator with rank-based weights to enhance the accuracy of sequence\ngeneration based on high scores. The subsequent stage involves bootstrapping,\nwhich augments the training dataset with self-generated data labeled by a proxy\nscore function. Our key idea is to align the score-based generation with a\nproxy score function, which distills the knowledge of the proxy score function\nto the generator. After training, we aggregate samples from multiple\nbootstrapped generators and proxies to produce a diverse design. Extensive\nexperiments show that our method outperforms competitive baselines on\nbiological sequential design tasks. We provide reproducible source code:\n\\href{https://github.com/kaist-silab/bootgen}{https://github.com/kaist-silab/bootgen}.\n", "date": "2023-06-05", "categories": []}
{"id": "2310.08774", "title": "PhyloGFN: Phylogenetic inference with generative flow networks", "abstract": "  Phylogenetics is a branch of computational biology that studies the\nevolutionary relationships among biological entities. Its long history and\nnumerous applications notwithstanding, inference of phylogenetic trees from\nsequence data remains challenging: the high complexity of tree space poses a\nsignificant obstacle for the current combinatorial and probabilistic\ntechniques. In this paper, we adopt the framework of generative flow networks\n(GFlowNets) to tackle two core problems in phylogenetics: parsimony-based and\nBayesian phylogenetic inference. Because GFlowNets are well-suited for sampling\ncomplex combinatorial structures, they are a natural choice for exploring and\nsampling from the multimodal posterior distribution over tree topologies and\nevolutionary distances. We demonstrate that our amortized posterior sampler,\nPhyloGFN, produces diverse and high-quality evolutionary hypotheses on real\nbenchmark datasets. PhyloGFN is competitive with prior works in marginal\nlikelihood estimation and achieves a closer fit to the target distribution than\nstate-of-the-art variational inference methods. Our code is available at\nhttps://github.com/zmy1116/phylogfn.\n", "date": "2023-10-12", "categories": []}
{"id": "2403.15076", "title": "Comprehensive Lipidomic Automation Workflow using Large Language Models", "abstract": "  Lipidomics generates large data that makes manual annotation and\ninterpretation challenging. Lipid chemical and structural diversity with\nstructural isomers further complicates annotation. Although, several commercial\nand open-source software for targeted lipid identification exists, it lacks\nautomated method generation workflows and integration with statistical and\nbioinformatics tools. We have developed the Comprehensive Lipidomic Automated\nWorkflow (CLAW) platform with integrated workflow for parsing, detailed\nstatistical analysis and lipid annotations based on custom multiple reaction\nmonitoring (MRM) precursor and product ion pair transitions. CLAW contains\nseveral modules including identification of carbon-carbon double bond\nposition(s) in unsaturated lipids when combined with ozone electrospray\nionization (OzESI)-MRM methodology. To demonstrate the utility of the automated\nworkflow in CLAW, large-scale lipidomics data was collected with traditional\nand OzESI-MRM profiling on biological and non-biological samples. Specifically,\na total of 1497 transitions organized into 10 MRM-based mass spectrometry\nmethods were used to profile lipid droplets isolated from different brain\nregions of 18-24 month-old Alzheimer's disease mice and age-matched wild-type\ncontrols. Additionally, triacyclglycerols (TGs) profiles with carbon-carbon\ndouble bond specificity were generated from canola oil samples using OzESI-MRM\nprofiling. We also developed an integrated language user interface with large\nlanguage models using artificially intelligent (AI) agents that permits users\nto interact with the CLAW platform using a chatbot terminal to perform\nstatistical and bioinformatic analyses. We envision CLAW pipeline to be used in\nhigh-throughput lipid structural identification tasks aiding users to generate\nautomated lipidomics workflows ranging from data acquisition to AI agent-based\nbioinformatic analysis.\n", "date": "2024-03-22", "categories": []}
{"id": "2403.12987", "title": "Rethinking Specificity in SBDD: Leveraging Delta Score and Energy-Guided\n  Diffusion", "abstract": "  In the field of Structure-based Drug Design (SBDD), deep learning-based\ngenerative models have achieved outstanding performance in terms of docking\nscore. However, further study shows that the existing molecular generative\nmethods and docking scores both have lacked consideration in terms of\nspecificity, which means that generated molecules bind to almost every protein\npocket with high affinity. To address this, we introduce the Delta Score, a new\nmetric for evaluating the specificity of molecular binding. To further\nincorporate this insight for generation, we develop an innovative energy-guided\napproach using contrastive learning, with active compounds as decoys, to direct\ngenerative models toward creating molecules with high specificity. Our\nempirical results show that this method not only enhances the delta score but\nalso maintains or improves traditional docking scores, successfully bridging\nthe gap between SBDD and real-world needs.\n", "date": "2024-03-04", "categories": []}
{"id": "2306.15006", "title": "DNABERT-2: Efficient Foundation Model and Benchmark For Multi-Species\n  Genome", "abstract": "  Decoding the linguistic intricacies of the genome is a crucial problem in\nbiology, and pre-trained foundational models such as DNABERT and Nucleotide\nTransformer have made significant strides in this area. Existing works have\nlargely hinged on k-mer, fixed-length permutations of A, T, C, and G, as the\ntoken of the genome language due to its simplicity. However, we argue that the\ncomputation and sample inefficiencies introduced by k-mer tokenization are\nprimary obstacles in developing large genome foundational models. We provide\nconceptual and empirical insights into genome tokenization, building on which\nwe propose to replace k-mer tokenization with Byte Pair Encoding (BPE), a\nstatistics-based data compression algorithm that constructs tokens by\niteratively merging the most frequent co-occurring genome segment in the\ncorpus. We demonstrate that BPE not only overcomes the limitations of k-mer\ntokenization but also benefits from the computational efficiency of\nnon-overlapping tokenization. Based on these insights, we introduce DNABERT-2,\na refined genome foundation model that adapts an efficient tokenizer and\nemploys multiple strategies to overcome input length constraints, reduce time\nand memory expenditure, and enhance model capability. Furthermore, we identify\nthe absence of a comprehensive and standardized benchmark for genome\nunderstanding as another significant impediment to fair comparative analysis.\nIn response, we propose the Genome Understanding Evaluation (GUE), a\ncomprehensive multi-species genome classification dataset that amalgamates $36$\ndistinct datasets across $9$ tasks, with input lengths ranging from $70$ to\n$10000$. Through comprehensive experiments on the GUE benchmark, we demonstrate\nthat DNABERT-2 achieves comparable performance to the state-of-the-art model\nwith $21 \\times$ fewer parameters and approximately $92 \\times$ less GPU time\nin pre-training.\n", "date": "2023-06-26", "categories": []}
{"id": "2403.09673", "title": "FoldToken: Learning Protein Language via Vector Quantization and Beyond", "abstract": "  Is there a foreign language describing protein sequences and structures\nsimultaneously? Protein structures, represented by continuous 3D points, have\nlong posed a challenge due to the contrasting modeling paradigms of discrete\nsequences. We introduce \\textbf{FoldTokenizer} to represent protein\nsequence-structure as discrete symbols. This innovative approach involves\nprojecting residue types and structures into a discrete space, guided by a\nreconstruction loss for information preservation. We refer to the learned\ndiscrete symbols as \\textbf{FoldToken}, and the sequence of FoldTokens serves\nas a new protein language, transforming the protein sequence-structure into a\nunified modality. We apply the created protein language on general backbone\ninpainting and antibody design tasks, building the first GPT-style model\n(\\textbf{FoldGPT}) for sequence-structure co-generation with promising results.\nKey to our success is the substantial enhancement of the vector quantization\nmodule, Soft Conditional Vector Quantization (\\textbf{SoftCVQ}).\n", "date": "2024-02-04", "categories": []}
{"id": "2310.18515", "title": "Learning to design protein-protein interactions with enhanced\n  generalization", "abstract": "  Discovering mutations enhancing protein-protein interactions (PPIs) is\ncritical for advancing biomedical research and developing improved\ntherapeutics. While machine learning approaches have substantially advanced the\nfield, they often struggle to generalize beyond training data in practical\nscenarios. The contributions of this work are three-fold. First, we construct\nPPIRef, the largest and non-redundant dataset of 3D protein-protein\ninteractions, enabling effective large-scale learning. Second, we leverage the\nPPIRef dataset to pre-train PPIformer, a new SE(3)-equivariant model\ngeneralizing across diverse protein-binder variants. We fine-tune PPIformer to\npredict effects of mutations on protein-protein interactions via a\nthermodynamically motivated adjustment of the pre-training loss function.\nFinally, we demonstrate the enhanced generalization of our new PPIformer\napproach by outperforming other state-of-the-art methods on new, non-leaking\nsplits of standard labeled PPI mutational data and independent case studies\noptimizing a human antibody against SARS-CoV-2 and increasing the thrombolytic\nactivity of staphylokinase.\n", "date": "2023-10-27", "categories": []}
{"id": "2403.11015", "title": "Identifying the Attractors of Gene Regulatory Networks from Expression\n  Data under Uncertainty: An Interpretable Approach", "abstract": "  In systems biology, attractor landscape analysis of gene regulatory networks\nis recognized as a powerful computational tool for studying various cellular\nstates from proliferation and differentiation to senescence and apoptosis.\nTherefore, accurate identification of attractors plays a critical role in\ndetermination of the cell fates. On the other hand, in a real biological\ncircuit, genetic/epigenetic alterations as well as varying environmental\nfactors drastically take effect on the location, characteristics, and even the\nnumber of attractors. The central question is: Given a temporal gene expression\nprofile of a real gene regulatory network, how can the attractors be robustly\nidentified in the presence of huge amount of uncertainty? This paper addresses\nthis question using a novel approach based on Zadeh Computing with Words. The\nproposed scheme could effectively identify the attractors from temporal gene\nexpression data in terms of both fuzzy logic-based and linguistic descriptions\nwhich are simply interpretable by human experts. Therefore, this method can be\nconsidered as an effective step towards interpretable artificial intelligence.\nWithout loss of generality, genetic toggle switch is considered as the case\nstudy. The nonlinear dynamics of this benchmark gene regulatory network is\ncomputationally modeled by the notion of uncertain stochastic differential\nequations. The results of in-silico study demonstrate the efficiency and\nrobustness of the proposed method.\n", "date": "2024-03-16", "categories": []}
{"id": "2312.07977", "title": "Modeling non-genetic information dynamics in cells using reservoir\n  computing", "abstract": "  Virtually all cells use energy and ion-specific membrane pumps to maintain\nlarge transmembrane gradients of Na$^+$, K$^+$, Cl$^-$, Mg$^{++}$, and\nCa$^{++}$. Although they consume up to 1/3 of a cell's energy budget, the\ncorresponding evolutionary benefit of transmembrane ion gradients remain\nunclear. Here, we propose that ion gradients enable a dynamic and versatile\nbiological system that acquires, analyzes, and responds to environmental\ninformation. We hypothesize environmental signals are transmitted into the cell\nby ion fluxes along pre-existing gradients through gated ion-specific membrane\nchannels. The consequent changes of cytoplasmic ion concentration can generate\na local response and orchestrate global or regional responses through wire-like\nion fluxes along pre-existing and self-assembling cytoskeleton to engage the\nendoplasmic reticulum, mitochondria, and nucleus.\n  Here, we frame our hypothesis through a quasi-physical (Cell-Reservoir) model\nthat treats intra-cellular ion-based information dynamics as a sub-cellular\nprocess permitting spatiotemporally resolved cellular response that is also\ncapable of learning complex nonlinear dynamical cellular behavior. We\ndemonstrate the proposed ion dynamics permits rapid dissemination of response\nto information extrinsic perturbations that is consistent with experimental\nobservations.\n", "date": "2023-12-13", "categories": []}
{"id": "2403.07013", "title": "AdaNovo: Adaptive \\emph{De Novo} Peptide Sequencing with Conditional\n  Mutual Information", "abstract": "  Tandem mass spectrometry has played a pivotal role in advancing proteomics,\nenabling the analysis of protein composition in biological samples. Despite the\ndevelopment of various deep learning methods for identifying amino acid\nsequences (peptides) responsible for observed spectra, challenges persist in\n\\emph{de novo} peptide sequencing. Firstly, prior methods struggle to identify\namino acids with post-translational modifications (PTMs) due to their lower\nfrequency in training data compared to canonical amino acids, further resulting\nin decreased peptide-level identification precision. Secondly, diverse types of\nnoise and missing peaks in mass spectra reduce the reliability of training data\n(peptide-spectrum matches, PSMs). To address these challenges, we propose\nAdaNovo, a novel framework that calculates conditional mutual information (CMI)\nbetween the spectrum and each amino acid/peptide, using CMI for adaptive model\ntraining. Extensive experiments demonstrate AdaNovo's state-of-the-art\nperformance on a 9-species benchmark, where the peptides in the training set\nare almost completely disjoint from the peptides of the test sets. Moreover,\nAdaNovo excels in identifying amino acids with PTMs and exhibits robustness\nagainst data noise. The supplementary materials contain the official code.\n", "date": "2024-03-09", "categories": []}
{"id": "2312.09236", "title": "A framework for conditional diffusion modelling with applications in\n  motif scaffolding for protein design", "abstract": "  Many protein design applications, such as binder or enzyme design, require\nscaffolding a structural motif with high precision. Generative modelling\nparadigms based on denoising diffusion processes emerged as a leading candidate\nto address this motif scaffolding problem and have shown early experimental\nsuccess in some cases. In the diffusion paradigm, motif scaffolding is treated\nas a conditional generation task, and several conditional generation protocols\nwere proposed or imported from the Computer Vision literature. However, most of\nthese protocols are motivated heuristically, e.g. via analogies to Langevin\ndynamics, and lack a unifying framework, obscuring connections between the\ndifferent approaches. In this work, we unify conditional training and\nconditional sampling procedures under one common framework based on the\nmathematically well-understood Doob's h-transform. This new perspective allows\nus to draw connections between existing methods and propose a new variation on\nexisting conditional training protocols. We illustrate the effectiveness of\nthis new protocol in both, image outpainting and motif scaffolding and find\nthat it outperforms standard methods.\n", "date": "2023-12-14", "categories": []}
{"id": "2306.03117", "title": "Str2Str: A Score-based Framework for Zero-shot Protein Conformation\n  Sampling", "abstract": "  The dynamic nature of proteins is crucial for determining their biological\nfunctions and properties, for which Monte Carlo (MC) and molecular dynamics\n(MD) simulations stand as predominant tools to study such phenomena. By\nutilizing empirically derived force fields, MC or MD simulations explore the\nconformational space through numerically evolving the system via Markov chain\nor Newtonian mechanics. However, the high-energy barrier of the force fields\ncan hamper the exploration of both methods by the rare event, resulting in\ninadequately sampled ensemble without exhaustive running. Existing\nlearning-based approaches perform direct sampling yet heavily rely on\ntarget-specific simulation data for training, which suffers from high data\nacquisition cost and poor generalizability. Inspired by simulated annealing, we\npropose Str2Str, a novel structure-to-structure translation framework capable\nof zero-shot conformation sampling with roto-translation equivariant property.\nOur method leverages an amortized denoising score matching objective trained on\ngeneral crystal structures and has no reliance on simulation data during both\ntraining and inference. Experimental results across several benchmarking\nprotein systems demonstrate that Str2Str outperforms previous state-of-the-art\ngenerative structure prediction models and can be orders of magnitude faster\ncompared to long MD simulations. Our open-source implementation is available at\nhttps://github.com/lujiarui/Str2Str\n", "date": "2023-06-05", "categories": []}
{"id": "2310.07990", "title": "Multi-View Variational Autoencoder for Missing Value Imputation in\n  Untargeted Metabolomics", "abstract": "  Background: Missing data is a common challenge in mass spectrometry-based\nmetabolomics, which can lead to biased and incomplete analyses. The integration\nof whole-genome sequencing (WGS) data with metabolomics data has emerged as a\npromising approach to enhance the accuracy of data imputation in metabolomics\nstudies. Method: In this study, we propose a novel method that leverages the\ninformation from WGS data and reference metabolites to impute unknown\nmetabolites. Our approach utilizes a multi-view variational autoencoder to\njointly model the burden score, polygenetic risk score (PGS), and linkage\ndisequilibrium (LD) pruned single nucleotide polymorphisms (SNPs) for feature\nextraction and missing metabolomics data imputation. By learning the latent\nrepresentations of both omics data, our method can effectively impute missing\nmetabolomics values based on genomic information. Results: We evaluate the\nperformance of our method on empirical metabolomics datasets with missing\nvalues and demonstrate its superiority compared to conventional imputation\ntechniques. Using 35 template metabolites derived burden scores, PGS and\nLD-pruned SNPs, the proposed methods achieved R^2-scores > 0.01 for 71.55% of\nmetabolites. Conclusion: The integration of WGS data in metabolomics imputation\nnot only improves data completeness but also enhances downstream analyses,\npaving the way for more comprehensive and accurate investigations of metabolic\npathways and disease associations. Our findings offer valuable insights into\nthe potential benefits of utilizing WGS data for metabolomics data imputation\nand underscore the importance of leveraging multi-modal data integration in\nprecision medicine research.\n", "date": "2023-10-12", "categories": []}
{"id": "2402.10433", "title": "Fusing Neural and Physical: Augment Protein Conformation Sampling with\n  Tractable Simulations", "abstract": "  The protein dynamics are common and important for their biological functions\nand properties, the study of which usually involves time-consuming molecular\ndynamics (MD) simulations in silico. Recently, generative models has been\nleveraged as a surrogate sampler to obtain conformation ensembles with orders\nof magnitude faster and without requiring any simulation data (a \"zero-shot\"\ninference). However, being agnostic of the underlying energy landscape, the\naccuracy of such generative model may still be limited. In this work, we\nexplore the few-shot setting of such pre-trained generative sampler which\nincorporates MD simulations in a tractable manner. Specifically, given a target\nprotein of interest, we first acquire some seeding conformations from the\npre-trained sampler followed by a number of physical simulations in parallel\nstarting from these seeding samples. Then we fine-tuned the generative model\nusing the simulation trajectories above to become a target-specific sampler.\nExperimental results demonstrated the superior performance of such few-shot\nconformation sampler at a tractable computational cost.\n", "date": "2024-02-16", "categories": []}
{"id": "2403.07144", "title": "Thought Graph: Generating Thought Process for Biological Reasoning", "abstract": "  We present the Thought Graph as a novel framework to support complex\nreasoning and use gene set analysis as an example to uncover semantic\nrelationships between biological processes. Our framework stands out for its\nability to provide a deeper understanding of gene sets, significantly\nsurpassing GSEA by 40.28% and LLM baselines by 5.38% based on cosine similarity\nto human annotations. Our analysis further provides insights into future\ndirections of biological processes naming, and implications for bioinformatics\nand precision medicine.\n", "date": "2024-03-11", "categories": []}
{"id": "2403.04837", "title": "Cell reprogramming design by transfer learning of functional\n  transcriptional networks", "abstract": "  Recent developments in synthetic biology, next-generation sequencing, and\nmachine learning provide an unprecedented opportunity to rationally design new\ndisease treatments based on measured responses to gene perturbations and drugs\nto reprogram cells. The main challenges to seizing this opportunity are the\nincomplete knowledge of the cellular network and the combinatorial explosion of\npossible interventions, both of which are insurmountable by experiments. To\naddress these challenges, we develop a transfer learning approach to control\ncell behavior that is pre-trained on transcriptomic data associated with human\ncell fates, thereby generating a model of the network dynamics that can be\ntransferred to specific reprogramming goals. The approach combines\ntranscriptional responses to gene perturbations to minimize the difference\nbetween a given pair of initial and target transcriptional states. We\ndemonstrate our approach's versatility by applying it to a microarray dataset\ncomprising >9,000 microarrays across 54 cell types and 227 unique\nperturbations, and an RNASeq dataset consisting of >10,000 sequencing runs\nacross 36 cell types and 138 perturbations. Our approach reproduces known\nreprogramming protocols with an AUROC of 0.91 while innovating over existing\nmethods by pre-training an adaptable model that can be tailored to specific\nreprogramming transitions. We show that the number of gene perturbations\nrequired to steer from one fate to another increases with decreasing\ndevelopmental relatedness and that fewer genes are needed to progress along\ndevelopmental paths than to regress. These findings establish a\nproof-of-concept for our approach to computationally design control strategies\nand provide insights into how gene regulatory networks govern phenotype.\n", "date": "2024-03-07", "categories": []}
{"id": "2403.05602", "title": "Extracting Protein-Protein Interactions (PPIs) from Biomedical\n  Literature using Attention-based Relational Context Information", "abstract": "  Because protein-protein interactions (PPIs) are crucial to understand living\nsystems, harvesting these data is essential to probe disease development and\ndiscern gene/protein functions and biological processes. Some curated datasets\ncontain PPI data derived from the literature and other sources (e.g., IntAct,\nBioGrid, DIP, and HPRD). However, they are far from exhaustive, and their\nmaintenance is a labor-intensive process. On the other hand, machine learning\nmethods to automate PPI knowledge extraction from the scientific literature\nhave been limited by a shortage of appropriate annotated data. This work\npresents a unified, multi-source PPI corpora with vetted interaction\ndefinitions augmented by binary interaction type labels and a Transformer-based\ndeep learning method that exploits entities' relational context information for\nrelation representation to improve relation classification performance. The\nmodel's performance is evaluated on four widely studied biomedical relation\nextraction datasets, as well as this work's target PPI datasets, to observe the\neffectiveness of the representation to relation extraction tasks in various\ndata. Results show the model outperforms prior state-of-the-art models. The\ncode and data are available at:\nhttps://github.com/BNLNLP/PPI-Relation-Extraction\n", "date": "2024-03-08", "categories": []}
{"id": "2301.10774", "title": "RDesign: Hierarchical Data-efficient Representation Learning for\n  Tertiary Structure-based RNA Design", "abstract": "  While artificial intelligence has made remarkable strides in revealing the\nrelationship between biological macromolecules' primary sequence and tertiary\nstructure, designing RNA sequences based on specified tertiary structures\nremains challenging. Though existing approaches in protein design have\nthoroughly explored structure-to-sequence dependencies in proteins, RNA design\nstill confronts difficulties due to structural complexity and data scarcity.\nMoreover, direct transplantation of protein design methodologies into RNA\ndesign fails to achieve satisfactory outcomes although sharing similar\nstructural components. In this study, we aim to systematically construct a\ndata-driven RNA design pipeline. We crafted a large, well-curated benchmark\ndataset and designed a comprehensive structural modeling approach to represent\nthe complex RNA tertiary structure. More importantly, we proposed a\nhierarchical data-efficient representation learning framework that learns\nstructural representations through contrastive learning at both cluster-level\nand sample-level to fully leverage the limited data. By constraining data\nrepresentations within a limited hyperspherical space, the intrinsic\nrelationships between data points could be explicitly imposed. Moreover, we\nincorporated extracted secondary structures with base pairs as prior knowledge\nto facilitate the RNA design process. Extensive experiments demonstrate the\neffectiveness of our proposed method, providing a reliable baseline for future\nRNA design tasks. The source code and benchmark dataset are available at\nhttps://github.com/A4Bio/RDesign.\n", "date": "2023-01-25", "categories": []}
{"id": "2403.04106", "title": "Understanding Biology in the Age of Artificial Intelligence", "abstract": "  Modern life sciences research is increasingly relying on artificial\nintelligence approaches to model biological systems, primarily centered around\nthe use of machine learning (ML) models. Although ML is undeniably useful for\nidentifying patterns in large, complex data sets, its widespread application in\nbiological sciences represents a significant deviation from traditional methods\nof scientific inquiry. As such, the interplay between these models and\nscientific understanding in biology is a topic with important implications for\nthe future of scientific research, yet it is a subject that has received little\nattention. Here, we draw from an epistemological toolkit to contextualize\nrecent applications of ML in biological sciences under modern philosophical\ntheories of understanding, identifying general principles that can guide the\ndesign and application of ML systems to model biological phenomena and advance\nscientific knowledge. We propose that conceptions of scientific understanding\nas information compression, qualitative intelligibility, and dependency\nrelation modelling provide a useful framework for interpreting ML-mediated\nunderstanding of biological systems. Through a detailed analysis of two key\napplication areas of ML in modern biological research - protein structure\nprediction and single cell RNA-sequencing - we explore how these features have\nthus far enabled ML systems to advance scientific understanding of their target\nphenomena, how they may guide the development of future ML models, and the key\nobstacles that remain in preventing ML from achieving its potential as a tool\nfor biological discovery. Consideration of the epistemological features of ML\napplications in biology will improve the prospects of these methods to solve\nimportant problems and advance scientific understanding of living systems.\n", "date": "2024-03-06", "categories": []}
{"id": "2403.04187", "title": "Preference optimization of protein language models as a multi-objective\n  binder design paradigm", "abstract": "  We present a multi-objective binder design paradigm based on instruction\nfine-tuning and direct preference optimization (DPO) of autoregressive protein\nlanguage models (pLMs). Multiple design objectives are encoded in the language\nmodel through direct optimization on expert curated preference sequence\ndatasets comprising preferred and dispreferred distributions. We show the\nproposed alignment strategy enables ProtGPT2 to effectively design binders\nconditioned on specified receptors and a drug developability criterion.\nGenerated binder samples demonstrate median isoelectric point (pI) improvements\nby $17\\%-60\\%$.\n", "date": "2024-03-07", "categories": []}
{"id": "2308.05027", "title": "AbDiffuser: Full-Atom Generation of in vitro Functioning Antibodies", "abstract": "  We introduce AbDiffuser, an equivariant and physics-informed diffusion model\nfor the joint generation of antibody 3D structures and sequences. AbDiffuser is\nbuilt on top of a new representation of protein structure, relies on a novel\narchitecture for aligned proteins, and utilizes strong diffusion priors to\nimprove the denoising process. Our approach improves protein diffusion by\ntaking advantage of domain knowledge and physics-based constraints; handles\nsequence-length changes; and reduces memory complexity by an order of\nmagnitude, enabling backbone and side chain generation. We validate AbDiffuser\nin silico and in vitro. Numerical experiments showcase the ability of\nAbDiffuser to generate antibodies that closely track the sequence and\nstructural properties of a reference set. Laboratory experiments confirm that\nall 16 HER2 antibodies discovered were expressed at high levels and that 57.1%\nof the selected designs were tight binders.\n", "date": "2023-07-28", "categories": []}
{"id": "2403.03231", "title": "Machine and deep learning methods for predicting 3D genome organization", "abstract": "  Three-Dimensional (3D) chromatin interactions, such as enhancer-promoter\ninteractions (EPIs), loops, Topologically Associating Domains (TADs), and A/B\ncompartments play critical roles in a wide range of cellular processes by\nregulating gene expression. Recent development of chromatin conformation\ncapture technologies has enabled genome-wide profiling of various 3D\nstructures, even with single cells. However, current catalogs of 3D structures\nremain incomplete and unreliable due to differences in technology, tools, and\nlow data resolution. Machine learning methods have emerged as an alternative to\nobtain missing 3D interactions and/or improve resolution. Such methods\nfrequently use genome annotation data (ChIP-seq, DNAse-seq, etc.), DNA\nsequencing information (k-mers, Transcription Factor Binding Site (TFBS)\nmotifs), and other genomic properties to learn the associations between genomic\nfeatures and chromatin interactions. In this review, we discuss computational\ntools for predicting three types of 3D interactions (EPIs, chromatin\ninteractions, TAD boundaries) and analyze their pros and cons. We also point\nout obstacles of computational prediction of 3D interactions and suggest future\nresearch directions.\n", "date": "2024-03-04", "categories": []}
{"id": "2403.02765", "title": "G4-Attention: Deep Learning Model with Attention for predicting DNA\n  G-Quadruplexes", "abstract": "  G-Quadruplexes are the four-stranded non-canonical nucleic acid secondary\nstructures, formed by the stacking arrangement of the guanine tetramers. They\nare involved in a wide range of biological roles because of their exceptionally\nunique and distinct structural characteristics. After the completion of the\nhuman genome sequencing project, a lot of bioinformatic algorithms were\nintroduced to predict the active G4s regions \\textit{in vitro} based on the\ncanonical G4 sequence elements, G-\\textit{richness}, and G-\\textit{skewness},\nas well as the non-canonical sequence features. Recently, sequencing techniques\nlike G4-seq and G4-ChIP-seq were developed to map the G4s \\textit{in vitro},\nand \\textit{in vivo} respectively at a few hundred base resolution.\nSubsequently, several machine learning approaches were developed for predicting\nthe G4 regions using the existing databases. However, their prediction models\nwere simplistic, and the prediction accuracy was notably poor. In response,\nhere, we propose a novel convolutional neural network with Bi-LSTM and\nattention layers, named G4-attention, to predict the G4 forming sequences with\nimproved accuracy. G4-attention achieves high accuracy and attains\nstate-of-the-art results in the G4 prediction task. Our model also predicts the\nG4 regions accurately in the highly class-imbalanced datasets. In addition, the\ndeveloped model trained on the human genome dataset can be applied to any\nnon-human genome DNA sequences to predict the G4 formation propensities.\n", "date": "2024-03-05", "categories": []}
{"id": "2403.03018", "title": "CRISPR: Ensemble Model", "abstract": "  Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) is a gene\nediting technology that has revolutionized the fields of biology and medicine.\nHowever, one of the challenges of using CRISPR is predicting the on-target\nefficacy and off-target sensitivity of single-guide RNAs (sgRNAs). This is\nbecause most existing methods are trained on separate datasets with different\ngenes and cells, which limits their generalizability. In this paper, we propose\na novel ensemble learning method for sgRNA design that is accurate and\ngeneralizable. Our method combines the predictions of multiple machine learning\nmodels to produce a single, more robust prediction. This approach allows us to\nlearn from a wider range of data, which improves the generalizability of our\nmodel. We evaluated our method on a benchmark dataset of sgRNA designs and\nfound that it outperformed existing methods in terms of both accuracy and\ngeneralizability. Our results suggest that our method can be used to design\nsgRNAs with high sensitivity and specificity, even for new genes or cells. This\ncould have important implications for the clinical use of CRISPR, as it would\nallow researchers to design more effective and safer treatments for a variety\nof diseases.\n", "date": "2024-03-05", "categories": []}
{"id": "2307.00494", "title": "Improving Protein Optimization with Smoothed Fitness Landscapes", "abstract": "  The ability to engineer novel proteins with higher fitness for a desired\nproperty would be revolutionary for biotechnology and medicine. Modeling the\ncombinatorially large space of sequences is infeasible; prior methods often\nconstrain optimization to a small mutational radius, but this drastically\nlimits the design space. Instead of heuristics, we propose smoothing the\nfitness landscape to facilitate protein optimization. First, we formulate\nprotein fitness as a graph signal then use Tikunov regularization to smooth the\nfitness landscape. We find optimizing in this smoothed landscape leads to\nimproved performance across multiple methods in the GFP and AAV benchmarks.\nSecond, we achieve state-of-the-art results utilizing discrete energy-based\nmodels and MCMC in the smoothed landscape. Our method, called Gibbs sampling\nwith Graph-based Smoothing (GGS), demonstrates a unique ability to achieve 2.5\nfold fitness improvement (with in-silico evaluation) over its training set. GGS\ndemonstrates potential to optimize proteins in the limited data regime. Code:\nhttps://github.com/kirjner/GGS\n", "date": "2023-07-02", "categories": []}
{"id": "2401.06182", "title": "Prediction of Cellular Identities from Trajectory and Cell Fate\n  Information", "abstract": "  Determining cell identities in imaging sequences is an important yet\nchallenging task. The conventional method for cell identification is via cell\ntracking, which is complex and can be time-consuming. In this study, we propose\nan innovative approach to cell identification during early $\\textit{C.\nelegans}$ embryogenesis using machine learning. Cell identification during\n$\\textit{C. elegans}$ embryogenesis would provide insights into neural\ndevelopment with implications for higher organisms including humans. We\nemployed random forest, MLP, and LSTM models, and tested cell classification\naccuracy on 3D time-lapse confocal datasets spanning the first 4 hours of\nembryogenesis. By leveraging a small number of spatial-temporal features of\nindividual cells, including cell trajectory and cell fate information, our\nmodels achieve an accuracy of over 91%, even with limited data. We also\ndetermine the most important feature contributions and can interpret these\nfeatures in the context of biological knowledge. Our research demonstrates the\nsuccess of predicting cell identities in time-lapse imaging sequences directly\nfrom simple spatio-temporal features.\n", "date": "2024-01-11", "categories": []}
{"id": "2403.00875", "title": "Enhancing Protein Predictive Models via Proteins Data Augmentation: A\n  Benchmark and New Directions", "abstract": "  Augmentation is an effective alternative to utilize the small amount of\nlabeled protein data. However, most of the existing work focuses on design-ing\nnew architectures or pre-training tasks, and relatively little work has studied\ndata augmentation for proteins. This paper extends data augmentation techniques\npreviously used for images and texts to proteins and then benchmarks these\ntechniques on a variety of protein-related tasks, providing the first\ncomprehensive evaluation of protein augmentation. Furthermore, we propose two\nnovel semantic-level protein augmentation methods, namely Integrated Gradients\nSubstitution and Back Translation Substitution, which enable protein\nsemantic-aware augmentation through saliency detection and biological\nknowledge. Finally, we integrate extended and proposed augmentations into an\naugmentation pool and propose a simple but effective framework, namely\nAutomated Protein Augmentation (APA), which can adaptively select the most\nsuitable augmentation combinations for different tasks. Extensive experiments\nhave shown that APA enhances the performance of five protein related tasks by\nan average of 10.55% across three architectures compared to vanilla\nimplementations without augmentation, highlighting its potential to make a\ngreat impact on the field.\n", "date": "2024-03-01", "categories": []}
{"id": "2308.01839", "title": "Is your data alignable? Principled and interpretable alignability\n  testing and integration of single-cell data", "abstract": "  Single-cell data integration can provide a comprehensive molecular view of\ncells, and many algorithms have been developed to remove unwanted technical or\nbiological variations and integrate heterogeneous single-cell datasets. Despite\ntheir wide usage, existing methods suffer from several fundamental limitations.\nIn particular, we lack a rigorous statistical test for whether two\nhigh-dimensional single-cell datasets are alignable (and therefore should even\nbe aligned). Moreover, popular methods can substantially distort the data\nduring alignment, making the aligned data and downstream analysis difficult to\ninterpret. To overcome these limitations, we present a spectral manifold\nalignment and inference (SMAI) framework, which enables principled and\ninterpretable alignability testing and structure-preserving integration of\nsingle-cell data with the same type of features. SMAI provides a statistical\ntest to robustly assess the alignability between datasets to avoid misleading\ninference, and is justified by high-dimensional statistical theory. On a\ndiverse range of real and simulated benchmark datasets, it outperforms commonly\nused alignment methods. Moreover, we show that SMAI improves various downstream\nanalyses such as identification of differentially expressed genes and\nimputation of single-cell spatial transcriptomics, providing further biological\ninsights. SMAI's interpretability also enables quantification and a deeper\nunderstanding of the sources of technical confounders in single-cell data.\n", "date": "2023-08-03", "categories": []}
{"id": "2402.18583", "title": "Binding-Adaptive Diffusion Models for Structure-Based Drug Design", "abstract": "  Structure-based drug design (SBDD) aims to generate 3D ligand molecules that\nbind to specific protein targets. Existing 3D deep generative models including\ndiffusion models have shown great promise for SBDD. However, it is complex to\ncapture the essential protein-ligand interactions exactly in 3D space for\nmolecular generation. To address this problem, we propose a novel framework,\nnamely Binding-Adaptive Diffusion Models (BindDM). In BindDM, we adaptively\nextract subcomplex, the essential part of binding sites responsible for\nprotein-ligand interactions. Then the selected protein-ligand subcomplex is\nprocessed with SE(3)-equivariant neural networks, and transmitted back to each\natom of the complex for augmenting the target-aware 3D molecule diffusion\ngeneration with binding interaction information. We iterate this hierarchical\ncomplex-subcomplex process with cross-hierarchy interaction node for adequately\nfusing global binding context between the complex and its corresponding\nsubcomplex. Empirical studies on the CrossDocked2020 dataset show BindDM can\ngenerate molecules with more realistic 3D structures and higher binding\naffinities towards the protein targets, with up to -5.92 Avg. Vina Score, while\nmaintaining proper molecular properties. Our code is available at\nhttps://github.com/YangLing0818/BindDM\n", "date": "2024-01-15", "categories": []}
{"id": "2402.17807", "title": "Exploring Gene Regulatory Interaction Networks and predicting\n  therapeutic molecules for Hypopharyngeal Cancer and EGFR-mutated lung\n  adenocarcinoma", "abstract": "  With the advent of Information technology, the Bioinformatics research field\nis becoming increasingly attractive to researchers and academicians. The recent\ndevelopment of various Bioinformatics toolkits has facilitated the rapid\nprocessing and analysis of vast quantities of biological data for human\nperception. Most studies focus on locating two connected diseases and making\nsome observations to construct diverse gene regulatory interaction networks, a\nforerunner to general drug design for curing illness. For instance,\nHypopharyngeal cancer is a disease that is associated with EGFR-mutated lung\nadenocarcinoma. In this study, we select EGFR-mutated lung adenocarcinoma and\nHypopharyngeal cancer by finding the Lung metastases in hypopharyngeal cancer.\nTo conduct this study, we collect Mircorarray datasets from GEO (Gene\nExpression Omnibus), an online database controlled by NCBI. Differentially\nexpressed genes, common genes, and hub genes between the selected two diseases\nare detected for the succeeding move. Our research findings have suggested\ncommon therapeutic molecules for the selected diseases based on 10 hub genes\nwith the highest interactions according to the degree topology method and the\nmaximum clique centrality (MCC). Our suggested therapeutic molecules will be\nfruitful for patients with those two diseases simultaneously.\n", "date": "2024-02-27", "categories": []}
{"id": "2402.18396", "title": "Deep Confident Steps to New Pockets: Strategies for Docking\n  Generalization", "abstract": "  Accurate blind docking has the potential to lead to new biological\nbreakthroughs, but for this promise to be realized, docking methods must\ngeneralize well across the proteome. Existing benchmarks, however, fail to\nrigorously assess generalizability. Therefore, we develop DockGen, a new\nbenchmark based on the ligand-binding domains of proteins, and we show that\nexisting machine learning-based docking models have very weak generalization\nabilities. We carefully analyze the scaling laws of ML-based docking and show\nthat, by scaling data and model size, as well as integrating synthetic data\nstrategies, we are able to significantly increase the generalization capacity\nand set new state-of-the-art performance across benchmarks. Further, we propose\nConfidence Bootstrapping, a new training paradigm that solely relies on the\ninteraction between diffusion and confidence models and exploits the\nmulti-resolution generation process of diffusion models. We demonstrate that\nConfidence Bootstrapping significantly improves the ability of ML-based docking\nmethods to dock to unseen protein classes, edging closer to accurate and\ngeneralizable blind docking methods.\n", "date": "2024-02-28", "categories": []}
{"id": "2402.16854", "title": "Attention Based Molecule Generation via Hierarchical Variational\n  Autoencoder", "abstract": "  Molecule generation is a task made very difficult by the complex ways in\nwhich we represent molecules computationally. A common technique used in\nmolecular generative modeling is to use SMILES strings with recurrent neural\nnetworks built into variational autoencoders - but these suffer from a myriad\nof issues: vanishing gradients, long-range forgetting, and invalid molecules.\nIn this work, we show that by combining recurrent neural networks with\nconvolutional networks in a hierarchical manner, we are able to both extract\nautoregressive information from SMILES strings while maintaining signal and\nlong-range dependencies. This allows for generations with very high validity\nrates on the order of 95% when reconstructing known molecules. We also observe\nan average Tanimoto similarity of .6 between test set and reconstructed\nmolecules, which suggests our method is able to map between SMILES strings and\ntheir learned representations in a more effective way than prior works using\nsimilar methods.\n", "date": "2024-01-18", "categories": []}
{"id": "2402.16930", "title": "TrustMol: Trustworthy Inverse Molecular Design via Alignment with\n  Molecular Dynamics", "abstract": "  Data-driven generation of molecules with desired properties, also known as\ninverse molecular design (IMD), has attracted significant attention in recent\nyears. Despite the significant progress in the accuracy and diversity of\nsolutions, existing IMD methods lag behind in terms of trustworthiness. The\nroot issue is that the design process of these methods is increasingly more\nimplicit and indirect, and this process is also isolated from the native\nforward process (NFP), the ground-truth function that models the molecular\ndynamics. Following this insight, we propose TrustMol, an IMD method built to\nbe trustworthy. For this purpose, TrustMol relies on a set of technical\nnovelties including a new variational autoencoder network. Moreover, we propose\na latent-property pairs acquisition method to effectively navigate the\ncomplexities of molecular latent optimization, a process that seems intuitive\nyet challenging due to the high-frequency and discontinuous nature of molecule\nspace. TrustMol also integrates uncertainty-awareness into molecular latent\noptimization. These lead to improvements in both explainability and reliability\nof the IMD process. We validate the trustworthiness of TrustMol through a wide\nrange of experiments.\n", "date": "2024-02-26", "categories": []}
{"id": "2311.15156", "title": "xTrimoGene: An Efficient and Scalable Representation Learner for\n  Single-Cell RNA-Seq Data", "abstract": "  Advances in high-throughput sequencing technology have led to significant\nprogress in measuring gene expressions at the single-cell level. The amount of\npublicly available single-cell RNA-seq (scRNA-seq) data is already surpassing\n50M records for humans with each record measuring 20,000 genes. This highlights\nthe need for unsupervised representation learning to fully ingest these data,\nyet classical transformer architectures are prohibitive to train on such data\nin terms of both computation and memory. To address this challenge, we propose\na novel asymmetric encoder-decoder transformer for scRNA-seq data, called\nxTrimoGene$^\\alpha$ (or xTrimoGene for short), which leverages the sparse\ncharacteristic of the data to scale up the pre-training. This scalable design\nof xTrimoGene reduces FLOPs by one to two orders of magnitude compared to\nclassical transformers while maintaining high accuracy, enabling us to train\nthe largest transformer models over the largest scRNA-seq dataset today. Our\nexperiments also show that the performance of xTrimoGene improves as we scale\nup the model sizes, and it also leads to SOTA performance over various\ndownstream tasks, such as cell type annotation, perturb-seq effect prediction,\nand drug combination prediction. xTrimoGene model is now available for use as a\nservice via the following link: https://api.biomap.com/xTrimoGene/apply.\n", "date": "2023-11-26", "categories": []}
{"id": "2306.05587", "title": "MC-NN: An End-to-End Multi-Channel Neural Network Approach for\n  Predicting Influenza A Virus Hosts and Antigenic Types", "abstract": "  Influenza poses a significant threat to public health, particularly among the\nelderly, young children, and people with underlying dis-eases. The\nmanifestation of severe conditions, such as pneumonia, highlights the\nimportance of preventing the spread of influenza. An accurate and\ncost-effective prediction of the host and antigenic sub-types of influenza A\nviruses is essential to addressing this issue, particularly in\nresource-constrained regions. In this study, we propose a multi-channel neural\nnetwork model to predict the host and antigenic subtypes of influenza A viruses\nfrom hemagglutinin and neuraminidase protein sequences. Our model was trained\non a comprehensive data set of complete protein sequences and evaluated on\nvarious test data sets of complete and incomplete sequences. The results\ndemonstrate the potential and practicality of using multi-channel neural\nnetworks in predicting the host and antigenic subtypes of influenza A viruses\nfrom both full and partial protein sequences.\n", "date": "2023-06-08", "categories": []}
{"id": "2401.12999", "title": "Quantum-Inspired Machine Learning for Molecular Docking", "abstract": "  Molecular docking is an important tool for structure-based drug design,\naccelerating the efficiency of drug development. Complex and dynamic binding\nprocesses between proteins and small molecules require searching and sampling\nover a wide spatial range. Traditional docking by searching for possible\nbinding sites and conformations is computationally complex and results poorly\nunder blind docking. Quantum-inspired algorithms combining quantum properties\nand annealing show great advantages in solving combinatorial optimization\nproblems. Inspired by this, we achieve an improved in blind docking by using\nquantum-inspired combined with gradients learned by deep learning in the\nencoded molecular space. Numerical simulation shows that our method outperforms\ntraditional docking algorithms and deep learning-based algorithms over 10\\%.\nCompared to the current state-of-the-art deep learning-based docking algorithm\nDiffDock, the success rate of Top-1 (RMSD<2) achieves an improvement from 33\\%\nto 35\\% in our same setup. In particular, a 6\\% improvement is realized in the\nhigh-precision region(RMSD<1) on molecules data unseen in DiffDock, which\ndemonstrates the well-generalized of our method.\n", "date": "2024-01-22", "categories": []}
{"id": "2402.14391", "title": "MAPE-PPI: Towards Effective and Efficient Protein-Protein Interaction\n  Prediction via Microenvironment-Aware Protein Embedding", "abstract": "  Protein-Protein Interactions (PPIs) are fundamental in various biological\nprocesses and play a key role in life activities. The growing demand and cost\nof experimental PPI assays require computational methods for efficient PPI\nprediction. While existing methods rely heavily on protein sequence for PPI\nprediction, it is the protein structure that is the key to determine the\ninteractions. To take both protein modalities into account, we define the\nmicroenvironment of an amino acid residue by its sequence and structural\ncontexts, which describe the surrounding chemical properties and geometric\nfeatures. In addition, microenvironments defined in previous work are largely\nbased on experimentally assayed physicochemical properties, for which the\n\"vocabulary\" is usually extremely small. This makes it difficult to cover the\ndiversity and complexity of microenvironments. In this paper, we propose\nMicroenvironment-Aware Protein Embedding for PPI prediction (MPAE-PPI), which\nencodes microenvironments into chemically meaningful discrete codes via a\nsufficiently large microenvironment \"vocabulary\" (i.e., codebook). Moreover, we\npropose a novel pre-training strategy, namely Masked Codebook Modeling (MCM),\nto capture the dependencies between different microenvironments by randomly\nmasking the codebook and reconstructing the input. With the learned\nmicroenvironment codebook, we can reuse it as an off-the-shelf tool to\nefficiently and effectively encode proteins of different sizes and functions\nfor large-scale PPI prediction. Extensive experiments show that MAPE-PPI can\nscale to PPI prediction with millions of PPIs with superior trade-offs between\neffectiveness and computational efficiency than the state-of-the-art\ncompetitors.\n", "date": "2024-02-22", "categories": []}
{"id": "2402.11459", "title": "Re-Dock: Towards Flexible and Realistic Molecular Docking with Diffusion\n  Bridge", "abstract": "  Accurate prediction of protein-ligand binding structures, a task known as\nmolecular docking is crucial for drug design but remains challenging. While\ndeep learning has shown promise, existing methods often depend on holo-protein\nstructures (docked, and not accessible in realistic tasks) or neglect pocket\nsidechain conformations, leading to limited practical utility and unrealistic\nconformation predictions. To fill these gaps, we introduce an under-explored\ntask, named flexible docking to predict poses of ligand and pocket sidechains\nsimultaneously and introduce Re-Dock, a novel diffusion bridge generative model\nextended to geometric manifolds. Specifically, we propose energy-to-geometry\nmapping inspired by the Newton-Euler equation to co-model the binding energy\nand conformations for reflecting the energy-constrained docking generative\nprocess. Comprehensive experiments on designed benchmark datasets including\napo-dock and cross-dock demonstrate our model's superior effectiveness and\nefficiency over current methods.\n", "date": "2024-02-18", "categories": []}
{"id": "2402.13297", "title": "Integrating Deep Learning and Synthetic Biology: A Co-Design Approach\n  for Enhancing Gene Expression via N-terminal Coding Sequences", "abstract": "  N-terminal coding sequence (NCS) influences gene expression by impacting the\ntranslation initiation rate. The NCS optimization problem is to find an NCS\nthat maximizes gene expression. The problem is important in genetic\nengineering. However, current methods for NCS optimization such as rational\ndesign and statistics-guided approaches are labor-intensive yield only\nrelatively small improvements. This paper introduces a deep learning/synthetic\nbiology co-designed few-shot training workflow for NCS optimization. Our method\nutilizes k-nearest encoding followed by word2vec to encode the NCS, then\nperforms feature extraction using attention mechanisms, before constructing a\ntime-series network for predicting gene expression intensity, and finally a\ndirect search algorithm identifies the optimal NCS with limited training data.\nWe took green fluorescent protein (GFP) expressed by Bacillus subtilis as a\nreporting protein of NCSs, and employed the fluorescence enhancement factor as\nthe metric of NCS optimization. Within just six iterative experiments, our\nmodel generated an NCS (MLD62) that increased average GFP expression by\n5.41-fold, outperforming the state-of-the-art NCS designs. Extending our\nfindings beyond GFP, we showed that our engineered NCS (MLD62) can effectively\nboost the production of N-acetylneuraminic acid by enhancing the expression of\nthe crucial rate-limiting GNA1 gene, demonstrating its practical utility. We\nhave open-sourced our NCS expression database and experimental procedures for\npublic use.\n", "date": "2024-02-20", "categories": []}
{"id": "2402.13714", "title": "An Evaluation of Large Language Models in Bioinformatics Research", "abstract": "  Large language models (LLMs) such as ChatGPT have gained considerable\ninterest across diverse research communities. Their notable ability for text\ncompletion and generation has inaugurated a novel paradigm for\nlanguage-interfaced problem solving. However, the potential and efficacy of\nthese models in bioinformatics remain incompletely explored. In this work, we\nstudy the performance LLMs on a wide spectrum of crucial bioinformatics tasks.\nThese tasks include the identification of potential coding regions, extraction\nof named entities for genes and proteins, detection of antimicrobial and\nanti-cancer peptides, molecular optimization, and resolution of educational\nbioinformatics problems. Our findings indicate that, given appropriate prompts,\nLLMs like GPT variants can successfully handle most of these tasks. In\naddition, we provide a thorough analysis of their limitations in the context of\ncomplicated bioinformatics tasks. In conclusion, we believe that this work can\nprovide new perspectives and motivate future research in the field of LLMs\napplications, AI for Science and bioinformatics.\n", "date": "2024-02-21", "categories": []}
{"id": "2306.01794", "title": "DiffPack: A Torsional Diffusion Model for Autoregressive Protein\n  Side-Chain Packing", "abstract": "  Proteins play a critical role in carrying out biological functions, and their\n3D structures are essential in determining their functions. Accurately\npredicting the conformation of protein side-chains given their backbones is\nimportant for applications in protein structure prediction, design and\nprotein-protein interactions. Traditional methods are computationally intensive\nand have limited accuracy, while existing machine learning methods treat the\nproblem as a regression task and overlook the restrictions imposed by the\nconstant covalent bond lengths and angles. In this work, we present DiffPack, a\ntorsional diffusion model that learns the joint distribution of side-chain\ntorsional angles, the only degrees of freedom in side-chain packing, by\ndiffusing and denoising on the torsional space. To avoid issues arising from\nsimultaneous perturbation of all four torsional angles, we propose\nautoregressively generating the four torsional angles from $\\chi_1$ to $\\chi_4$\nand training diffusion models for each torsional angle. We evaluate the method\non several benchmarks for protein side-chain packing and show that our method\nachieves improvements of $11.9\\%$ and $13.5\\%$ in angle accuracy on CASP13 and\nCASP14, respectively, with a significantly smaller model size ($60\\times$ fewer\nparameters). Additionally, we show the effectiveness of our method in enhancing\nside-chain predictions in the AlphaFold2 model. Code is available at\nhttps://github.com/DeepGraphLearning/DiffPack.\n", "date": "2023-06-01", "categories": []}
{"id": "2402.10227", "title": "Correlational Lagrangian Schr\\\"odinger Bridge: Learning Dynamics with\n  Population-Level Regularization", "abstract": "  Accurate modeling of system dynamics holds intriguing potential in broad\nscientific fields including cytodynamics and fluid mechanics. This task often\npresents significant challenges when (i) observations are limited to\ncross-sectional samples (where individual trajectories are inaccessible for\nlearning), and moreover, (ii) the behaviors of individual particles are\nheterogeneous (especially in biological systems due to biodiversity). To\naddress them, we introduce a novel framework dubbed correlational Lagrangian\nSchr\\\"odinger bridge (CLSB), aiming to seek for the evolution \"bridging\" among\ncross-sectional observations, while regularized for the minimal population\n\"cost\". In contrast to prior methods relying on \\textit{individual}-level\nregularizers for all particles \\textit{homogeneously} (e.g. restraining\nindividual motions), CLSB operates at the population level admitting the\nheterogeneity nature, resulting in a more generalizable modeling in practice.\nTo this end, our contributions include (1) a new class of population\nregularizers capturing the temporal variations in multivariate relations, with\nthe tractable formulation derived, (2) three domain-informed instantiations\nbased on genetic co-expression stability, and (3) an integration of population\nregularizers into data-driven generative models as constrained optimization,\nand a numerical solution, with further extension to conditional generative\nmodels. Empirically, we demonstrate the superiority of CLSB in single-cell\nsequencing data analyses such as simulating cell development over time and\npredicting cellular responses to drugs of varied doses.\n", "date": "2024-02-04", "categories": []}
{"id": "2402.10551", "title": "Personalised Drug Identifier for Cancer Treatment with Transformers\n  using Auxiliary Information", "abstract": "  Cancer remains a global challenge due to its growing clinical and economic\nburden. Its uniquely personal manifestation, which makes treatment difficult,\nhas fuelled the quest for personalized treatment strategies. Thus, genomic\nprofiling is increasingly becoming part of clinical diagnostic panels.\nEffective use of such panels requires accurate drug response prediction (DRP)\nmodels, which are challenging to build due to limited labelled patient data.\nPrevious methods to address this problem have used various forms of transfer\nlearning. However, they do not explicitly model the variable length sequential\nstructure of the list of mutations in such diagnostic panels. Further, they do\nnot utilize auxiliary information (like patient survival) for model training.\nWe address these limitations through a novel transformer based method, which\nsurpasses the performance of state-of-the-art DRP models on benchmark data. We\nalso present the design of a treatment recommendation system (TRS), which is\ncurrently deployed at the National University Hospital, Singapore and is being\nevaluated in a clinical trial.\n", "date": "2024-02-16", "categories": []}
{"id": "2402.09381", "title": "GraSSRep: Graph-Based Self-Supervised Learning for Repeat Detection in\n  Metagenomic Assembly", "abstract": "  Repetitive DNA (repeats) poses significant challenges for accurate and\nefficient genome assembly and sequence alignment. This is particularly true for\nmetagenomic data, where genome dynamics such as horizontal gene transfer, gene\nduplication, and gene loss/gain complicate accurate genome assembly from\nmetagenomic communities. Detecting repeats is a crucial first step in\novercoming these challenges. To address this issue, we propose GraSSRep, a\nnovel approach that leverages the assembly graph's structure through graph\nneural networks (GNNs) within a self-supervised learning framework to classify\nDNA sequences into repetitive and non-repetitive categories. Specifically, we\nframe this problem as a node classification task within a metagenomic assembly\ngraph. In a self-supervised fashion, we rely on a high-precision (but\nlow-recall) heuristic to generate pseudo-labels for a small proportion of the\nnodes. We then use those pseudo-labels to train a GNN embedding and a random\nforest classifier to propagate the labels to the remaining nodes. In this way,\nGraSSRep combines sequencing features with pre-defined and learned graph\nfeatures to achieve state-of-the-art performance in repeat detection. We\nevaluate our method using simulated and synthetic metagenomic datasets. The\nresults on the simulated data highlight our GraSSRep's robustness to repeat\nattributes, demonstrating its effectiveness in handling the complexity of\nrepeated sequences. Additionally, our experiments with synthetic metagenomic\ndatasets reveal that incorporating the graph structure and the GNN enhances our\ndetection performance. Finally, in comparative analyses, GraSSRep outperforms\nexisting repeat detection tools with respect to precision and recall.\n", "date": "2024-02-14", "categories": []}
{"id": "2402.07948", "title": "evolSOM: an R Package for evolutionary conservation analysis with SOMs", "abstract": "  Motivation: Unraveling the connection between genes and traits is crucial for\nsolving many biological puzzles. Genes provide instructions for building\ncellular machinery, directing the processes that sustain life. RNA molecules\nand proteins, derived from these genetic instructions, play crucial roles in\nshaping cell structures, influencing reactions, and guiding behavior. This\nfundamental biological principle links genetic makeup to observable traits, but\nintegrating and extracting meaningful relationships from this complex,\nmultimodal data presents a significant challenge. Results: We introduce\nevolSOM, a novel R package that utilizes Self-Organizing Maps (SOMs) to explore\nand visualize the conservation of biological variables, easing the integration\nof phenotypic and genotypic attributes. By constructing species-specific or\ncondition-specific SOMs that capture non-redundant patterns, evolSOM allows the\nanalysis of displacement of biological variables between species or conditions.\nVariables displaced together suggest membership in the same regulatory network,\nand the nature of the displacement may hold biological significance. The\npackage automatically calculates and graphically presents these displacements,\nenabling efficient comparison and revealing conserved and displaced variables.\nThe package facilitates the integration of diverse phenotypic data types,\nenabling the exploration of potential gene drivers underlying observed\nphenotypic changes. Its user-friendly interface and visualization capabilities\nenhance the accessibility of complex network analyses. Illustratively, we\nemployed evolSOM to study the displacement of genes and phenotypic traits,\nsuccessfully identifying potential drivers of phenotypic differentiation in\ngrass leaves. Availability: The package is open-source and is available at\nhttps://github.com/sanprochetto/evolSOM.\n", "date": "2024-02-09", "categories": []}
{"id": "2402.07955", "title": "ProtIR: Iterative Refinement between Retrievers and Predictors for\n  Protein Function Annotation", "abstract": "  Protein function annotation is an important yet challenging task in biology.\nRecent deep learning advancements show significant potential for accurate\nfunction prediction by learning from protein sequences and structures.\nNevertheless, these predictor-based methods often overlook the modeling of\nprotein similarity, an idea commonly employed in traditional approaches using\nsequence or structure retrieval tools. To fill this gap, we first study the\neffect of inter-protein similarity modeling by benchmarking retriever-based\nmethods against predictors on protein function annotation tasks. Our results\nshow that retrievers can match or outperform predictors without large-scale\npre-training. Building on these insights, we introduce a novel variational\npseudo-likelihood framework, ProtIR, designed to improve function predictors by\nincorporating inter-protein similarity modeling. This framework iteratively\nrefines knowledge between a function predictor and retriever, thereby combining\nthe strengths of both predictors and retrievers. ProtIR showcases around 10%\nimprovement over vanilla predictor-based methods. Besides, it achieves\nperformance on par with protein language model-based methods, yet without the\nneed for massive pre-training, highlighting the efficacy of our framework. Code\nwill be released upon acceptance.\n", "date": "2024-02-10", "categories": []}
{"id": "2402.08117", "title": "A Universal Non-Parametric Approach For Improved Molecular Sequence\n  Analysis", "abstract": "  In the field of biological research, it is essential to comprehend the\ncharacteristics and functions of molecular sequences. The classification of\nmolecular sequences has seen widespread use of neural network-based techniques.\nDespite their astounding accuracy, these models often require a substantial\nnumber of parameters and more data collection. In this work, we present a novel\napproach based on the compression-based Model, motivated from\n\\cite{jiang2023low}, which combines the simplicity of basic compression\nalgorithms like Gzip and Bz2, with Normalized Compression Distance (NCD)\nalgorithm to achieve better performance on classification tasks without relying\non handcrafted features or pre-trained models. Firstly, we compress the\nmolecular sequence using well-known compression algorithms, such as Gzip and\nBz2. By leveraging the latent structure encoded in compressed files, we compute\nthe Normalized Compression Distance between each pair of molecular sequences,\nwhich is derived from the Kolmogorov complexity. This gives us a distance\nmatrix, which is the input for generating a kernel matrix using a Gaussian\nkernel. Next, we employ kernel Principal Component Analysis (PCA) to get the\nvector representations for the corresponding molecular sequence, capturing\nimportant structural and functional information. The resulting vector\nrepresentations provide an efficient yet effective solution for molecular\nsequence analysis and can be used in ML-based downstream tasks. The proposed\napproach eliminates the need for computationally intensive Deep Neural Networks\n(DNNs), with their large parameter counts and data requirements. Instead, it\nleverages a lightweight and universally accessible compression-based model.\n", "date": "2024-02-12", "categories": []}
{"id": "2402.08198", "title": "PSC-CPI: Multi-Scale Protein Sequence-Structure Contrasting for\n  Efficient and Generalizable Compound-Protein Interaction Prediction", "abstract": "  Compound-Protein Interaction (CPI) prediction aims to predict the pattern and\nstrength of compound-protein interactions for rational drug discovery. Existing\ndeep learning-based methods utilize only the single modality of protein\nsequences or structures and lack the co-modeling of the joint distribution of\nthe two modalities, which may lead to significant performance drops in complex\nreal-world scenarios due to various factors, e.g., modality missing and domain\nshifting. More importantly, these methods only model protein sequences and\nstructures at a single fixed scale, neglecting more fine-grained multi-scale\ninformation, such as those embedded in key protein fragments. In this paper, we\npropose a novel multi-scale Protein Sequence-structure Contrasting framework\nfor CPI prediction (PSC-CPI), which captures the dependencies between protein\nsequences and structures through both intra-modality and cross-modality\ncontrasting. We further apply length-variable protein augmentation to allow\ncontrasting to be performed at different scales, from the amino acid level to\nthe sequence level. Finally, in order to more fairly evaluate the model\ngeneralizability, we split the test data into four settings based on whether\ncompounds and proteins have been observed during the training stage. Extensive\nexperiments have shown that PSC-CPI generalizes well in all four settings,\nparticularly in the more challenging ``Unseen-Both\" setting, where neither\ncompounds nor proteins have been observed during training. Furthermore, even\nwhen encountering a situation of modality missing, i.e., inference with only\nsingle-modality protein data, PSC-CPI still exhibits comparable or even better\nperformance than previous approaches.\n", "date": "2024-02-13", "categories": []}
{"id": "2310.04017", "title": "PGraphDTA: Improving Drug Target Interaction Prediction using Protein\n  Language Models and Contact Maps", "abstract": "  Developing and discovering new drugs is a complex and resource-intensive\nendeavor that often involves substantial costs, time investment, and safety\nconcerns. A key aspect of drug discovery involves identifying novel drug-target\n(DT) interactions. Existing computational methods for predicting DT\ninteractions have primarily focused on binary classification tasks, aiming to\ndetermine whether a DT pair interacts or not. However, protein-ligand\ninteractions exhibit a continuum of binding strengths, known as binding\naffinity, presenting a persistent challenge for accurate prediction. In this\nstudy, we investigate various techniques employed in Drug Target Interaction\n(DTI) prediction and propose novel enhancements to enhance their performance.\nOur approaches include the integration of Protein Language Models (PLMs) and\nthe incorporation of Contact Map information as an inductive bias within\ncurrent models. Through extensive experimentation, we demonstrate that our\nproposed approaches outperform the baseline models considered in this study,\npresenting a compelling case for further development in this direction. We\nanticipate that the insights gained from this work will significantly narrow\nthe search space for potential drugs targeting specific proteins, thereby\naccelerating drug discovery. Code and data for PGraphDTA are available at\nhttps://github.com/Yijia-Xiao/PgraphDTA/.\n", "date": "2023-10-06", "categories": []}
{"id": "2402.07268", "title": "Highly Accurate Disease Diagnosis and Highly Reproducible Biomarker\n  Identification with PathFormer", "abstract": "  Biomarker identification is critical for precise disease diagnosis and\nunderstanding disease pathogenesis in omics data analysis, like using fold\nchange and regression analysis. Graph neural networks (GNNs) have been the\ndominant deep learning model for analyzing graph-structured data. However, we\nfound two major limitations of existing GNNs in omics data analysis, i.e.,\nlimited-prediction (diagnosis) accuracy and limited-reproducible biomarker\nidentification capacity across multiple datasets. The root of the challenges is\nthe unique graph structure of biological signaling pathways, which consists of\na large number of targets and intensive and complex signaling interactions\namong these targets. To resolve these two challenges, in this study, we\npresented a novel GNN model architecture, named PathFormer, which\nsystematically integrate signaling network, priori knowledge and omics data to\nrank biomarkers and predict disease diagnosis. In the comparison results,\nPathFormer outperformed existing GNN models significantly in terms of highly\naccurate prediction capability ( 30% accuracy improvement in disease diagnosis\ncompared with existing GNN models) and high reproducibility of biomarker\nranking across different datasets. The improvement was confirmed using two\nindependent Alzheimer's Disease (AD) and cancer transcriptomic datasets. The\nPathFormer model can be directly applied to other omics data analysis studies.\n", "date": "2024-02-11", "categories": []}
{"id": "2309.03631", "title": "Insights Into the Inner Workings of Transformer Models for Protein\n  Function Prediction", "abstract": "  Motivation: We explored how explainable artificial intelligence (XAI) can\nhelp to shed light into the inner workings of neural networks for protein\nfunction prediction, by extending the widely used XAI method of integrated\ngradients such that latent representations inside of transformer models, which\nwere finetuned to Gene Ontology term and Enzyme Commission number prediction,\ncan be inspected too. Results: The approach enabled us to identify amino acids\nin the sequences that the transformers pay particular attention to, and to show\nthat these relevant sequence parts reflect expectations from biology and\nchemistry, both in the embedding layer and inside of the model, where we\nidentified transformer heads with a statistically significant correspondence of\nattribution maps with ground truth sequence annotations (e.g. transmembrane\nregions, active sites) across many proteins. Availability and Implementation:\nSource code can be accessed at https://github.com/markuswenzel/xai-proteins .\n", "date": "2023-09-07", "categories": []}
{"id": "2402.05953", "title": "idMotif: An Interactive Motif Identification in Protein Sequences", "abstract": "  This article introduces idMotif, a visual analytics framework designed to aid\ndomain experts in the identification of motifs within protein sequences.\nMotifs, short sequences of amino acids, are critical for understanding the\ndistinct functions of proteins. Identifying these motifs is pivotal for\npredicting diseases or infections. idMotif employs a deep learning-based method\nfor the categorization of protein sequences, enabling the discovery of\npotential motif candidates within protein groups through local explanations of\ndeep learning model decisions. It offers multiple interactive views for the\nanalysis of protein clusters or groups and their sequences. A case study,\ncomplemented by expert feedback, illustrates idMotif's utility in facilitating\nthe analysis and identification of protein sequences and motifs.\n", "date": "2024-02-04", "categories": []}
{"id": "2402.05543", "title": "Machine learning applied to omics data", "abstract": "  In this chapter we illustrate the use of some Machine Learning techniques in\nthe context of omics data. More precisely, we review and evaluate the use of\nRandom Forest and Penalized Multinomial Logistic Regression for integrative\nanalysis of genomics and immunomics in pancreatic cancer. Furthermore, we\npropose the use of association rules with predictive purposes to overcome the\nlow predictive power of the previously mentioned models. Finally, we apply the\nreviewed methods to a real data set from TCGA made of 107 tumoral pancreatic\nsamples and 117,486 germline SNPs, showing the good performance of the proposed\nmethods to predict the immunological infiltration in pancreatic cancer.\n", "date": "2024-02-08", "categories": []}
{"id": "2402.04268", "title": "ProtAgents: Protein discovery via large language model multi-agent\n  collaborations combining physics and machine learning", "abstract": "  Designing de novo proteins beyond those found in nature holds significant\npromise for advancements in both scientific and engineering applications.\nCurrent methodologies for protein design often rely on AI-based models, such as\nsurrogate models that address end-to-end problems by linking protein structure\nto material properties or vice versa. However, these models frequently focus on\nspecific material objectives or structural properties, limiting their\nflexibility when incorporating out-of-domain knowledge into the design process\nor comprehensive data analysis is required. In this study, we introduce\nProtAgents, a platform for de novo protein design based on Large Language\nModels (LLMs), where multiple AI agents with distinct capabilities\ncollaboratively address complex tasks within a dynamic environment. The\nversatility in agent development allows for expertise in diverse domains,\nincluding knowledge retrieval, protein structure analysis, physics-based\nsimulations, and results analysis. The dynamic collaboration between agents,\nempowered by LLMs, provides a versatile approach to tackling protein design and\nanalysis problems, as demonstrated through diverse examples in this study. The\nproblems of interest encompass designing new proteins, analyzing protein\nstructures and obtaining new first-principles data -- natural vibrational\nfrequencies -- via physics simulations. The concerted effort of the system\nallows for powerful automated and synergistic design of de novo proteins with\ntargeted mechanical properties. The flexibility in designing the agents, on one\nhand, and their capacity in autonomous collaboration through the dynamic\nLLM-based multi-agent environment on the other hand, unleashes great potentials\nof LLMs in addressing multi-objective materials problems and opens up new\navenues for autonomous materials discovery and design.\n", "date": "2024-01-27", "categories": []}
{"id": "2402.04286", "title": "Progress and Opportunities of Foundation Models in Bioinformatics", "abstract": "  Bioinformatics has witnessed a paradigm shift with the increasing integration\nof artificial intelligence (AI), particularly through the adoption of\nfoundation models (FMs). These AI techniques have rapidly advanced, addressing\nhistorical challenges in bioinformatics such as the scarcity of annotated data\nand the presence of data noise. FMs are particularly adept at handling\nlarge-scale, unlabeled data, a common scenario in biological contexts due to\nthe time-consuming and costly nature of experimentally determining labeled\ndata. This characteristic has allowed FMs to excel and achieve notable results\nin various downstream validation tasks, demonstrating their ability to\nrepresent diverse biological entities effectively. Undoubtedly, FMs have\nushered in a new era in computational biology, especially in the realm of deep\nlearning. The primary goal of this survey is to conduct a systematic\ninvestigation and summary of FMs in bioinformatics, tracing their evolution,\ncurrent research status, and the methodologies employed. Central to our focus\nis the application of FMs to specific biological problems, aiming to guide the\nresearch community in choosing appropriate FMs for their research needs. We\ndelve into the specifics of the problem at hand including sequence analysis,\nstructure prediction, function annotation, and multimodal integration,\ncomparing the structures and advancements against traditional methods.\nFurthermore, the review analyses challenges and limitations faced by FMs in\nbiology, such as data noise, model explainability, and potential biases.\nFinally, we outline potential development paths and strategies for FMs in\nfuture biological research, setting the stage for continued innovation and\napplication in this rapidly evolving field. This comprehensive review serves\nnot only as an academic resource but also as a roadmap for future explorations\nand applications of FMs in biology.\n", "date": "2024-02-06", "categories": []}
{"id": "2402.03675", "title": "Effective Protein-Protein Interaction Exploration with PPIretrieval", "abstract": "  Protein-protein interactions (PPIs) are crucial in regulating numerous\ncellular functions, including signal transduction, transportation, and immune\ndefense. As the accuracy of multi-chain protein complex structure prediction\nimproves, the challenge has shifted towards effectively navigating the vast\ncomplex universe to identify potential PPIs. Herein, we propose PPIretrieval,\nthe first deep learning-based model for protein-protein interaction\nexploration, which leverages existing PPI data to effectively search for\npotential PPIs in an embedding space, capturing rich geometric and chemical\ninformation of protein surfaces. When provided with an unseen query protein\nwith its associated binding site, PPIretrieval effectively identifies a\npotential binding partner along with its corresponding binding site in an\nembedding space, facilitating the formation of protein-protein complexes.\n", "date": "2024-02-06", "categories": []}
{"id": "2402.00926", "title": "A Comparative Analysis of Gene Expression Profiling by Statistical and\n  Machine Learning Approaches", "abstract": "  Many machine learning models have been proposed to classify phenotypes from\ngene expression data. In addition to their good performance, these models can\npotentially provide some understanding of phenotypes by extracting explanations\nfor their decisions. These explanations often take the form of a list of genes\nranked in order of importance for the predictions, the highest-ranked genes\nbeing interpreted as linked to the phenotype. We discuss the biological and the\nmethodological limitations of such explanations. Experiments are performed on\nseveral datasets gathering cancer and healthy tissue samples from the TCGA,\nGTEx and TARGET databases. A collection of machine learning models including\nlogistic regression, multilayer perceptron, and graph neural network are\ntrained to classify samples according to their cancer type. Gene rankings are\nobtained from explainability methods adapted to these models, and compared to\nthe ones from classical statistical feature selection methods such as mutual\ninformation, DESeq2, and EdgeR. Interestingly, on simple tasks, we observe that\nthe information learned by black-box neural networks is related to the notion\nof differential expression. In all cases, a small set containing the\nbest-ranked genes is sufficient to achieve a good classification. However,\nthese genes differ significantly between the methods and similar classification\nperformance can be achieved with numerous lower ranked genes. In conclusion,\nalthough these methods enable the identification of biomarkers characteristic\nof certain pathologies, our results question the completeness of the selected\ngene sets and thus of explainability by the identification of the underlying\nbiological processes.\n", "date": "2024-02-01", "categories": []}
{"id": "2401.16771", "title": "MolPLA: A Molecular Pretraining Framework for Learning Cores, R-Groups\n  and their Linker Joints", "abstract": "  Molecular core structures and R-groups are essential concepts in drug\ndevelopment. Integration of these concepts with conventional graph pre-training\napproaches can promote deeper understanding in molecules. We propose MolPLA, a\nnovel pre-training framework that employs masked graph contrastive learning in\nunderstanding the underlying decomposable parts inmolecules that implicate\ntheir core structure and peripheral R-groups. Furthermore, we formulate an\nadditional framework that grants MolPLA the ability to help chemists find\nreplaceable R-groups in lead optimization scenarios. Experimental results on\nmolecular property prediction show that MolPLA exhibits predictability\ncomparable to current state-of-the-art models. Qualitative analysis implicate\nthat MolPLA is capable of distinguishing core and R-group sub-structures,\nidentifying decomposable regions in molecules and contributing to lead\noptimization scenarios by rationally suggesting R-group replacements given\nvarious query core templates. The code implementation for MolPLA and its\npre-trained model checkpoint is available at https://github.com/dmis-lab/MolPLA\n", "date": "2024-01-30", "categories": []}
{"id": "2310.07276", "title": "BioT5: Enriching Cross-modal Integration in Biology with Chemical\n  Knowledge and Natural Language Associations", "abstract": "  Recent advancements in biological research leverage the integration of\nmolecules, proteins, and natural language to enhance drug discovery. However,\ncurrent models exhibit several limitations, such as the generation of invalid\nmolecular SMILES, underutilization of contextual information, and equal\ntreatment of structured and unstructured knowledge. To address these issues, we\npropose $\\mathbf{BioT5}$, a comprehensive pre-training framework that enriches\ncross-modal integration in biology with chemical knowledge and natural language\nassociations. $\\mathbf{BioT5}$ utilizes SELFIES for $100%$ robust molecular\nrepresentations and extracts knowledge from the surrounding context of\nbio-entities in unstructured biological literature. Furthermore,\n$\\mathbf{BioT5}$ distinguishes between structured and unstructured knowledge,\nleading to more effective utilization of information. After fine-tuning, BioT5\nshows superior performance across a wide range of tasks, demonstrating its\nstrong capability of capturing underlying relations and properties of\nbio-entities. Our code is available at\n$\\href{https://github.com/QizhiPei/BioT5}{Github}$.\n", "date": "2023-10-11", "categories": []}
{"id": "2401.16094", "title": "Federated unsupervised random forest for privacy-preserving patient\n  stratification", "abstract": "  In the realm of precision medicine, effective patient stratification and\ndisease subtyping demand innovative methodologies tailored for multi-omics\ndata. Clustering techniques applied to multi-omics data have become\ninstrumental in identifying distinct subgroups of patients, enabling a\nfiner-grained understanding of disease variability. This work establishes a\npowerful framework for advancing precision medicine through unsupervised\nrandom-forest-based clustering and federated computing. We introduce a novel\nmulti-omics clustering approach utilizing unsupervised random-forests. The\nunsupervised nature of the random forest enables the determination of\ncluster-specific feature importance, unraveling key molecular contributors to\ndistinct patient groups. Moreover, our methodology is designed for federated\nexecution, a crucial aspect in the medical domain where privacy concerns are\nparamount. We have validated our approach on machine learning benchmark data\nsets as well as on cancer data from The Cancer Genome Atlas (TCGA). Our method\nis competitive with the state-of-the-art in terms of disease subtyping, but at\nthe same time substantially improves the cluster interpretability. Experiments\nindicate that local clustering performance can be improved through federated\ncomputing.\n", "date": "2024-01-29", "categories": []}
{"id": "2401.14665", "title": "PepGB: Facilitating peptide drug discovery via graph neural networks", "abstract": "  Peptides offer great biomedical potential and serve as promising drug\ncandidates. Currently, the majority of approved peptide drugs are directly\nderived from well-explored natural human peptides. It is quite necessary to\nutilize advanced deep learning techniques to identify novel peptide drugs in\nthe vast, unexplored biochemical space. Despite various in silico methods\nhaving been developed to accelerate peptide early drug discovery, existing\nmodels face challenges of overfitting and lacking generalizability due to the\nlimited size, imbalanced distribution and inconsistent quality of experimental\ndata. In this study, we propose PepGB, a deep learning framework to facilitate\npeptide early drug discovery by predicting peptide-protein interactions\n(PepPIs). Employing graph neural networks, PepGB incorporates a fine-grained\nperturbation module and a dual-view objective with contrastive learning-based\npeptide pre-trained representation to predict PepPIs. Through rigorous\nevaluations, we demonstrated that PepGB greatly outperforms baselines and can\naccurately identify PepPIs for novel targets and peptide hits, thereby\ncontributing to the target identification and hit discovery processes. Next, we\nderive an extended version, diPepGB, to tackle the bottleneck of modeling\nhighly imbalanced data prevalent in lead generation and optimization processes.\nUtilizing directed edges to represent relative binding strength between two\npeptide nodes, diPepGB achieves superior performance in real-world assays. In\nsummary, our proposed frameworks can serve as potent tools to facilitate\npeptide early drug discovery.\n", "date": "2024-01-26", "categories": []}
{"id": "2401.14025", "title": "DNA Sequence Classification with Compressors", "abstract": "  Recent studies in DNA sequence classification have leveraged sophisticated\nmachine learning techniques, achieving notable accuracy in categorizing complex\ngenomic data. Among these, methods such as k-mer counting have proven effective\nin distinguishing sequences from varied species like chimpanzees, dogs, and\nhumans, becoming a staple in contemporary genomic research. However, these\napproaches often demand extensive computational resources, posing a challenge\nin terms of scalability and efficiency. Addressing this issue, our study\nintroduces a novel adaptation of Jiang et al.'s compressor-based,\nparameter-free classification method, specifically tailored for DNA sequence\nanalysis. This innovative approach utilizes a variety of compression\nalgorithms, such as Gzip, Brotli, and LZMA, to efficiently process and classify\ngenomic sequences. Not only does this method align with the current\nstate-of-the-art in terms of accuracy, but it also offers a more\nresource-efficient alternative to traditional machine learning methods. Our\ncomprehensive evaluation demonstrates the proposed method's effectiveness in\naccurately classifying DNA sequences from multiple species. We present a\ndetailed analysis of the performance of each algorithm used, highlighting the\nstrengths and limitations of our approach in various genomic contexts.\nFurthermore, we discuss the broader implications of our findings for\nbioinformatics, particularly in genomic data processing and analysis. The\nresults of our study pave the way for more efficient and scalable DNA sequence\nclassification methods, offering significant potential for advancements in\ngenomic research and applications.\n", "date": "2024-01-25", "categories": []}
{"id": "2302.13711", "title": "Internal-Coordinate Density Modelling of Protein Structure: Covariance\n  Matters", "abstract": "  After the recent ground-breaking advances in protein structure prediction,\none of the remaining challenges in protein machine learning is to reliably\npredict distributions of structural states. Parametric models of fluctuations\nare difficult to fit due to complex covariance structures between degrees of\nfreedom in the protein chain, often causing models to either violate local or\nglobal structural constraints. In this paper, we present a new strategy for\nmodelling protein densities in internal coordinates, which uses constraints in\n3D space to induce covariance structure between the internal degrees of\nfreedom. We illustrate the potential of the procedure by constructing a\nvariational autoencoder with full covariance output induced by the constraints\nimplied by the conditional mean in 3D, and demonstrate that our approach makes\nit possible to scale density models of internal coordinates to full protein\nbackbones in two settings: 1) a unimodal setting for proteins exhibiting small\nfluctuations and limited amounts of available data, and 2) a multimodal setting\nfor larger conformational changes in a high data regime.\n", "date": "2023-02-27", "categories": []}
{"id": "2306.10456", "title": "Advancing Biomedicine with Graph Representation Learning: Recent\n  Progress, Challenges, and Future Directions", "abstract": "  Graph representation learning (GRL) has emerged as a pivotal field that has\ncontributed significantly to breakthroughs in various fields, including\nbiomedicine. The objective of this survey is to review the latest advancements\nin GRL methods and their applications in the biomedical field. We also\nhighlight key challenges currently faced by GRL and outline potential\ndirections for future research.\n", "date": "2023-06-18", "categories": []}
{"id": "2401.10608", "title": "M2ORT: Many-To-One Regression Transformer for Spatial Transcriptomics\n  Prediction from Histopathology Images", "abstract": "  The advancement of Spatial Transcriptomics (ST) has facilitated the\nspatially-aware profiling of gene expressions based on histopathology images.\nAlthough ST data offers valuable insights into the micro-environment of tumors,\nits acquisition cost remains expensive. Therefore, directly predicting the ST\nexpressions from digital pathology images is desired. Current methods usually\nadopt existing regression backbones for this task, which ignore the inherent\nmulti-scale hierarchical data structure of digital pathology images. To address\nthis limit, we propose M2ORT, a many-to-one regression Transformer that can\naccommodate the hierarchical structure of the pathology images through a\ndecoupled multi-scale feature extractor. Different from traditional models that\nare trained with one-to-one image-label pairs, M2ORT accepts multiple pathology\nimages of different magnifications at a time to jointly predict the gene\nexpressions at their corresponding common ST spot, aiming at learning a\nmany-to-one relationship through training. We have tested M2ORT on three public\nST datasets and the experimental results show that M2ORT can achieve\nstate-of-the-art performance with fewer parameters and floating-point\noperations (FLOPs). The code is available at:\nhttps://github.com/Dootmaan/M2ORT/.\n", "date": "2024-01-19", "categories": []}
{"id": "2401.13219", "title": "TEPI: Taxonomy-aware Embedding and Pseudo-Imaging for Scarcely-labeled\n  Zero-shot Genome Classification", "abstract": "  A species' genetic code or genome encodes valuable evolutionary, biological,\nand phylogenetic information that aids in species recognition, taxonomic\nclassification, and understanding genetic predispositions like drug resistance\nand virulence. However, the vast number of potential species poses significant\nchallenges in developing a general-purpose whole genome classification tool.\nTraditional bioinformatics tools have made notable progress but lack\nscalability and are computationally expensive. Machine learning-based\nframeworks show promise but must address the issue of large classification\nvocabularies with long-tail distributions. In this study, we propose addressing\nthis problem through zero-shot learning using TEPI, Taxonomy-aware Embedding\nand Pseudo-Imaging. We represent each genome as pseudo-images and map them to a\ntaxonomy-aware embedding space for reasoning and classification. This embedding\nspace captures compositional and phylogenetic relationships of species,\nenabling predictions in extensive search spaces. We evaluate TEPI using two\nrigorous zero-shot settings and demonstrate its generalization capabilities\nqualitatively on curated, large-scale, publicly sourced data.\n", "date": "2024-01-24", "categories": []}
{"id": "2306.05739", "title": "Leaping through tree space: continuous phylogenetic inference for rooted\n  and unrooted trees", "abstract": "  Phylogenetics is now fundamental in life sciences, providing insights into\nthe earliest branches of life and the origins and spread of epidemics. However,\nfinding suitable phylogenies from the vast space of possible trees remains\nchallenging. To address this problem, for the first time, we perform both tree\nexploration and inference in a continuous space where the computation of\ngradients is possible. This continuous relaxation allows for major leaps across\ntree space in both rooted and unrooted trees, and is less susceptible to\nconvergence to local minima. Our approach outperforms the current best methods\nfor inference on unrooted trees and, in simulation, accurately infers the tree\nand root in ultrametric cases. The approach is effective in cases of empirical\ndata with negligible amounts of data, which we demonstrate on the phylogeny of\njawed vertebrates. Indeed, only a few genes with an ultrametric signal were\ngenerally sufficient for resolving the major lineages of vertebrates.\nOptimisation is possible via automatic differentiation and our method presents\nan effective way forwards for exploring the most difficult, data-deficient\nphylogenetic questions.\n", "date": "2023-06-09", "categories": []}
{"id": "2401.12232", "title": "Machine Learning Modeling Of SiRNA Structure-Potency Relationship With\n  Applications Against Sars-Cov-2 Spike Gene", "abstract": "  The pharmaceutical Research and development (R&D) process is lengthy and\ncostly, taking nearly a decade to bring a new drug to the market. However,\nadvancements in biotechnology, computational methods, and machine learning\nalgorithms have the potential to revolutionize drug discovery, speeding up the\nprocess and improving patient outcomes. The COVID-19 pandemic has further\naccelerated and deepened the recognition of the potential of these techniques,\nespecially in the areas of drug repurposing and efficacy predictions.\nMeanwhile, non-small molecule therapeutic modalities such as cell therapies,\nmonoclonal antibodies, and RNA interference (RNAi) technology have gained\nimportance due to their ability to target specific disease pathways and/or\npatient populations. In the field of RNAi, many experiments have been carried\nout to design and select highly efficient siRNAs. However, the established\npatterns for efficient siRNAs are sometimes contradictory and unable to\nconsistently determine the most potent siRNA molecules against a target mRNA.\nThus, this paper focuses on developing machine learning models based on the\ncheminformatics representation of the nucleotide composition (i.e. AUTGC) of\nsiRNA to predict their potency and aid the selection of the most efficient\nsiRNAs for further development. The PLS (Partial Least Square) and SVR (Support\nVector Regression) machine learning models built in this work outperformed\npreviously published models. These models can help in predicting siRNA potency\nand aid in selecting the best siRNA molecules for experimental validation and\nfurther clinical development. The study has demonstrated the potential of\nAI/machine learning models to help expedite siRNA-based drug discovery\nincluding the discovery of potent siRNAs against SARS-CoV-2.\n", "date": "2024-01-18", "categories": []}
{"id": "2312.00842", "title": "ESM-NBR: fast and accurate nucleic acid-binding residue prediction via\n  protein language model feature representation and multi-task learning", "abstract": "  Protein-nucleic acid interactions play a very important role in a variety of\nbiological activities. Accurate identification of nucleic acid-binding residues\nis a critical step in understanding the interaction mechanisms. Although many\ncomputationally based methods have been developed to predict nucleic\nacid-binding residues, challenges remain. In this study, a fast and accurate\nsequence-based method, called ESM-NBR, is proposed. In ESM-NBR, we first use\nthe large protein language model ESM2 to extract discriminative biological\nproperties feature representation from protein primary sequences; then, a\nmulti-task deep learning model composed of stacked bidirectional long\nshort-term memory (BiLSTM) and multi-layer perceptron (MLP) networks is\nemployed to explore common and private information of DNA- and RNA-binding\nresidues with ESM2 feature as input. Experimental results on benchmark data\nsets demonstrate that the prediction performance of ESM2 feature representation\ncomprehensively outperforms evolutionary information-based hidden Markov model\n(HMM) features. Meanwhile, the ESM-NBR obtains the MCC values for DNA-binding\nresidues prediction of 0.427 and 0.391 on two independent test sets, which are\n18.61 and 10.45% higher than those of the second-best methods, respectively.\nMoreover, by completely discarding the time-cost multiple sequence alignment\nprocess, the prediction speed of ESM-NBR far exceeds that of existing methods\n(5.52s for a protein sequence of length 500, which is about 16 times faster\nthan the second-fastest method). A user-friendly standalone package and the\ndata of ESM-NBR are freely available for academic use at:\nhttps://github.com/wwzll123/ESM-NBR.\n", "date": "2023-12-01", "categories": []}
{"id": "2401.07379", "title": "Inference of dynamical gene regulatory networks from single-cell data\n  with physics informed neural networks", "abstract": "  One of the main goals of developmental biology is to reveal the gene\nregulatory networks (GRNs) underlying the robust differentiation of multipotent\nprogenitors into precisely specified cell types. Most existing methods to infer\nGRNs from experimental data have limited predictive power as the inferred GRNs\nmerely reflect gene expression similarity or correlation. Here, we demonstrate,\nhow physics-informed neural networks (PINNs) can be used to infer the\nparameters of predictive, dynamical GRNs that provide mechanistic understanding\nof biological processes. Specifically we study GRNs that exhibit bifurcation\nbehavior and can therefore model cell differentiation. We show that PINNs\noutperform regular feed-forward neural networks on the parameter inference task\nand analyze two relevant experimental scenarios: 1. a system with cell\ncommunication for which gene expression trajectories are available and 2.\nsnapshot measurements of a cell population in which cell communication is\nabsent. Our analysis will inform the design of future experiments to be\nanalyzed with PINNs and provides a starting point to explore this powerful\nclass of neural network models further.\n", "date": "2024-01-14", "categories": []}
{"id": "2401.11360", "title": "PepHarmony: A Multi-View Contrastive Learning Framework for Integrated\n  Sequence and Structure-Based Peptide Encoding", "abstract": "  Recent advances in protein language models have catalyzed significant\nprogress in peptide sequence representation. Despite extensive exploration in\nthis field, pre-trained models tailored for peptide-specific needs remain\nlargely unaddressed due to the difficulty in capturing the complex and\nsometimes unstable structures of peptides. This study introduces a novel\nmulti-view contrastive learning framework PepHarmony for the sequence-based\npeptide encoding task. PepHarmony innovatively combines both sequence- and\nstructure-level information into a sequence-level encoding module through\ncontrastive learning. We carefully select datasets from the Protein Data Bank\n(PDB) and AlphaFold database to encompass a broad spectrum of peptide sequences\nand structures. The experimental data highlights PepHarmony's exceptional\ncapability in capturing the intricate relationship between peptide sequences\nand structures compared with the baseline and fine-tuned models. The robustness\nof our model is confirmed through extensive ablation studies, which emphasize\nthe crucial roles of contrastive loss and strategic data sorting in enhancing\npredictive performance. The proposed PepHarmony framework serves as a notable\ncontribution to peptide representations, and offers valuable insights for\nfuture applications in peptide drug discovery and peptide engineering. We have\nmade all the source code utilized in this study publicly accessible via GitHub\nat https://github.com/zhangruochi/PepHarmony or\nhttp://www.healthinformaticslab.org/supp/.\n", "date": "2024-01-21", "categories": []}
{"id": "2302.06120", "title": "Knowledge from Large-Scale Protein Contact Prediction Models Can Be\n  Transferred to the Data-Scarce RNA Contact Prediction Task", "abstract": "  RNA, whose functionality is largely determined by its structure, plays an\nimportant role in many biological activities. The prediction of pairwise\nstructural proximity between each nucleotide of an RNA sequence can\ncharacterize the structural information of the RNA. Historically, this problem\nhas been tackled by machine learning models using expert-engineered features\nand trained on scarce labeled datasets. Here, we find that the knowledge\nlearned by a protein-coevolution Transformer-based deep neural network can be\ntransferred to the RNA contact prediction task. As protein datasets are orders\nof magnitude larger than those for RNA contact prediction, our findings and the\nsubsequent framework greatly reduce the data scarcity bottleneck. Experiments\nconfirm that RNA contact prediction through transfer learning using a publicly\navailable protein model is greatly improved. Our findings indicate that the\nlearned structural patterns of proteins can be transferred to RNAs, opening up\npotential new avenues for research.\n", "date": "2023-02-13", "categories": []}
{"id": "2310.03320", "title": "BioBridge: Bridging Biomedical Foundation Models via Knowledge Graphs", "abstract": "  Foundation models (FMs) are able to leverage large volumes of unlabeled data\nto demonstrate superior performance across a wide range of tasks. However, FMs\ndeveloped for biomedical domains have largely remained unimodal, i.e.,\nindependently trained and used for tasks on protein sequences alone, small\nmolecule structures alone, or clinical data alone. To overcome this limitation\nof biomedical FMs, we present BioBridge, a novel parameter-efficient learning\nframework, to bridge independently trained unimodal FMs to establish multimodal\nbehavior. BioBridge achieves it by utilizing Knowledge Graphs (KG) to learn\ntransformations between one unimodal FM and another without fine-tuning any\nunderlying unimodal FMs. Our empirical results demonstrate that BioBridge can\nbeat the best baseline KG embedding methods (on average by around 76.3%) in\ncross-modal retrieval tasks. We also identify BioBridge demonstrates\nout-of-domain generalization ability by extrapolating to unseen modalities or\nrelations. Additionally, we also show that BioBridge presents itself as a\ngeneral purpose retriever that can aid biomedical multimodal question answering\nas well as enhance the guided generation of novel drugs.\n", "date": "2023-10-05", "categories": []}
{"id": "2401.09840", "title": "FREED++: Improving RL Agents for Fragment-Based Molecule Generation by\n  Thorough Reproduction", "abstract": "  A rational design of new therapeutic drugs aims to find a molecular structure\nwith desired biological functionality, e.g., an ability to activate or suppress\na specific protein via binding to it. Molecular docking is a common technique\nfor evaluating protein-molecule interactions. Recently, Reinforcement Learning\n(RL) has emerged as a promising approach to generating molecules with the\ndocking score (DS) as a reward. In this work, we reproduce, scrutinize and\nimprove the recent RL model for molecule generation called FREED\n(arXiv:2110.01219). Extensive evaluation of the proposed method reveals several\nlimitations and challenges despite the outstanding results reported for three\ntarget proteins. Our contributions include fixing numerous implementation bugs\nand simplifying the model while increasing its quality, significantly extending\nexperiments, and conducting an accurate comparison with current\nstate-of-the-art methods for protein-conditioned molecule generation. We show\nthat the resulting fixed model is capable of producing molecules with superior\ndocking scores compared to alternative approaches.\n", "date": "2024-01-18", "categories": []}
{"id": "2401.10144", "title": "Exploiting Hierarchical Interactions for Protein Surface Learning", "abstract": "  Predicting interactions between proteins is one of the most important yet\nchallenging problems in structural bioinformatics. Intrinsically, potential\nfunction sites in protein surfaces are determined by both geometric and\nchemical features. However, existing works only consider handcrafted or\nindividually learned chemical features from the atom type and extract geometric\nfeatures independently. Here, we identify two key properties of effective\nprotein surface learning: 1) relationship among atoms: atoms are linked with\neach other by covalent bonds to form biomolecules instead of appearing alone,\nleading to the significance of modeling the relationship among atoms in\nchemical feature learning. 2) hierarchical feature interaction: the neighboring\nresidue effect validates the significance of hierarchical feature interaction\namong atoms and between surface points and atoms (or residues). In this paper,\nwe present a principled framework based on deep learning techniques, namely\nHierarchical Chemical and Geometric Feature Interaction Network (HCGNet), for\nprotein surface analysis by bridging chemical and geometric features with\nhierarchical interactions. Extensive experiments demonstrate that our method\noutperforms the prior state-of-the-art method by 2.3% in site prediction task\nand 3.2% in interaction matching task, respectively. Our code is available at\nhttps://github.com/xmed-lab/HCGNet.\n", "date": "2024-01-17", "categories": []}
{"id": "2401.08986", "title": "Rigid Protein-Protein Docking via Equivariant Elliptic-Paraboloid\n  Interface Prediction", "abstract": "  The study of rigid protein-protein docking plays an essential role in a\nvariety of tasks such as drug design and protein engineering. Recently, several\nlearning-based methods have been proposed for the task, exhibiting much faster\ndocking speed than those computational methods. In this paper, we propose a\nnovel learning-based method called ElliDock, which predicts an elliptic\nparaboloid to represent the protein-protein docking interface. To be specific,\nour model estimates elliptic paraboloid interfaces for the two input proteins\nrespectively, and obtains the roto-translation transformation for docking by\nmaking two interfaces coincide. By its design, ElliDock is independently\nequivariant with respect to arbitrary rotations/translations of the proteins,\nwhich is an indispensable property to ensure the generalization of the docking\nprocess. Experimental evaluations show that ElliDock achieves the fastest\ninference time among all compared methods and is strongly competitive with\ncurrent state-of-the-art learning-based models such as DiffDock-PP and Multimer\nparticularly for antibody-antigen docking.\n", "date": "2024-01-17", "categories": []}
{"id": "2401.09176", "title": "ADCNet: a unified framework for predicting the activity of antibody-drug\n  conjugates", "abstract": "  Antibody-drug conjugate (ADC) has revolutionized the field of cancer\ntreatment in the era of precision medicine due to their ability to precisely\ntarget cancer cells and release highly effective drug. Nevertheless, the\nrealization of rational design of ADC is very difficult because the\nrelationship between their structures and activities is difficult to\nunderstand. In the present study, we introduce a unified deep learning\nframework called ADCNet to help design potential ADCs. The ADCNet highly\nintegrates the protein representation learning language model ESM-2 and\nsmall-molecule representation learning language model FG-BERT models to achieve\nactivity prediction through learning meaningful features from antigen and\nantibody protein sequences of ADC, SMILES strings of linker and payload, and\ndrug-antibody ratio (DAR) value. Based on a carefully designed and manually\ntailored ADC data set, extensive evaluation results reveal that ADCNet performs\nbest on the test set compared to baseline machine learning models across all\nevaluation metrics. For example, it achieves an average prediction accuracy of\n87.12%, a balanced accuracy of 0.8689, and an area under receiver operating\ncharacteristic curve of 0.9293 on the test set. In addition, cross-validation,\nablation experiments, and external independent testing results further prove\nthe stability, advancement, and robustness of the ADCNet architecture. For the\nconvenience of the community, we develop the first online platform\n(https://ADCNet.idruglab.cn) for the prediction of ADCs activity based on the\noptimal ADCNet model, and the source code is publicly available at\nhttps://github.com/idrugLab/ADCNet.\n", "date": "2024-01-17", "categories": []}
{"id": "2208.10962", "title": "Prediction of good reaction coordinates and future evolution of MD\n  trajectories using Regularized Sparse Autoencoders: A novel deep learning\n  approach", "abstract": "  Identifying reaction coordinates(RCs) is an active area of research, given\nthe crucial role RCs play in determining the progress of a chemical reaction.\nThe choice of the reaction coordinate is often based on heuristic knowledge.\nHowever, an essential criterion for the choice is that the coordinate should\ncapture both the reactant and product states unequivocally. Also, the\ncoordinate should be the slowest one so that all the other degrees of freedom\ncan easily equilibrate along the reaction coordinate. Also, the coordinate\nshould be the slowest one so that all the other degrees of freedom can easily\nequilibrate along the reaction coordinate. We used a regularised sparse\nautoencoder, an energy-based model, to discover a crucial set of reaction\ncoordinates. Along with discovering reaction coordinates, our model also\npredicts the evolution of a molecular dynamics(MD) trajectory. We showcased\nthat including sparsity enforcing regularisation helps in choosing a small but\nimportant set of reaction coordinates. We used two model systems to demonstrate\nour approach: alanine dipeptide system and proflavine and DNA system, which\nexhibited intercalation of proflavine into DNA minor groove in an aqueous\nenvironment. We model MD trajectory as a multivariate time series, and our\nlatent variable model performs the task of multi-step time series prediction.\nThis idea is inspired by the popular sparse coding approach - to represent each\ninput sample as a linear combination of few elements taken from a set of\nrepresentative patterns.\n", "date": "2022-08-22", "categories": []}
{"id": "2303.15228", "title": "Interpretable machine learning of amino acid patterns in proteins: a\n  statistical ensemble approach", "abstract": "  Explainable and interpretable unsupervised machine learning helps understand\nthe underlying structure of data. We introduce an ensemble analysis of machine\nlearning models to consolidate their interpretation. Its application shows that\nrestricted Boltzmann machines compress consistently into a few bits the\ninformation stored in a sequence of five amino acids at the start or end of\n$\\alpha$-helices or $\\beta$-sheets. The weights learned by the machines reveal\nunexpected properties of the amino acids and the secondary structure of\nproteins: (i) His and Thr have a negligible contribution to the amphiphilic\npattern of $\\alpha$-helices; (ii) there is a class of $\\alpha$-helices\nparticularly rich in Ala at their end; (iii) Pro occupies most often slots\notherwise occupied by polar or charged amino acids, and its presence at the\nstart of helices is relevant; (iv) Glu and especially Asp on one side, and Val,\nLeu, Iso, and Phe on the other, display the strongest tendency to mark\namphiphilic patterns, i.e., extreme values of an \"effective hydrophobicity\",\nthough they are not the most powerful (non) hydrophobic amino acids.\n", "date": "2023-03-27", "categories": []}
{"id": "2311.02794", "title": "Modelling Cellular Perturbations with the Sparse Additive Mechanism\n  Shift Variational Autoencoder", "abstract": "  Generative models of observations under interventions have been a vibrant\ntopic of interest across machine learning and the sciences in recent years. For\nexample, in drug discovery, there is a need to model the effects of diverse\ninterventions on cells in order to characterize unknown biological mechanisms\nof action. We propose the Sparse Additive Mechanism Shift Variational\nAutoencoder, SAMS-VAE, to combine compositionality, disentanglement, and\ninterpretability for perturbation models. SAMS-VAE models the latent state of a\nperturbed sample as the sum of a local latent variable capturing\nsample-specific variation and sparse global variables of latent intervention\neffects. Crucially, SAMS-VAE sparsifies these global latent variables for\nindividual perturbations to identify disentangled, perturbation-specific latent\nsubspaces that are flexibly composable. We evaluate SAMS-VAE both\nquantitatively and qualitatively on a range of tasks using two popular single\ncell sequencing datasets. In order to measure perturbation-specific\nmodel-properties, we also introduce a framework for evaluation of perturbation\nmodels based on average treatment effects with links to posterior predictive\nchecks. SAMS-VAE outperforms comparable models in terms of generalization\nacross in-distribution and out-of-distribution tasks, including a combinatorial\nreasoning task under resource paucity, and yields interpretable latent\nstructures which correlate strongly to known biological mechanisms. Our results\nsuggest SAMS-VAE is an interesting addition to the modeling toolkit for machine\nlearning-driven scientific discovery.\n", "date": "2023-11-05", "categories": []}
{"id": "2312.07899", "title": "Morphological Profiling for Drug Discovery in the Era of Deep Learning", "abstract": "  Morphological profiling is a valuable tool in phenotypic drug discovery. The\nadvent of high-throughput automated imaging has enabled the capturing of a wide\nrange of morphological features of cells or organisms in response to\nperturbations at the single-cell resolution. Concurrently, significant advances\nin machine learning and deep learning, especially in computer vision, have led\nto substantial improvements in analyzing large-scale high-content images at\nhigh-throughput. These efforts have facilitated understanding of compound\nmechanism-of-action (MOA), drug repurposing, characterization of cell\nmorphodynamics under perturbation, and ultimately contributing to the\ndevelopment of novel therapeutics. In this review, we provide a comprehensive\noverview of the recent advances in the field of morphological profiling. We\nsummarize the image profiling analysis workflow, survey a broad spectrum of\nanalysis strategies encompassing feature engineering- and deep learning-based\napproaches, and introduce publicly available benchmark datasets. We place a\nparticular emphasis on the application of deep learning in this pipeline,\ncovering cell segmentation, image representation learning, and multimodal\nlearning. Additionally, we illuminate the application of morphological\nprofiling in phenotypic drug discovery and highlight potential challenges and\nopportunities in this field.\n", "date": "2023-12-13", "categories": []}
{"id": "2401.07543", "title": "Must: Maximizing Latent Capacity of Spatial Transcriptomics Data", "abstract": "  Spatial transcriptomics (ST) technologies have revolutionized the study of\ngene expression patterns in tissues by providing multimodality data in\ntranscriptomic, spatial, and morphological, offering opportunities for\nunderstanding tissue biology beyond transcriptomics. However, we identify the\nmodality bias phenomenon in ST data species, i.e., the inconsistent\ncontribution of different modalities to the labels leads to a tendency for the\nanalysis methods to retain the information of the dominant modality. How to\nmitigate the adverse effects of modality bias to satisfy various downstream\ntasks remains a fundamental challenge. This paper introduces Multiple-modality\nStructure Transformation, named MuST, a novel methodology to tackle the\nchallenge. MuST integrates the multi-modality information contained in the ST\ndata effectively into a uniform latent space to provide a foundation for all\nthe downstream tasks. It learns intrinsic local structures by topology\ndiscovery strategy and topology fusion loss function to solve the\ninconsistencies among different modalities. Thus, these topology-based and deep\nlearning techniques provide a solid foundation for a variety of analytical\ntasks while coordinating different modalities. The effectiveness of MuST is\nassessed by performance metrics and biological significance. The results show\nthat it outperforms existing state-of-the-art methods with clear advantages in\nthe precision of identifying and preserving structures of tissues and\nbiomarkers. MuST offers a versatile toolkit for the intricate analysis of\ncomplex biological systems.\n", "date": "2024-01-15", "categories": []}
{"id": "2401.07937", "title": "Integrate Any Omics: Towards genome-wide data integration for patient\n  stratification", "abstract": "  High-throughput omics profiling advancements have greatly enhanced cancer\npatient stratification. However, incomplete data in multi-omics integration\npresents a significant challenge, as traditional methods like sample exclusion\nor imputation often compromise biological diversity and dependencies.\nFurthermore, the critical task of accurately classifying new patients with\npartial omics data into existing subtypes is commonly overlooked. To address\nthese issues, we introduce IntegrAO (Integrate Any Omics), an unsupervised\nframework for integrating incomplete multi-omics data and classifying new\nsamples. IntegrAO first combines partially overlapping patient graphs from\ndiverse omics sources and utilizes graph neural networks to produce unified\npatient embeddings. Our systematic evaluation across five cancer cohorts\ninvolving six omics modalities demonstrates IntegrAO's robustness to missing\ndata and its accuracy in classifying new samples with partial profiles. An\nacute myeloid leukemia case study further validates its capability to uncover\nbiological and clinical heterogeneity in incomplete datasets. IntegrAO's\nability to handle heterogeneous and incomplete data makes it an essential tool\nfor precision oncology, offering a holistic approach to patient\ncharacterization.\n", "date": "2024-01-15", "categories": []}
{"id": "2309.01036", "title": "SEPAL: Spatial Gene Expression Prediction from Local Graphs", "abstract": "  Spatial transcriptomics is an emerging technology that aligns histopathology\nimages with spatially resolved gene expression profiling. It holds the\npotential for understanding many diseases but faces significant bottlenecks\nsuch as specialized equipment and domain expertise. In this work, we present\nSEPAL, a new model for predicting genetic profiles from visual tissue\nappearance. Our method exploits the biological biases of the problem by\ndirectly supervising relative differences with respect to mean expression, and\nleverages local visual context at every coordinate to make predictions using a\ngraph neural network. This approach closes the gap between complete locality\nand complete globality in current methods. In addition, we propose a novel\nbenchmark that aims to better define the task by following current best\npractices in transcriptomics and restricting the prediction variables to only\nthose with clear spatial patterns. Our extensive evaluation in two different\nhuman breast cancer datasets indicates that SEPAL outperforms previous\nstate-of-the-art methods and other mechanisms of including spatial context.\n", "date": "2023-09-02", "categories": []}
{"id": "2305.09480", "title": "Cross-Gate MLP with Protein Complex Invariant Embedding is A One-Shot\n  Antibody Designer", "abstract": "  Antibodies are crucial proteins produced by the immune system in response to\nforeign substances or antigens. The specificity of an antibody is determined by\nits complementarity-determining regions (CDRs), which are located in the\nvariable domains of the antibody chains and form the antigen-binding site.\nPrevious studies have utilized complex techniques to generate CDRs, but they\nsuffer from inadequate geometric modeling. Moreover, the common iterative\nrefinement strategies lead to an inefficient inference. In this paper, we\npropose a \\textit{simple yet effective} model that can co-design 1D sequences\nand 3D structures of CDRs in a one-shot manner. To achieve this, we decouple\nthe antibody CDR design problem into two stages: (i) geometric modeling of\nprotein complex structures and (ii) sequence-structure co-learning. We develop\na novel macromolecular structure invariant embedding, typically for protein\ncomplexes, that captures both intra- and inter-component interactions among the\nbackbone atoms, including C$\\alpha$, N, C, and O atoms, to achieve\ncomprehensive geometric modeling. Then, we introduce a simple cross-gate MLP\nfor sequence-structure co-learning, allowing sequence and structure\nrepresentations to implicitly refine each other. This enables our model to\ndesign desired sequences and structures in a one-shot manner. Extensive\nexperiments are conducted to evaluate our results at both the sequence and\nstructure levels, which demonstrate that our model achieves superior\nperformance compared to the state-of-the-art antibody CDR design methods.\n", "date": "2023-04-21", "categories": []}
{"id": "2310.04343", "title": "Functional Geometry Guided Protein Sequence and Backbone Structure\n  Co-Design", "abstract": "  Proteins are macromolecules responsible for essential functions in almost all\nliving organisms. Designing reasonable proteins with desired functions is\ncrucial. A protein's sequence and structure are strongly correlated and they\ntogether determine its function. In this paper, we propose NAEPro, a model to\njointly design Protein sequence and structure based on automatically detected\nfunctional sites. NAEPro is powered by an interleaving network of attention and\nequivariant layers, which can capture global correlation in a whole sequence\nand local influence from nearest amino acids in three dimensional (3D) space.\nSuch an architecture facilitates effective yet economic message passing at two\nlevels. We evaluate our model and several strong baselines on two protein\ndatasets, $\\beta$-lactamase and myoglobin. Experimental results show that our\nmodel consistently achieves the highest amino acid recovery rate, TM-score, and\nthe lowest RMSD among all competitors. These findings prove the capability of\nour model to design protein sequences and structures that closely resemble\ntheir natural counterparts. Furthermore, in-depth analysis further confirms our\nmodel's ability to generate highly effective proteins capable of binding to\ntheir target metallocofactors. We provide code, data and models in Github.\n", "date": "2023-10-06", "categories": []}
{"id": "2401.04246", "title": "Scalable Normalizing Flows Enable Boltzmann Generators for\n  Macromolecules", "abstract": "  The Boltzmann distribution of a protein provides a roadmap to all of its\nfunctional states. Normalizing flows are a promising tool for modeling this\ndistribution, but current methods are intractable for typical pharmacological\ntargets; they become computationally intractable due to the size of the system,\nheterogeneity of intra-molecular potential energy, and long-range interactions.\nTo remedy these issues, we present a novel flow architecture that utilizes\nsplit channels and gated attention to efficiently learn the conformational\ndistribution of proteins defined by internal coordinates. We show that by\nutilizing a 2-Wasserstein loss, one can smooth the transition from maximum\nlikelihood training to energy-based training, enabling the training of\nBoltzmann Generators for macromolecules. We evaluate our model and training\nstrategy on villin headpiece HP35(nle-nle), a 35-residue subdomain, and protein\nG, a 56-residue protein. We demonstrate that standard architectures and\ntraining strategies, such as maximum likelihood alone, fail while our novel\narchitecture and multi-stage training strategy are able to model the\nconformational distributions of protein G and HP35.\n", "date": "2024-01-08", "categories": []}
{"id": "2401.03571", "title": "{\\alpha}-HMM: A Graphical Model for RNA Folding", "abstract": "  RNA secondary structure is modeled with the novel arbitrary-order hidden\nMarkov model ({\\alpha}-HMM). The {\\alpha}-HMM extends over the traditional HMM\nwith capability to model stochastic events that may be in influenced by\nhistorically distant ones, making it suitable to account for long-range\ncanonical base pairings between nucleotides, which constitute the RNA secondary\nstructure. Unlike previous heavy-weight extensions over HMM, the {\\alpha}-HMM\nhas the flexibility to apply restrictions on how one event may influence\nanother in stochastic processes, enabling efficient prediction of RNA secondary\nstructure including pseudoknots.\n", "date": "2024-01-07", "categories": []}
{"id": "2306.06156", "title": "PoET: A generative model of protein families as sequences-of-sequences", "abstract": "  Generative protein language models are a natural way to design new proteins\nwith desired functions. However, current models are either difficult to direct\nto produce a protein from a specific family of interest, or must be trained on\na large multiple sequence alignment (MSA) from the specific family of interest,\nmaking them unable to benefit from transfer learning across families. To\naddress this, we propose $\\textbf{P}$r$\\textbf{o}$tein $\\textbf{E}$volutionary\n$\\textbf{T}$ransformer (PoET), an autoregressive generative model of whole\nprotein families that learns to generate sets of related proteins as\nsequences-of-sequences across tens of millions of natural protein sequence\nclusters. PoET can be used as a retrieval-augmented language model to generate\nand score arbitrary modifications conditioned on any protein family of\ninterest, and can extrapolate from short context lengths to generalize well\neven for small families. This is enabled by a unique Transformer layer; we\nmodel tokens sequentially within sequences while attending between sequences\norder invariantly, allowing PoET to scale to context lengths beyond those used\nduring training. In extensive experiments on deep mutational scanning datasets,\nwe show that PoET outperforms existing protein language models and evolutionary\nsequence models for variant function prediction across proteins of all MSA\ndepths. We also demonstrate PoET's ability to controllably generate new protein\nsequences.\n", "date": "2023-06-09", "categories": []}
{"id": "2401.02789", "title": "Large Language Models in Plant Biology", "abstract": "  Large Language Models (LLMs), such as ChatGPT, have taken the world by storm\nand have passed certain forms of the Turing test. However, LLMs are not limited\nto human language and analyze sequential data, such as DNA, protein, and gene\nexpression. The resulting foundation models can be repurposed to identify the\ncomplex patterns within the data, resulting in powerful, multi-purpose\nprediction tools able to explain cellular systems. This review outlines the\ndifferent types of LLMs and showcases their recent uses in biology. Since LLMs\nhave not yet been embraced by the plant community, we also cover how these\nmodels can be deployed for the plant kingdom.\n", "date": "2024-01-05", "categories": []}
{"id": "2312.15665", "title": "A Multi-Modal Contrastive Diffusion Model for Therapeutic Peptide\n  Generation", "abstract": "  Therapeutic peptides represent a unique class of pharmaceutical agents\ncrucial for the treatment of human diseases. Recently, deep generative models\nhave exhibited remarkable potential for generating therapeutic peptides, but\nthey only utilize sequence or structure information alone, which hinders the\nperformance in generation. In this study, we propose a Multi-Modal Contrastive\nDiffusion model (MMCD), fusing both sequence and structure modalities in a\ndiffusion framework to co-generate novel peptide sequences and structures.\nSpecifically, MMCD constructs the sequence-modal and structure-modal diffusion\nmodels, respectively, and devises a multi-modal contrastive learning strategy\nwith intercontrastive and intra-contrastive in each diffusion timestep, aiming\nto capture the consistency between two modalities and boost model performance.\nThe inter-contrastive aligns sequences and structures of peptides by maximizing\nthe agreement of their embeddings, while the intra-contrastive differentiates\ntherapeutic and non-therapeutic peptides by maximizing the disagreement of\ntheir sequence/structure embeddings simultaneously. The extensive experiments\ndemonstrate that MMCD performs better than other state-of-theart deep\ngenerative methods in generating therapeutic peptides across various metrics,\nincluding antimicrobial/anticancer score, diversity, and peptide-docking.\n", "date": "2023-12-25", "categories": []}
{"id": "2401.02124", "title": "ACP-ESM: A novel framework for classification of anticancer peptides\n  using protein-oriented transformer approach", "abstract": "  Anticancer peptides (ACPs) are a class of molecules that have gained\nsignificant attention in the field of cancer research and therapy. ACPs are\nshort chains of amino acids, the building blocks of proteins, and they possess\nthe ability to selectively target and kill cancer cells. One of the key\nadvantages of ACPs is their ability to selectively target cancer cells while\nsparing healthy cells to a greater extent. This selectivity is often attributed\nto differences in the surface properties of cancer cells compared to normal\ncells. That is why ACPs are being investigated as potential candidates for\ncancer therapy. ACPs may be used alone or in combination with other treatment\nmodalities like chemotherapy and radiation therapy. While ACPs hold promise as\na novel approach to cancer treatment, there are challenges to overcome,\nincluding optimizing their stability, improving selectivity, and enhancing\ntheir delivery to cancer cells, continuous increasing in number of peptide\nsequences, developing a reliable and precise prediction model. In this work, we\npropose an efficient transformer-based framework to identify anticancer\npeptides for by performing accurate a reliable and precise prediction model.\nFor this purpose, four different transformer models, namely ESM, ProtBert,\nBioBERT, and SciBERT are employed to detect anticancer peptides from amino acid\nsequences. To demonstrate the contribution of the proposed framework, extensive\nexperiments are carried on widely-used datasets in the literature, two versions\nof AntiCp2, cACP-DeepGram, ACP-740. Experiment results show the usage of\nproposed model enhances classification accuracy when compared to the\nstate-of-the-art studies. The proposed framework, ESM, exhibits 96.45 of\naccuracy for AntiCp2 dataset, 97.66 of accuracy for cACP-DeepGram dataset, and\n88.51 of accuracy for ACP-740 dataset, thence determining new state-of-the-art.\n", "date": "2024-01-04", "categories": []}
{"id": "2310.02508", "title": "Ophiuchus: Scalable Modeling of Protein Structures through Hierarchical\n  Coarse-graining SO(3)-Equivariant Autoencoders", "abstract": "  Three-dimensional native states of natural proteins display recurring and\nhierarchical patterns. Yet, traditional graph-based modeling of protein\nstructures is often limited to operate within a single fine-grained resolution,\nand lacks hourglass neural architectures to learn those high-level building\nblocks. We narrow this gap by introducing Ophiuchus, an SO(3)-equivariant\ncoarse-graining model that efficiently operates on all-atom protein structures.\nOur model departs from current approaches that employ graph modeling, instead\nfocusing on local convolutional coarsening to model sequence-motif interactions\nwith efficient time complexity in protein length. We measure the reconstruction\ncapabilities of Ophiuchus across different compression rates, and compare it to\nexisting models. We examine the learned latent space and demonstrate its\nutility through conformational interpolation. Finally, we leverage denoising\ndiffusion probabilistic models (DDPM) in the latent space to efficiently sample\nprotein structures. Our experiments demonstrate Ophiuchus to be a scalable\nbasis for efficient protein modeling and generation.\n", "date": "2023-10-04", "categories": []}
{"id": "2310.15202", "title": "Predicting Transcription Factor Binding Sites using Transformer based\n  Capsule Network", "abstract": "  Prediction of binding sites for transcription factors is important to\nunderstand how they regulate gene expression and how this regulation can be\nmodulated for therapeutic purposes. Although in the past few years there are\nsignificant works addressing this issue, there is still space for improvement.\nIn this regard, a transformer based capsule network viz. DNABERT-Cap is\nproposed in this work to predict transcription factor binding sites mining\nChIP-seq datasets. DNABERT-Cap is a bidirectional encoder pre-trained with\nlarge number of genomic DNA sequences, empowered with a capsule layer\nresponsible for the final prediction. The proposed model builds a predictor for\ntranscription factor binding sites using the joint optimisation of features\nencompassing both bidirectional encoder and capsule layer, along with\nconvolutional and bidirectional long-short term memory layers. To evaluate the\nefficiency of the proposed approach, we use a benchmark ChIP-seq datasets of\nfive cell lines viz. A549, GM12878, Hep-G2, H1-hESC and Hela, available in the\nENCODE repository. The results show that the average area under the receiver\noperating characteristic curve score exceeds 0.91 for all such five cell lines.\nDNABERT-Cap is also compared with existing state-of-the-art deep learning based\npredictors viz. DeepARC, DeepTF, CNN-Zeng and DeepBind, and is seen to\noutperform them.\n", "date": "2023-10-23", "categories": []}
{"id": "2311.02594", "title": "scBeacon: single-cell biomarker extraction via identifying paired cell\n  clusters across biological conditions with contrastive siamese networks", "abstract": "  Despite the breakthroughs in biomarker discovery facilitated by differential\ngene analysis, challenges remain, particularly at the single-cell level.\nTraditional methodologies heavily rely on user-supplied cell annotations,\nfocusing on individually expressed data, often neglecting the critical\ninteractions between biological conditions, such as healthy versus diseased\nstates. In response, here we introduce scBeacon, an innovative framework built\nupon a deep contrastive siamese network. scBeacon pioneers an unsupervised\napproach, adeptly identifying matched cell populations across varied\nconditions, enabling a refined differential gene analysis. By utilizing a\nVQ-VAE framework, a contrastive siamese network, and a greedy iterative\nstrategy, scBeacon effectively pinpoints differential genes that hold potential\nas key biomarkers. Comprehensive evaluations on a diverse array of datasets\nvalidate scBeacon's superiority over existing single-cell differential gene\nanalysis tools. Its precision and adaptability underscore its significant role\nin enhancing diagnostic accuracy in biomarker discovery. With the emphasis on\nthe importance of biomarkers in diagnosis, scBeacon is positioned to be a\npivotal asset in the evolution of personalized medicine and targeted\ntreatments.\n", "date": "2023-11-05", "categories": []}
{"id": "2310.06150", "title": "Latent Diffusion Model for DNA Sequence Generation", "abstract": "  The harnessing of machine learning, especially deep generative models, has\nopened up promising avenues in the field of synthetic DNA sequence generation.\nWhilst Generative Adversarial Networks (GANs) have gained traction for this\napplication, they often face issues such as limited sample diversity and mode\ncollapse. On the other hand, Diffusion Models are a promising new class of\ngenerative models that are not burdened with these problems, enabling them to\nreach the state-of-the-art in domains such as image generation. In light of\nthis, we propose a novel latent diffusion model, DiscDiff, tailored for\ndiscrete DNA sequence generation. By simply embedding discrete DNA sequences\ninto a continuous latent space using an autoencoder, we are able to leverage\nthe powerful generative abilities of continuous diffusion models for the\ngeneration of discrete data. Additionally, we introduce Fr\\'echet\nReconstruction Distance (FReD) as a new metric to measure the sample quality of\nDNA sequence generations. Our DiscDiff model demonstrates an ability to\ngenerate synthetic DNA sequences that align closely with real DNA in terms of\nMotif Distribution, Latent Embedding Distribution (FReD), and Chromatin\nProfiles. Additionally, we contribute a comprehensive cross-species dataset of\n150K unique promoter-gene sequences from 15 species, enriching resources for\nfuture generative modelling in genomics. We will make our code public upon\npublication.\n", "date": "2023-10-09", "categories": []}
{"id": "2312.15055", "title": "Deep Learning for Efficient GWAS Feature Selection", "abstract": "  Genome-Wide Association Studies (GWAS) face unique challenges in the era of\nbig genomics data, particularly when dealing with ultra-high-dimensional\ndatasets where the number of genetic features significantly exceeds the\navailable samples. This paper introduces an extension to the feature selection\nmethodology proposed by Mirzaei et al. (2020), specifically tailored to tackle\nthe intricacies associated with ultra-high-dimensional GWAS data. Our extended\napproach enhances the original method by introducing a Frobenius norm penalty\ninto the student network, augmenting its capacity to adapt to scenarios\ncharacterized by a multitude of features and limited samples. Operating\nseamlessly in both supervised and unsupervised settings, our method employs two\nkey neural networks. The first leverages an autoencoder or supervised\nautoencoder for dimension reduction, extracting salient features from the\nultra-high-dimensional genomic data. The second network, a regularized\nfeed-forward model with a single hidden layer, is designed for precise feature\nselection. The introduction of the Frobenius norm penalty in the student\nnetwork significantly boosts the method's resilience to the challenges posed by\nultra-high-dimensional GWAS datasets. Experimental results showcase the\nefficacy of our approach in feature selection for GWAS data. The method not\nonly handles the inherent complexities of ultra-high-dimensional settings but\nalso demonstrates superior adaptability to the nuanced structures present in\ngenomics data. The flexibility and versatility of our proposed methodology are\nunderscored by its successful performance across a spectrum of experiments.\n", "date": "2023-12-22", "categories": []}
{"id": "2312.14220", "title": "Single-Cell RNA-seq Synthesis with Latent Diffusion Model", "abstract": "  The single-cell RNA sequencing (scRNA-seq) technology enables researchers to\nstudy complex biological systems and diseases with high resolution. The central\nchallenge is synthesizing enough scRNA-seq samples; insufficient samples can\nimpede downstream analysis and reproducibility. While various methods have been\nattempted in past research, the resulting scRNA-seq samples were often of poor\nquality or limited in terms of useful specific cell subpopulations. To address\nthese issues, we propose a novel method called Single-Cell Latent Diffusion\n(SCLD) based on the Diffusion Model. This method is capable of synthesizing\nlarge-scale, high-quality scRNA-seq samples, including both 'holistic' or\ntargeted specific cellular subpopulations within a unified framework. A\npre-guidance mechanism is designed for synthesizing specific cellular\nsubpopulations, while a post-guidance mechanism aims to enhance the quality of\nscRNA-seq samples. The SCLD can synthesize large-scale and high-quality\nscRNA-seq samples for various downstream tasks. Our experimental results\ndemonstrate state-of-the-art performance in cell classification and data\ndistribution distances when evaluated on two scRNA-seq benchmarks.\nAdditionally, visualization experiments show the SCLD's capability in\nsynthesizing specific cellular subpopulations.\n", "date": "2023-12-21", "categories": []}
{"id": "2312.14432", "title": "Scalable 3D Reconstruction From Single Particle X-Ray Diffraction Images\n  Based on Online Machine Learning", "abstract": "  X-ray free-electron lasers (XFELs) offer unique capabilities for measuring\nthe structure and dynamics of biomolecules, helping us understand the basic\nbuilding blocks of life. Notably, high-repetition-rate XFELs enable single\nparticle imaging (X-ray SPI) where individual, weakly scattering biomolecules\nare imaged under near-physiological conditions with the opportunity to access\nfleeting states that cannot be captured in cryogenic or crystallized\nconditions. Existing X-ray SPI reconstruction algorithms, which estimate the\nunknown orientation of a particle in each captured image as well as its shared\n3D structure, are inadequate in handling the massive datasets generated by\nthese emerging XFELs. Here, we introduce X-RAI, an online reconstruction\nframework that estimates the structure of a 3D macromolecule from large X-ray\nSPI datasets. X-RAI consists of a convolutional encoder, which amortizes pose\nestimation over large datasets, as well as a physics-based decoder, which\nemploys an implicit neural representation to enable high-quality 3D\nreconstruction in an end-to-end, self-supervised manner. We demonstrate that\nX-RAI achieves state-of-the-art performance for small-scale datasets in\nsimulation and challenging experimental settings and demonstrate its\nunprecedented ability to process large datasets containing millions of\ndiffraction images in an online fashion. These abilities signify a paradigm\nshift in X-ray SPI towards real-time capture and reconstruction.\n", "date": "2023-12-22", "categories": []}
{"id": "2208.05745", "title": "A biology-driven deep generative model for cell-type annotation in\n  cytometry", "abstract": "  Cytometry enables precise single-cell phenotyping within heterogeneous\npopulations. These cell types are traditionally annotated via manual gating,\nbut this method suffers from a lack of reproducibility and sensitivity to\nbatch-effect. Also, the most recent cytometers - spectral flow or mass\ncytometers - create rich and high-dimensional data whose analysis via manual\ngating becomes challenging and time-consuming. To tackle these limitations, we\nintroduce Scyan (https://github.com/MICS-Lab/scyan), a Single-cell Cytometry\nAnnotation Network that automatically annotates cell types using only prior\nexpert knowledge about the cytometry panel. We demonstrate that Scyan\nsignificantly outperforms the related state-of-the-art models on multiple\npublic datasets while being faster and interpretable. In addition, Scyan\novercomes several complementary tasks such as batch-effect removal,\ndebarcoding, and population discovery. Overall, this model accelerates and\neases cell population characterisation, quantification, and discovery in\ncytometry.\n", "date": "2022-08-11", "categories": []}
{"id": "2306.04886", "title": "Multi-task Bioassay Pre-training for Protein-ligand Binding Affinity\n  Prediction", "abstract": "  Protein-ligand binding affinity (PLBA) prediction is the fundamental task in\ndrug discovery. Recently, various deep learning-based models predict binding\naffinity by incorporating the three-dimensional structure of protein-ligand\ncomplexes as input and achieving astounding progress. However, due to the\nscarcity of high-quality training data, the generalization ability of current\nmodels is still limited. In addition, different bioassays use varying affinity\nmeasurement labels (i.e., IC50, Ki, Kd), and different experimental conditions\ninevitably introduce systematic noise, which poses a significant challenge to\nconstructing high-precision affinity prediction models. To address these\nissues, we (1) propose Multi-task Bioassay Pre-training (MBP), a pre-training\nframework for structure-based PLBA prediction; (2) construct a pre-training\ndataset called ChEMBL-Dock with more than 300k experimentally measured affinity\nlabels and about 2.8M docked three-dimensional structures. By introducing\nmulti-task pre-training to treat the prediction of different affinity labels as\ndifferent tasks and classifying relative rankings between samples from the same\nbioassay, MBP learns robust and transferrable structural knowledge from our new\nChEMBL-Dock dataset with varied and noisy labels. Experiments substantiate the\ncapability of MBP as a general framework that can improve and be tailored to\nmainstream structure-based PLBA prediction tasks. To the best of our knowledge,\nMBP is the first affinity pre-training model and shows great potential for\nfuture development.\n", "date": "2023-06-08", "categories": []}
{"id": "2312.11584", "title": "ContraNovo: A Contrastive Learning Approach to Enhance De Novo Peptide\n  Sequencing", "abstract": "  De novo peptide sequencing from mass spectrometry (MS) data is a critical\ntask in proteomics research. Traditional de novo algorithms have encountered a\nbottleneck in accuracy due to the inherent complexity of proteomics data. While\ndeep learning-based methods have shown progress, they reduce the problem to a\ntranslation task, potentially overlooking critical nuances between spectra and\npeptides. In our research, we present ContraNovo, a pioneering algorithm that\nleverages contrastive learning to extract the relationship between spectra and\npeptides and incorporates the mass information into peptide decoding, aiming to\naddress these intricacies more efficiently. Through rigorous evaluations on two\nbenchmark datasets, ContraNovo consistently outshines contemporary\nstate-of-the-art solutions, underscoring its promising potential in enhancing\nde novo peptide sequencing. The source code is available at\nhttps://github.com/BEAM-Labs/ContraNovo.\n", "date": "2023-12-18", "categories": []}
{"id": "2310.10605", "title": "ForceGen: End-to-end de novo protein generation based on nonlinear\n  mechanical unfolding responses using a protein language diffusion model", "abstract": "  Through evolution, nature has presented a set of remarkable protein\nmaterials, including elastins, silks, keratins and collagens with superior\nmechanical performances that play crucial roles in mechanobiology. However,\ngoing beyond natural designs to discover proteins that meet specified\nmechanical properties remains challenging. Here we report a generative model\nthat predicts protein designs to meet complex nonlinear mechanical\nproperty-design objectives. Our model leverages deep knowledge on protein\nsequences from a pre-trained protein language model and maps mechanical\nunfolding responses to create novel proteins. Via full-atom molecular\nsimulations for direct validation, we demonstrate that the designed proteins\nare novel, and fulfill the targeted mechanical properties, including unfolding\nenergy and mechanical strength, as well as the detailed unfolding\nforce-separation curves. Our model offers rapid pathways to explore the\nenormous mechanobiological protein sequence space unconstrained by biological\nsynthesis, using mechanical features as target to enable the discovery of\nprotein materials with superior mechanical properties.\n", "date": "2023-10-16", "categories": []}
{"id": "2311.17401", "title": "Gene-MOE: A sparsely gated prognosis and classification framework\n  exploiting pan-cancer genomic information", "abstract": "  Benefiting from the advancements in deep learning, various genomic analytical\ntechniques, such as survival analysis, classification of tumors and their\nsubtypes, and exploration of specific pathways, have significantly enhanced our\nunderstanding of the biological mechanisms driving cancer. However, the\noverfitting issue, arising from the limited number of patient samples, poses a\nchallenge in improving the accuracy of genome analysis by deepening the neural\nnetwork. Furthermore, it remains uncertain whether novel approaches such as the\nsparsely gated mixture of expert (MOE) and self-attention mechanisms can\nimprove the accuracy of genomic analysis. In this paper, we introduce a novel\nsparsely gated RNA-seq analysis framework called Gene-MOE. This framework\nexploits the potential of the MOE layers and the proposed mixture of attention\nexpert (MOAE) layers to enhance the analysis accuracy. Additionally, it\naddresses overfitting challenges by integrating pan-cancer information from 33\ndistinct cancer types through pre-training.We pre-trained Gene-MOE on TCGA\npan-cancer RNA-seq dataset with 33 cancer types. Subsequently, we conducted\nexperiments involving cancer classification and survival analysis based on the\npre-trained Gene-MOE. According to the survival analysis results on 14 cancer\ntypes, Gene-MOE outperformed state-of-the-art models on 12 cancer types.\nThrough detailed feature analysis, we found that the Gene-MOE model could learn\nrich feature representations of high-dimensional genes. According to the\nclassification results, the total accuracy of the classification model for 33\ncancer classifications reached 95.8%, representing the best performance\ncompared to state-of-the-art models. These results indicate that Gene-MOE holds\nstrong potential for use in cancer classification and survival analysis.\n", "date": "2023-11-29", "categories": []}
{"id": "2312.08470", "title": "Best practices for machine learning in antibody discovery and\n  development", "abstract": "  Over the past 40 years, the discovery and development of therapeutic\nantibodies to treat disease has become common practice. However, as therapeutic\nantibody constructs are becoming more sophisticated (e.g., multi-specifics),\nconventional approaches to optimisation are increasingly inefficient. Machine\nlearning (ML) promises to open up an in silico route to antibody discovery and\nhelp accelerate the development of drug products using a reduced number of\nexperiments and hence cost. Over the past few years, we have observed rapid\ndevelopments in the field of ML-guided antibody discovery and development\n(D&D). However, many of the results are difficult to compare or hard to assess\nfor utility by other experts in the field due to the high diversity in the\ndatasets and evaluation techniques and metrics that are across industry and\nacademia. This limitation of the literature curtails the broad adoption of ML\nacross the industry and slows down overall progress in the field, highlighting\nthe need to develop standards and guidelines that may help improve the\nreproducibility of ML models across different research groups. To address these\nchallenges, we set out in this perspective to critically review current\npractices, explain common pitfalls, and clearly define a set of method\ndevelopment and evaluation guidelines that can be applied to different types of\nML-based techniques for therapeutic antibody D&D. Specifically, we address in\nan end-to-end analysis, challenges associated with all aspects of the ML\nprocess and recommend a set of best practices for each stage.\n", "date": "2023-12-13", "categories": []}
{"id": "2312.08987", "title": "Unbiased organism-agnostic and highly sensitive signal peptide predictor\n  with deep protein language model", "abstract": "  Signal peptide (SP) is a short peptide located in the N-terminus of proteins.\nIt is essential to target and transfer transmembrane and secreted proteins to\ncorrect positions. Compared with traditional experimental methods to identify\nsignal peptides, computational methods are faster and more efficient, which are\nmore practical for analyzing thousands or even millions of protein sequences,\nespecially for metagenomic data. Here we present Unbiased Organism-agnostic\nSignal Peptide Network (USPNet), a signal peptide classification and cleavage\nsite prediction deep learning method that takes advantage of protein language\nmodels. We propose to apply label distribution-aware margin loss to handle data\nimbalance problems and use evolutionary information of protein to enrich\nrepresentation and overcome species information dependence.\n", "date": "2023-12-14", "categories": []}
{"id": "2301.10814", "title": "Unsupervised Protein-Ligand Binding Energy Prediction via Neural Euler's\n  Rotation Equation", "abstract": "  Protein-ligand binding prediction is a fundamental problem in AI-driven drug\ndiscovery. Prior work focused on supervised learning methods using a large set\nof binding affinity data for small molecules, but it is hard to apply the same\nstrategy to other drug classes like antibodies as labelled data is limited. In\nthis paper, we explore unsupervised approaches and reformulate binding energy\nprediction as a generative modeling task. Specifically, we train an\nenergy-based model on a set of unlabelled protein-ligand complexes using SE(3)\ndenoising score matching and interpret its log-likelihood as binding affinity.\nOur key contribution is a new equivariant rotation prediction network called\nNeural Euler's Rotation Equations (NERE) for SE(3) score matching. It predicts\na rotation by modeling the force and torque between protein and ligand atoms,\nwhere the force is defined as the gradient of an energy function with respect\nto atom coordinates. We evaluate NERE on protein-ligand and antibody-antigen\nbinding affinity prediction benchmarks. Our model outperforms all unsupervised\nbaselines (physics-based and statistical potentials) and matches supervised\nlearning methods in the antibody case.\n", "date": "2023-01-25", "categories": []}
{"id": "2303.17728", "title": "Evaluation of GPT and BERT-based models on identifying protein-protein\n  interactions in biomedical text", "abstract": "  Detecting protein-protein interactions (PPIs) is crucial for understanding\ngenetic mechanisms, disease pathogenesis, and drug design. However, with the\nfast-paced growth of biomedical literature, there is a growing need for\nautomated and accurate extraction of PPIs to facilitate scientific knowledge\ndiscovery. Pre-trained language models, such as generative pre-trained\ntransformers (GPT) and bidirectional encoder representations from transformers\n(BERT), have shown promising results in natural language processing (NLP)\ntasks. We evaluated the performance of PPI identification of multiple GPT and\nBERT models using three manually curated gold-standard corpora: Learning\nLanguage in Logic (LLL) with 164 PPIs in 77 sentences, Human Protein Reference\nDatabase with 163 PPIs in 145 sentences, and Interaction Extraction Performance\nAssessment with 335 PPIs in 486 sentences. BERT-based models achieved the best\noverall performance, with BioBERT achieving the highest recall (91.95%) and\nF1-score (86.84%) and PubMedBERT achieving the highest precision (85.25%).\nInterestingly, despite not being explicitly trained for biomedical texts, GPT-4\nachieved commendable performance, comparable to the top-performing BERT models.\nIt achieved a precision of 88.37%, a recall of 85.14%, and an F1-score of\n86.49% on the LLL dataset. These results suggest that GPT models can\neffectively detect PPIs from text data, offering promising avenues for\napplication in biomedical literature mining. Further research could explore how\nthese models might be fine-tuned for even more specialized tasks within the\nbiomedical domain.\n", "date": "2023-03-30", "categories": []}
{"id": "2312.07316", "title": "GateNet: A novel Neural Network Architecture for Automated Flow\n  Cytometry Gating", "abstract": "  Flow cytometry is widely used to identify cell populations in patient-derived\nfluids such as peripheral blood (PB) or cerebrospinal fluid (CSF). While\nubiquitous in research and clinical practice, flow cytometry requires gating,\ni.e. cell type identification which requires labor-intensive and error-prone\nmanual adjustments. To facilitate this process, we designed GateNet, the first\nneural network architecture enabling full end-to-end automated gating without\nthe need to correct for batch effects. We train GateNet with over 8,000,000\nevents based on N=127 PB and CSF samples which were manually labeled\nindependently by four experts. We show that for novel, unseen samples, GateNet\nachieves human-level performance (F1 score ranging from 0.910 to 0.997). In\naddition we apply GateNet to a publicly available dataset confirming\ngeneralization with an F1 score of 0.936. As our implementation utilizes\ngraphics processing units (GPU), gating only needs 15 microseconds per event.\nImportantly, we also show that GateNet only requires ~10 samples to reach\nhuman-level performance, rendering it widely applicable in all domains of flow\ncytometry.\n", "date": "2023-12-12", "categories": []}
{"id": "2312.05273", "title": "Evaluating Zero-Shot Scoring for In Vitro Antibody Binding Prediction\n  with Experimental Validation", "abstract": "  The success of therapeutic antibodies relies on their ability to selectively\nbind antigens. AI-based antibody design protocols have shown promise in\ngenerating epitope-specific designs. Many of these protocols use an inverse\nfolding step to generate diverse sequences given a backbone structure. Due to\nprohibitive screening costs, it is key to identify candidate sequences likely\nto bind in vitro. Here, we compare the efficacy of 8 common scoring paradigms\nbased on open-source models to classify antibody designs as binders or\nnon-binders. We evaluate these approaches on a novel surface plasmon resonance\n(SPR) dataset, spanning 5 antigens. Our results show that existing methods\nstruggle to detect binders, and performance is highly variable across antigens.\nWe find that metrics computed on flexibly docked antibody-antigen complexes are\nmore robust, and ensembles scores are more consistent than individual metrics.\nWe provide experimental insight to analyze current scoring techniques,\nhighlighting that the development of robust, zero-shot filters is an important\nresearch gap.\n", "date": "2023-12-07", "categories": []}
{"id": "2312.06082", "title": "XAI meets Biology: A Comprehensive Review of Explainable AI in\n  Bioinformatics Applications", "abstract": "  Artificial intelligence (AI), particularly machine learning and deep learning\nmodels, has significantly impacted bioinformatics research by offering powerful\ntools for analyzing complex biological data. However, the lack of\ninterpretability and transparency of these models presents challenges in\nleveraging these models for deeper biological insights and for generating\ntestable hypotheses. Explainable AI (XAI) has emerged as a promising solution\nto enhance the transparency and interpretability of AI models in\nbioinformatics. This review provides a comprehensive analysis of various XAI\ntechniques and their applications across various bioinformatics domains\nincluding DNA, RNA, and protein sequence analysis, structural analysis, gene\nexpression and genome analysis, and bioimaging analysis. We introduce the most\npertinent machine learning and XAI methods, then discuss their diverse\napplications and address the current limitations of available XAI tools. By\noffering insights into XAI's potential and challenges, this review aims to\nfacilitate its practical implementation in bioinformatics research and help\nresearchers navigate the landscape of XAI tools.\n", "date": "2023-12-11", "categories": []}
{"id": "2310.06725", "title": "Growing ecosystem of deep learning methods for modeling\n  protein$\\unicode{x2013}$protein interactions", "abstract": "  Numerous cellular functions rely on protein$\\unicode{x2013}$protein\ninteractions. Efforts to comprehensively characterize them remain challenged\nhowever by the diversity of molecular recognition mechanisms employed within\nthe proteome. Deep learning has emerged as a promising approach for tackling\nthis problem by exploiting both experimental data and basic biophysical\nknowledge about protein interactions. Here, we review the growing ecosystem of\ndeep learning methods for modeling protein interactions, highlighting the\ndiversity of these biophysically-informed models and their respective\ntrade-offs. We discuss recent successes in using representation learning to\ncapture complex features pertinent to predicting protein interactions and\ninteraction sites, geometric deep learning to reason over protein structures\nand predict complex structures, and generative modeling to design de novo\nprotein assemblies. We also outline some of the outstanding challenges and\npromising new directions. Opportunities abound to discover novel interactions,\nelucidate their physical mechanisms, and engineer binders to modulate their\nfunctions using deep learning and, ultimately, unravel how protein interactions\norchestrate complex cellular behaviors.\n", "date": "2023-10-10", "categories": []}
{"id": "2312.04019", "title": "Efficiently Predicting Protein Stability Changes Upon Single-point\n  Mutation with Large Language Models", "abstract": "  Predicting protein stability changes induced by single-point mutations has\nbeen a persistent challenge over the years, attracting immense interest from\nnumerous researchers. The ability to precisely predict protein thermostability\nis pivotal for various subfields and applications in biochemistry, including\ndrug development, protein evolution analysis, and enzyme synthesis. Despite the\nproposition of multiple methodologies aimed at addressing this issue, few\napproaches have successfully achieved optimal performance coupled with high\ncomputational efficiency. Two principal hurdles contribute to the existing\nchallenges in this domain. The first is the complexity of extracting and\naggregating sufficiently representative features from proteins. The second\nrefers to the limited availability of experimental data for protein mutation\nanalysis, further complicating the comprehensive evaluation of model\nperformance on unseen data samples. With the advent of Large Language\nModels(LLM), such as the ESM models in protein research, profound\ninterpretation of protein features is now accessibly aided by enormous training\ndata. Therefore, LLMs are indeed to facilitate a wide range of protein\nresearch. In our study, we introduce an ESM-assisted efficient approach that\nintegrates protein sequence and structural features to predict the\nthermostability changes in protein upon single-point mutations. Furthermore, we\nhave curated a dataset meticulously designed to preclude data leakage,\ncorresponding to two extensively employed test datasets, to facilitate a more\nequitable model comparison.\n", "date": "2023-12-07", "categories": []}
{"id": "2312.04064", "title": "DiscoBAX: Discovery of Optimal Intervention Sets in Genomic Experiment\n  Design", "abstract": "  The discovery of therapeutics to treat genetically-driven pathologies relies\non identifying genes involved in the underlying disease mechanisms. Existing\napproaches search over the billions of potential interventions to maximize the\nexpected influence on the target phenotype. However, to reduce the risk of\nfailure in future stages of trials, practical experiment design aims to find a\nset of interventions that maximally change a target phenotype via diverse\nmechanisms. We propose DiscoBAX, a sample-efficient method for maximizing the\nrate of significant discoveries per experiment while simultaneously probing for\na wide range of diverse mechanisms during a genomic experiment campaign. We\nprovide theoretical guarantees of approximate optimality under standard\nassumptions, and conduct a comprehensive experimental evaluation covering both\nsynthetic as well as real-world experimental design tasks. DiscoBAX outperforms\nexisting state-of-the-art methods for experimental design, selecting effective\nand diverse perturbations in biological systems.\n", "date": "2023-12-07", "categories": []}
{"id": "2312.02248", "title": "Towards early diagnosis of Alzheimer's disease: Advances in\n  immune-related blood biomarkers and computational modeling approaches", "abstract": "  Alzheimer's disease has an increasing prevalence in the population\nworld-wide, yet current diagnostic methods based on recommended biomarkers are\nonly available in specialized clinics. Due to these circumstances, Alzheimer's\ndisease is usually diagnosed late, which contrasts with the currently available\ntreatment options that are only effective for patients at an early stage.\nBlood-based biomarkers could fill in the gap of easily accessible and low-cost\nmethods for early diagnosis of the disease. In particular, immune-based\nblood-biomarkers might be a promising option, given the recently discovered\ncross-talk of immune cells of the central nervous system with those in the\nperipheral immune system. With the help of machine learning algorithms and\nmechanistic modeling approaches, such as agent-based modeling, an in-depth\nanalysis of the simulation of cell dynamics is possible as well as of\nhigh-dimensional omics resources indicative of pathway signaling changes. Here,\nwe give a background on advances in research on brain-immune system cross-talk\nin Alzheimer's disease and review recent machine learning and mechanistic\nmodeling approaches which leverage modern omics technologies for blood-based\nimmune system-related biomarker discovery.\n", "date": "2023-12-04", "categories": []}
{"id": "2312.03016", "title": "Protein Language Model-Powered 3D Ligand Binding Site Prediction from\n  Protein Sequence", "abstract": "  Prediction of ligand binding sites of proteins is a fundamental and important\ntask for understanding the function of proteins and screening potential drugs.\nMost existing methods require experimentally determined protein holo-structures\nas input. However, such structures can be unavailable on novel or less-studied\nproteins. To tackle this limitation, we propose LaMPSite, which only takes\nprotein sequences and ligand molecular graphs as input for ligand binding site\npredictions. The protein sequences are used to retrieve residue-level\nembeddings and contact maps from the pre-trained ESM-2 protein language model.\nThe ligand molecular graphs are fed into a graph neural network to compute\natom-level embeddings. Then we compute and update the protein-ligand\ninteraction embedding based on the protein residue-level embeddings and ligand\natom-level embeddings, and the geometric constraints in the inferred protein\ncontact map and ligand distance map. A final pooling on protein-ligand\ninteraction embedding would indicate which residues belong to the binding\nsites. Without any 3D coordinate information of proteins, our proposed model\nachieves competitive performance compared to baseline methods that require 3D\nprotein structures when predicting binding sites. Given that less than 50% of\nproteins have reliable structure information in the current stage, LaMPSite\nwill provide new opportunities for drug discovery.\n", "date": "2023-12-05", "categories": []}
{"id": "2312.03292", "title": "Enhancing Molecular Property Prediction via Mixture of Collaborative\n  Experts", "abstract": "  Molecular Property Prediction (MPP) task involves predicting biochemical\nproperties based on molecular features, such as molecular graph structures,\ncontributing to the discovery of lead compounds in drug development. To address\ndata scarcity and imbalance in MPP, some studies have adopted Graph Neural\nNetworks (GNN) as an encoder to extract commonalities from molecular graphs.\nHowever, these approaches often use a separate predictor for each task,\nneglecting the shared characteristics among predictors corresponding to\ndifferent tasks. In response to this limitation, we introduce the GNN-MoCE\narchitecture. It employs the Mixture of Collaborative Experts (MoCE) as\npredictors, exploiting task commonalities while confronting the homogeneity\nissue in the expert pool and the decision dominance dilemma within the expert\ngroup. To enhance expert diversity for collaboration among all experts, the\nExpert-Specific Projection method is proposed to assign a unique projection\nperspective to each expert. To balance decision-making influence for\ncollaboration within the expert group, the Expert-Specific Loss is presented to\nintegrate individual expert loss into the weighted decision loss of the group\nfor more equitable training. Benefiting from the enhancements of MoCE in expert\ncreation, dynamic expert group formation, and experts' collaboration, our model\ndemonstrates superior performance over traditional methods on 24 MPP datasets,\nespecially in tasks with limited data or high imbalance.\n", "date": "2023-12-06", "categories": []}
{"id": "2309.09126", "title": "How much can ChatGPT really help Computational Biologists in\n  Programming?", "abstract": "  ChatGPT, a recently developed product by openAI, is successfully leaving its\nmark as a multi-purpose natural language based chatbot. In this paper, we are\nmore interested in analyzing its potential in the field of computational\nbiology. A major share of work done by computational biologists these days\ninvolve coding up bioinformatics algorithms, analyzing data, creating\npipelining scripts and even machine learning modeling and feature extraction.\nThis paper focuses on the potential influence (both positive and negative) of\nChatGPT in the mentioned aspects with illustrative examples from different\nperspectives. Compared to other fields of computer science, computational\nbiology has - (1) less coding resources, (2) more sensitivity and bias issues\n(deals with medical data) and (3) more necessity of coding assistance (people\nfrom diverse background come to this field). Keeping such issues in mind, we\ncover use cases such as code writing, reviewing, debugging, converting,\nrefactoring and pipelining using ChatGPT from the perspective of computational\nbiologists in this paper.\n", "date": "2023-09-17", "categories": []}
{"id": "2312.01272", "title": "Multiscale Topology in Interactomic Network: From Transcriptome to\n  Antiaddiction Drug Repurposing", "abstract": "  The escalating drug addiction crisis in the United States underscores the\nurgent need for innovative therapeutic strategies. This study embarked on an\ninnovative and rigorous strategy to unearth potential drug repurposing\ncandidates for opioid and cocaine addiction treatment, bridging the gap between\ntranscriptomic data analysis and drug discovery. We initiated our approach by\nconducting differential gene expression analysis on addiction-related\ntranscriptomic data to identify key genes. We propose a novel topological\ndifferentiation to identify key genes from a protein-protein interaction (PPI)\nnetwork derived from DEGs. This method utilizes persistent Laplacians to\naccurately single out pivotal nodes within the network, conducting this\nanalysis in a multiscale manner to ensure high reliability. Through rigorous\nliterature validation, pathway analysis, and data-availability scrutiny, we\nidentified three pivotal molecular targets, mTOR, mGluR5, and NMDAR, for drug\nrepurposing from DrugBank. We crafted machine learning models employing two\nnatural language processing (NLP)-based embeddings and a traditional 2D\nfingerprint, which demonstrated robust predictive ability in gauging binding\naffinities of DrugBank compounds to selected targets. Furthermore, we\nelucidated the interactions of promising drugs with the targets and evaluated\ntheir drug-likeness. This study delineates a multi-faceted and comprehensive\nanalytical framework, amalgamating bioinformatics, topological data analysis\nand machine learning, for drug repurposing in addiction treatment, setting the\nstage for subsequent experimental validation. The versatility of the methods we\ndeveloped allows for applications across a range of diseases and transcriptomic\ndatasets.\n", "date": "2023-12-03", "categories": []}
{"id": "2312.01275", "title": "A Review of Link Prediction Applications in Network Biology", "abstract": "  In the domain of network biology, the interactions among heterogeneous\ngenomic and molecular entities are represented through networks. Link\nprediction (LP) methodologies are instrumental in inferring missing or\nprospective associations within these biological networks. In this review, we\nsystematically dissect the attributes of local, centrality, and embedding-based\nLP approaches, applied to static and dynamic biological networks. We undertake\nan examination of the current applications of LP metrics for predicting links\nbetween diseases, genes, proteins, RNA, microbiomes, drugs, and neurons. We\ncarry out comprehensive performance evaluations on established biological\nnetwork datasets to show the practical applications of standard LP models.\nMoreover, we compare the similarity in prediction trends among the models and\nthe specific network attributes that contribute to effective link prediction,\nbefore underscoring the role of LP in addressing the formidable challenges\nprevalent in biological systems, ranging from noise, bias, and data sparseness\nto interpretability. We conclude the review with an exploration of the\nessential characteristics expected from future LP models, poised to advance our\ncomprehension of the intricate interactions governing biological systems.\n", "date": "2023-12-03", "categories": []}
{"id": "2307.03675", "title": "GeoPhy: Differentiable Phylogenetic Inference via Geometric Gradients of\n  Tree Topologies", "abstract": "  Phylogenetic inference, grounded in molecular evolution models, is essential\nfor understanding the evolutionary relationships in biological data. Accounting\nfor the uncertainty of phylogenetic tree variables, which include tree\ntopologies and evolutionary distances on branches, is crucial for accurately\ninferring species relationships from molecular data and tasks requiring\nvariable marginalization. Variational Bayesian methods are key to developing\nscalable, practical models; however, it remains challenging to conduct\nphylogenetic inference without restricting the combinatorially vast number of\npossible tree topologies. In this work, we introduce a novel, fully\ndifferentiable formulation of phylogenetic inference that leverages a unique\nrepresentation of topological distributions in continuous geometric spaces.\nThrough practical considerations on design spaces and control variates for\ngradient estimations, our approach, GeoPhy, enables variational inference\nwithout limiting the topological candidates. In experiments using real\nbenchmark datasets, GeoPhy significantly outperformed other approximate\nBayesian methods that considered whole topologies.\n", "date": "2023-07-07", "categories": []}
{"id": "2312.00080", "title": "PDB-Struct: A Comprehensive Benchmark for Structure-based Protein Design", "abstract": "  Structure-based protein design has attracted increasing interest, with\nnumerous methods being introduced in recent years. However, a universally\naccepted method for evaluation has not been established, since the wet-lab\nvalidation can be overly time-consuming for the development of new algorithms,\nand the $\\textit{in silico}$ validation with recovery and perplexity metrics is\nefficient but may not precisely reflect true foldability. To address this gap,\nwe introduce two novel metrics: refoldability-based metric, which leverages\nhigh-accuracy protein structure prediction models as a proxy for wet lab\nexperiments, and stability-based metric, which assesses whether models can\nassign high likelihoods to experimentally stable proteins. We curate datasets\nfrom high-quality CATH protein data, high-throughput $\\textit{de novo}$\ndesigned proteins, and mega-scale experimental mutagenesis experiments, and in\ndoing so, present the $\\textbf{PDB-Struct}$ benchmark that evaluates both\nrecent and previously uncompared protein design methods. Experimental results\nindicate that ByProt, ProteinMPNN, and ESM-IF perform exceptionally well on our\nbenchmark, while ESM-Design and AF-Design fall short on the refoldability\nmetric. We also show that while some methods exhibit high sequence recovery,\nthey do not perform as well on our new benchmark. Our proposed benchmark paves\nthe way for a fair and comprehensive evaluation of protein design methods in\nthe future. Code is available at https://github.com/WANG-CR/PDB-Struct.\n", "date": "2023-11-30", "categories": []}
{"id": "2312.00189", "title": "HeTriNet: Heterogeneous Graph Triplet Attention Network for\n  Drug-Target-Disease Interaction", "abstract": "  Modeling the interactions between drugs, targets, and diseases is paramount\nin drug discovery and has significant implications for precision medicine and\npersonalized treatments. Current approaches frequently consider drug-target or\ndrug-disease interactions individually, ignoring the interdependencies among\nall three entities. Within human metabolic systems, drugs interact with protein\ntargets in cells, influencing target activities and subsequently impacting\nbiological pathways to promote healthy functions and treat diseases. Moving\nbeyond binary relationships and exploring tighter triple relationships is\nessential to understanding drugs' mechanism of action (MoAs). Moreover,\nidentifying the heterogeneity of drugs, targets, and diseases, along with their\ndistinct characteristics, is critical to model these complex interactions\nappropriately. To address these challenges, we effectively model the\ninterconnectedness of all entities in a heterogeneous graph and develop a novel\nHeterogeneous Graph Triplet Attention Network (\\texttt{HeTriNet}).\n\\texttt{HeTriNet} introduces a novel triplet attention mechanism within this\nheterogeneous graph structure. Beyond pairwise attention as the importance of\nan entity for the other one, we define triplet attention to model the\nimportance of pairs for entities in the drug-target-disease triplet prediction\nproblem. Experimental results on real-world datasets show that\n\\texttt{HeTriNet} outperforms several baselines, demonstrating its remarkable\nproficiency in uncovering novel drug-target-disease relationships.\n", "date": "2023-11-30", "categories": []}
{"id": "2312.00191", "title": "Enhancing Ligand Pose Sampling for Molecular Docking", "abstract": "  Deep learning promises to dramatically improve scoring functions for\nmolecular docking, leading to substantial advances in binding pose prediction\nand virtual screening. To train scoring functions-and to perform molecular\ndocking-one must generate a set of candidate ligand binding poses.\nUnfortunately, the sampling protocols currently used to generate candidate\nposes frequently fail to produce any poses close to the correct, experimentally\ndetermined pose, unless information about the correct pose is provided. This\nlimits the accuracy of learned scoring functions and molecular docking. Here,\nwe describe two improved protocols for pose sampling: GLOW (auGmented sampLing\nwith sOftened vdW potential) and a novel technique named IVES (IteratiVe\nEnsemble Sampling). Our benchmarking results demonstrate the effectiveness of\nour methods in improving the likelihood of sampling accurate poses, especially\nfor binding pockets whose shape changes substantially when different ligands\nbind. This improvement is observed across both experimentally determined and\nAlphaFold-generated protein structures. Additionally, we present datasets of\ncandidate ligand poses generated using our methods for each of around 5,000\nprotein-ligand cross-docking pairs, for training and testing scoring functions.\nTo benefit the research community, we provide these cross-docking datasets and\nan open-source Python implementation of GLOW and IVES at\nhttps://github.com/drorlab/GLOW_IVES .\n", "date": "2023-11-30", "categories": []}
{"id": "2312.00671", "title": "CellMixer: Annotation-free Semantic Cell Segmentation of Heterogeneous\n  Cell Populations", "abstract": "  In recent years, several unsupervised cell segmentation methods have been\npresented, trying to omit the requirement of laborious pixel-level annotations\nfor the training of a cell segmentation model. Most if not all of these methods\nhandle the instance segmentation task by focusing on the detection of different\ncell instances ignoring their type. While such models prove adequate for\ncertain tasks, like cell counting, other applications require the\nidentification of each cell's type. In this paper, we present CellMixer, an\ninnovative annotation-free approach for the semantic segmentation of\nheterogeneous cell populations. Our augmentation-based method enables the\ntraining of a segmentation model from image-level labels of homogeneous cell\npopulations. Our results show that CellMixer can achieve competitive\nsegmentation performance across multiple cell types and imaging modalities,\ndemonstrating the method's scalability and potential for broader applications\nin medical imaging, cellular biology, and diagnostics.\n", "date": "2023-12-01", "categories": []}
{"id": "2312.00718", "title": "Removing Biases from Molecular Representations via Information\n  Maximization", "abstract": "  High-throughput drug screening -- using cell imaging or gene expression\nmeasurements as readouts of drug effect -- is a critical tool in biotechnology\nto assess and understand the relationship between the chemical structure and\nbiological activity of a drug. Since large-scale screens have to be divided\ninto multiple experiments, a key difficulty is dealing with batch effects,\nwhich can introduce systematic errors and non-biological associations in the\ndata. We propose InfoCORE, an Information maximization approach for COnfounder\nREmoval, to effectively deal with batch effects and obtain refined molecular\nrepresentations. InfoCORE establishes a variational lower bound on the\nconditional mutual information of the latent representations given a batch\nidentifier. It adaptively reweighs samples to equalize their implied batch\ndistribution. Extensive experiments on drug screening data reveal InfoCORE's\nsuperior performance in a multitude of tasks including molecular property\nprediction and molecule-phenotype retrieval. Additionally, we show results for\nhow InfoCORE offers a versatile framework and resolves general distribution\nshifts and issues of data fairness by minimizing correlation with spurious\nfeatures or removing sensitive attributes. The code is available at\nhttps://github.com/uhlerlab/InfoCORE.\n", "date": "2023-12-01", "categories": []}
{"id": "2310.03121", "title": "OpenMM 8: Molecular Dynamics Simulation with Machine Learning Potentials", "abstract": "  Machine learning plays an important and growing role in molecular simulation.\nThe newest version of the OpenMM molecular dynamics toolkit introduces new\nfeatures to support the use of machine learning potentials. Arbitrary PyTorch\nmodels can be added to a simulation and used to compute forces and energy. A\nhigher-level interface allows users to easily model their molecules of interest\nwith general purpose, pretrained potential functions. A collection of optimized\nCUDA kernels and custom PyTorch operations greatly improves the speed of\nsimulations. We demonstrate these features on simulations of cyclin-dependent\nkinase 8 (CDK8) and the green fluorescent protein (GFP) chromophore in water.\nTaken together, these features make it practical to use machine learning to\nimprove the accuracy of simulations at only a modest increase in cost.\n", "date": "2023-10-04", "categories": []}
{"id": "2311.17964", "title": "Linear normalised hash function for clustering gene sequences and\n  identifying reference sequences from multiple sequence alignments", "abstract": "  The aim of this study was to develop a method that would identify the cluster\ncentroids and the optimal number of clusters for a given sensitivity level and\ncould work equally well for the different sequence datasets. A novel method\nthat combines the linear mapping hash function and multiple sequence alignment\n(MSA) was developed. This method takes advantage of the already sorted by\nsimilarity sequences from the MSA output, and identifies the optimal number of\nclusters, clusters cut-offs, and clusters centroids that can represent\nreference gene vouchers for the different species. The linear mapping hash\nfunction can map an already ordered by similarity distance matrix to indices to\nreveal gaps in the values around which the optimal cut-offs of the different\nclusters can be identified. The method was evaluated using sets of closely\nrelated (16S rRNA gene sequences of Nocardia species) and highly variable (VP1\ngenomic region of Enterovirus 71) sequences and outperformed existing\nunsupervised machine learning clustering methods and dimensionality reduction\nmethods. This method does not require prior knowledge of the number of clusters\nor the distance between clusters, handles clusters of different sizes and\nshapes, and scales linearly with the dataset. The combination of MSA with the\nlinear mapping hash function is a computationally efficient way of gene\nsequence clustering and can be a valuable tool for the assessment of\nsimilarity, clustering of different microbial genomes, identifying reference\nsequences, and for the study of evolution of bacteria and viruses.\n", "date": "2023-11-29", "categories": []}
{"id": "2311.17965", "title": "Defining Reference Sequences for Nocardia Species by Similarity and\n  Clustering Analyses of 16S rRNA Gene Sequence Data", "abstract": "  The intra- and inter-species genetic diversity of bacteria and the absence of\n'reference', or the most representative, sequences of individual species\npresent a significant challenge for sequence-based identification. The aims of\nthis study were to determine the utility, and compare the performance of\nseveral clustering and classification algorithms to identify the species of 364\nsequences of 16S rRNA gene with a defined species in GenBank, and 110 sequences\nof 16S rRNA gene with no defined species, all within the genus Nocardia. A\ntotal of 364 16S rRNA gene sequences of Nocardia species were studied. In\naddition, 110 16S rRNA gene sequences assigned only to the Nocardia genus level\nat the time of submission to GenBank were used for machine learning\nclassification experiments. Different clustering algorithms were compared with\na novel algorithm or the linear mapping (LM) of the distance matrix. Principal\nComponents Analysis was used for the dimensionality reduction and\nvisualization. Results: The LM algorithm achieved the highest performance and\nclassified the set of 364 16S rRNA sequences into 80 clusters, the majority of\nwhich (83.52%) corresponded with the original species. The most representative\n16S rRNA sequences for individual Nocardia species have been identified as\n'centroids' in respective clusters from which the distances to all other\nsequences were minimized; 110 16S rRNA gene sequences with identifications\nrecorded only at the genus level were classified using machine learning\nmethods. Simple kNN machine learning demonstrated the highest performance and\nclassified Nocardia species sequences with an accuracy of 92.7% and a mean\nfrequency of 0.578.\n", "date": "2023-11-29", "categories": []}
{"id": "2311.17969", "title": "Generation of a Compendium of Transcription Factor Cascades and\n  Identification of Potential Therapeutic Targets using Graph Machine Learning", "abstract": "  Transcription factors (TFs) play a vital role in the regulation of gene\nexpression thereby making them critical to many cellular processes. In this\nstudy, we used graph machine learning methods to create a compendium of TF\ncascades using data extracted from the STRING database. A TF cascade is a\nsequence of TFs that regulate each other, forming a directed path in the TF\nnetwork. We constructed a knowledge graph of 81,488 unique TF cascades, with\nthe longest cascade consisting of 62 TFs. Our results highlight the complex and\nintricate nature of TF interactions, where multiple TFs work together to\nregulate gene expression. We also identified 10 TFs with the highest regulatory\ninfluence based on centrality measurements, providing valuable information for\nresearchers interested in studying specific TFs. Furthermore, our pathway\nenrichment analysis revealed significant enrichment of various pathways and\nfunctional categories, including those involved in cancer and other diseases,\nas well as those involved in development, differentiation, and cell signaling.\nThe enriched pathways identified in this study may have potential as targets\nfor therapeutic intervention in diseases associated with dysregulation of\ntranscription factors. We have released the dataset, knowledge graph, and\ngraphML methods for the TF cascades, and created a website to display the\nresults, which can be accessed by researchers interested in using this dataset.\nOur study provides a valuable resource for understanding the complex network of\ninteractions between TFs and their regulatory roles in cellular processes.\n", "date": "2023-11-29", "categories": []}
{"id": "2311.17970", "title": "Description Generation using Variational Auto-Encoders for precursor\n  microRNA", "abstract": "  Micro RNAs (miRNA) are a type of non-coding RNA, which are involved in gene\nregulation and can be associated with diseases such as cancer, cardiovascular\nand neurological diseases. As such, identifying the entire genome of miRNA can\nbe of great relevance. Since experimental methods for novel precursor miRNA\n(pre-miRNA) detection are complex and expensive, computational detection using\nML could be useful. Existing ML methods are often complex black boxes, which do\nnot create an interpretable structural description of pre-miRNA. In this paper,\nwe propose a novel framework, which makes use of generative modeling through\nVariational Auto-Encoders to uncover the generative factors of pre-miRNA. After\ntraining the VAE, the pre-miRNA description is developed using a decision tree\non the lower dimensional latent space. Applying the framework to miRNA\nclassification, we obtain a high reconstruction and classification performance,\nwhile also developing an accurate miRNA description.\n", "date": "2023-11-29", "categories": []}
{"id": "2310.01768", "title": "Backdiff: a diffusion model for generalized transferable protein\n  backmapping", "abstract": "  Coarse-grained (CG) models play a crucial role in the study of protein\nstructures, protein thermodynamic properties, and protein conformation\ndynamics. Due to the information loss in the coarse-graining process,\nbackmapping from CG to all-atom configurations is essential in many protein\ndesign and drug discovery applications when detailed atomic representations are\nneeded for in-depth studies. Despite recent progress in data-driven backmapping\napproaches, devising a backmapping method that can be universally applied\nacross various CG models and proteins remains unresolved. In this work, we\npropose BackDiff, a new generative model designed to achieve generalization and\nreliability in the protein backmapping problem. BackDiff leverages the\nconditional score-based diffusion model with geometric representations. Since\ndifferent CG models can contain different coarse-grained sites which include\nselected atoms (CG atoms) and simple CG auxiliary functions of atomistic\ncoordinates (CG auxiliary variables), we design a self-supervised training\nframework to adapt to different CG atoms, and constrain the diffusion sampling\npaths with arbitrary CG auxiliary variables as conditions. Our method\nfacilitates end-to-end training and allows efficient sampling across different\nproteins and diverse CG models without the need for retraining. Comprehensive\nexperiments over multiple popular CG models demonstrate BackDiff's superior\nperformance to existing state-of-the-art approaches, and generalization and\nflexibility that these approaches cannot achieve. A pretrained BackDiff model\ncan offer a convenient yet reliable plug-and-play solution for protein\nresearchers, enabling them to investigate further from their own CG models.\n", "date": "2023-10-03", "categories": []}
{"id": "2311.17521", "title": "Spinal Muscle Atrophy Disease Modelling as Bayesian Network", "abstract": "  We investigate the molecular gene expressions studies and public databases\nfor disease modelling using Probabilistic Graphical Models and Bayesian\nInference. A case study on Spinal Muscle Atrophy Genome-Wide Association Study\nresults is modelled and analyzed. The genes up and down-regulated in two stages\nof the disease development are linked to prior knowledge published in the\npublic domain and co-expressions network is created and analyzed. The Molecular\nPathways triggered by these genes are identified. The Bayesian inference\nposteriors distributions are estimated using a variational analytical algorithm\nand a Markov chain Monte Carlo sampling algorithm. Assumptions, limitations and\npossible future work are concluded.\n", "date": "2023-11-29", "categories": []}
{"id": "2311.09312", "title": "H-Packer: Holographic Rotationally Equivariant Convolutional Neural\n  Network for Protein Side-Chain Packing", "abstract": "  Accurately modeling protein 3D structure is essential for the design of\nfunctional proteins. An important sub-task of structure modeling is protein\nside-chain packing: predicting the conformation of side-chains (rotamers) given\nthe protein's backbone structure and amino-acid sequence. Conventional\napproaches for this task rely on expensive sampling procedures over\nhand-crafted energy functions and rotamer libraries. Recently, several deep\nlearning methods have been developed to tackle the problem in a data-driven\nway, albeit with vastly different formulations (from image-to-image translation\nto directly predicting atomic coordinates). Here, we frame the problem as a\njoint regression over the side-chains' true degrees of freedom: the dihedral\n$\\chi$ angles. We carefully study possible objective functions for this task,\nwhile accounting for the underlying symmetries of the task. We propose\nHolographic Packer (H-Packer), a novel two-stage algorithm for side-chain\npacking built on top of two light-weight rotationally equivariant neural\nnetworks. We evaluate our method on CASP13 and CASP14 targets. H-Packer is\ncomputationally efficient and shows favorable performance against conventional\nphysics-based algorithms and is competitive against alternative deep learning\nsolutions.\n", "date": "2023-11-15", "categories": []}
{"id": "2311.16126", "title": "A Hierarchical Training Paradigm for Antibody Structure-sequence\n  Co-design", "abstract": "  Therapeutic antibodies are an essential and rapidly expanding drug modality.\nThe binding specificity between antibodies and antigens is decided by\ncomplementarity-determining regions (CDRs) at the tips of these Y-shaped\nproteins. In this paper, we propose a hierarchical training paradigm (HTP) for\nthe antibody sequence-structure co-design. HTP consists of four levels of\ntraining stages, each corresponding to a specific protein modality within a\nparticular protein domain. Through carefully crafted tasks in different stages,\nHTP seamlessly and effectively integrates geometric graph neural networks\n(GNNs) with large-scale protein language models to excavate evolutionary\ninformation from not only geometric structures but also vast antibody and\nnon-antibody sequence databases, which determines ligand binding pose and\nstrength. Empirical experiments show that HTP sets the new state-of-the-art\nperformance in the co-design problem as well as the fix-backbone design. Our\nresearch offers a hopeful path to unleash the potential of deep generative\narchitectures and seeks to illuminate the way forward for the antibody sequence\nand structure co-design challenge.\n", "date": "2023-10-30", "categories": []}
{"id": "2311.16132", "title": "A novel RNA pseudouridine site prediction model using Utility Kernel and\n  data-driven parameters", "abstract": "  RNA protein Interactions (RPIs) play an important role in biological systems.\nRecently, we have enumerated the RPIs at the residue level and have elucidated\nthe minimum structural unit (MSU) in these interactions to be a stretch of five\nresidues (Nucleotides/amino acids). Pseudouridine is the most frequent\nmodification in RNA. The conversion of uridine to pseudouridine involves\ninteractions between pseudouridine synthase and RNA. The existing models to\npredict the pseudouridine sites in a given RNA sequence mainly depend on\nuser-defined features such as mono and dinucleotide composition/propensities of\nRNA sequences. Predicting pseudouridine sites is a non-linear classification\nproblem with limited data points. Deep Learning models are efficient\ndiscriminators when the data set size is reasonably large and fail when there\nis a paucity of data ($<1000$ samples). To mitigate this problem, we propose a\nSupport Vector Machine (SVM) Kernel based on utility theory from Economics, and\nusing data-driven parameters (i.e. MSU) as features. For this purpose, we have\nused position-specific tri/quad/pentanucleotide composition/propensity\n(PSPC/PSPP) besides nucleotide and dineculeotide composition as features. SVMs\nare known to work well in small data regimes and kernels in SVM are designed to\nclassify non-linear data. The proposed model outperforms the existing\nstate-of-the-art models significantly (10%-15% on average).\n", "date": "2023-11-02", "categories": []}
{"id": "2311.16160", "title": "Protein-ligand binding representation learning from fine-grained\n  interactions", "abstract": "  The binding between proteins and ligands plays a crucial role in the realm of\ndrug discovery. Previous deep learning approaches have shown promising results\nover traditional computationally intensive methods, but resulting in poor\ngeneralization due to limited supervised data. In this paper, we propose to\nlearn protein-ligand binding representation in a self-supervised learning\nmanner. Different from existing pre-training approaches which treat proteins\nand ligands individually, we emphasize to discern the intricate binding\npatterns from fine-grained interactions. Specifically, this self-supervised\nlearning problem is formulated as a prediction of the conclusive binding\ncomplex structure given a pocket and ligand with a Transformer based\ninteraction module, which naturally emulates the binding process. To ensure the\nrepresentation of rich binding information, we introduce two pre-training\ntasks, i.e.~atomic pairwise distance map prediction and mask ligand\nreconstruction, which comprehensively model the fine-grained interactions from\nboth structure and feature space. Extensive experiments have demonstrated the\nsuperiority of our method across various binding tasks, including\nprotein-ligand affinity prediction, virtual screening and protein-ligand\ndocking.\n", "date": "2023-11-09", "categories": []}
{"id": "2311.15696", "title": "Peptide Binding Classification on Quantum Computers", "abstract": "  We conduct an extensive study on using near-term quantum computers for a task\nin the domain of computational biology. By constructing quantum models based on\nparameterised quantum circuits we perform sequence classification on a task\nrelevant to the design of therapeutic proteins, and find competitive\nperformance with classical baselines of similar scale. To study the effect of\nnoise, we run some of the best-performing quantum models with favourable\nresource requirements on emulators of state-of-the-art noisy quantum\nprocessors. We then apply error mitigation methods to improve the signal. We\nfurther execute these quantum models on the Quantinuum H1-1 trapped-ion quantum\nprocessor and observe very close agreement with noiseless exact simulation.\nFinally, we perform feature attribution methods and find that the quantum\nmodels indeed identify sensible relationships, at least as well as the\nclassical baselines. This work constitutes the first proof-of-concept\napplication of near-term quantum computing to a task critical to the design of\ntherapeutic proteins, opening the route toward larger-scale applications in\nthis and related fields, in line with the hardware development roadmaps of\nnear-term quantum technologies.\n", "date": "2023-11-27", "categories": []}
{"id": "2311.12670", "title": "Towards a more inductive world for drug repurposing approaches", "abstract": "  Drug-target interaction (DTI) prediction is a challenging, albeit essential\ntask in drug repurposing. Learning on graph models have drawn special attention\nas they can significantly reduce drug repurposing costs and time commitment.\nHowever, many current approaches require high-demanding additional information\nbesides DTIs that complicates their evaluation process and usability.\nAdditionally, structural differences in the learning architecture of current\nmodels hinder their fair benchmarking. In this work, we first perform an\nin-depth evaluation of current DTI datasets and prediction models through a\nrobust benchmarking process, and show that DTI prediction methods based on\ntransductive models lack generalization and lead to inflated performance when\nevaluated as previously done in the literature, hence not being suited for drug\nrepurposing approaches. We then propose a novel biologically-driven strategy\nfor negative edge subsampling and show through in vitro validation that newly\ndiscovered interactions are indeed true. We envision this work as the\nunderpinning for future fair benchmarking and robust model design. All\ngenerated resources and tools are publicly available as a python package.\n", "date": "2023-11-21", "categories": []}
{"id": "2311.12884", "title": "Identifying DNA Sequence Motifs Using Deep Learning", "abstract": "  Splice sites play a crucial role in gene expression, and accurate prediction\nof these sites in DNA sequences is essential for diagnosing and treating\ngenetic disorders. We address the challenge of splice site prediction by\nintroducing DeepDeCode, an attention-based deep learning sequence model to\ncapture the long-term dependencies in the nucleotides in DNA sequences. We\nfurther propose using visualization techniques for accurate identification of\nsequence motifs, which enhance the interpretability and trustworthiness of\nDeepDeCode. We compare DeepDeCode to other state-of-the-art methods for splice\nsite prediction and demonstrate its accuracy, explainability and efficiency.\nGiven the results of our methodology, we expect that it can used for healthcare\napplications to reason about genomic processes and be extended to discover new\nsplice sites and genomic regulatory elements.\n", "date": "2023-11-20", "categories": []}
{"id": "2301.02748", "title": "Generative Antibody Design for Complementary Chain Pairing Sequences\n  through Encoder-Decoder Language Model", "abstract": "  Current protein language models (pLMs) predominantly focus on single-chain\nprotein sequences and often have not accounted for constraints on generative\ndesign imposed by protein-protein interactions. To address this gap, we present\npaired Antibody T5 (pAbT5), an encoder-decoder model to generate complementary\nheavy or light chain from its pairing partner. We show that our model respects\nconservation in framework regions and variability in hypervariable domains,\ndemonstrated by agreement with sequence alignment and variable-length CDR\nloops. We also show that our model captures chain pairing preferences through\nthe recovery of ground-truth chain type and gene families. Our results showcase\nthe potential of pAbT5 in generative antibody design, incorporating biological\nconstraints from chain pairing preferences.\n", "date": "2023-01-06", "categories": []}
{"id": "2311.09264", "title": "Cross-domain feature disentanglement for interpretable modeling of tumor\n  microenvironment impact on drug response", "abstract": "  High-throughput screening technology has facilitated the generation of\nlarge-scale drug responses across hundreds of cancer cell lines. However, there\nexists significant discrepancy between in vitro cell lines and actual tumors in\nvivo in terms of their response to drug treatments, because of tumors comprise\nof complex cellular compositions and histopathology structure, known as tumor\nmicroenvironment (TME), which greatly influences the drug cytotoxicity against\ntumor cells. To date, no study has focused on modeling the impact of the TME on\nclinical drug response. This paper proposed a domain adaptation network for\nfeature disentanglement to separate representations of cancer cells and TME of\na tumor in patients. Two denoising autoencoders were separately used to extract\nfeatures from cell lines (source domain) and tumors (target domain) for partial\ndomain alignment and feature decoupling. The specific encoder was enforced to\nextract information only about TME. Moreover, to ensure generalizability to\nnovel drugs, we applied a graph attention network to learn the latent\nrepresentation of drugs, allowing us to linearly model the drug perturbation on\ncellular state in latent space. We calibrated our model on a benchmark dataset\nand demonstrated its superior performance in predicting clinical drug response\nand dissecting the influence of the TME on drug efficacy.\n", "date": "2023-11-15", "categories": []}
{"id": "2311.07636", "title": "Attention-based Multi-task Learning for Base Editor Outcome Prediction", "abstract": "  Human genetic diseases often arise from point mutations, emphasizing the\ncritical need for precise genome editing techniques. Among these, base editing\nstands out as it allows targeted alterations at the single nucleotide level.\nHowever, its clinical application is hindered by low editing efficiency and\nunintended mutations, necessitating extensive trial-and-error experimentation\nin the laboratory. To speed up this process, we present an attention-based\ntwo-stage machine learning model that learns to predict the likelihood of all\npossible editing outcomes for a given genomic target sequence. We further\npropose a multi-task learning schema to jointly learn multiple base editors\n(i.e. variants) at once. Our model's predictions consistently demonstrated a\nstrong correlation with the actual experimental results on multiple datasets\nand base editor variants. These results provide further validation for the\nmodels' capacity to enhance and accelerate the process of refining base editing\ndesigns.\n", "date": "2023-11-13", "categories": []}
{"id": "2306.14809", "title": "Tanimoto Random Features for Scalable Molecular Machine Learning", "abstract": "  The Tanimoto coefficient is commonly used to measure the similarity between\nmolecules represented as discrete fingerprints, either as a distance metric or\na positive definite kernel. While many kernel methods can be accelerated using\nrandom feature approximations, at present there is a lack of such\napproximations for the Tanimoto kernel. In this paper we propose two kinds of\nnovel random features to allow this kernel to scale to large datasets, and in\nthe process discover a novel extension of the kernel to real-valued vectors. We\ntheoretically characterize these random features, and provide error bounds on\nthe spectral norm of the Gram matrix. Experimentally, we show that these random\nfeatures are effective at approximating the Tanimoto coefficient of real-world\ndatasets and are useful for molecular property prediction and optimization\ntasks.\n", "date": "2023-06-26", "categories": []}
{"id": "2306.15794", "title": "HyenaDNA: Long-Range Genomic Sequence Modeling at Single Nucleotide\n  Resolution", "abstract": "  Genomic (DNA) sequences encode an enormous amount of information for gene\nregulation and protein synthesis. Similar to natural language models,\nresearchers have proposed foundation models in genomics to learn generalizable\nfeatures from unlabeled genome data that can then be fine-tuned for downstream\ntasks such as identifying regulatory elements. Due to the quadratic scaling of\nattention, previous Transformer-based genomic models have used 512 to 4k tokens\nas context (<0.001% of the human genome), significantly limiting the modeling\nof long-range interactions in DNA. In addition, these methods rely on\ntokenizers or fixed k-mers to aggregate meaningful DNA units, losing single\nnucleotide resolution where subtle genetic variations can completely alter\nprotein function via single nucleotide polymorphisms (SNPs). Recently, Hyena, a\nlarge language model based on implicit convolutions was shown to match\nattention in quality while allowing longer context lengths and lower time\ncomplexity. Leveraging Hyena's new long-range capabilities, we present\nHyenaDNA, a genomic foundation model pretrained on the human reference genome\nwith context lengths of up to 1 million tokens at the single nucleotide-level -\nan up to 500x increase over previous dense attention-based models. HyenaDNA\nscales sub-quadratically in sequence length (training up to 160x faster than\nTransformer), uses single nucleotide tokens, and has full global context at\neach layer. We explore what longer context enables - including the first use of\nin-context learning in genomics. On fine-tuned benchmarks from the Nucleotide\nTransformer, HyenaDNA reaches state-of-the-art (SotA) on 12 of 18 datasets\nusing a model with orders of magnitude less parameters and pretraining data. On\nthe GenomicBenchmarks, HyenaDNA surpasses SotA on 7 of 8 datasets on average by\n+10 accuracy points. Code at https://github.com/HazyResearch/hyena-dna.\n", "date": "2023-06-27", "categories": []}
{"id": "2311.07621", "title": "To Transformers and Beyond: Large Language Models for the Genome", "abstract": "  In the rapidly evolving landscape of genomics, deep learning has emerged as a\nuseful tool for tackling complex computational challenges. This review focuses\non the transformative role of Large Language Models (LLMs), which are mostly\nbased on the transformer architecture, in genomics. Building on the foundation\nof traditional convolutional neural networks and recurrent neural networks, we\nexplore both the strengths and limitations of transformers and other LLMs for\ngenomics. Additionally, we contemplate the future of genomic modeling beyond\nthe transformer architecture based on current trends in research. The paper\naims to serve as a guide for computational biologists and computer scientists\ninterested in LLMs for genomic data. We hope the paper can also serve as an\neducational introduction and discussion for biologists to a fundamental shift\nin how we will be analyzing genomic data in the future.\n", "date": "2023-11-13", "categories": []}
{"id": "2208.09652", "title": "Unsupervisedly Prompting AlphaFold2 for Few-Shot Learning of Accurate\n  Folding Landscape and Protein Structure Prediction", "abstract": "  Data-driven predictive methods which can efficiently and accurately transform\nprotein sequences into biologically active structures are highly valuable for\nscientific research and medical development. Determining accurate folding\nlandscape using co-evolutionary information is fundamental to the success of\nmodern protein structure prediction methods. As the state of the art,\nAlphaFold2 has dramatically raised the accuracy without performing explicit\nco-evolutionary analysis. Nevertheless, its performance still shows strong\ndependence on available sequence homologs. Based on the interrogation on the\ncause of such dependence, we presented EvoGen, a meta generative model, to\nremedy the underperformance of AlphaFold2 for poor MSA targets. By prompting\nthe model with calibrated or virtually generated homologue sequences, EvoGen\nhelps AlphaFold2 fold accurately in low-data regime and even achieve\nencouraging performance with single-sequence predictions. Being able to make\naccurate predictions with few-shot MSA not only generalizes AlphaFold2 better\nfor orphan sequences, but also democratizes its use for high-throughput\napplications. Besides, EvoGen combined with AlphaFold2 yields a probabilistic\nstructure generation method which could explore alternative conformations of\nprotein sequences, and the task-aware differentiable algorithm for sequence\ngeneration will benefit other related tasks including protein design.\n", "date": "2022-08-20", "categories": []}
{"id": "2305.14575", "title": "Towards Early Prediction of Human iPSC Reprogramming Success", "abstract": "  This paper presents advancements in automated early-stage prediction of the\nsuccess of reprogramming human induced pluripotent stem cells (iPSCs) as a\npotential source for regenerative cell therapies.The minuscule success rate of\niPSC-reprogramming of around $ 0.01% $ to $ 0.1% $ makes it labor-intensive,\ntime-consuming, and exorbitantly expensive to generate a stable iPSC line.\nSince that requires culturing of millions of cells and intense biological\nscrutiny of multiple clones to identify a single optimal clone. The ability to\nreliably predict which cells are likely to establish as an optimal iPSC line at\nan early stage of pluripotency would therefore be ground-breaking in rendering\nthis a practical and cost-effective approach to personalized medicine. Temporal\ninformation about changes in cellular appearance over time is crucial for\npredicting its future growth outcomes. In order to generate this data, we first\nperformed continuous time-lapse imaging of iPSCs in culture using an ultra-high\nresolution microscope. We then annotated the locations and identities of cells\nin late-stage images where reliable manual identification is possible. Next, we\npropagated these labels backwards in time using a semi-automated tracking\nsystem to obtain labels for early stages of growth. Finally, we used this data\nto train deep neural networks to perform automatic cell segmentation and\nclassification. Our code and data are available at\nhttps://github.com/abhineet123/ipsc_prediction.\n", "date": "2023-05-23", "categories": []}
{"id": "2311.06699", "title": "An Investigation of Hepatitis B Virus Genome using Markov Models", "abstract": "  The human genome encodes a family of editing enzymes known as APOBEC3\n(apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like 3). Several\nfamily members, such as APO-BEC3G, APOBEC3F, and APOBEC3H haplotype II, exhibit\nactivity against viruses such as HIV. These enzymes induce C-to-U mutations in\nthe negative strand of viral genomes, resulting in multiple G-to-A changes,\ncommonly referred to as 'hypermutation.' Mutations catalyzed by these enzymes\nare sequence context-dependent in the HIV genome; for instance, APOBEC3G\npreferen-tially mutates G within GG, TGG, and TGGG contexts, while other\nmembers mutate G within GA, TGA, and TGAA contexts. However, the same sequence\ncontext has not been explored in relation to these enzymes and HBV. In this\nstudy, our objective is to identify the mutational footprint of APOBEC3 enzymes\nin the HBV genome. To achieve this, we employ a multivariable data analytics\ntechnique to investigate motif preferences and potential sequence hierarchies\nof mutation by APOBEC3 enzymes using full genome HBV sequences from a diverse\nrange of naturally infected patients. This approach allows us to distinguish\nbetween normal and hypermutated sequences based on the representation of mono-\nto tetra-nucleotide motifs. Additionally, we aim to identify motifs associated\nwith hypermutation induced by different APOBEC3 enzymes in HBV genomes. Our\nanalyses reveal that either APOBEC3 enzymes are not active against HBV, or the\ninduction of G-to-A mutations by these enzymes is not sequence\ncontext-dependent in the HBV genome.\n", "date": "2023-11-12", "categories": []}
{"id": "2310.02919", "title": "Attention-based Multi-task Learning for Base Editor Outcome Prediction", "abstract": "  Human genetic diseases often arise from point mutations, emphasizing the\ncritical need for precise genome editing techniques. Among these, base editing\nstands out as it allows targeted alterations at the single nucleotide level.\nHowever, its clinical application is hindered by low editing efficiency and\nunintended mutations, necessitating extensive trial-and-error experimentation\nin the laboratory. To speed up this process, we present an attention-based\ntwo-stage machine learning model that learns to predict the likelihood of all\npossible editing outcomes for a given genomic target sequence. We further\npropose a multi-task learning schema to jointly learn multiple base editors\n(i.e. variants) at once. Our model's predictions consistently demonstrated a\nstrong correlation with the actual experimental results on multiple datasets\nand base editor variants. These results provide further validation for the\nmodels' capacity to enhance and accelerate the process of refining base editing\ndesigns.\n", "date": "2023-10-04", "categories": []}
{"id": "2311.05486", "title": "Disease Gene Prioritization With Quantum Walks", "abstract": "  Disease gene prioritization assigns scores to genes or proteins according to\ntheir likely relevance for a given disease based on a provided set of seed\ngenes. Here, we describe a new algorithm for disease gene prioritization based\non continuous-time quantum walks using the adjacency matrix of a\nprotein-protein interaction (PPI) network. Our algorithm can be seen as a\nquantum version of a previous method known as the diffusion kernel, but,\nimportantly, has higher performance in predicting disease genes, and also\npermits the encoding of seed node self-loops into the underlying Hamiltonian,\nwhich offers yet another boost in performance. We demonstrate the success of\nour proposed method by comparing it to several well-known gene prioritization\nmethods on three disease sets, across seven different PPI networks. In order to\ncompare these methods, we use cross-validation and examine the mean reciprocal\nranks and recall values. We further validate our method by performing an\nenrichment analysis of the predicted genes for coronary artery disease. We also\ninvestigate the impact of adding self-loops to the seeds, and argue that they\nallow the quantum walker to remain more local to low-degree seed nodes.\n", "date": "2023-11-09", "categories": []}
{"id": "2311.04419", "title": "PepLand: a large-scale pre-trained peptide representation model for a\n  comprehensive landscape of both canonical and non-canonical amino acids", "abstract": "  In recent years, the scientific community has become increasingly interested\non peptides with non-canonical amino acids due to their superior stability and\nresistance to proteolytic degradation. These peptides present promising\nmodifications to biological, pharmacological, and physiochemical attributes in\nboth endogenous and engineered peptides. Notwithstanding their considerable\nadvantages, the scientific community exhibits a conspicuous absence of an\neffective pre-trained model adept at distilling feature representations from\nsuch complex peptide sequences. We herein propose PepLand, a novel pre-training\narchitecture for representation and property analysis of peptides spanning both\ncanonical and non-canonical amino acids. In essence, PepLand leverages a\ncomprehensive multi-view heterogeneous graph neural network tailored to unveil\nthe subtle structural representations of peptides. Empirical validations\nunderscore PepLand's effectiveness across an array of peptide property\npredictions, encompassing protein-protein interactions, permeability,\nsolubility, and synthesizability. The rigorous evaluation confirms PepLand's\nunparalleled capability in capturing salient synthetic peptide features,\nthereby laying a robust foundation for transformative advances in\npeptide-centric research domains. We have made all the source code utilized in\nthis study publicly accessible via GitHub at\nhttps://github.com/zhangruochi/pepland\n", "date": "2023-11-08", "categories": []}
{"id": "2311.03409", "title": "Visualizing DNA reaction trajectories with deep graph embedding\n  approaches", "abstract": "  Synthetic biologists and molecular programmers design novel nucleic acid\nreactions, with many potential applications. Good visualization tools are\nneeded to help domain experts make sense of the complex outputs of folding\npathway simulations of such reactions. Here we present ViDa, a new approach for\nvisualizing DNA reaction folding trajectories over the energy landscape of\nsecondary structures. We integrate a deep graph embedding model with common\ndimensionality reduction approaches, to map high-dimensional data onto 2D\nEuclidean space. We assess ViDa on two well-studied and contrasting DNA\nhybridization reactions. Our preliminary results suggest that ViDa's\nvisualization successfully separates trajectories with different folding\nmechanisms, thereby providing useful insight to users, and is a big improvement\nover the current state-of-the-art in DNA kinetics visualization.\n", "date": "2023-11-06", "categories": []}
{"id": "2311.03411", "title": "ViDa: Visualizing DNA hybridization trajectories with\n  biophysics-informed deep graph embeddings", "abstract": "  Visualization tools can help synthetic biologists and molecular programmers\nunderstand the complex reactive pathways of nucleic acid reactions, which can\nbe designed for many potential applications and can be modelled using a\ncontinuous-time Markov chain (CTMC). Here we present ViDa, a new visualization\napproach for DNA reaction trajectories that uses a 2D embedding of the\nsecondary structure state space underlying the CTMC model. To this end, we\nintegrate a scattering transform of the secondary structure adjacency, a\nvariational autoencoder, and a nonlinear dimensionality reduction method. We\naugment the training loss with domain-specific supervised terms that capture\nboth thermodynamic and kinetic features. We assess ViDa on two well-studied DNA\nhybridization reactions. Our results demonstrate that the domain-specific\nfeatures lead to significant quality improvements over the state-of-the-art in\nDNA state space visualization, successfully separating different folding\npathways and thus providing useful insights into dominant reaction mechanisms.\n", "date": "2023-11-06", "categories": []}
{"id": "2302.08284", "title": "ClaPIM: Scalable Sequence CLAssification using Processing-In-Memory", "abstract": "  DNA sequence classification is a fundamental task in computational biology\nwith vast implications for applications such as disease prevention and drug\ndesign. Therefore, fast high-quality sequence classifiers are significantly\nimportant. This paper introduces ClaPIM, a scalable DNA sequence classification\narchitecture based on the emerging concept of hybrid in-crossbar and\nnear-crossbar memristive processing-in-memory (PIM). We enable efficient and\nhigh-quality classification by uniting the filter and search stages within a\nsingle algorithm. Specifically, we propose a custom filtering technique that\ndrastically narrows the search space and a search approach that facilitates\napproximate string matching through a distance function. ClaPIM is the first\nPIM architecture for scalable approximate string matching that benefits from\nthe high density of memristive crossbar arrays and the massive computational\nparallelism of PIM. Compared with Kraken2, a state-of-the-art software\nclassifier, ClaPIM provides significantly higher classification quality (up to\n20x improvement in F1 score) and also demonstrates a 1.8x throughput\nimprovement. Compared with EDAM, a recently-proposed SRAM-based accelerator\nthat is restricted to small datasets, we observe both a 30.4x improvement in\nnormalized throughput per area and a 7% increase in classification precision.\n", "date": "2023-02-16", "categories": []}
{"id": "2311.02326", "title": "FragXsiteDTI: Revealing Responsible Segments in Drug-Target Interaction\n  with Transformer-Driven Interpretation", "abstract": "  Drug-Target Interaction (DTI) prediction is vital for drug discovery, yet\nchallenges persist in achieving model interpretability and optimizing\nperformance. We propose a novel transformer-based model, FragXsiteDTI, that\naims to address these challenges in DTI prediction. Notably, FragXsiteDTI is\nthe first DTI model to simultaneously leverage drug molecule fragments and\nprotein pockets. Our information-rich representations for both proteins and\ndrugs offer a detailed perspective on their interaction. Inspired by the\nPerceiver IO framework, our model features a learnable latent array, initially\ninteracting with protein binding site embeddings using cross-attention and\nlater refined through self-attention and used as a query to the drug fragments\nin the drug's cross-attention transformer block. This learnable query array\nserves as a mediator and enables seamless information translation, preserving\ncritical nuances in drug-protein interactions. Our computational results on\nthree benchmarking datasets demonstrate the superior predictive power of our\nmodel over several state-of-the-art models. We also show the interpretability\nof our model in terms of the critical components of both target proteins and\ndrug molecules within drug-target pairs.\n", "date": "2023-11-04", "categories": []}
{"id": "2311.03314", "title": "FATE: Feature-Agnostic Transformer-based Encoder for learning\n  generalized embedding spaces in flow cytometry data", "abstract": "  While model architectures and training strategies have become more generic\nand flexible with respect to different data modalities over the past years, a\npersistent limitation lies in the assumption of fixed quantities and\narrangements of input features. This limitation becomes particularly relevant\nin scenarios where the attributes captured during data acquisition vary across\ndifferent samples. In this work, we aim at effectively leveraging data with\nvarying features, without the need to constrain the input space to the\nintersection of potential feature sets or to expand it to their union. We\npropose a novel architecture that can directly process data without the\nnecessity of aligned feature modalities by learning a general embedding space\nthat captures the relationship between features across data samples with\nvarying sets of features. This is achieved via a set-transformer architecture\naugmented by feature-encoder layers, thereby enabling the learning of a shared\nlatent feature space from data originating from heterogeneous feature spaces.\nThe advantages of the model are demonstrated for automatic cancer cell\ndetection in acute myeloid leukemia in flow cytometry data, where the features\nmeasured during acquisition often vary between samples. Our proposed\narchitecture's capacity to operate seamlessly across incongruent feature spaces\nis particularly relevant in this context, where data scarcity arises from the\nlow prevalence of the disease. The code is available for research purposes at\nhttps://github.com/lisaweijler/FATE.\n", "date": "2023-11-06", "categories": []}
{"id": "2307.08452", "title": "SBMLtoODEjax: Efficient Simulation and Optimization of Biological\n  Network Models in JAX", "abstract": "  Advances in bioengineering and biomedicine demand a deep understanding of the\ndynamic behavior of biological systems, ranging from protein pathways to\ncomplex cellular processes. Biological networks like gene regulatory networks\nand protein pathways are key drivers of embryogenesis and physiological\nprocesses. Comprehending their diverse behaviors is essential for tackling\ndiseases, including cancer, as well as for engineering novel biological\nconstructs. Despite the availability of extensive mathematical models\nrepresented in Systems Biology Markup Language (SBML), researchers face\nsignificant challenges in exploring the full spectrum of behaviors and\noptimizing interventions to efficiently shape those behaviors. Existing tools\ndesigned for simulation of biological network models are not tailored to\nfacilitate interventions on network dynamics nor to facilitate automated\ndiscovery. Leveraging recent developments in machine learning (ML), this paper\nintroduces SBMLtoODEjax, a lightweight library designed to seamlessly integrate\nSBML models with ML-supported pipelines, powered by JAX. SBMLtoODEjax\nfacilitates the reuse and customization of SBML-based models, harnessing JAX's\ncapabilities for efficient parallel simulations and optimization, with the aim\nto accelerate research in biological network analysis.\n", "date": "2023-07-17", "categories": []}
{"id": "2310.10893", "title": "Active Learning Framework for Cost-Effective TCR-Epitope Binding\n  Affinity Prediction", "abstract": "  T cell receptors (TCRs) are critical components of adaptive immune systems,\nresponsible for responding to threats by recognizing epitope sequences\npresented on host cell surface. Computational prediction of binding affinity\nbetween TCRs and epitope sequences using machine/deep learning has attracted\nintense attention recently. However, its success is hindered by the lack of\nlarge collections of annotated TCR-epitope pairs. Annotating their binding\naffinity requires expensive and time-consuming wet-lab evaluation. To reduce\nannotation cost, we present ActiveTCR, a framework that incorporates active\nlearning and TCR-epitope binding affinity prediction models. Starting with a\nsmall set of labeled training pairs, ActiveTCR iteratively searches for\nunlabeled TCR-epitope pairs that are ''worth'' for annotation. It aims to\nmaximize performance gains while minimizing the cost of annotation. We compared\nfour query strategies with a random sampling baseline and demonstrated that\nActiveTCR reduces annotation costs by approximately 40%. Furthermore, we showed\nthat providing ground truth labels of TCR-epitope pairs to query strategies can\nhelp identify and reduce more than 40% redundancy among already annotated pairs\nwithout compromising model performance, enabling users to train equally\npowerful prediction models with less training data. Our work is the first\nsystematic investigation of data optimization for TCR-epitope binding affinity\nprediction.\n", "date": "2023-10-16", "categories": []}
{"id": "2310.19513", "title": "Inverse folding for antibody sequence design using deep learning", "abstract": "  We consider the problem of antibody sequence design given 3D structural\ninformation. Building on previous work, we propose a fine-tuned inverse folding\nmodel that is specifically optimised for antibody structures and outperforms\ngeneric protein models on sequence recovery and structure robustness when\napplied on antibodies, with notable improvement on the hypervariable CDR-H3\nloop. We study the canonical conformations of complementarity-determining\nregions and find improved encoding of these loops into known clusters. Finally,\nwe consider the applications of our model to drug discovery and binder design\nand evaluate the quality of proposed sequences using physics-based methods.\n", "date": "2023-10-30", "categories": []}
{"id": "2310.13725", "title": "Enhancing drug and cell line representations via contrastive learning\n  for improved anti-cancer drug prioritization", "abstract": "  Due to cancer's complex nature and variable response to therapy, precision\noncology informed by omics sequence analysis has become the current standard of\ncare. However, the amount of data produced for each patients makes it difficult\nto quickly identify the best treatment regimen. Moreover, limited data\navailability has hindered computational methods' abilities to learn patterns\nassociated with effective drug-cell line pairs. In this work, we propose the\nuse of contrastive learning to improve learned drug and cell line\nrepresentations by preserving relationship structures associated with drug\nmechanism of action and cell line cancer types. In addition to achieving\nenhanced performance relative to a state-of-the-art method, we find that\nclassifiers using our learned representations exhibit a more balances reliance\non drug- and cell line-derived features when making predictions. This\nfacilitates more personalized drug prioritizations that are informed by signals\nrelated to drug resistance.\n", "date": "2023-10-20", "categories": []}
{"id": "2310.15211", "title": "Modeling Path Importance for Effective Alzheimer's Disease Drug\n  Repurposing", "abstract": "  Recently, drug repurposing has emerged as an effective and resource-efficient\nparadigm for AD drug discovery. Among various methods for drug repurposing,\nnetwork-based methods have shown promising results as they are capable of\nleveraging complex networks that integrate multiple interaction types, such as\nprotein-protein interactions, to more effectively identify candidate drugs.\nHowever, existing approaches typically assume paths of the same length in the\nnetwork have equal importance in identifying the therapeutic effect of drugs.\nOther domains have found that same length paths do not necessarily have the\nsame importance. Thus, relying on this assumption may be deleterious to drug\nrepurposing attempts. In this work, we propose MPI (Modeling Path Importance),\na novel network-based method for AD drug repurposing. MPI is unique in that it\nprioritizes important paths via learned node embeddings, which can effectively\ncapture a network's rich structural information. Thus, leveraging learned\nembeddings allows MPI to effectively differentiate the importance among paths.\nWe evaluate MPI against a commonly used baseline method that identifies anti-AD\ndrug candidates primarily based on the shortest paths between drugs and AD in\nthe network. We observe that among the top-50 ranked drugs, MPI prioritizes\n20.0% more drugs with anti-AD evidence compared to the baseline. Finally, Cox\nproportional-hazard models produced from insurance claims data aid us in\nidentifying the use of etodolac, nicotine, and BBB-crossing ACE-INHs as having\na reduced risk of AD, suggesting such drugs may be viable candidates for\nrepurposing and should be explored further in future studies.\n", "date": "2023-10-23", "categories": []}
{"id": "2310.17164", "title": "Bridging Phylogeny and Taxonomy with Protein-protein Interaction\n  Networks", "abstract": "  The protein-protein interaction (PPI) network provides an overview of the\ncomplex biological reactions vital to an organism's metabolism and survival.\nEven though in the past PPI network were compared across organisms in detail,\nthere has not been large-scale research on how individual PPI networks reflect\non the species relationships. In this study we aim to increase our\nunderstanding of the tree of life and taxonomy by gleaming information from the\nPPI networks. We successful created (1) a predictor of network statistics based\non known traits of existing species in the phylogeny, and (2) a taxonomic\nclassifier of organism using the known protein network statistics, whether\nexperimentally determined or predicted de novo. With the knowledge of protein\ninteractions at its core, our two models effectively connects two field with\nwidely diverging methodologies - the phylogeny and taxonomy of species.\n", "date": "2023-10-26", "categories": []}
{"id": "2310.15201", "title": "A Skin Microbiome Model with AMP interactions and Analysis of\n  Quasi-Stability vs Stability in Population Dynamics", "abstract": "  The skin microbiome plays an important role in the maintenance of a healthy\nskin. It is an ecosystem, composed of several species, competing for resources\nand interacting with the skin cells. Imbalance in the cutaneous microbiome,\nalso called dysbiosis, has been correlated with several skin conditions,\nincluding acne and atopic dermatitis. Generally, dysbiosis is linked to\ncolonization of the skin by a population of opportunistic pathogenic bacteria.\nTreatments consisting in non-specific elimination of cutaneous microflora have\nshown conflicting results. In this article, we introduce a mathematical model\nbased on ordinary differential equations, with 2 types of bacteria populations\n(skin commensals and opportunistic pathogens) and including the production of\nantimicrobial peptides to study the mechanisms driving the dominance of one\npopulation over the other. By using published experimental data, assumed to\ncorrespond to the observation of stable states in our model, we reduce the\nnumber of parameters of the model from 13 to 5. We then use a formal\nspecification in quantitative temporal logic to calibrate our model by global\nparameter optimization and perform sensitivity analyses. On the time scale of 2\ndays of the experiments, the model predicts that certain changes of the\nenvironment, like the elevation of skin surface pH, create favorable conditions\nfor the emergence and colonization of the skin by the opportunistic pathogen\npopulation, while the production of human AMPs has non-linear effect on the\nbalance between pathogens and commensals. Surprisingly, simulations on longer\ntime scales reveal that the equilibrium reached around 2 days can in fact be a\nquasi-stable state followed by the reaching of a reversed stable state after 12\ndays or more. We analyse the conditions of quasi-stability observed in this\nmodel using tropical algebraic methods, and show their non-generic character in\ncontrast to slow-fast systems. These conditions are then generalized to a large\nclass of population dynamics models over any number of species.\n", "date": "2023-10-23", "categories": []}
{"id": "2310.15744", "title": "Analyzing Single Cell RNA Sequencing with Topological Nonnegative Matrix\n  Factorization", "abstract": "  Single-cell RNA sequencing (scRNA-seq) is a relatively new technology that\nhas stimulated enormous interest in statistics, data science, and computational\nbiology due to the high dimensionality, complexity, and large scale associated\nwith scRNA-seq data. Nonnegative matrix factorization (NMF) offers a unique\napproach due to its meta-gene interpretation of resulting low-dimensional\ncomponents. However, NMF approaches suffer from the lack of multiscale\nanalysis. This work introduces two persistent Laplacian regularized NMF\nmethods, namely, topological NMF (TNMF) and robust topological NMF (rTNMF). By\nemploying a total of 12 datasets, we demonstrate that the proposed TNMF and\nrTNMF significantly outperform all other NMF-based methods. We have also\nutilized TNMF and rTNMF for the visualization of popular Uniform Manifold\nApproximation and Projection (UMAP) and t-distributed stochastic neighbor\nembedding (t-SNE).\n", "date": "2023-10-24", "categories": []}
{"id": "2310.13806", "title": "RoseNet: Predicting Energy Metrics of Double InDel Mutants Using Deep\n  Learning", "abstract": "  An amino acid insertion or deletion, or InDel, can have profound and varying\nfunctional impacts on a protein's structure. InDel mutations in the\ntransmembrane conductor regulator protein for example give rise to cystic\nfibrosis. Unfortunately performing InDel mutations on physical proteins and\nstudying their effects is a time prohibitive process. Consequently, modeling\nInDels computationally can supplement and inform wet lab experiments. In this\nwork, we make use of our data sets of exhaustive double InDel mutations for\nthree proteins which we computationally generated using a robotics inspired\ninverse kinematics approach available in Rosetta. We develop and train a neural\nnetwork, RoseNet, on several structural and energetic metrics output by Rosetta\nduring the mutant generation process. We explore and present how RoseNet is\nable to emulate the exhaustive data set using deep learning methods, and show\nto what extent it can predict Rosetta metrics for unseen mutant sequences with\ntwo InDels. RoseNet achieves a Pearson correlation coefficient median accuracy\nof 0.775 over all Rosetta scores for the largest protein. Furthermore, a\nsensitivity analysis is performed to determine the necessary quantity of data\nrequired to accurately emulate the structural scores for computationally\ngenerated mutants. We show that the model can be trained on minimal data (<50%)\nand still retain a high level of accuracy.\n", "date": "2023-10-20", "categories": []}
{"id": "2310.14782", "title": "Towards equilibrium molecular conformation generation with GFlowNets", "abstract": "  Sampling diverse, thermodynamically feasible molecular conformations plays a\ncrucial role in predicting properties of a molecule. In this paper we propose\nto use GFlowNet for sampling conformations of small molecules from the\nBoltzmann distribution, as determined by the molecule's energy. The proposed\napproach can be used in combination with energy estimation methods of different\nfidelity and discovers a diverse set of low-energy conformations for highly\nflexible drug-like molecules. We demonstrate that GFlowNet can reproduce\nmolecular potential energy surfaces by sampling proportionally to the Boltzmann\ndistribution.\n", "date": "2023-10-20", "categories": []}
{"id": "2310.13259", "title": "Domain-specific optimization and diverse evaluation of self-supervised\n  models for histopathology", "abstract": "  Task-specific deep learning models in histopathology offer promising\nopportunities for improving diagnosis, clinical research, and precision\nmedicine. However, development of such models is often limited by availability\nof high-quality data. Foundation models in histopathology that learn general\nrepresentations across a wide range of tissue types, diagnoses, and\nmagnifications offer the potential to reduce the data, compute, and technical\nexpertise necessary to develop task-specific deep learning models with the\nrequired level of model performance. In this work, we describe the development\nand evaluation of foundation models for histopathology via self-supervised\nlearning (SSL). We first establish a diverse set of benchmark tasks involving\n17 unique tissue types and 12 unique cancer types and spanning different\noptimal magnifications and task types. Next, we use this benchmark to explore\nand evaluate histopathology-specific SSL methods followed by further evaluation\non held out patch-level and weakly supervised tasks. We found that standard SSL\nmethods thoughtfully applied to histopathology images are performant across our\nbenchmark tasks and that domain-specific methodological improvements can\nfurther increase performance. Our findings reinforce the value of using\ndomain-specific SSL methods in pathology, and establish a set of high quality\nfoundation models to enable further research across diverse applications.\n", "date": "2023-10-20", "categories": []}
{"id": "2305.04934", "title": "Generative Pretrained Autoregressive Transformer Graph Neural Network\n  applied to the Analysis and Discovery of Novel Proteins", "abstract": "  We report a flexible language-model based deep learning strategy, applied\nhere to solve complex forward and inverse problems in protein modeling, based\non an attention neural network that integrates transformer and graph\nconvolutional architectures in a causal multi-headed graph mechanism, to\nrealize a generative pretrained model. The model is applied to predict\nsecondary structure content (per-residue level and overall content), protein\nsolubility, and sequencing tasks. Further trained on inverse tasks, the model\nis rendered capable of designing proteins with these properties as target\nfeatures. The model is formulated as a general framework, completely\nprompt-based, and can be adapted for a variety of downstream tasks. We find\nthat adding additional tasks yields emergent synergies that the model exploits\nin improving overall performance, beyond what would be possible by training a\nmodel on each dataset alone. Case studies are presented to validate the method,\nyielding protein designs specifically focused on structural proteins, but also\nexploring the applicability in the design of soluble, antimicrobial\nbiomaterials. While our model is trained to ultimately perform 8 distinct\ntasks, with available datasets it can be extended to solve additional problems.\nIn a broader sense, this work illustrates a form of multiscale modeling that\nrelates a set of ultimate building blocks (here, byte-level utf8 characters\nthat define the nature of the physical system at hand) to complex output. This\nmateriomic scheme captures complex emergent relationships between universal\nbuilding block and resulting properties via a synergizing learning capacity to\nexpress a set of potentialities embedded in the knowledge used in training, via\nthe interplay of universality and diversity.\n", "date": "2023-05-07", "categories": []}
{"id": "2309.10170", "title": "Generative modeling, design and analysis of spider silk protein\n  sequences for enhanced mechanical properties", "abstract": "  Spider silks are remarkable materials characterized by superb mechanical\nproperties such as strength, extensibility and lightweightedness. Yet, to date,\nlimited models are available to fully explore sequence-property relationships\nfor analysis and design. Here we propose a custom generative large-language\nmodel to enable design of novel spider silk protein sequences to meet complex\ncombinations of target mechanical properties. The model, pretrained on a large\nset of protein sequences, is fine-tuned on ~1,000 major ampullate spidroin\n(MaSp) sequences for which associated fiber-level mechanical properties exist,\nto yield an end-to-end forward and inverse generative strategy. Performance is\nassessed through: (1), a novelty analysis and protein type classification for\ngenerated spidroin sequences through BLAST searches, (2) property evaluation\nand comparison with similar sequences, (3) comparison of molecular structures,\nas well as, and (4) a detailed sequence motif analyses. We generate silk\nsequences with property combinations that do not exist in nature, and develop a\ndeep understanding the mechanistic roles of sequence patterns in achieving\noverarching key mechanical properties (elastic modulus, strength, toughness,\nfailure strain). The model provides an efficient approach to expand the silkome\ndataset, facilitating further sequence-structure analyses of silks, and\nestablishes a foundation for synthetic silk design and optimization.\n", "date": "2023-09-18", "categories": []}
{"id": "2310.12304", "title": "Preference Optimization for Molecular Language Models", "abstract": "  Molecular language modeling is an effective approach to generating novel\nchemical structures. However, these models do not \\emph{a priori} encode\ncertain preferences a chemist may desire. We investigate the use of fine-tuning\nusing Direct Preference Optimization to better align generated molecules with\nchemist preferences. Our findings suggest that this approach is simple,\nefficient, and highly effective.\n", "date": "2023-10-18", "categories": []}
{"id": "2207.13921", "title": "HelixFold-Single: MSA-free Protein Structure Prediction by Using Protein\n  Language Model as an Alternative", "abstract": "  AI-based protein structure prediction pipelines, such as AlphaFold2, have\nachieved near-experimental accuracy. These advanced pipelines mainly rely on\nMultiple Sequence Alignments (MSAs) as inputs to learn the co-evolution\ninformation from the homologous sequences. Nonetheless, searching MSAs from\nprotein databases is time-consuming, usually taking dozens of minutes.\nConsequently, we attempt to explore the limits of fast protein structure\nprediction by using only primary sequences of proteins. HelixFold-Single is\nproposed to combine a large-scale protein language model with the superior\ngeometric learning capability of AlphaFold2. Our proposed method,\nHelixFold-Single, first pre-trains a large-scale protein language model (PLM)\nwith thousands of millions of primary sequences utilizing the self-supervised\nlearning paradigm, which will be used as an alternative to MSAs for learning\nthe co-evolution information. Then, by combining the pre-trained PLM and the\nessential components of AlphaFold2, we obtain an end-to-end differentiable\nmodel to predict the 3D coordinates of atoms from only the primary sequence.\nHelixFold-Single is validated in datasets CASP14 and CAMEO, achieving\ncompetitive accuracy with the MSA-based methods on the targets with large\nhomologous families. Furthermore, HelixFold-Single consumes much less time than\nthe mainstream pipelines for protein structure prediction, demonstrating its\npotential in tasks requiring many predictions. The code of HelixFold-Single is\navailable at\nhttps://github.com/PaddlePaddle/PaddleHelix/tree/dev/apps/protein_folding/helixfold-single,\nand we also provide stable web services on\nhttps://paddlehelix.baidu.com/app/drug/protein-single/forecast.\n", "date": "2022-07-28", "categories": []}
{"id": "2310.04292", "title": "Towards Foundational Models for Molecular Learning on Large-Scale\n  Multi-Task Datasets", "abstract": "  Recently, pre-trained foundation models have enabled significant advancements\nin multiple fields. In molecular machine learning, however, where datasets are\noften hand-curated, and hence typically small, the lack of datasets with\nlabeled features, and codebases to manage those datasets, has hindered the\ndevelopment of foundation models. In this work, we present seven novel datasets\ncategorized by size into three distinct categories: ToyMix, LargeMix and\nUltraLarge. These datasets push the boundaries in both the scale and the\ndiversity of supervised labels for molecular learning. They cover nearly 100\nmillion molecules and over 3000 sparsely defined tasks, totaling more than 13\nbillion individual labels of both quantum and biological nature. In comparison,\nour datasets contain 300 times more data points than the widely used OGB-LSC\nPCQM4Mv2 dataset, and 13 times more than the quantum-only QM1B dataset. In\naddition, to support the development of foundational models based on our\nproposed datasets, we present the Graphium graph machine learning library which\nsimplifies the process of building and training molecular machine learning\nmodels for multi-task and multi-level molecular datasets. Finally, we present a\nrange of baseline results as a starting point of multi-task and multi-level\ntraining on these datasets. Empirically, we observe that performance on\nlow-resource biological datasets show improvement by also training on large\namounts of quantum data. This indicates that there may be potential in\nmulti-task and multi-level training of a foundation model and fine-tuning it to\nresource-constrained downstream tasks.\n", "date": "2023-10-06", "categories": []}
{"id": "2310.11802", "title": "De novo protein design using geometric vector field networks", "abstract": "  Innovations like protein diffusion have enabled significant progress in de\nnovo protein design, which is a vital topic in life science. These methods\ntypically depend on protein structure encoders to model residue backbone\nframes, where atoms do not exist. Most prior encoders rely on atom-wise\nfeatures, such as angles and distances between atoms, which are not available\nin this context. Thus far, only several simple encoders, such as IPA, have been\nproposed for this scenario, exposing the frame modeling as a bottleneck. In\nthis work, we proffer the Vector Field Network (VFN), which enables network\nlayers to perform learnable vector computations between coordinates of\nframe-anchored virtual atoms, thus achieving a higher capability for modeling\nframes. The vector computation operates in a manner similar to a linear layer,\nwith each input channel receiving 3D virtual atom coordinates instead of scalar\nvalues. The multiple feature vectors output by the vector computation are then\nused to update the residue representations and virtual atom coordinates via\nattention aggregation. Remarkably, VFN also excels in modeling both frames and\natoms, as the real atoms can be treated as the virtual atoms for modeling,\npositioning VFN as a potential universal encoder. In protein diffusion (frame\nmodeling), VFN exhibits an impressive performance advantage over IPA, excelling\nin terms of both designability (67.04% vs. 53.58%) and diversity (66.54% vs.\n51.98%). In inverse folding (frame and atom modeling), VFN outperforms the\nprevious SoTA model, PiFold (54.7% vs. 51.66%), on sequence recovery rate. We\nalso propose a method of equipping VFN with the ESM model, which significantly\nsurpasses the previous ESM-based SoTA (62.67% vs. 55.65%), LM-Design, by a\nsubstantial margin.\n", "date": "2023-10-18", "categories": []}
{"id": "2308.13877", "title": "Applications of machine Learning to improve the efficiency and range of\n  microbial biosynthesis: a review of state-of-art techniques", "abstract": "  In the modern world, technology is at its peak. Different avenues in\nprogramming and technology have been explored for data analysis, automation,\nand robotics. Machine learning is key to optimize data analysis, make accurate\npredictions, and hasten/improve existing functions. Thus, presently, the field\nof machine learning in artificial intelligence is being developed and its uses\nin varying fields are being explored. One field in which its uses stand out is\nthat of microbial biosynthesis. In this paper, a comprehensive overview of the\ndiffering machine learning programs used in biosynthesis is provided, alongside\nbrief descriptions of the fields of machine learning and microbial biosynthesis\nseparately. This information includes past trends, modern developments, future\nimprovements, explanations of processes, and current problems they face. Thus,\nthis paper's main contribution is to distill developments in, and provide a\nholistic explanation of, 2 key fields and their applicability to improve\nindustry/research. It also highlights challenges and research directions,\nacting to instigate more research and development in the growing fields.\nFinally, the paper aims to act as a reference for academics performing\nresearch, industry professionals improving their processes, and students\nlooking to understand the concept of machine learning in biosynthesis.\n", "date": "2023-08-26", "categories": []}
{"id": "2310.11082", "title": "Multi-omics Sampling-based Graph Transformer for Synthetic Lethality\n  Prediction", "abstract": "  Synthetic lethality (SL) prediction is used to identify if the co-mutation of\ntwo genes results in cell death. The prevalent strategy is to abstract SL\nprediction as an edge classification task on gene nodes within SL data and\nachieve it through graph neural networks (GNNs). However, GNNs suffer from\nlimitations in their message passing mechanisms, including over-smoothing and\nover-squashing issues. Moreover, harnessing the information of non-SL gene\nrelationships within large-scale multi-omics data to facilitate SL prediction\nposes a non-trivial challenge. To tackle these issues, we propose a new\nmulti-omics sampling-based graph transformer for SL prediction (MSGT-SL).\nConcretely, we introduce a shallow multi-view GNN to acquire local structural\npatterns from both SL and multi-omics data. Further, we input gene features\nthat encode multi-view information into the standard self-attention to capture\nlong-range dependencies. Notably, starting with batch genes from SL data, we\nadopt parallel random walk sampling across multiple omics gene graphs\nencompassing them. Such sampling effectively and modestly incorporates genes\nfrom omics in a structure-aware manner before using self-attention. We showcase\nthe effectiveness of MSGT-SL on real-world SL tasks, demonstrating the\nempirical benefits gained from the graph transformer and multi-omics data.\n", "date": "2023-10-17", "categories": []}
{"id": "2211.04020", "title": "Generating counterfactual explanations of tumor spatial proteomes to\n  discover effective strategies for enhancing immune infiltration", "abstract": "  The tumor microenvironment (TME) significantly impacts cancer prognosis due\nto its immune composition. While therapies for altering the immune composition,\nincluding immunotherapies, have shown exciting results for treating\nhematological cancers, they are less effective for immunologically-cold, solid\ntumors. Spatial omics technologies capture the spatial organization of the TME\nwith unprecedented molecular detail, revealing the relationship between immune\ncell localization and molecular signals. Here, we formulate T-cell infiltration\nprediction as a self-supervised machine learning problem and develop a\ncounterfactual optimization strategy that leverages large scale spatial omics\nprofiles of patient tumors to design tumor perturbations predicted to boost\nT-cell infiltration. A convolutional neural network predicts T-cell\ndistribution based on signaling molecules in the TME provided by imaging mass\ncytometry. Gradient-based counterfactual generation, then, computes\nperturbations predicted to boost T-cell abundance. We apply our framework to\nmelanoma, colorectal cancer liver metastases, and breast tumor data,\ndiscovering combinatorial perturbations predicted to support T-cell\ninfiltration across tens to hundreds of patients. This work presents a paradigm\nfor counterfactual-based prediction and design of cancer therapeutics using\nspatial omics data.\n", "date": "2022-11-08", "categories": []}
{"id": "2310.09685", "title": "Generative artificial intelligence for de novo protein design", "abstract": "  Engineering new molecules with desirable functions and properties has the\npotential to extend our ability to engineer proteins beyond what nature has so\nfar evolved. Advances in the so-called \"de novo\" design problem have recently\nbeen brought forward by developments in artificial intelligence. Generative\narchitectures, such as language models and diffusion processes, seem adept at\ngenerating novel, yet realistic proteins that display desirable properties and\nperform specified functions. State-of-the-art design protocols now achieve\nexperimental success rates nearing 20%, thus widening the access to de novo\ndesigned proteins. Despite extensive progress, there are clear field-wide\nchallenges, for example in determining the best in silico metrics to prioritise\ndesigns for experimental testing, and in designing proteins that can undergo\nlarge conformational changes or be regulated by post-translational\nmodifications and other cellular processes. With an increase in the number of\nmodels being developed, this review provides a framework to understand how\nthese tools fit into the overall process of de novo protein design. Throughout,\nwe highlight the power of incorporating biochemical knowledge to improve\nperformance and interpretability.\n", "date": "2023-10-15", "categories": []}
{"id": "2307.15933", "title": "GeneMask: Fast Pretraining of Gene Sequences to Enable Few-Shot Learning", "abstract": "  Large-scale language models such as DNABert and LOGO aim to learn optimal\ngene representations and are trained on the entire Human Reference Genome.\nHowever, standard tokenization schemes involve a simple sliding window of\ntokens like k-mers that do not leverage any gene-based semantics and thus may\nlead to (trivial) masking of easily predictable sequences and subsequently\ninefficient Masked Language Modeling (MLM) training. Therefore, we propose a\nnovel masking algorithm, GeneMask, for MLM training of gene sequences, where we\nrandomly identify positions in a gene sequence as mask centers and locally\nselect the span around the mask center with the highest Normalized Pointwise\nMutual Information (NPMI) to mask. We observe that in the absence of\nhuman-understandable semantics in the genomics domain (in contrast, semantic\nunits like words and phrases are inherently available in NLP), GeneMask-based\nmodels substantially outperform the SOTA models (DNABert and LOGO) over four\nbenchmark gene sequence classification datasets in five few-shot settings (10\nto 1000-shot). More significantly, the GeneMask-based DNABert model is trained\nfor less than one-tenth of the number of epochs of the original SOTA model. We\nalso observe a strong correlation between top-ranked PMI tokens and conserved\nDNA sequence motifs, which may indicate the incorporation of latent genomic\ninformation. The codes (including trained models) and datasets are made\npublicly available at https://github.com/roysoumya/GeneMask.\n", "date": "2023-07-29", "categories": []}
{"id": "2310.09267", "title": "Genetic algorithms are strong baselines for molecule generation", "abstract": "  Generating molecules, both in a directed and undirected fashion, is a huge\npart of the drug discovery pipeline. Genetic algorithms (GAs) generate\nmolecules by randomly modifying known molecules. In this paper we show that GAs\nare very strong algorithms for such tasks, outperforming many complicated\nmachine learning methods: a result which many researchers may find surprising.\nWe therefore propose insisting during peer review that new algorithms must have\nsome clear advantage over GAs, which we call the GA criterion. Ultimately our\nwork suggests that a lot of research in molecule generation should be\nre-assessed.\n", "date": "2023-10-13", "categories": []}
{"id": "2306.03329", "title": "AVIDa-hIL6: A Large-Scale VHH Dataset Produced from an Immunized Alpaca\n  for Predicting Antigen-Antibody Interactions", "abstract": "  Antibodies have become an important class of therapeutic agents to treat\nhuman diseases. To accelerate therapeutic antibody discovery, computational\nmethods, especially machine learning, have attracted considerable interest for\npredicting specific interactions between antibody candidates and target\nantigens such as viruses and bacteria. However, the publicly available datasets\nin existing works have notable limitations, such as small sizes and the lack of\nnon-binding samples and exact amino acid sequences. To overcome these\nlimitations, we have developed AVIDa-hIL6, a large-scale dataset for predicting\nantigen-antibody interactions in the variable domain of heavy chain of heavy\nchain antibodies (VHHs), produced from an alpaca immunized with the human\ninterleukin-6 (IL-6) protein, as antigens. By leveraging the simple structure\nof VHHs, which facilitates identification of full-length amino acid sequences\nby DNA sequencing technology, AVIDa-hIL6 contains 573,891 antigen-VHH pairs\nwith amino acid sequences. All the antigen-VHH pairs have reliable labels for\nbinding or non-binding, as generated by a novel labeling method. Furthermore,\nvia introduction of artificial mutations, AVIDa-hIL6 contains 30 different\nmutants in addition to wild-type IL-6 protein. This characteristic provides\nopportunities to develop machine learning models for predicting changes in\nantibody binding by antigen mutations. We report experimental benchmark results\non AVIDa-hIL6 by using machine learning models. The results indicate that the\nexisting models have potential, but further research is needed to generalize\nthem to predict effective antibodies against unknown mutants. The dataset is\navailable at https://avida-hil6.cognanous.com.\n", "date": "2023-06-06", "categories": []}
{"id": "2310.04895", "title": "Cell Tracking-by-detection using Elliptical Bounding Boxes", "abstract": "  Cell detection and tracking are paramount for bio-analysis. Recent approaches\nrely on the tracking-by-model evolution paradigm, which usually consists of\ntraining end-to-end deep learning models to detect and track the cells on the\nframes with promising results. However, such methods require extensive amounts\nof annotated data, which is time-consuming to obtain and often requires\nspecialized annotators. This work proposes a new approach based on the\nclassical tracking-by-detection paradigm that alleviates the requirement of\nannotated data. More precisely, it approximates the cell shapes as oriented\nellipses and then uses generic-purpose oriented object detectors to identify\nthe cells in each frame. We then rely on a global data association algorithm\nthat explores temporal cell similarity using probability distance metrics,\nconsidering that the ellipses relate to two-dimensional Gaussian distributions.\nOur results show that our method can achieve detection and tracking results\ncompetitively with state-of-the-art techniques that require considerably more\nextensive data annotation. Our code is available at:\nhttps://github.com/LucasKirsten/Deep-Cell-Tracking-EBB.\n", "date": "2023-10-07", "categories": []}
{"id": "2310.07351", "title": "Atom-Motif Contrastive Transformer for Molecular Property Prediction", "abstract": "  Recently, Graph Transformer (GT) models have been widely used in the task of\nMolecular Property Prediction (MPP) due to their high reliability in\ncharacterizing the latent relationship among graph nodes (i.e., the atoms in a\nmolecule). However, most existing GT-based methods usually explore the basic\ninteractions between pairwise atoms, and thus they fail to consider the\nimportant interactions among critical motifs (e.g., functional groups consisted\nof several atoms) of molecules. As motifs in a molecule are significant\npatterns that are of great importance for determining molecular properties\n(e.g., toxicity and solubility), overlooking motif interactions inevitably\nhinders the effectiveness of MPP. To address this issue, we propose a novel\nAtom-Motif Contrastive Transformer (AMCT), which not only explores the\natom-level interactions but also considers the motif-level interactions. Since\nthe representations of atoms and motifs for a given molecule are actually two\ndifferent views of the same instance, they are naturally aligned to generate\nthe self-supervisory signals for model training. Meanwhile, the same motif can\nexist in different molecules, and hence we also employ the contrastive loss to\nmaximize the representation agreement of identical motifs across different\nmolecules. Finally, in order to clearly identify the motifs that are critical\nin deciding the properties of each molecule, we further construct a\nproperty-aware attention mechanism into our learning framework. Our proposed\nAMCT is extensively evaluated on seven popular benchmark datasets, and both\nquantitative and qualitative results firmly demonstrate its effectiveness when\ncompared with the state-of-the-art methods.\n", "date": "2023-10-11", "categories": []}
{"id": "2310.07387", "title": "Linguistic laws in biology", "abstract": "  Linguistic laws, the common statistical patterns of human language, have been\ninvestigated by quantitative linguists for nearly a century. Recently,\nbiologists from a range of disciplines have started to explore the prevalence\nof these laws beyond language, finding patterns consistent with linguistic laws\nacross multiple levels of biological organisation, from molecular (genomes,\ngenes, and proteins) to organismal (animal behaviour) to ecological\n(populations and ecosystems). We propose a new conceptual framework for the\nstudy of linguistic laws in biology, comprising and integrating distinct levels\nof analysis, from description to prediction to theory building. Adopting this\nframework will provide critical new insights into the fundamental rules of\norganisation underpinning natural systems, unifying linguistic laws and core\ntheory in biology.\n", "date": "2023-10-11", "categories": []}
{"id": "2306.11375", "title": "Top-down machine learning of coarse-grained protein force-fields", "abstract": "  Developing accurate and efficient coarse-grained representations of proteins\nis crucial for understanding their folding, function, and interactions over\nextended timescales. Our methodology involves simulating proteins with\nmolecular dynamics and utilizing the resulting trajectories to train a neural\nnetwork potential through differentiable trajectory reweighting. Remarkably,\nthis method requires only the native conformation of proteins, eliminating the\nneed for labeled data derived from extensive simulations or memory-intensive\nend-to-end differentiable simulations. Once trained, the model can be employed\nto run parallel molecular dynamics simulations and sample folding events for\nproteins both within and beyond the training distribution, showcasing its\nextrapolation capabilities. By applying Markov State Models, native-like\nconformations of the simulated proteins can be predicted from the\ncoarse-grained simulations. Owing to its theoretical transferability and\nability to use solely experimental static structures as training data, we\nanticipate that this approach will prove advantageous for developing new\nprotein force fields and further advancing the study of protein dynamics,\nfolding, and interactions.\n", "date": "2023-06-20", "categories": []}
{"id": "2308.15116", "title": "Mixup-Augmented Meta-Learning for Sample-Efficient Fine-Tuning of\n  Protein Simulators", "abstract": "  Molecular dynamics simulations have emerged as a fundamental instrument for\nstudying biomolecules. At the same time, it is desirable to perform simulations\nof a collection of particles under various conditions in which the molecules\ncan fluctuate. In this paper, we explore and adapt the soft prompt-based\nlearning method to molecular dynamics tasks. Our model can remarkably\ngeneralize to unseen and out-of-distribution scenarios with limited training\ndata. While our work focuses on temperature as a test case, the versatility of\nour approach allows for efficient simulation through any continuous dynamic\nconditions, such as pressure and volumes. Our framework has two stages: 1)\nPre-trains with data mixing technique, augments molecular structure data and\ntemperature prompts, then applies a curriculum learning method by increasing\nthe ratio of them smoothly. 2) Meta-learning-based fine-tuning framework\nimproves sample-efficiency of fine-tuning process and gives the soft\nprompt-tuning better initialization points. Comprehensive experiments reveal\nthat our framework excels in accuracy for in-domain data and demonstrates\nstrong generalization capabilities for unseen and out-of-distribution samples.\n", "date": "2023-08-29", "categories": []}
{"id": "2310.06177", "title": "DockGame: Cooperative Games for Multimeric Rigid Protein Docking", "abstract": "  Protein interactions and assembly formation are fundamental to most\nbiological processes. Predicting the assembly structure from constituent\nproteins -- referred to as the protein docking task -- is thus a crucial step\nin protein design applications. Most traditional and deep learning methods for\ndocking have focused mainly on binary docking, following either a search-based,\nregression-based, or generative modeling paradigm. In this paper, we focus on\nthe less-studied multimeric (i.e., two or more proteins) docking problem. We\nintroduce DockGame, a novel game-theoretic framework for docking -- we view\nprotein docking as a cooperative game between proteins, where the final\nassembly structure(s) constitute stable equilibria w.r.t. the underlying game\npotential. Since we do not have access to the true potential, we consider two\napproaches - i) learning a surrogate game potential guided by physics-based\nenergy functions and computing equilibria by simultaneous gradient updates, and\nii) sampling from the Gibbs distribution of the true potential by learning a\ndiffusion generative model over the action spaces (rotations and translations)\nof all proteins. Empirically, on the Docking Benchmark 5.5 (DB5.5) dataset,\nDockGame has much faster runtimes than traditional docking methods, can\ngenerate multiple plausible assembly structures, and achieves comparable\nperformance to existing binary docking baselines, despite solving the harder\ntask of coordinating multiple protein chains.\n", "date": "2023-10-09", "categories": []}
{"id": "2310.06661", "title": "Tertiary Lymphoid Structures Generation through Graph-based Diffusion", "abstract": "  Graph-based representation approaches have been proven to be successful in\nthe analysis of biomedical data, due to their capability of capturing intricate\ndependencies between biological entities, such as the spatial organization of\ndifferent cell types in a tumor tissue. However, to further enhance our\nunderstanding of the underlying governing biological mechanisms, it is\nimportant to accurately capture the actual distributions of such complex data.\nGraph-based deep generative models are specifically tailored to accomplish\nthat. In this work, we leverage state-of-the-art graph-based diffusion models\nto generate biologically meaningful cell-graphs. In particular, we show that\nthe adopted graph diffusion model is able to accurately learn the distribution\nof cells in terms of their tertiary lymphoid structures (TLS) content, a\nwell-established biomarker for evaluating the cancer progression in oncology\nresearch. Additionally, we further illustrate the utility of the learned\ngenerative models for data augmentation in a TLS classification task. To the\nbest of our knowledge, this is the first work that leverages the power of graph\ndiffusion models in generating meaningful biological cell structures.\n", "date": "2023-10-10", "categories": []}
{"id": "2310.04424", "title": "Stability Analysis of Non-Linear Classifiers using Gene Regulatory\n  Neural Network for Biological AI", "abstract": "  The Gene Regulatory Network (GRN) of biological cells governs a number of key\nfunctionalities that enables them to adapt and survive through different\nenvironmental conditions. Close observation of the GRN shows that the structure\nand operational principles resembles an Artificial Neural Network (ANN), which\ncan pave the way for the development of Biological Artificial Intelligence. In\nparticular, a gene's transcription and translation process resembles a\nsigmoidal-like property based on transcription factor inputs. In this paper, we\ndevelop a mathematical model of gene-perceptron using a dual-layered\ntranscription-translation chemical reaction model, enabling us to transform a\nGRN into a Gene Regulatory Neural Network (GRNN). We perform stability analysis\nfor each gene-perceptron within the fully-connected GRNN sub network to\ndetermine temporal as well as stable concentration outputs that will result in\nreliable computing performance. We focus on a non-linear classifier application\nfor the GRNN, where we analyzed generic multi-layer GRNNs as well as E.Coli\nGRNN that is derived from trans-omic experimental data. Our analysis found that\nvarying the parameters of the chemical reactions can allow us shift the\nboundaries of the classification region, laying the platform for programmable\nGRNNs that suit diverse application requirements.\n", "date": "2023-09-14", "categories": []}
{"id": "2310.03281", "title": "A 5' UTR Language Model for Decoding Untranslated Regions of mRNA and\n  Function Predictions", "abstract": "  The 5' UTR, a regulatory region at the beginning of an mRNA molecule, plays a\ncrucial role in regulating the translation process and impacts the protein\nexpression level. Language models have showcased their effectiveness in\ndecoding the functions of protein and genome sequences. Here, we introduced a\nlanguage model for 5' UTR, which we refer to as the UTR-LM. The UTR-LM is\npre-trained on endogenous 5' UTRs from multiple species and is further\naugmented with supervised information including secondary structure and minimum\nfree energy. We fine-tuned the UTR-LM in a variety of downstream tasks. The\nmodel outperformed the best-known benchmark by up to 42% for predicting the\nMean Ribosome Loading, and by up to 60% for predicting the Translation\nEfficiency and the mRNA Expression Level. The model also applies to identifying\nunannotated Internal Ribosome Entry Sites within the untranslated region and\nimproves the AUPR from 0.37 to 0.52 compared to the best baseline. Further, we\ndesigned a library of 211 novel 5' UTRs with high predicted values of\ntranslation efficiency and evaluated them via a wet-lab assay. Experiment\nresults confirmed that our top designs achieved a 32.5% increase in protein\nproduction level relative to well-established 5' UTR optimized for\ntherapeutics.\n", "date": "2023-10-05", "categories": []}
{"id": "2310.03611", "title": "GENER: A Parallel Layer Deep Learning Network To Detect Gene-Gene\n  Interactions From Gene Expression Data", "abstract": "  Detecting and discovering new gene interactions based on known gene\nexpressions and gene interaction data presents a significant challenge. Various\nstatistical and deep learning methods have attempted to tackle this challenge\nby leveraging the topological structure of gene interactions and gene\nexpression patterns to predict novel gene interactions. In contrast, some\napproaches have focused exclusively on utilizing gene expression profiles. In\nthis context, we introduce GENER, a parallel-layer deep learning network\ndesigned exclusively for the identification of gene-gene relationships using\ngene expression data. We conducted two training experiments and compared the\nperformance of our network with that of existing statistical and deep learning\napproaches. Notably, our model achieved an average AUROC score of 0.834 on the\ncombined BioGRID&DREAM5 dataset, outperforming competing methods in predicting\ngene-gene interactions.\n", "date": "2023-10-05", "categories": []}
{"id": "2310.03899", "title": "CrysFormer: Protein Structure Prediction via 3d Patterson Maps and\n  Partial Structure Attention", "abstract": "  Determining the structure of a protein has been a decades-long open question.\nA protein's three-dimensional structure often poses nontrivial computation\ncosts, when classical simulation algorithms are utilized. Advances in the\ntransformer neural network architecture -- such as AlphaFold2 -- achieve\nsignificant improvements for this problem, by learning from a large dataset of\nsequence information and corresponding protein structures. Yet, such methods\nonly focus on sequence information; other available prior knowledge, such as\nprotein crystallography and partial structure of amino acids, could be\npotentially utilized. To the best of our knowledge, we propose the first\ntransformer-based model that directly utilizes protein crystallography and\npartial structure information to predict the electron density maps of proteins.\nVia two new datasets of peptide fragments (2-residue and 15-residue) , we\ndemonstrate our method, dubbed \\texttt{CrysFormer}, can achieve accurate\npredictions, based on a much smaller dataset size and with reduced computation\ncosts.\n", "date": "2023-10-05", "categories": []}
{"id": "2310.02964", "title": "Co-modeling the Sequential and Graphical Routes for Peptide\n  Representation Learning", "abstract": "  Peptides are formed by the dehydration condensation of multiple amino acids.\nThe primary structure of a peptide can be represented either as an amino acid\nsequence or as a molecular graph consisting of atoms and chemical bonds.\nPrevious studies have indicated that deep learning routes specific to\nsequential and graphical peptide forms exhibit comparable performance on\ndownstream tasks. Despite the fact that these models learn representations of\nthe same modality of peptides, we find that they explain their predictions\ndifferently. Considering sequential and graphical models as two experts making\ninferences from different perspectives, we work on fusing expert knowledge to\nenrich the learned representations for improving the discriminative\nperformance. To achieve this, we propose a peptide co-modeling method, RepCon,\nwhich employs a contrastive learning-based framework to enhance the mutual\ninformation of representations from decoupled sequential and graphical\nend-to-end models. It considers representations from the sequential encoder and\nthe graphical encoder for the same peptide sample as a positive pair and learns\nto enhance the consistency of representations between positive sample pairs and\nto repel representations between negative pairs. Empirical studies of RepCon\nand other co-modeling methods are conducted on open-source discriminative\ndatasets, including aggregation propensity, retention time, antimicrobial\npeptide prediction, and family classification from Peptide Database. Our\nresults demonstrate the superiority of the co-modeling approach over\nindependent modeling, as well as the superiority of RepCon over other methods\nunder the co-modeling framework. In addition, the attribution on RepCon further\ncorroborates the validity of the approach at the level of model explanation.\n", "date": "2023-10-04", "categories": []}
{"id": "2310.03086", "title": "Deep Learning in Computational Biology: Advancements, Challenges, and\n  Future Outlook", "abstract": "  Deep learning has become a powerful tool in computational biology,\nrevolutionising the analysis and interpretation of biological data over time.\nIn our article review, we delve into various aspects of deep learning in\ncomputational biology. Specifically, we examine its history, advantages, and\nchallenges. Our focus is on two primary applications: DNA sequence\nclassification and prediction, as well as protein structure prediction from\nsequence data. Additionally, we provide insights into the outlook for this\nfield. To fully harness the potential of deep learning in computational\nbiology, it is crucial to address the challenges that come with it. These\nchallenges include the requirement for large, labelled datasets and the\ninterpretability of deep learning models. The use of deep learning in the\nanalysis of DNA sequences has brought about a significant transformation in the\ndetection of genomic variants and the analysis of gene expression. This has\ngreatly contributed to the advancement of personalised medicine and drug\ndiscovery. Convolutional neural networks (CNNs) have been shown to be highly\naccurate in predicting genetic variations and gene expression levels. Deep\nlearning techniques are used for analysing epigenetic data, including DNA\nmethylation and histone modifications. This provides valuable insights into\nmetabolic conditions and gene regulation. The field of protein structure\nprediction has been significantly impacted by deep learning, which has enabled\naccurate determination of the three-dimensional shape of proteins and\nprediction of their interactions. The future of deep learning in computational\nbiology looks promising. With the development of advanced deep learning models\nand interpretation techniques, there is potential to overcome current\nchallenges and further our understanding of biological systems.\n", "date": "2023-10-02", "categories": []}
{"id": "2310.03579", "title": "Causal Inference in Gene Regulatory Networks with GFlowNet: Towards\n  Scalability in Large Systems", "abstract": "  Understanding causal relationships within Gene Regulatory Networks (GRNs) is\nessential for unraveling the gene interactions in cellular processes. However,\ncausal discovery in GRNs is a challenging problem for multiple reasons\nincluding the existence of cyclic feedback loops and uncertainty that yields\ndiverse possible causal structures. Previous works in this area either ignore\ncyclic dynamics (assume acyclic structure) or struggle with scalability. We\nintroduce Swift-DynGFN as a novel framework that enhances causal structure\nlearning in GRNs while addressing scalability concerns. Specifically,\nSwift-DynGFN exploits gene-wise independence to boost parallelization and to\nlower computational cost. Experiments on real single-cell RNA velocity and\nsynthetic GRN datasets showcase the advancement in learning causal structure in\nGRNs and scalability in larger systems.\n", "date": "2023-10-05", "categories": []}
{"id": "2310.02546", "title": "Joint Design of Protein Sequence and Structure based on Motifs", "abstract": "  Designing novel proteins with desired functions is crucial in biology and\nchemistry. However, most existing work focus on protein sequence design,\nleaving protein sequence and structure co-design underexplored. In this paper,\nwe propose GeoPro, a method to design protein backbone structure and sequence\njointly. Our motivation is that protein sequence and its backbone structure\nconstrain each other, and thus joint design of both can not only avoid\nnonfolding and misfolding but also produce more diverse candidates with desired\nfunctions. To this end, GeoPro is powered by an equivariant encoder for\nthree-dimensional (3D) backbone structure and a protein sequence decoder guided\nby 3D geometry. Experimental results on two biologically significant\nmetalloprotein datasets, including $\\beta$-lactamases and myoglobins, show that\nour proposed GeoPro outperforms several strong baselines on most metrics.\nRemarkably, our method discovers novel $\\beta$-lactamases and myoglobins which\nare not present in protein data bank (PDB) and UniProt. These proteins exhibit\nstable folding and active site environments reminiscent of those of natural\nproteins, demonstrating their excellent potential to be biologically\nfunctional.\n", "date": "2023-10-04", "categories": []}
{"id": "2310.02713", "title": "scHyena: Foundation Model for Full-Length Single-Cell RNA-Seq Analysis\n  in Brain", "abstract": "  Single-cell RNA sequencing (scRNA-seq) has made significant strides in\nunraveling the intricate cellular diversity within complex tissues. This is\nparticularly critical in the brain, presenting a greater diversity of cell\ntypes than other tissue types, to gain a deeper understanding of brain function\nwithin various cellular contexts. However, analyzing scRNA-seq data remains a\nchallenge due to inherent measurement noise stemming from dropout events and\nthe limited utilization of extensive gene expression information. In this work,\nwe introduce scHyena, a foundation model designed to address these challenges\nand enhance the accuracy of scRNA-seq analysis in the brain. Specifically,\ninspired by the recent Hyena operator, we design a novel Transformer\narchitecture called singe-cell Hyena (scHyena) that is equipped with a linear\nadaptor layer, the positional encoding via gene-embedding, and a\n{bidirectional} Hyena operator. This enables us to process full-length\nscRNA-seq data without losing any information from the raw data. In particular,\nour model learns generalizable features of cells and genes through pre-training\nscHyena using the full length of scRNA-seq data. We demonstrate the superior\nperformance of scHyena compared to other benchmark methods in downstream tasks,\nincluding cell type classification and scRNA-seq imputation.\n", "date": "2023-10-04", "categories": []}
{"id": "2310.01426", "title": "REMEDI: REinforcement learning-driven adaptive MEtabolism modeling of\n  primary sclerosing cholangitis DIsease progression", "abstract": "  Primary sclerosing cholangitis (PSC) is a rare disease wherein altered bile\nacid metabolism contributes to sustained liver injury. This paper introduces\nREMEDI, a framework that captures bile acid dynamics and the body's adaptive\nresponse during PSC progression that can assist in exploring treatments. REMEDI\nmerges a differential equation (DE)-based mechanistic model that describes bile\nacid metabolism with reinforcement learning (RL) to emulate the body's\nadaptations to PSC continuously. An objective of adaptation is to maintain\nhomeostasis by regulating enzymes involved in bile acid metabolism. These\nenzymes correspond to the parameters of the DEs. REMEDI leverages RL to\napproximate adaptations in PSC, treating homeostasis as a reward signal and the\nadjustment of the DE parameters as the corresponding actions. On real-world\ndata, REMEDI generated bile acid dynamics and parameter adjustments consistent\nwith published findings. Also, our results support discussions in the\nliterature that early administration of drugs that suppress bile acid synthesis\nmay be effective in PSC treatment.\n", "date": "2023-10-02", "categories": []}
{"id": "2310.00941", "title": "Improved Variational Bayesian Phylogenetic Inference using Mixtures", "abstract": "  We present VBPI-Mixtures, an algorithm designed to enhance the accuracy of\nphylogenetic posterior distributions, particularly for tree-topology and\nbranch-length approximations. Despite the Variational Bayesian Phylogenetic\nInference (VBPI), a leading-edge black-box variational inference (BBVI)\nframework, achieving remarkable approximations of these distributions, the\nmultimodality of the tree-topology posterior presents a formidable challenge to\nsampling-based learning techniques such as BBVI. Advanced deep learning\nmethodologies such as normalizing flows and graph neural networks have been\nexplored to refine the branch-length posterior approximation, yet efforts to\nameliorate the posterior approximation over tree topologies have been lacking.\nOur novel VBPI-Mixtures algorithm bridges this gap by harnessing the latest\nbreakthroughs in mixture learning within the BBVI domain. As a result,\nVBPI-Mixtures is capable of capturing distributions over tree-topologies that\nVBPI fails to model. We deliver state-of-the-art performance on difficult\ndensity estimation tasks across numerous real phylogenetic datasets.\n", "date": "2023-10-02", "categories": []}
{"id": "2305.11917", "title": "Interpretable neural architecture search and transfer learning for\n  understanding CRISPR/Cas9 off-target enzymatic reactions", "abstract": "  Finely-tuned enzymatic pathways control cellular processes, and their\ndysregulation can lead to disease. Creating predictive and interpretable models\nfor these pathways is challenging because of the complexity of the pathways and\nof the cellular and genomic contexts. Here we introduce Elektrum, a deep\nlearning framework which addresses these challenges with data-driven and\nbiophysically interpretable models for determining the kinetics of biochemical\nsystems. First, it uses in vitro kinetic assays to rapidly hypothesize an\nensemble of high-quality Kinetically Interpretable Neural Networks (KINNs) that\npredict reaction rates. It then employs a novel transfer learning step, where\nthe KINNs are inserted as intermediary layers into deeper convolutional neural\nnetworks, fine-tuning the predictions for reaction-dependent in vivo outcomes.\nElektrum makes effective use of the limited, but clean in vitro data and the\ncomplex, yet plentiful in vivo data that captures cellular context. We apply\nElektrum to predict CRISPR-Cas9 off-target editing probabilities and\ndemonstrate that Elektrum achieves state-of-the-art performance, regularizes\nneural network architectures, and maintains physical interpretability.\n", "date": "2023-05-18", "categories": []}
{"id": "2309.16773", "title": "Neural scaling laws for phenotypic drug discovery", "abstract": "  Recent breakthroughs by deep neural networks (DNNs) in natural language\nprocessing (NLP) and computer vision have been driven by a scale-up of models\nand data rather than the discovery of novel computing paradigms. Here, we\ninvestigate if scale can have a similar impact for models designed to aid small\nmolecule drug discovery. We address this question through a large-scale and\nsystematic analysis of how DNN size, data diet, and learning routines interact\nto impact accuracy on our Phenotypic Chemistry Arena (Pheno-CA) benchmark: a\ndiverse set of drug development tasks posed on image-based high content\nscreening data. Surprisingly, we find that DNNs explicitly supervised to solve\ntasks in the Pheno-CA do not continuously improve as their data and model size\nis scaled-up. To address this issue, we introduce a novel precursor task, the\nInverse Biological Process (IBP), which is designed to resemble the causal\nobjective functions that have proven successful for NLP. We indeed find that\nDNNs first trained with IBP then probed for performance on the Pheno-CA\nsignificantly outperform task-supervised DNNs. More importantly, the\nperformance of these IBP-trained DNNs monotonically improves with data and\nmodel scale. Our findings reveal that the DNN ingredients needed to accurately\nsolve small molecule drug development tasks are already in our hands, and\nproject how much more experimental data is needed to achieve any desired level\nof improvement. We release our Pheno-CA benchmark and code to encourage further\nstudy of neural scaling laws for small molecule drug discovery.\n", "date": "2023-09-28", "categories": []}
{"id": "2309.15132", "title": "Genetic InfoMax: Exploring Mutual Information Maximization in\n  High-Dimensional Imaging Genetics Studies", "abstract": "  Genome-wide association studies (GWAS) are used to identify relationships\nbetween genetic variations and specific traits. When applied to\nhigh-dimensional medical imaging data, a key step is to extract\nlower-dimensional, yet informative representations of the data as traits.\nRepresentation learning for imaging genetics is largely under-explored due to\nthe unique challenges posed by GWAS in comparison to typical visual\nrepresentation learning. In this study, we tackle this problem from the mutual\ninformation (MI) perspective by identifying key limitations of existing\nmethods. We introduce a trans-modal learning framework Genetic InfoMax (GIM),\nincluding a regularized MI estimator and a novel genetics-informed transformer\nto address the specific challenges of GWAS. We evaluate GIM on human brain 3D\nMRI data and establish standardized evaluation protocols to compare it to\nexisting approaches. Our results demonstrate the effectiveness of GIM and a\nsignificantly improved performance on GWAS.\n", "date": "2023-09-26", "categories": []}
{"id": "2309.14404", "title": "pLMFPPred: a novel approach for accurate prediction of functional\n  peptides integrating embedding from pre-trained protein language model and\n  imbalanced learning", "abstract": "  Functional peptides have the potential to treat a variety of diseases. Their\ngood therapeutic efficacy and low toxicity make them ideal therapeutic agents.\nArtificial intelligence-based computational strategies can help quickly\nidentify new functional peptides from collections of protein sequences and\ndiscover their different functions.Using protein language model-based\nembeddings (ESM-2), we developed a tool called pLMFPPred (Protein Language\nModel-based Functional Peptide Predictor) for predicting functional peptides\nand identifying toxic peptides. We also introduced SMOTE-TOMEK data synthesis\nsampling and Shapley value-based feature selection techniques to relieve data\nimbalance issues and reduce computational costs. On a validated independent\ntest set, pLMFPPred achieved accuracy, Area under the curve - Receiver\nOperating Characteristics, and F1-Score values of 0.974, 0.99, and 0.974,\nrespectively. Comparative experiments show that pLMFPPred outperforms current\nmethods for predicting functional peptides.The experimental results suggest\nthat the proposed method (pLMFPPred) can provide better performance in terms of\nAccuracy, Area under the curve - Receiver Operating Characteristics, and\nF1-Score than existing methods. pLMFPPred has achieved good performance in\npredicting functional peptides and represents a new computational method for\npredicting functional peptides.\n", "date": "2023-09-25", "categories": []}
{"id": "2309.14954", "title": "Addressing preferred orientation in single-particle cryo-EM through\n  AI-generated auxiliary particles", "abstract": "  The single-particle cryo-EM field faces the persistent challenge of preferred\norientation, lacking general computational solutions. We introduce cryoPROS, an\nAI-based approach designed to address the above issue. By generating the\nauxiliary particles with a conditional deep generative model, cryoPROS\naddresses the intrinsic bias in orientation estimation for the observed\nparticles. We effectively employed cryoPROS in the cryo-EM single particle\nanalysis of the hemagglutinin trimer, showing the ability to restore the\nnear-atomic resolution structure on non-tilt data. Moreover, the enhanced\nversion named cryoPROS-MP significantly improves the resolution of the membrane\nprotein NaX using the no-tilted data that contains the effects of micelles.\nCompared to the classical approaches, cryoPROS does not need special\nexperimental or image acquisition techniques, providing a purely computational\nyet effective solution for the preferred orientation problem. Finally, we\nconduct extensive experiments that establish the low risk of model bias and the\nhigh robustness of cryoPROS.\n", "date": "2023-09-26", "categories": []}
{"id": "2212.10614", "title": "MolCPT: Molecule Continuous Prompt Tuning to Generalize Molecular\n  Representation Learning", "abstract": "  Molecular representation learning is crucial for the problem of molecular\nproperty prediction, where graph neural networks (GNNs) serve as an effective\nsolution due to their structure modeling capabilities. Since labeled data is\noften scarce and expensive to obtain, it is a great challenge for GNNs to\ngeneralize in the extensive molecular space. Recently, the training paradigm of\n\"pre-train, fine-tune\" has been leveraged to improve the generalization\ncapabilities of GNNs. It uses self-supervised information to pre-train the GNN,\nand then performs fine-tuning to optimize the downstream task with just a few\nlabels. However, pre-training does not always yield statistically significant\nimprovement, especially for self-supervised learning with random structural\nmasking. In fact, the molecular structure is characterized by motif subgraphs,\nwhich are frequently occurring and influence molecular properties. To leverage\nthe task-related motifs, we propose a novel paradigm of \"pre-train, prompt,\nfine-tune\" for molecular representation learning, named molecule continuous\nprompt tuning (MolCPT). MolCPT defines a motif prompting function that uses the\npre-trained model to project the standalone input into an expressive prompt.\nThe prompt effectively augments the molecular graph with meaningful motifs in\nthe continuous representation space; this provides more structural patterns to\naid the downstream classifier in identifying molecular properties. Extensive\nexperiments on several benchmark datasets show that MolCPT efficiently\ngeneralizes pre-trained GNNs for molecular property prediction, with or without\na few fine-tuning steps.\n", "date": "2022-12-20", "categories": []}
{"id": "2303.06945", "title": "CoGANPPIS: A Coevolution-enhanced Global Attention Neural Network for\n  Protein-Protein Interaction Site Prediction", "abstract": "  Protein-protein interactions are of great importance in biochemical\nprocesses. Accurate prediction of protein-protein interaction sites (PPIs) is\ncrucial for our understanding of biological mechanism. Although numerous\napproaches have been developed recently and achieved gratifying results, there\nare still two limitations: (1) Most existing models have excavated a number of\nuseful input features, but failed to take coevolutionary features into account,\nwhich could provide clues for inter-residue relationships; (2) The\nattention-based models only allocate attention weights for neighboring\nresidues, instead of doing it globally, which may limit the model's prediction\nperformance since some residues being far away from the target residues might\nalso matter.\n  We propose a coevolution-enhanced global attention neural network, a\nsequence-based deep learning model for PPIs prediction, called CoGANPPIS.\nSpecifically, CoGANPPIS utilizes three layers in parallel for feature\nextraction: (1) Local-level representation aggregation layer, which aggregates\nthe neighboring residues' features as the local feature representation; (2)\nGlobal-level representation learning layer, which employs a novel\ncoevolution-enhanced global attention mechanism to allocate attention weights\nto all residues on the same protein sequences; (3) Coevolutionary information\nlearning layer, which applies CNN & pooling to coevolutionary information to\nobtain the coevolutionary profile representation. Then, the three outputs are\nconcatenated and passed into several fully connected layers for the final\nprediction. Extensive experiments on two benchmark datasets have been\nconducted, demonstrating that our proposed model achieves the state-of-the-art\nperformance.\n", "date": "2023-03-13", "categories": []}
{"id": "2309.03033", "title": "Deep Learning for Polycystic Kidney Disease: Utilizing Neural Networks\n  for Accurate and Early Detection through Gene Expression Analysis", "abstract": "  With Polycystic Kidney Disease (PKD) potentially leading to fatal\ncomplications in patients due to the formation of cysts in kidneys, early\ndetection of PKD is crucial for effective management of the condition. However,\nthe various patient-specific factors that play a role in the diagnosis make it\nan intricate puzzle for clinicians to solve, leading to possible kidney\nfailure. Therefore, in this study we aim to utilize a deep learning-based\napproach for early disease detection through gene expression analysis. The\ndevised neural network is able to achieve accurate and robust prediction\nresults for possible PKD in kidneys, thereby improving patient outcomes.\nFurthermore, by conducting a gene ontology analysis, we were able to predict\nthe top gene processes and functions that PKD may affect.\n", "date": "2023-09-06", "categories": []}
{"id": "2309.13129", "title": "AntiBARTy Diffusion for Property Guided Antibody Design", "abstract": "  Over the past decade, antibodies have steadily grown in therapeutic\nimportance thanks to their high specificity and low risk of adverse effects\ncompared to other drug modalities. While traditional antibody discovery is\nprimarily wet lab driven, the rapid improvement of ML-based generative modeling\nhas made in-silico approaches an increasingly viable route for discovery and\nengineering. To this end, we train an antibody-specific language model,\nAntiBARTy, based on BART (Bidirectional and Auto-Regressive Transformer) and\nuse its latent space to train a property-conditional diffusion model for guided\nIgG de novo design. As a test case, we show that we can effectively generate\nnovel antibodies with improved in-silico solubility while maintaining antibody\nvalidity and controlling sequence diversity.\n", "date": "2023-09-22", "categories": []}
{"id": "2302.00600", "title": "Two for One: Diffusion Models and Force Fields for Coarse-Grained\n  Molecular Dynamics", "abstract": "  Coarse-grained (CG) molecular dynamics enables the study of biological\nprocesses at temporal and spatial scales that would be intractable at an\natomistic resolution. However, accurately learning a CG force field remains a\nchallenge. In this work, we leverage connections between score-based generative\nmodels, force fields and molecular dynamics to learn a CG force field without\nrequiring any force inputs during training. Specifically, we train a diffusion\ngenerative model on protein structures from molecular dynamics simulations, and\nwe show that its score function approximates a force field that can directly be\nused to simulate CG molecular dynamics. While having a vastly simplified\ntraining setup compared to previous work, we demonstrate that our approach\nleads to improved performance across several small- to medium-sized protein\nsimulations, reproducing the CG equilibrium distribution, and preserving\ndynamics of all-atom simulations such as protein folding events.\n", "date": "2023-02-01", "categories": []}
{"id": "2309.12058", "title": "An Efficient Consolidation of Word Embedding and Deep Learning\n  Techniques for Classifying Anticancer Peptides: FastText+BiLSTM", "abstract": "  Anticancer peptides (ACPs) are a group of peptides that exhibite\nantineoplastic properties. The utilization of ACPs in cancer prevention can\npresent a viable substitute for conventional cancer therapeutics, as they\npossess a higher degree of selectivity and safety. Recent scientific\nadvancements generate an interest in peptide-based therapies which offer the\nadvantage of efficiently treating intended cells without negatively impacting\nnormal cells. However, as the number of peptide sequences continues to increase\nrapidly, developing a reliable and precise prediction model becomes a\nchallenging task. In this work, our motivation is to advance an efficient model\nfor categorizing anticancer peptides employing the consolidation of word\nembedding and deep learning models. First, Word2Vec and FastText are evaluated\nas word embedding techniques for the purpose of extracting peptide sequences.\nThen, the output of word embedding models are fed into deep learning approaches\nCNN, LSTM, BiLSTM. To demonstrate the contribution of proposed framework,\nextensive experiments are carried on widely-used datasets in the literature,\nACPs250 and Independent. Experiment results show the usage of proposed model\nenhances classification accuracy when compared to the state-of-the-art studies.\nThe proposed combination, FastText+BiLSTM, exhibits 92.50% of accuracy for\nACPs250 dataset, and 96.15% of accuracy for Independent dataset, thence\ndetermining new state-of-the-art.\n", "date": "2023-09-21", "categories": []}
{"id": "2309.10008", "title": "DeepHEN: quantitative prediction essential lncRNA genes and rethinking\n  essentialities of lncRNA genes", "abstract": "  Gene essentiality refers to the degree to which a gene is necessary for the\nsurvival and reproductive efficacy of a living organism. Although the\nessentiality of non-coding genes has been documented, there are still aspects\nof non-coding genes' essentiality that are unknown to us. For example, We do\nnot know the contribution of sequence features and network spatial features to\nessentiality. As a consequence, in this work, we propose DeepHEN that could\nanswer the above question. By buidling a new lncRNA-proteion-protein network\nand utilizing both representation learning and graph neural network, we\nsuccessfully build our DeepHEN models that could predict the essentiality of\nlncRNA genes. Compared to other methods for predicting the essentiality of\nlncRNA genes, our DeepHEN model not only tells whether sequence features or\nnetwork spatial features have a greater influence on essentiality but also\naddresses the overfitting issue of those methods caused by the low number of\nessential lncRNA genes, as evidenced by the results of enrichment analysis.\n", "date": "2023-09-18", "categories": []}
{"id": "2303.09987", "title": "Breast Cancer Histopathology Image based Gene Expression Prediction\n  using Spatial Transcriptomics data and Deep Learning", "abstract": "  Tumour heterogeneity in breast cancer poses challenges in predicting outcome\nand response to therapy. Spatial transcriptomics technologies may address these\nchallenges, as they provide a wealth of information about gene expression at\nthe cell level, but they are expensive, hindering their use in large-scale\nclinical oncology studies. Predicting gene expression from hematoxylin and\neosin stained histology images provides a more affordable alternative for such\nstudies. Here we present BrST-Net, a deep learning framework for predicting\ngene expression from histopathology images using spatial transcriptomics data.\nUsing this framework, we trained and evaluated 10 state-of-the-art deep\nlearning models without utilizing pretrained weights for the prediction of 250\ngenes. To enhance the generalisation performance of the main network, we\nintroduce an auxiliary network into the framework. Our methodology outperforms\nprevious studies, with 237 genes identified with positive correlation,\nincluding 24 genes with a median correlation coefficient greater than 0.50.\nThis is a notable improvement over previous studies, which could predict only\n102 genes with positive correlation, with the highest correlation values\nranging from 0.29 to 0.34.\n", "date": "2023-03-17", "categories": []}
{"id": "2309.09191", "title": "End-to-End Optimized Pipeline for Prediction of Protein Folding Kinetics", "abstract": "  Protein folding is the intricate process by which a linear sequence of amino\nacids self-assembles into a unique three-dimensional structure. Protein folding\nkinetics is the study of pathways and time-dependent mechanisms a protein\nundergoes when it folds. Understanding protein kinetics is essential as a\nprotein needs to fold correctly for it to perform its biological functions\noptimally, and a misfolded protein can sometimes be contorted into shapes that\nare not ideal for a cellular environment giving rise to many degenerative,\nneuro-degenerative disorders and amyloid diseases. Monitoring at-risk\nindividuals and detecting protein discrepancies in a protein's folding kinetics\nat the early stages could majorly result in public health benefits, as\npreventive measures can be taken. This research proposes an efficient pipeline\nfor predicting protein folding kinetics with high accuracy and low memory\nfootprint. The deployed machine learning (ML) model outperformed the\nstate-of-the-art ML models by 4.8% in terms of accuracy while consuming 327x\nlesser memory and being 7.3% faster.\n", "date": "2023-09-17", "categories": []}
{"id": "2308.13066", "title": "Objective-Agnostic Enhancement of Molecule Properties via Multi-Stage\n  VAE", "abstract": "  Variational autoencoder (VAE) is a popular method for drug discovery and\nvarious architectures and pipelines have been proposed to improve its\nperformance. However, VAE approaches are known to suffer from poor manifold\nrecovery when the data lie on a low-dimensional manifold embedded in a higher\ndimensional ambient space [Dai and Wipf, 2019]. The consequences of it in drug\ndiscovery are somewhat under-explored. In this paper, we explore applying a\nmulti-stage VAE approach, that can improve manifold recovery on a synthetic\ndataset, to the field of drug discovery. We experimentally evaluate our\nmulti-stage VAE approach using the ChEMBL dataset and demonstrate its ability\nto improve the property statistics of generated molecules substantially from\npre-existing methods without incorporating property predictors into the\ntraining pipeline. We further fine-tune our models on two curated and much\nsmaller molecule datasets that target different proteins. Our experiments show\nan increase in the number of active molecules generated by the multi-stage VAE\nin comparison to their one-stage equivalent. For each of the two tasks, our\nbaselines include methods that use learned property predictors to incorporate\ntarget metrics directly into the training objective and we discuss\ncomplications that arise with this methodology.\n", "date": "2023-08-24", "categories": []}
{"id": "2309.03242", "title": "Automated Bioinformatics Analysis via AutoBA", "abstract": "  With the fast-growing and evolving omics data, the demand for streamlined and\nadaptable tools to handle the analysis continues to grow. In response to this\nneed, we introduce Auto Bioinformatics Analysis (AutoBA), an autonomous AI\nagent based on a large language model designed explicitly for conventional\nomics data analysis. AutoBA simplifies the analytical process by requiring\nminimal user input while delivering detailed step-by-step plans for various\nbioinformatics tasks. Through rigorous validation by expert bioinformaticians,\nAutoBA's robustness and adaptability are affirmed across a diverse range of\nomics analysis cases, including whole genome sequencing (WGS), RNA sequencing\n(RNA-seq), single-cell RNA-seq, ChIP-seq, and spatial transcriptomics. AutoBA's\nunique capacity to self-design analysis processes based on input data\nvariations further underscores its versatility. Compared with online\nbioinformatic services, AutoBA deploys the analysis locally, preserving data\nprivacy. Moreover, different from the predefined pipeline, AutoBA has\nadaptability in sync with emerging bioinformatics tools. Overall, AutoBA\nrepresents a convenient tool, offering robustness and adaptability for complex\nomics data analysis.\n", "date": "2023-09-06", "categories": []}
{"id": "2309.03099", "title": "PeptideBERT: A Language Model based on Transformers for Peptide Property\n  Prediction", "abstract": "  Recent advances in Language Models have enabled the protein modeling\ncommunity with a powerful tool since protein sequences can be represented as\ntext. Specifically, by taking advantage of Transformers, sequence-to-property\nprediction will be amenable without the need for explicit structural data. In\nthis work, inspired by recent progress in Large Language Models (LLMs), we\nintroduce PeptideBERT, a protein language model for predicting three key\nproperties of peptides (hemolysis, solubility, and non-fouling). The\nPeptideBert utilizes the ProtBERT pretrained transformer model with 12\nattention heads and 12 hidden layers. We then finetuned the pretrained model\nfor the three downstream tasks. Our model has achieved state of the art (SOTA)\nfor predicting Hemolysis, which is a task for determining peptide's potential\nto induce red blood cell lysis. Our PeptideBert non-fouling model also achieved\nremarkable accuracy in predicting peptide's capacity to resist non-specific\ninteractions. This model, trained predominantly on shorter sequences, benefits\nfrom the dataset where negative examples are largely associated with insoluble\npeptides. Codes, models, and data used in this study are freely available at:\nhttps://github.com/ChakradharG/PeptideBERT\n", "date": "2023-08-28", "categories": []}
{"id": "2209.13492", "title": "Unraveling Key Elements Underlying Molecular Property Prediction: A\n  Systematic Study", "abstract": "  Artificial intelligence (AI) has been widely applied in drug discovery with a\nmajor task as molecular property prediction. Despite booming techniques in\nmolecular representation learning, key elements underlying molecular property\nprediction remain largely unexplored, which impedes further advancements in\nthis field. Herein, we conduct an extensive evaluation of representative models\nusing various representations on the MoleculeNet datasets, a suite of\nopioids-related datasets and two additional activity datasets from the\nliterature. To investigate the predictive power in low-data and high-data\nspace, a series of descriptors datasets of varying sizes are also assembled to\nevaluate the models. In total, we have trained 62,820 models, including 50,220\nmodels on fixed representations, 4,200 models on SMILES sequences and 8,400\nmodels on molecular graphs. Based on extensive experimentation and rigorous\ncomparison, we show that representation learning models exhibit limited\nperformance in molecular property prediction in most datasets. Besides,\nmultiple key elements underlying molecular property prediction can affect the\nevaluation results. Furthermore, we show that activity cliffs can significantly\nimpact model prediction. Finally, we explore into potential causes why\nrepresentation learning models can fail and show that dataset size is essential\nfor representation learning models to excel.\n", "date": "2022-09-26", "categories": []}
{"id": "2309.01122", "title": "AI driven B-cell Immunotherapy Design", "abstract": "  Antibodies, a prominent class of approved biologics, play a crucial role in\ndetecting foreign antigens. The effectiveness of antigen neutralisation and\nelimination hinges upon the strength, sensitivity, and specificity of the\nparatope-epitope interaction, which demands resource-intensive experimental\ntechniques for characterisation. In recent years, artificial intelligence and\nmachine learning methods have made significant strides, revolutionising the\nprediction of protein structures and their complexes. The past decade has also\nwitnessed the evolution of computational approaches aiming to support\nimmunotherapy design. This review focuses on the progress of machine\nlearning-based tools and their frameworks in the domain of B-cell immunotherapy\ndesign, encompassing linear and conformational epitope prediction, paratope\nprediction, and antibody design. We mapped the most commonly used data sources,\nevaluation metrics, and method availability and thoroughly assessed their\nsignificance and limitations, discussing the main challenges ahead.\n", "date": "2023-09-03", "categories": []}
{"id": "2309.01670", "title": "Blind Biological Sequence Denoising with Self-Supervised Set Learning", "abstract": "  Biological sequence analysis relies on the ability to denoise the imprecise\noutput of sequencing platforms. We consider a common setting where a short\nsequence is read out repeatedly using a high-throughput long-read platform to\ngenerate multiple subreads, or noisy observations of the same sequence.\nDenoising these subreads with alignment-based approaches often fails when too\nfew subreads are available or error rates are too high. In this paper, we\npropose a novel method for blindly denoising sets of sequences without directly\nobserving clean source sequence labels. Our method, Self-Supervised Set\nLearning (SSSL), gathers subreads together in an embedding space and estimates\na single set embedding as the midpoint of the subreads in both the latent and\nsequence spaces. This set embedding represents the \"average\" of the subreads\nand can be decoded into a prediction of the clean sequence. In experiments on\nsimulated long-read DNA data, SSSL methods denoise small reads of $\\leq 6$\nsubreads with 17% fewer errors and large reads of $>6$ subreads with 8% fewer\nerrors compared to the best baseline. On a real dataset of antibody sequences,\nSSSL improves over baselines on two self-supervised metrics, with a significant\nimprovement on difficult small reads that comprise over 60% of the test set. By\naccurately denoising these reads, SSSL promises to better realize the potential\nof high-throughput DNA sequencing data for downstream scientific applications.\n", "date": "2023-09-04", "categories": []}
{"id": "2308.12740", "title": "Human Comprehensible Active Learning of Genome-Scale Metabolic Networks", "abstract": "  An important application of Synthetic Biology is the engineering of the host\ncell system to yield useful products. However, an increase in the scale of the\nhost system leads to huge design space and requires a large number of\nvalidation trials with high experimental costs. A comprehensible machine\nlearning approach that efficiently explores the hypothesis space and guides\nexperimental design is urgently needed for the Design-Build-Test-Learn (DBTL)\ncycle of the host cell system. We introduce a novel machine learning framework\nILP-iML1515 based on Inductive Logic Programming (ILP) that performs abductive\nlogical reasoning and actively learns from training examples. In contrast to\nnumerical models, ILP-iML1515 is built on comprehensible logical\nrepresentations of a genome-scale metabolic model and can update the model by\nlearning new logical structures from auxotrophic mutant trials. The ILP-iML1515\nframework 1) allows high-throughput simulations and 2) actively selects\nexperiments that reduce the experimental cost of learning gene functions in\ncomparison to randomly selected experiments.\n", "date": "2023-08-24", "categories": []}
{"id": "2309.00408", "title": "Boosting AND/OR-Based Computational Protein Design: Dynamic Heuristics\n  and Generalizable UFO", "abstract": "  Scientific computing has experienced a surge empowered by advancements in\ntechnologies such as neural networks. However, certain important tasks are less\namenable to these technologies, benefiting from innovations to traditional\ninference schemes. One such task is protein re-design. Recently a new re-design\nalgorithm, AOBB-K*, was introduced and was competitive with state-of-the-art\nBBK* on small protein re-design problems. However, AOBB-K* did not scale well.\nIn this work we focus on scaling up AOBB-K* and introduce three new versions:\nAOBB-K*-b (boosted), AOBB-K*-DH (with dynamic heuristics), and AOBB-K*-UFO\n(with underflow optimization) that significantly enhance scalability.\n", "date": "2023-08-31", "categories": []}
{"id": "2309.00483", "title": "Geometry-aware Line Graph Transformer Pre-training for Molecular\n  Property Prediction", "abstract": "  Molecular property prediction with deep learning has gained much attention\nover the past years. Owing to the scarcity of labeled molecules, there has been\ngrowing interest in self-supervised learning methods that learn generalizable\nmolecular representations from unlabeled data. Molecules are typically treated\nas 2D topological graphs in modeling, but it has been discovered that their 3D\ngeometry is of great importance in determining molecular functionalities. In\nthis paper, we propose the Geometry-aware line graph transformer (Galformer)\npre-training, a novel self-supervised learning framework that aims to enhance\nmolecular representation learning with 2D and 3D modalities. Specifically, we\nfirst design a dual-modality line graph transformer backbone to encode the\ntopological and geometric information of a molecule. The designed backbone\nincorporates effective structural encodings to capture graph structures from\nboth modalities. Then we devise two complementary pre-training tasks at the\ninter and intra-modality levels. These tasks provide properly supervised\ninformation and extract discriminative 2D and 3D knowledge from unlabeled\nmolecules. Finally, we evaluate Galformer against six state-of-the-art\nbaselines on twelve property prediction benchmarks via downstream fine-tuning.\nExperimental results show that Galformer consistently outperforms all baselines\non both classification and regression tasks, demonstrating its effectiveness.\n", "date": "2023-09-01", "categories": []}
{"id": "2307.05628", "title": "DNAGPT: A Generalized Pre-trained Tool for Versatile DNA Sequence\n  Analysis Tasks", "abstract": "  Pre-trained large language models demonstrate potential in extracting\ninformation from DNA sequences, yet adapting to a variety of tasks and data\nmodalities remains a challenge. To address this, we propose DNAGPT, a\ngeneralized DNA pre-training model trained on over 200 billion base pairs from\nall mammals. By enhancing the classic GPT model with a binary classification\ntask (DNA sequence order), a numerical regression task (guanine-cytosine\ncontent prediction), and a comprehensive token language, DNAGPT can handle\nversatile DNA analysis tasks while processing both sequence and numerical data.\nOur evaluation of genomic signal and region recognition, mRNA abundance\nregression, and artificial genomes generation tasks demonstrates DNAGPT's\nsuperior performance compared to existing models designed for specific\ndownstream tasks, benefiting from pre-training using the newly designed model\nstructure.\n", "date": "2023-07-11", "categories": []}
{"id": "2308.14769", "title": "XVir: A Transformer-Based Architecture for Identifying Viral Reads from\n  Cancer Samples", "abstract": "  It is estimated that approximately 15% of cancers worldwide can be linked to\nviral infections. The viruses that can cause or increase the risk of cancer\ninclude human papillomavirus, hepatitis B and C viruses, Epstein-Barr virus,\nand human immunodeficiency virus, to name a few. The computational analysis of\nthe massive amounts of tumor DNA data, whose collection is enabled by the\nrecent advancements in sequencing technologies, have allowed studies of the\npotential association between cancers and viral pathogens. However, the high\ndiversity of oncoviral families makes reliable detection of viral DNA difficult\nand thus, renders such analysis challenging. In this paper, we introduce XVir,\na data pipeline that relies on a transformer-based deep learning architecture\nto reliably identify viral DNA present in human tumors. In particular, XVir is\ntrained on genomic sequencing reads from viral and human genomes and may be\nused with tumor sequence information to find evidence of viral DNA in human\ncancers. Results on semi-experimental data demonstrate that XVir is capable of\nachieving high detection accuracy, generally outperforming state-of-the-art\ncompeting methods while being more compact and less computationally demanding.\n", "date": "2023-08-28", "categories": []}
{"id": "2207.08162", "title": "Natural language processing for clusterization of genes according to\n  their functions", "abstract": "  There are hundreds of methods for analysis of data obtained in\nmRNA-sequencing. The most of them are focused on small number of genes. In this\nstudy, we propose an approach that reduces the analysis of several thousand\ngenes to analysis of several clusters. The list of genes is enriched with\ninformation from open databases. Then, the descriptions are encoded as vectors\nusing the pretrained language model (BERT) and some text processing approaches.\nThe encoded gene function pass through the dimensionality reduction and\nclusterization. Aiming to find the most efficient pipeline, 180 cases of\npipeline with different methods in the major pipeline steps were analyzed. The\nperformance was evaluated with clusterization indexes and expert review of the\nresults.\n", "date": "2022-07-17", "categories": []}
{"id": "2308.11671", "title": "Generalising sequence models for epigenome predictions with tissue and\n  assay embeddings", "abstract": "  Sequence modelling approaches for epigenetic profile prediction have recently\nexpanded in terms of sequence length, model size, and profile diversity.\nHowever, current models cannot infer on many experimentally feasible tissue and\nassay pairs due to poor usage of contextual information, limiting $\\textit{in\nsilico}$ understanding of regulatory genomics. We demonstrate that strong\ncorrelation can be achieved across a large range of experimental conditions by\nintegrating tissue and assay embeddings into a Contextualised Genomic Network\n(CGN). In contrast to previous approaches, we enhance long-range sequence\nembeddings with contextual information in the input space, rather than\nexpanding the output space. We exhibit the efficacy of our approach across a\nbroad set of epigenetic profiles and provide the first insights into the effect\nof genetic variants on epigenetic sequence model training. Our general approach\nto context integration exceeds state of the art in multiple settings while\nemploying a more rigorous validation procedure.\n", "date": "2023-08-22", "categories": []}
{"id": "2308.10917", "title": "PACS: Prediction and analysis of cancer subtypes from multi-omics data\n  based on a multi-head attention mechanism model", "abstract": "  Due to the high heterogeneity and clinical characteristics of cancer, there\nare significant differences in multi-omic data and clinical characteristics\namong different cancer subtypes. Therefore, accurate classification of cancer\nsubtypes can help doctors choose the most appropriate treatment options,\nimprove treatment outcomes, and provide more accurate patient survival\npredictions. In this study, we propose a supervised multi-head attention\nmechanism model (SMA) to classify cancer subtypes successfully. The attention\nmechanism and feature sharing module of the SMA model can successfully learn\nthe global and local feature information of multi-omics data. Second, it\nenriches the parameters of the model by deeply fusing multi-head attention\nencoders from Siamese through the fusion module. Validated by extensive\nexperiments, the SMA model achieves the highest accuracy, F1 macroscopic, F1\nweighted, and accurate classification of cancer subtypes in simulated,\nsingle-cell, and cancer multiomics datasets compared to AE, CNN, and GNN-based\nmodels. Therefore, we contribute to future research on multiomics data using\nour attention-based approach.\n", "date": "2023-08-21", "categories": []}
{"id": "2308.10921", "title": "Artificial intelligence-driven antimicrobial peptide discovery", "abstract": "  Antimicrobial peptides (AMPs) emerge as promising agents against\nantimicrobial resistance, providing an alternative to conventional antibiotics.\nArtificial intelligence (AI) revolutionized AMP discovery through both\ndiscrimination and generation approaches. The discriminators aid the\nidentification of promising candidates by predicting key peptide properties\nsuch as activity and toxicity, while the generators learn the distribution over\npeptides and enable sampling novel AMP candidates, either de novo, or as\nanalogues of a prototype peptide. Moreover, the controlled generation of AMPs\nwith desired properties is achieved by discriminator-guided filtering,\npositive-only learning, latent space sampling, as well as conditional and\noptimized generation. Here we review recent achievements in AI-driven AMP\ndiscovery, highlighting the most exciting directions.\n", "date": "2023-08-21", "categories": []}
{"id": "2308.11162", "title": "A Preliminary Investigation into Search and Matching for Tumour\n  Discrimination in WHO Breast Taxonomy Using Deep Networks", "abstract": "  Breast cancer is one of the most common cancers affecting women worldwide.\nThey include a group of malignant neoplasms with a variety of biological,\nclinical, and histopathological characteristics. There are more than 35\ndifferent histological forms of breast lesions that can be classified and\ndiagnosed histologically according to cell morphology, growth, and architecture\npatterns. Recently, deep learning, in the field of artificial intelligence, has\ndrawn a lot of attention for the computerized representation of medical images.\nSearchable digital atlases can provide pathologists with patch matching tools\nallowing them to search among evidently diagnosed and treated archival cases, a\ntechnology that may be regarded as computational second opinion. In this study,\nwe indexed and analyzed the WHO breast taxonomy (Classification of Tumours 5th\nEd.) spanning 35 tumour types. We visualized all tumour types using deep\nfeatures extracted from a state-of-the-art deep learning model, pre-trained on\nmillions of diagnostic histopathology images from the TCGA repository.\nFurthermore, we test the concept of a digital \"atlas\" as a reference for search\nand matching with rare test cases. The patch similarity search within the WHO\nbreast taxonomy data reached over 88% accuracy when validating through\n\"majority vote\" and more than 91% accuracy when validating using top-n tumour\ntypes. These results show for the first time that complex relationships among\ncommon and rare breast lesions can be investigated using an indexed digital\narchive.\n", "date": "2023-08-22", "categories": []}
{"id": "2212.09450", "title": "Accelerating Antimicrobial Peptide Discovery with Latent Structure", "abstract": "  Antimicrobial peptides (AMPs) are promising therapeutic approaches against\ndrug-resistant pathogens. Recently, deep generative models are used to discover\nnew AMPs. However, previous studies mainly focus on peptide sequence attributes\nand do not consider crucial structure information. In this paper, we propose a\nlatent sequence-structure model for designing AMPs (LSSAMP). LSSAMP exploits\nmulti-scale vector quantization in the latent space to represent secondary\nstructures (e.g. alpha helix and beta sheet). By sampling in the latent space,\nLSSAMP can simultaneously generate peptides with ideal sequence attributes and\nsecondary structures. Experimental results show that the peptides generated by\nLSSAMP have a high probability of antimicrobial activity. Our wet laboratory\nexperiments verified that two of the 21 candidates exhibit strong antimicrobial\nactivity. The code is released at https://github.com/dqwang122/LSSAMP.\n", "date": "2022-11-28", "categories": []}
{"id": "2308.09086", "title": "Embracing assay heterogeneity with neural processes for markedly\n  improved bioactivity predictions", "abstract": "  Predicting the bioactivity of a ligand is one of the hardest and most\nimportant challenges in computer-aided drug discovery. Despite years of data\ncollection and curation efforts by research organizations worldwide,\nbioactivity data remains sparse and heterogeneous, thus hampering efforts to\nbuild predictive models that are accurate, transferable and robust. The\nintrinsic variability of the experimental data is further compounded by data\naggregation practices that neglect heterogeneity to overcome sparsity. Here we\ndiscuss the limitations of these practices and present a hierarchical\nmeta-learning framework that exploits the information synergy across disparate\nassays by successfully accounting for assay heterogeneity. We show that the\nmodel achieves a drastic improvement in affinity prediction across diverse\nprotein targets and assay types compared to conventional baselines. It can\nquickly adapt to new target contexts using very few observations, thus enabling\nlarge-scale virtual screening in early-phase drug discovery.\n", "date": "2023-08-17", "categories": []}
{"id": "2308.09482", "title": "Atom-by-atom protein generation and beyond with language models", "abstract": "  Protein language models learn powerful representations directly from\nsequences of amino acids. However, they are constrained to generate proteins\nwith only the set of amino acids represented in their vocabulary. In contrast,\nchemical language models learn atom-level representations of smaller molecules\nthat include every atom, bond, and ring. In this work, we show that chemical\nlanguage models can learn atom-level representations of proteins enabling\nprotein generation unconstrained to the standard genetic code and far beyond\nit. In doing so, we show that language models can generate entire proteins atom\nby atom -- effectively learning the multiple hierarchical layers of molecular\ninformation that define proteins from their primary sequence to their\nsecondary, and tertiary structure. We demonstrate language models are able to\nexplore beyond protein space -- generating proteins with modified sidechains\nthat form unnatural amino acids. Even further, we find that language models can\nexplore chemical space and protein space simultaneously and generate novel\nexamples of protein-drug conjugates. The results demonstrate the potential for\nbiomolecular design at the atom level using language models.\n", "date": "2023-08-16", "categories": []}
{"id": "2308.06292", "title": "The divergence time of protein structures modelled by Markov matrices\n  and its relation to the divergence of sequences", "abstract": "  A complete time-parameterized statistical model quantifying the divergent\nevolution of protein structures in terms of the patterns of conservation of\ntheir secondary structures is inferred from a large collection of protein 3D\nstructure alignments. This provides a better alternative to time-parameterized\nsequence-based models of protein relatedness, that have clear limitations\ndealing with twilight and midnight zones of sequence relationships. Since\nprotein structures are far more conserved due to the selection pressure\ndirectly placed on their function, divergence time estimates can be more\naccurate when inferred from structures. We use the Bayesian and\ninformation-theoretic framework of Minimum Message Length to infer a\ntime-parameterized stochastic matrix (accounting for perturbed structural\nstates of related residues) and associated Dirichlet models (accounting for\ninsertions and deletions during the evolution of protein domains). These are\nused in concert to estimate the Markov time of divergence of tertiary\nstructures, a task previously only possible using proxies (like RMSD). By\nanalyzing one million pairs of homologous structures, we yield a relationship\nbetween the Markov divergence time of structures and of sequences. Using these\ninferred models and the relationship between the divergence of sequences and\nstructures, we demonstrate a competitive performance in secondary structure\nprediction against neural network architectures commonly employed for this\ntask. The source code and supplementary information are downloadable from\n\\url{http://lcb.infotech.monash.edu.au/sstsum}.\n", "date": "2023-08-11", "categories": []}
{"id": "2308.06429", "title": "Genetic heterogeneity analysis using genetic algorithm and network\n  science", "abstract": "  Through genome-wide association studies (GWAS), disease susceptible genetic\nvariables can be identified by comparing the genetic data of individuals with\nand without a specific disease. However, the discovery of these associations\nposes a significant challenge due to genetic heterogeneity and feature\ninteractions. Genetic variables intertwined with these effects often exhibit\nlower effect-size, and thus can be difficult to be detected using machine\nlearning feature selection methods. To address these challenges, this paper\nintroduces a novel feature selection mechanism for GWAS, named Feature\nCo-selection Network (FCSNet). FCS-Net is designed to extract heterogeneous\nsubsets of genetic variables from a network constructed from multiple\nindependent feature selection runs based on a genetic algorithm (GA), an\nevolutionary learning algorithm. We employ a non-linear machine learning\nalgorithm to detect feature interaction. We introduce the Community Risk Score\n(CRS), a synthetic feature designed to quantify the collective disease\nassociation of each variable subset. Our experiment showcases the effectiveness\nof the utilized GA-based feature selection method in identifying feature\ninteractions through synthetic data analysis. Furthermore, we apply our novel\napproach to a case-control colorectal cancer GWAS dataset. The resulting\nsynthetic features are then used to explain the genetic heterogeneity in an\nadditional case-only GWAS dataset.\n", "date": "2023-08-12", "categories": []}
{"id": "2308.05118", "title": "Vector Embeddings by Sequence Similarity and Context for Improved\n  Compression, Similarity Search, Clustering, Organization, and Manipulation of\n  cDNA Libraries", "abstract": "  This paper demonstrates the utility of organized numerical representations of\ngenes in research involving flat string gene formats (i.e., FASTA/FASTQ5).\nFASTA/FASTQ files have several current limitations, such as their large file\nsizes, slow processing speeds for mapping and alignment, and contextual\ndependencies. These challenges significantly hinder investigations and tasks\nthat involve finding similar sequences. The solution lies in transforming\nsequences into an alternative representation that facilitates easier clustering\ninto similar groups compared to the raw sequences themselves. By assigning a\nunique vector embedding to each short sequence, it is possible to more\nefficiently cluster and improve upon compression performance for the string\nrepresentations of cDNA libraries. Furthermore, through learning alternative\ncoordinate vector embeddings based on the contexts of codon triplets, we can\ndemonstrate clustering based on amino acid properties. Finally, using this\nsequence embedding method to encode barcodes and cDNA sequences, we can improve\nthe time complexity of the similarity search by coupling vector embeddings with\nan algorithm that determines the proximity of vectors in Euclidean space; this\nallows us to perform sequence similarity searches in a quicker and more modular\nfashion.\n", "date": "2023-08-08", "categories": []}
{"id": "2308.05125", "title": "Two Novel Approaches to Detect Community: A Case Study of Omicron\n  Lineage Variants PPI Network", "abstract": "  The capacity to identify and analyze protein-protein interactions, along with\ntheir internal modular organization, plays a crucial role in comprehending the\nintricate mechanisms underlying biological processes at the molecular level. We\ncan learn a lot about the structure and dynamics of these interactions by using\nnetwork analysis. We can improve our understanding of the biological roots of\ndisease pathogenesis by recognizing network communities. This knowledge, in\nturn, holds significant potential for driving advancements in drug discovery\nand facilitating personalized medicine approaches for disease treatment. In\nthis study, we aimed to uncover the communities within the variant B.1.1.529\n(Omicron virus) using two proposed novel algorithm (ABCDE and ALCDE) and four\nwidely recognized algorithms: Girvan-Newman, Louvain, Leiden, and Label\nPropagation algorithm. Each of these algorithms has established prominence in\nthe field and offers unique perspectives on identifying communities within\ncomplex networks. We also compare the networks by the global properties,\nstatistic summary, subgraph count, graphlet and validate by the modulaity. By\nemploying these approaches, we sought to gain deeper insights into the\nstructural organization and interconnections present within the Omicron virus\nnetwork.\n", "date": "2023-08-09", "categories": []}
{"id": "2308.05326", "title": "OpenProteinSet: Training data for structural biology at scale", "abstract": "  Multiple sequence alignments (MSAs) of proteins encode rich biological\ninformation and have been workhorses in bioinformatic methods for tasks like\nprotein design and protein structure prediction for decades. Recent\nbreakthroughs like AlphaFold2 that use transformers to attend directly over\nlarge quantities of raw MSAs have reaffirmed their importance. Generation of\nMSAs is highly computationally intensive, however, and no datasets comparable\nto those used to train AlphaFold2 have been made available to the research\ncommunity, hindering progress in machine learning for proteins. To remedy this\nproblem, we introduce OpenProteinSet, an open-source corpus of more than 16\nmillion MSAs, associated structural homologs from the Protein Data Bank, and\nAlphaFold2 protein structure predictions. We have previously demonstrated the\nutility of OpenProteinSet by successfully retraining AlphaFold2 on it. We\nexpect OpenProteinSet to be broadly useful as training and validation data for\n1) diverse tasks focused on protein structure, function, and design and 2)\nlarge-scale multimodal machine learning research.\n", "date": "2023-08-10", "categories": []}
{"id": "2307.00365", "title": "Understanding recent deep-learning techniques for identifying collective\n  variables of molecular dynamics", "abstract": "  High-dimensional metastable molecular system can often be characterised by a\nfew features of the system, i.e. collective variables (CVs). Thanks to the\nrapid advance in the area of machine learning and deep learning, various deep\nlearning-based CV identification techniques have been developed in recent\nyears, allowing accurate modelling and efficient simulation of complex\nmolecular systems. In this paper, we look at two different categories of deep\nlearning-based approaches for finding CVs, either by computing leading\neigenfunctions of infinitesimal generator or transfer operator associated to\nthe underlying dynamics, or by learning an autoencoder via minimisation of\nreconstruction error. We present a concise overview of the mathematics behind\nthese two approaches and conduct a comparative numerical study of these two\napproaches on illustrative examples.\n", "date": "2023-07-01", "categories": []}
{"id": "2301.04492", "title": "Fast conformational clustering of extensive molecular dynamics\n  simulation data", "abstract": "  We present an unsupervised data processing workflow that is specifically\ndesigned to obtain a fast conformational clustering of long molecular dynamics\nsimulation trajectories. In this approach we combine two dimensionality\nreduction algorithms (cc\\_analysis and encodermap) with a density-based spatial\nclustering algorithm (HDBSCAN). The proposed scheme benefits from the strengths\nof the three algorithms while avoiding most of the drawbacks of the individual\nmethods. Here the cc\\_analysis algorithm is for the first time applied to\nmolecular simulation data. Encodermap complements cc\\_analysis by providing an\nefficient way to process and assign large amounts of data to clusters. The main\ngoal of the procedure is to maximize the number of assigned frames of a given\ntrajectory, while keeping a clear conformational identity of the clusters that\nare found. In practice we achieve this by using an iterative clustering\napproach and a tunable root-mean-square-deviation-based criterion in the final\ncluster assignment. This allows to find clusters of different densities as well\nas different degrees of structural identity. With the help of four test systems\nwe illustrate the capability and performance of this clustering workflow:\nwild-type and thermostable mutant of the Trp-cage protein (TC5b and TC10b),\nNTL9 and Protein B. Each of these systems poses individual challenges to the\nscheme, which in total give a nice overview of the advantages, as well as\npotential difficulties that can arise when using the proposed method.\n", "date": "2023-01-11", "categories": []}
{"id": "2308.03654", "title": "FFF: Fragments-Guided Flexible Fitting for Building Complete Protein\n  Structures", "abstract": "  Cryo-electron microscopy (cryo-EM) is a technique for reconstructing the\n3-dimensional (3D) structure of biomolecules (especially large protein\ncomplexes and molecular assemblies). As the resolution increases to the\nnear-atomic scale, building protein structures de novo from cryo-EM maps\nbecomes possible. Recently, recognition-based de novo building methods have\nshown the potential to streamline this process. However, it cannot build a\ncomplete structure due to the low signal-to-noise ratio (SNR) problem. At the\nsame time, AlphaFold has led to a great breakthrough in predicting protein\nstructures. This has inspired us to combine fragment recognition and structure\nprediction methods to build a complete structure. In this paper, we propose a\nnew method named FFF that bridges protein structure prediction and protein\nstructure recognition with flexible fitting. First, a multi-level recognition\nnetwork is used to capture various structural features from the input 3D\ncryo-EM map. Next, protein structural fragments are generated using pseudo\npeptide vectors and a protein sequence alignment method based on these\nextracted features. Finally, a complete structural model is constructed using\nthe predicted protein fragments via flexible fitting. Based on our benchmark\ntests, FFF outperforms the baseline methods for building complete protein\nstructures.\n", "date": "2023-08-07", "categories": []}
{"id": "2308.01920", "title": "Sequence-Based Nanobody-Antigen Binding Prediction", "abstract": "  Nanobodies (Nb) are monomeric heavy-chain fragments derived from heavy-chain\nonly antibodies naturally found in Camelids and Sharks. Their considerably\nsmall size (~3-4 nm; 13 kDa) and favorable biophysical properties make them\nattractive targets for recombinant production. Furthermore, their unique\nability to bind selectively to specific antigens, such as toxins, chemicals,\nbacteria, and viruses, makes them powerful tools in cell biology, structural\nbiology, medical diagnostics, and future therapeutic agents in treating cancer\nand other serious illnesses. However, a critical challenge in nanobodies\nproduction is the unavailability of nanobodies for a majority of antigens.\nAlthough some computational methods have been proposed to screen potential\nnanobodies for given target antigens, their practical application is highly\nrestricted due to their reliance on 3D structures. Moreover, predicting\nnanobodyantigen interactions (binding) is a time-consuming and labor-intensive\ntask. This study aims to develop a machine-learning method to predict\nNanobody-Antigen binding solely based on the sequence data. We curated a\ncomprehensive dataset of Nanobody-Antigen binding and nonbinding data and\ndevised an embedding method based on gapped k-mers to predict binding based\nonly on sequences of nanobody and antigen. Our approach achieves up to 90%\naccuracy in binding prediction and is significantly more efficient compared to\nthe widely-used computational docking technique.\n", "date": "2023-07-15", "categories": []}
{"id": "2307.04603", "title": "Solvent: A Framework for Protein Folding", "abstract": "  Consistency and reliability are crucial for conducting AI research. Many\nfamous research fields, such as object detection, have been compared and\nvalidated with solid benchmark frameworks. After AlphaFold2, the protein\nfolding task has entered a new phase, and many methods are proposed based on\nthe component of AlphaFold2. The importance of a unified research framework in\nprotein folding contains implementations and benchmarks to consistently and\nfairly compare various approaches. To achieve this, we present Solvent, a\nprotein folding framework that supports significant components of\nstate-of-the-art models in the manner of an off-the-shelf interface Solvent\ncontains different models implemented in a unified codebase and supports\ntraining and evaluation for defined models on the same dataset. We benchmark\nwell-known algorithms and their components and provide experiments that give\nhelpful insights into the protein structure modeling field. We hope that\nSolvent will increase the reliability and consistency of proposed models and\ngive efficiency in both speed and costs, resulting in acceleration on protein\nfolding modeling research. The code is available at\nhttps://github.com/kakaobrain/solvent, and the project will continue to be\ndeveloped.\n", "date": "2023-07-07", "categories": []}
{"id": "2307.16182", "title": "Redundancy-aware unsupervised rankings for collections of gene sets", "abstract": "  The biological roles of gene sets are used to group them into collections.\nThese collections are often characterized by being high-dimensional,\noverlapping, and redundant families of sets, thus precluding a straightforward\ninterpretation and study of their content. Bioinformatics looked for solutions\nto reduce their dimension or increase their intepretability. One possibility\nlies in aggregating overlapping gene sets to create larger pathways, but the\nmodified biological pathways are hardly biologically justifiable. We propose to\nuse importance scores to rank the pathways in the collections studying the\ncontext from a set covering perspective. The proposed Shapley values-based\nscores consider the distribution of the singletons and the size of the sets in\nthe families; Furthermore, a trick allows us to circumvent the usual\nexponential complexity of Shapley values' computation. Finally, we address the\nchallenge of including a redundancy awareness in the obtained rankings where,\nin our case, sets are redundant if they show prominent intersections.\n  The rankings can be used to reduce the dimension of collections of gene sets,\nsuch that they show lower redundancy and still a high coverage of the genes. We\nfurther investigate the impact of our selection on Gene Sets Enrichment\nAnalysis. The proposed method shows a practical utility in bioinformatics to\nincrease the interpretability of the collections of gene sets and a step\nforward to include redundancy into Shapley values computations.\n", "date": "2023-07-30", "categories": []}
{"id": "2206.14566", "title": "Using Interpretable Machine Learning to Massively Increase the Number of\n  Antibody-Virus Interactions Across Studies", "abstract": "  A central challenge in every field of biology is to use existing measurements\nto predict the outcomes of future experiments. In this work, we consider the\nwealth of antibody inhibition data against variants of the influenza virus. Due\nto this viru's genetic diversity and evolvability, the variants examined in one\nstudy will often have little-to-no overlap with other studies, making it\ndifficult to discern common patterns or unify datasets for further analysis. To\nthat end, we develop a computational framework that predicts how an antibody or\nserum would inhibit any variant from any other study. We use this framework to\ngreatly expand seven influenza datasets utilizing hemagglutination inhibition,\nvalidating our method upon 200,000 existing measurements and predicting\n2,000,000 new values along with their uncertainties. With these new values, we\nquantify the transferability between seven vaccination and infection studies in\nhumans and ferrets, show that the serum potency is negatively correlated with\nbreadth, and present a tool for pandemic preparedness. This data-driven\napproach does not require any information beyond each virus's name and\nmeasurements, and even datasets with as few as 5 viruses can be expanded,\nmaking this approach widely applicable. Future influenza studies using\nhemagglutination inhibition can directly utilize our curated datasets to\npredict newly measured antibody responses against ~80 H3N2 influenza viruses\nfrom 1968-2011, whereas immunological studies utilizing other viruses or a\ndifferent assay only need a single partially-overlapping dataset to extend\ntheir work. In essence, this approach enables a shift in perspective when\nanalyzing data from \"what you see is what you get\" into \"what anyone sees is\nwhat everyone gets.\"\n", "date": "2022-06-10", "categories": []}
{"id": "2307.08800", "title": "regulAS: A Bioinformatics Tool for the Integrative Analysis of\n  Alternative Splicing Regulome using RNA-Seq data", "abstract": "  The regulAS software package is a bioinformatics tool designed to support\ncomputational biology researchers in investigating regulatory mechanisms of\nsplicing alterations through integrative analysis of large-scale RNA-Seq data\nfrom cancer and healthy human donors, characterized by TCGA and GTEx projects.\nThis technical report provides a comprehensive overview of regulAS, focusing on\nits core functionality, basic modules, experiment configuration, further\nextensibility and customisation.\n  The core functionality of regulAS enables the automation of computational\nexperiments, efficient results storage and processing, and streamlined workflow\nmanagement. Integrated basic modules extend regulAS with features such as\nRNA-Seq data retrieval from the public multi-omics UCSC Xena data repository,\npredictive modeling and feature ranking capabilities using the scikit-learn\npackage, and flexible reporting generation for analysing gene expression\nprofiles and relevant modulations of alternative splicing aberrations across\ntissues and cancer types. Experiment configuration is handled through YAML\nfiles with the Hydra and OmegaConf libraries, offering a user-friendly\napproach. Additionally, regulAS allows for the development and integration of\ncustom modules to handle specialized tasks.\n  In conclusion, regulAS provides an automated solution for alternative\nsplicing and cancer biology studies, enhancing efficiency, reproducibility, and\ncustomization of experimental design, while the extensibility of the pipeline\nenables researchers to further tailor the software package to their specific\nneeds. Source code is available under the MIT license at\nhttps://github.com/slipnitskaya/regulAS.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.13004", "title": "DeepGATGO: A Hierarchical Pretraining-Based Graph-Attention Model for\n  Automatic Protein Function Prediction", "abstract": "  Automatic protein function prediction (AFP) is classified as a large-scale\nmulti-label classification problem aimed at automating protein enrichment\nanalysis to eliminate the current reliance on labor-intensive wet-lab methods.\nCurrently, popular methods primarily combine protein-related information and\nGene Ontology (GO) terms to generate final functional predictions. For example,\nprotein sequences, structural information, and protein-protein interaction\nnetworks are integrated as prior knowledge to fuse with GO term embeddings and\ngenerate the ultimate prediction results. However, these methods are limited by\nthe difficulty in obtaining structural information or network topology\ninformation, as well as the accuracy of such data. Therefore, more and more\nmethods that only use protein sequences for protein function prediction have\nbeen proposed, which is a more reliable and computationally cheaper approach.\nHowever, the existing methods fail to fully extract feature information from\nprotein sequences or label data because they do not adequately consider the\nintrinsic characteristics of the data itself. Therefore, we propose a\nsequence-based hierarchical prediction method, DeepGATGO, which processes\nprotein sequences and GO term labels hierarchically, and utilizes graph\nattention networks (GATs) and contrastive learning for protein function\nprediction. Specifically, we compute embeddings of the sequence and label data\nusing pre-trained models to reduce computational costs and improve the\nembedding accuracy. Then, we use GATs to dynamically extract the structural\ninformation of non-Euclidean data, and learn general features of the label\ndataset with contrastive learning by constructing positive and negative example\nsamples. Experimental results demonstrate that our proposed model exhibits\nbetter scalability in GO term enrichment analysis on large-scale datasets.\n", "date": "2023-07-24", "categories": []}
{"id": "2307.13010", "title": "Deep neural network improves the estimation of polygenic risk scores for\n  breast cancer", "abstract": "  Polygenic risk scores (PRS) estimate the genetic risk of an individual for a\ncomplex disease based on many genetic variants across the whole genome. In this\nstudy, we compared a series of computational models for estimation of breast\ncancer PRS. A deep neural network (DNN) was found to outperform alternative\nmachine learning techniques and established statistical algorithms, including\nBLUP, BayesA and LDpred. In the test cohort with 50% prevalence, the Area Under\nthe receiver operating characteristic Curve (AUC) were 67.4% for DNN, 64.2% for\nBLUP, 64.5% for BayesA, and 62.4% for LDpred. BLUP, BayesA, and LPpred all\ngenerated PRS that followed a normal distribution in the case population.\nHowever, the PRS generated by DNN in the case population followed a bi-modal\ndistribution composed of two normal distributions with distinctly different\nmeans. This suggests that DNN was able to separate the case population into a\nhigh-genetic-risk case sub-population with an average PRS significantly higher\nthan the control population and a normal-genetic-risk case sub-population with\nan average PRS similar to the control population. This allowed DNN to achieve\n18.8% recall at 90% precision in the test cohort with 50% prevalence, which can\nbe extrapolated to 65.4% recall at 20% precision in a general population with\n12% prevalence. Interpretation of the DNN model identified salient variants\nthat were assigned insignificant p-values by association studies, but were\nimportant for DNN prediction. These variants may be associated with the\nphenotype through non-linear relationships.\n", "date": "2023-07-24", "categories": []}
{"id": "2307.13608", "title": "Geometric Epitope and Paratope Prediction", "abstract": "  Antibody-antigen interactions play a crucial role in identifying and\nneutralizing harmful foreign molecules. In this paper, we investigate the\noptimal representation for predicting the binding sites in the two molecules\nand emphasize the importance of geometric information. Specifically, we compare\ndifferent geometric deep learning methods applied to proteins' inner (I-GEP)\nand outer (O-GEP) structures. We incorporate 3D coordinates and spectral\ngeometric descriptors as input features to fully leverage the geometric\ninformation. Our research suggests that surface-based models are more efficient\nthan other methods, and our O-GEP experiments have achieved state-of-the-art\nresults with significant performance improvements.\n", "date": "2023-05-28", "categories": []}
{"id": "2210.16392", "title": "Physics-aware Graph Neural Network for Accurate RNA 3D Structure\n  Prediction", "abstract": "  Biological functions of RNAs are determined by their three-dimensional (3D)\nstructures. Thus, given the limited number of experimentally determined RNA\nstructures, the prediction of RNA structures will facilitate elucidating RNA\nfunctions and RNA-targeted drug discovery, but remains a challenging task. In\nthis work, we propose a Graph Neural Network (GNN)-based scoring function\ntrained only with the atomic types and coordinates on limited solved RNA 3D\nstructures for distinguishing accurate structural models. The proposed\nPhysics-aware Multiplex Graph Neural Network (PaxNet) separately models the\nlocal and non-local interactions inspired by molecular mechanics. Furthermore,\nPaxNet contains an attention-based fusion module that learns the individual\ncontribution of each interaction type for the final prediction. We rigorously\nevaluate the performance of PaxNet on two benchmarks and compare it with\nseveral state-of-the-art baselines. The results show that PaxNet significantly\noutperforms all the baselines overall, and demonstrate the potential of PaxNet\nfor improving the 3D structure modeling of RNA and other macromolecules. Our\ncode is available at https://github.com/zetayue/Physics-aware-Multiplex-GNN.\n", "date": "2022-10-28", "categories": []}
{"id": "2307.12451", "title": "DiAMoNDBack: Diffusion-denoising Autoregressive Model for\n  Non-Deterministic Backmapping of C{\\alpha} Protein Traces", "abstract": "  Coarse-grained molecular models of proteins permit access to length and time\nscales unattainable by all-atom models and the simulation of processes that\noccur on long-time scales such as aggregation and folding. The reduced\nresolution realizes computational accelerations but an atomistic representation\ncan be vital for a complete understanding of mechanistic details. Backmapping\nis the process of restoring all-atom resolution to coarse-grained molecular\nmodels. In this work, we report DiAMoNDBack (Diffusion-denoising Autoregressive\nModel for Non-Deterministic Backmapping) as an autoregressive denoising\ndiffusion probability model to restore all-atom details to coarse-grained\nprotein representations retaining only C{\\alpha} coordinates. The\nautoregressive generation process proceeds from the protein N-terminus to\nC-terminus in a residue-by-residue fashion conditioned on the C{\\alpha} trace\nand previously backmapped backbone and side chain atoms within the local\nneighborhood. The local and autoregressive nature of our model makes it\ntransferable between proteins. The stochastic nature of the denoising diffusion\nprocess means that the model generates a realistic ensemble of backbone and\nside chain all-atom configurations consistent with the coarse-grained C{\\alpha}\ntrace. We train DiAMoNDBack over 65k+ structures from Protein Data Bank (PDB)\nand validate it in applications to a hold-out PDB test set,\nintrinsically-disordered protein structures from the Protein Ensemble Database\n(PED), molecular dynamics simulations of fast-folding mini-proteins from DE\nShaw Research, and coarse-grained simulation data. We achieve state-of-the-art\nreconstruction performance in terms of correct bond formation, avoidance of\nside chain clashes, and diversity of the generated side chain configurational\nstates. We make DiAMoNDBack model publicly available as a free and open source\nPython package.\n", "date": "2023-07-23", "categories": []}
{"id": "2307.09484", "title": "MolFM: A Multimodal Molecular Foundation Model", "abstract": "  Molecular knowledge resides within three different modalities of information\nsources: molecular structures, biomedical documents, and knowledge bases.\nEffective incorporation of molecular knowledge from these modalities holds\nparamount significance in facilitating biomedical research. However, existing\nmultimodal molecular foundation models exhibit limitations in capturing\nintricate connections between molecular structures and texts, and more\nimportantly, none of them attempt to leverage a wealth of molecular expertise\nderived from knowledge graphs. In this study, we introduce MolFM, a multimodal\nmolecular foundation model designed to facilitate joint representation learning\nfrom molecular structures, biomedical texts, and knowledge graphs. We propose\ncross-modal attention between atoms of molecular structures, neighbors of\nmolecule entities and semantically related texts to facilitate cross-modal\ncomprehension. We provide theoretical analysis that our cross-modal\npre-training captures local and global molecular knowledge by minimizing the\ndistance in the feature space between different modalities of the same\nmolecule, as well as molecules sharing similar structures or functions. MolFM\nachieves state-of-the-art performance on various downstream tasks. On\ncross-modal retrieval, MolFM outperforms existing models with 12.13% and 5.04%\nabsolute gains under the zero-shot and fine-tuning settings, respectively.\nFurthermore, qualitative analysis showcases MolFM's implicit ability to provide\ngrounding from molecular substructures and knowledge graphs. Code and models\nare available on https://github.com/BioFM/OpenBioMed.\n", "date": "2023-06-06", "categories": []}
{"id": "2307.10343", "title": "ProtiGeno: a prokaryotic short gene finder using protein language models", "abstract": "  Prokaryotic gene prediction plays an important role in understanding the\nbiology of organisms and their function with applications in medicine and\nbiotechnology. Although the current gene finders are highly sensitive in\nfinding long genes, their sensitivity decreases noticeably in finding shorter\ngenes (<180 nts). The culprit is insufficient annotated gene data to identify\ndistinguishing features in short open reading frames (ORFs). We develop a deep\nlearning-based method called ProtiGeno, specifically targeting short\nprokaryotic genes using a protein language model trained on millions of evolved\nproteins. In systematic large-scale experiments on 4,288 prokaryotic genomes,\nwe demonstrate that ProtiGeno predicts short coding and noncoding genes with\nhigher accuracy and recall than the current state-of-the-art gene finders. We\ndiscuss the predictive features of ProtiGeno and possible limitations by\nvisualizing the three-dimensional structure of the predicted short genes. Data,\ncodes, and models are available at https://github.com/tonytu16/protigeno.\n", "date": "2023-07-19", "categories": []}
{"id": "2307.10073", "title": "Scalable Deep Learning for RNA Secondary Structure Prediction", "abstract": "  The field of RNA secondary structure prediction has made significant progress\nwith the adoption of deep learning techniques. In this work, we present the\nRNAformer, a lean deep learning model using axial attention and recycling in\nthe latent space. We gain performance improvements by designing the\narchitecture for modeling the adjacency matrix directly in the latent space and\nby scaling the size of the model. Our approach achieves state-of-the-art\nperformance on the popular TS0 benchmark dataset and even outperforms methods\nthat use external information. Further, we show experimentally that the\nRNAformer can learn a biophysical model of the RNA folding process.\n", "date": "2023-07-14", "categories": []}
{"id": "2307.08801", "title": "Towards Automated Design of Riboswitches", "abstract": "  Experimental screening and selection pipelines for the discovery of novel\nriboswitches are expensive, time-consuming, and inefficient. Using\ncomputational methods to reduce the number of candidates for the screen could\ndrastically decrease these costs. However, existing computational approaches do\nnot fully satisfy all requirements for the design of such initial screening\nlibraries. In this work, we present a new method, libLEARNA, capable of\nproviding RNA focus libraries of diverse variable-length qualified candidates.\nOur novel structure-based design approach considers global properties as well\nas desired sequence and structure features. We demonstrate the benefits of our\nmethod by designing theophylline riboswitch libraries, following a previously\npublished protocol, and yielding 30% more unique high-quality candidates.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.08893", "title": "Evaluating unsupervised disentangled representation learning for genomic\n  discovery and disease risk prediction", "abstract": "  High-dimensional clinical data have become invaluable resources for genetic\nstudies, due to their accessibility in biobank-scale datasets and the\ndevelopment of high performance modeling techniques especially using deep\nlearning. Recent work has shown that low dimensional embeddings of these\nclinical data learned by variational autoencoders (VAE) can be used for\ngenome-wide association studies and polygenic risk prediction. In this work, we\nconsider multiple unsupervised learning methods for learning disentangled\nrepresentations, namely autoencoders, VAE, beta-VAE, and FactorVAE, in the\ncontext of genetic association studies. Using spirograms from UK Biobank as a\nrunning example, we observed improvements in the number of genome-wide\nsignificant loci, heritability, and performance of polygenic risk scores for\nasthma and chronic obstructive pulmonary disease by using FactorVAE or\nbeta-VAE, compared to standard VAE or non-variational autoencoders. FactorVAEs\nperformed effectively across multiple values of the regularization\nhyperparameter, while beta-VAEs were much more sensitive to the hyperparameter\nvalues.\n", "date": "2023-07-17", "categories": []}
{"id": "2307.09169", "title": "Efficient Prediction of Peptide Self-assembly through Sequential and\n  Graphical Encoding", "abstract": "  In recent years, there has been an explosion of research on the application\nof deep learning to the prediction of various peptide properties, due to the\nsignificant development and market potential of peptides. Molecular dynamics\nhas enabled the efficient collection of large peptide datasets, providing\nreliable training data for deep learning. However, the lack of systematic\nanalysis of the peptide encoding, which is essential for AI-assisted\npeptide-related tasks, makes it an urgent problem to be solved for the\nimprovement of prediction accuracy. To address this issue, we first collect a\nhigh-quality, colossal simulation dataset of peptide self-assembly containing\nover 62,000 samples generated by coarse-grained molecular dynamics (CGMD).\nThen, we systematically investigate the effect of peptide encoding of amino\nacids into sequences and molecular graphs using state-of-the-art sequential\n(i.e., RNN, LSTM, and Transformer) and structural deep learning models (i.e.,\nGCN, GAT, and GraphSAGE), on the accuracy of peptide self-assembly prediction,\nan essential physiochemical process prior to any peptide-related applications.\nExtensive benchmarking studies have proven Transformer to be the most powerful\nsequence-encoding-based deep learning model, pushing the limit of peptide\nself-assembly prediction to decapeptides. In summary, this work provides a\ncomprehensive benchmark analysis of peptide encoding with advanced deep\nlearning models, serving as a guide for a wide range of peptide-related\npredictions such as isoelectric points, hydration free energy, etc.\n", "date": "2023-07-17", "categories": []}
{"id": "2212.01495", "title": "iEnhancer-ELM: improve enhancer identification by extracting\n  position-related multiscale contextual information based on enhancer language\n  models", "abstract": "  Motivation: Enhancers are important cis-regulatory elements that regulate a\nwide range of biological functions and enhance the transcription of target\ngenes. Although many feature extraction methods have been proposed to improve\nthe performance of enhancer identification, they cannot learn position-related\nmultiscale contextual information from raw DNA sequences.\n  Results: In this article, we propose a novel enhancer identification method\n(iEnhancer-ELM) based on BERT-like enhancer language models. iEnhancer-ELM\ntokenizes DNA sequences with multi-scale k-mers and extracts contextual\ninformation of different scale k-mers related with their positions via an\nmulti-head attention mechanism. We first evaluate the performance of different\nscale k-mers, then ensemble them to improve the performance of enhancer\nidentification. The experimental results on two popular benchmark datasets show\nthat our model outperforms stateof-the-art methods. We further illustrate the\ninterpretability of iEnhancer-ELM. For a case study, we discover 30 enhancer\nmotifs via a 3-mer-based model, where 12 of motifs are verified by STREME and\nJASPAR, demonstrating our model has a potential ability to unveil the\nbiological mechanism of enhancer.\n  Availability and implementation: The models and associated code are available\nat https://github.com/chen-bioinfo/iEnhancer-ELM\n  Contact: junjiechen@hit.edu.cn\n  Supplementary information: Supplementary data are available at Bioinformatics\nAdvances online.\n", "date": "2022-12-03", "categories": []}
{"id": "2307.01066", "title": "PIGNet2: A Versatile Deep Learning-based Protein-Ligand Interaction\n  Prediction Model for Binding Affinity Scoring and Virtual Screening", "abstract": "  Prediction of protein-ligand interactions (PLI) plays a crucial role in drug\ndiscovery as it guides the identification and optimization of molecules that\neffectively bind to target proteins. Despite remarkable advances in deep\nlearning-based PLI prediction, the development of a versatile model capable of\naccurately scoring binding affinity and conducting efficient virtual screening\nremains a challenge. The main obstacle in achieving this lies in the scarcity\nof experimental structure-affinity data, which limits the generalization\nability of existing models. Here, we propose a viable solution to address this\nchallenge by introducing a novel data augmentation strategy combined with a\nphysics-informed graph neural network. The model showed significant\nimprovements in both scoring and screening, outperforming task-specific deep\nlearning models in various tests including derivative benchmarks, and notably\nachieving results comparable to the state-of-the-art performance based on\ndistance likelihood learning. This demonstrates the potential of this approach\nto drug discovery.\n", "date": "2023-07-03", "categories": []}
{"id": "2307.05537", "title": "NLP Meets RNA: Unsupervised Embedding Learning for Ribozymes with\n  Word2Vec", "abstract": "  Ribozymes, RNA molecules with distinct 3D structures and catalytic activity,\nhave widespread applications in synthetic biology and therapeutics. However,\nrelatively little research has focused on leveraging deep learning to enhance\nour understanding of ribozymes. This study implements Word2Vec, an unsupervised\nlearning technique for natural language processing, to learn ribozyme\nembeddings. Ribo2Vec was trained on over 9,000 diverse ribozymes, learning to\nmap sequences to 128 and 256-dimensional vector spaces. Using Ribo2Vec,\nsequence embeddings for five classes of ribozymes (hatchet, pistol, hairpin,\nhovlinc, and twister sister) were calculated. Principal component analysis\ndemonstrated the ability of these embeddings to distinguish between ribozyme\nclasses. Furthermore, a simple SVM classifier trained on ribozyme embeddings\nshowed promising results in accurately classifying ribozyme types. Our results\nsuggest that the embedding vectors contained meaningful information about\nribozymes. Interestingly, 256-dimensional embeddings behaved similarly to\n128-dimensional embeddings, suggesting that a lower dimension vector space is\ngenerally sufficient to capture ribozyme features. This approach demonstrates\nthe potential of Word2Vec for bioinformatics, opening new avenues for ribozyme\nresearch. Future research includes using a Transformer-based method to learn\nRNA embeddings, which can capture long-range interactions between nucleotides.\n", "date": "2023-07-08", "categories": []}
{"id": "2307.00033", "title": "Application of data engineering approaches to address challenges in\n  microbiome data for optimal medical decision-making", "abstract": "  The human gut microbiota is known to contribute to numerous physiological\nfunctions of the body and also implicated in a myriad of pathological\nconditions. Prolific research work in the past few decades have yielded\nvaluable information regarding the relative taxonomic distribution of gut\nmicrobiota. Unfortunately, the microbiome data suffers from class imbalance and\nhigh dimensionality issues that must be addressed. In this study, we have\nimplemented data engineering algorithms to address the above-mentioned issues\ninherent to microbiome data. Four standard machine learning classifiers\n(logistic regression (LR), support vector machines (SVM), random forests (RF),\nand extreme gradient boosting (XGB) decision trees) were implemented on a\npreviously published dataset. The issue of class imbalance and high\ndimensionality of the data was addressed through synthetic minority\noversampling technique (SMOTE) and principal component analysis (PCA). Our\nresults indicate that ensemble classifiers (RF and XGB decision trees) exhibit\nsuperior classification accuracy in predicting the host phenotype. The\napplication of PCA significantly reduced testing time while maintaining high\nclassification accuracy. The highest classification accuracy was obtained at\nthe levels of species for most classifiers. The prototype employed in the study\naddresses the issues inherent to microbiome datasets and could be highly\nbeneficial for providing personalized medicine.\n", "date": "2023-06-30", "categories": []}
{"id": "2301.12068", "title": "Pre-Training Protein Encoder via Siamese Sequence-Structure Diffusion\n  Trajectory Prediction", "abstract": "  Self-supervised pre-training methods on proteins have recently gained\nattention, with most approaches focusing on either protein sequences or\nstructures, neglecting the exploration of their joint distribution, which is\ncrucial for a comprehensive understanding of protein functions by integrating\nco-evolutionary information and structural characteristics. In this work,\ninspired by the success of denoising diffusion models in generative tasks, we\npropose the DiffPreT approach to pre-train a protein encoder by\nsequence-structure joint diffusion modeling. DiffPreT guides the encoder to\nrecover the native protein sequences and structures from the perturbed ones\nalong the joint diffusion trajectory, which acquires the joint distribution of\nsequences and structures. Considering the essential protein conformational\nvariations, we enhance DiffPreT by a method called Siamese Diffusion Trajectory\nPrediction (SiamDiff) to capture the correlation between different conformers\nof a protein. SiamDiff attains this goal by maximizing the mutual information\nbetween representations of diffusion trajectories of structurally-correlated\nconformers. We study the effectiveness of DiffPreT and SiamDiff on both atom-\nand residue-level structure-based protein understanding tasks. Experimental\nresults show that the performance of DiffPreT is consistently competitive on\nall tasks, and SiamDiff achieves new state-of-the-art performance, considering\nthe mean ranks on all tasks. Our implementation is available at\nhttps://github.com/DeepGraphLearning/SiamDiff.\n", "date": "2023-01-28", "categories": []}
{"id": "2307.02502", "title": "Math Agents: Computational Infrastructure, Mathematical Embedding, and\n  Genomics", "abstract": "  The advancement in generative AI could be boosted with more accessible\nmathematics. Beyond human-AI chat, large language models (LLMs) are emerging in\nprogramming, algorithm discovery, and theorem proving, yet their genomics\napplication is limited. This project introduces Math Agents and mathematical\nembedding as fresh entries to the \"Moore's Law of Mathematics\", using a\nGPT-based workflow to convert equations from literature into LaTeX and Python\nformats. While many digital equation representations exist, there's a lack of\nautomated large-scale evaluation tools. LLMs are pivotal as linguistic user\ninterfaces, providing natural language access for human-AI chat and formal\nlanguages for large-scale AI-assisted computational infrastructure. Given the\ninfinite formal possibility spaces, Math Agents, which interact with math,\ncould potentially shift us from \"big data\" to \"big math\". Math, unlike the more\nflexible natural language, has properties subject to proof, enabling its use\nbeyond traditional applications like high-validation math-certified icons for\nAI alignment aims. This project aims to use Math Agents and mathematical\nembeddings to address the ageing issue in information systems biology by\napplying multiscalar physics mathematics to disease models and genomic data.\nGenerative AI with episodic memory could help analyse causal relations in\nlongitudinal health records, using SIR Precision Health models. Genomic data is\nsuggested for addressing the unsolved Alzheimer's disease problem.\n", "date": "2023-07-04", "categories": []}
{"id": "2304.13230", "title": "UNADON: Transformer-based model to predict genome-wide chromosome\n  spatial position", "abstract": "  The spatial positioning of chromosomes relative to functional nuclear bodies\nis intertwined with genome functions such as transcription. However, the\nsequence patterns and epigenomic features that collectively influence chromatin\nspatial positioning in a genome-wide manner are not well understood. Here, we\ndevelop a new transformer-based deep learning model called UNADON, which\npredicts the genome-wide cytological distance to a specific type of nuclear\nbody, as measured by TSA-seq, using both sequence features and epigenomic\nsignals. Evaluations of UNADON in four cell lines (K562, H1, HFFc6, HCT116)\nshow high accuracy in predicting chromatin spatial positioning to nuclear\nbodies when trained on a single cell line. UNADON also performed well in an\nunseen cell type. Importantly, we reveal potential sequence and epigenomic\nfactors that affect large-scale chromatin compartmentalization to nuclear\nbodies. Together, UNADON provides new insights into the principles between\nsequence features and large-scale chromatin spatial localization, which has\nimportant implications for understanding nuclear structure and function.\n", "date": "2023-04-26", "categories": []}
{"id": "2306.17702", "title": "Why Deep Models Often cannot Beat Non-deep Counterparts on Molecular\n  Property Prediction?", "abstract": "  Molecular property prediction (MPP) is a crucial task in the drug discovery\npipeline, which has recently gained considerable attention thanks to advances\nin deep neural networks. However, recent research has revealed that deep models\nstruggle to beat traditional non-deep ones on MPP. In this study, we benchmark\n12 representative models (3 non-deep models and 9 deep models) on 14 molecule\ndatasets. Through the most comprehensive study to date, we make the following\nkey observations: \\textbf{(\\romannumeral 1)} Deep models are generally unable\nto outperform non-deep ones; \\textbf{(\\romannumeral 2)} The failure of deep\nmodels on MPP cannot be solely attributed to the small size of molecular\ndatasets. What matters is the irregular molecule data pattern;\n\\textbf{(\\romannumeral 3)} In particular, tree models using molecular\nfingerprints as inputs tend to perform better than other competitors.\nFurthermore, we conduct extensive empirical investigations into the unique\npatterns of molecule data and inductive biases of various models underlying\nthese phenomena.\n", "date": "2023-06-30", "categories": []}
{"id": "2306.17771", "title": "Precision Anti-Cancer Drug Selection via Neural Ranking", "abstract": "  Personalized cancer treatment requires a thorough understanding of complex\ninteractions between drugs and cancer cell lines in varying genetic and\nmolecular contexts. To address this, high-throughput screening has been used to\ngenerate large-scale drug response data, facilitating data-driven computational\nmodels. Such models can capture complex drug-cell line interactions across\nvarious contexts in a fully data-driven manner. However, accurately\nprioritizing the most sensitive drugs for each cell line still remains a\nsignificant challenge. To address this, we developed neural ranking approaches\nthat leverage large-scale drug response data across multiple cell lines from\ndiverse cancer types. Unlike existing approaches that primarily utilize\nregression and classification techniques for drug response prediction, we\nformulated the objective of drug selection and prioritization as a drug ranking\nproblem. In this work, we proposed two neural listwise ranking methods that\nlearn latent representations of drugs and cell lines, and then use those\nrepresentations to score drugs in each cell line via a learnable scoring\nfunction. Specifically, we developed a neural listwise ranking method,\nList-One, on top of the existing method ListNet. Additionally, we proposed a\nnovel listwise ranking method, List-All, that focuses on all the sensitive\ndrugs instead of the top sensitive drug, unlike List-One. Our results\ndemonstrate that List-All outperforms the best baseline with significant\nimprovements of as much as 8.6% in hit@20 across 50% test cell lines.\nFurthermore, our analyses suggest that the learned latent spaces from our\nproposed methods demonstrate informative clustering structures and capture\nrelevant underlying biological features. Moreover, our comprehensive empirical\nevaluation provides a thorough and objective comparison of the performance of\ndifferent methods (including our proposed ones).\n", "date": "2023-06-30", "categories": []}
{"id": "2305.09790", "title": "Molecule-Morphology Contrastive Pretraining for Transferable Molecular\n  Representation", "abstract": "  Image-based profiling techniques have become increasingly popular over the\npast decade for their applications in target identification,\nmechanism-of-action inference, and assay development. These techniques have\ngenerated large datasets of cellular morphologies, which are typically used to\ninvestigate the effects of small molecule perturbagens. In this work, we extend\nthe impact of such dataset to improving quantitative structure-activity\nrelationship (QSAR) models by introducing Molecule-Morphology Contrastive\nPretraining (MoCoP), a framework for learning multi-modal representation of\nmolecular graphs and cellular morphologies. We scale MoCoP to approximately\n100K molecules and 600K morphological profiles using data from the JUMP-CP\nConsortium and show that MoCoP consistently improves performances of graph\nneural networks (GNNs) on molecular property prediction tasks in ChEMBL20\nacross all dataset sizes. The pretrained GNNs are also evaluated on internal\nGSK pharmacokinetic data and show an average improvement of 2.6% and 6.3% in\nAUPRC for full and low data regimes, respectively. Our findings suggest that\nintegrating cellular morphologies with molecular graphs using MoCoP can\nsignificantly improve the performance of QSAR models, ultimately expanding the\ndeep learning toolbox available for QSAR applications.\n", "date": "2023-04-27", "categories": []}
{"id": "2305.15559", "title": "Inverse square Levy walk emerging universally in goal-oriented tasks", "abstract": "  The Levy walk in which the frequency of occurrence of step lengths follows a\npower-law distribution, can be observed in the migratory behavior of organisms\nat various levels. Levy walks with power exponents close to 2 are observed, and\nthe reasons are unclear. This study aims to propose a model that universally\ngenerates inverse square Levy walks (called Cauchy walks) and to identify the\nconditions under which Cauchy walks appear. We demonstrate that Cauchy walks\nemerge universally in goal-oriented tasks. We use the term \"goal-oriented\" when\nthe goal is clear, but this can be achieved in different ways, which cannot be\nuniquely determined. We performed a simulation in which an agent observed the\ndata generated from a probability distribution in a two-dimensional space and\nsuccessively estimated the central coordinates of that probability\ndistribution. The agent has a model of probability distribution as a hypothesis\nfor data-generating distribution and can modify the model such that each time a\ndata point is observed, thereby increasing the estimated probability of\noccurrence of the observed data. To achieve this, the center coordinates of the\nmodel must be moved closer to those of the observed data. However, in the case\nof a two-dimensional space, arbitrariness arises in the direction of correction\nof the center; this task is goal oriented. We analyze two cases: a strategy\nthat allocates the amount of modification randomly in the x- and y-directions,\nand a strategy that determines allocation such that movement is minimized. The\nresults reveal that when a random strategy is used, the Cauchy walk appears.\nWhen the minimum strategy is used, the Brownian walk appears. The presence or\nabsence of the constraint of minimizing the amount of movement may be a factor\nthat causes the difference between Brownian and Levy walks.\n", "date": "2023-05-23", "categories": []}
{"id": "2306.13750", "title": "Analyzing scRNA-seq data by CCP-assisted UMAP and t-SNE", "abstract": "  Single-cell RNA sequencing (scRNA-seq) is widely used to reveal heterogeneity\nin cells, which has given us insights into cell-cell communication, cell\ndifferentiation, and differential gene expression. However, analyzing scRNA-seq\ndata is a challenge due to sparsity and the large number of genes involved.\nTherefore, dimensionality reduction and feature selection are important for\nremoving spurious signals and enhancing downstream analysis. Correlated\nclustering and projection (CCP) was recently introduced as an effective method\nfor preprocessing scRNA-seq data. CCP utilizes gene-gene correlations to\npartition the genes and, based on the partition, employs cell-cell interactions\nto obtain super-genes. Because CCP is a data-domain approach that does not\nrequire matrix diagonalization, it can be used in many downstream machine\nlearning tasks. In this work, we utilize CCP as an initialization tool for\nuniform manifold approximation and projection (UMAP) and t-distributed\nstochastic neighbor embedding (t-SNE). By using eight publicly available\ndatasets, we have found that CCP significantly improves UMAP and t-SNE\nvisualization and dramatically improve their accuracy.\n", "date": "2023-06-23", "categories": []}
{"id": "2306.13957", "title": "DiffDTM: A conditional structure-free framework for bioactive molecules\n  generation targeted for dual proteins", "abstract": "  Advances in deep generative models shed light on de novo molecule generation\nwith desired properties. However, molecule generation targeted for dual protein\ntargets still faces formidable challenges including protein 3D structure data\nrequisition for model training, auto-regressive sampling, and model\ngeneralization for unseen targets. Here, we proposed DiffDTM, a novel\nconditional structure-free deep generative model based on a diffusion model for\ndual targets based molecule generation to address the above issues.\nSpecifically, DiffDTM receives protein sequences and molecular graphs as inputs\ninstead of protein and molecular conformations and incorporates an information\nfusion module to achieve conditional generation in a one-shot manner. We have\nconducted comprehensive multi-view experiments to demonstrate that DiffDTM can\ngenerate drug-like, synthesis-accessible, novel, and high-binding affinity\nmolecules targeting specific dual proteins, outperforming the state-of-the-art\n(SOTA) models in terms of multiple evaluation metrics. Furthermore, we utilized\nDiffDTM to generate molecules towards dopamine receptor D2 and\n5-hydroxytryptamine receptor 1A as new antipsychotics. The experimental results\nindicate that DiffDTM can be easily plugged into unseen dual targets to\ngenerate bioactive molecules, addressing the issues of requiring insufficient\nactive molecule data for training as well as the need to retrain when\nencountering new targets.\n", "date": "2023-06-24", "categories": []}
{"id": "2302.03294", "title": "Linear-scaling kernels for protein sequences and small molecules\n  outperform deep learning while providing uncertainty quantitation and\n  improved interpretability", "abstract": "  Gaussian process (GP) is a Bayesian model which provides several advantages\nfor regression tasks in machine learning such as reliable quantitation of\nuncertainty and improved interpretability. Their adoption has been precluded by\ntheir excessive computational cost and by the difficulty in adapting them for\nanalyzing sequences (e.g. amino acid and nucleotide sequences) and graphs (e.g.\nones representing small molecules). In this study, we develop efficient and\nscalable approaches for fitting GP models as well as fast convolution kernels\nwhich scale linearly with graph or sequence size. We implement these\nimprovements by building an open-source Python library called xGPR. We compare\nthe performance of xGPR with the reported performance of various deep learning\nmodels on 20 benchmarks, including small molecule, protein sequence and tabular\ndata. We show that xGRP achieves highly competitive performance with much\nshorter training time. Furthermore, we also develop new kernels for sequence\nand graph data and show that xGPR generally outperforms convolutional neural\nnetworks on predicting key properties of proteins and small molecules.\nImportantly, xGPR provides uncertainty information not available from typical\ndeep learning models. Additionally, xGPR provides a representation of the input\ndata that can be used for clustering and data visualization. These results\ndemonstrate that xGPR provides a powerful and generic tool that can be broadly\nuseful in protein engineering and drug discovery.\n", "date": "2023-02-07", "categories": []}
{"id": "2305.10699", "title": "Dirichlet Diffusion Score Model for Biological Sequence Generation", "abstract": "  Designing biological sequences is an important challenge that requires\nsatisfying complex constraints and thus is a natural problem to address with\ndeep generative modeling. Diffusion generative models have achieved\nconsiderable success in many applications. Score-based generative stochastic\ndifferential equations (SDE) model is a continuous-time diffusion model\nframework that enjoys many benefits, but the originally proposed SDEs are not\nnaturally designed for modeling discrete data. To develop generative SDE models\nfor discrete data such as biological sequences, here we introduce a diffusion\nprocess defined in the probability simplex space with stationary distribution\nbeing the Dirichlet distribution. This makes diffusion in continuous space\nnatural for modeling discrete data. We refer to this approach as Dirchlet\ndiffusion score model. We demonstrate that this technique can generate samples\nthat satisfy hard constraints using a Sudoku generation task. This generative\nmodel can also solve Sudoku, including hard puzzles, without additional\ntraining. Finally, we applied this approach to develop the first human promoter\nDNA sequence design model and showed that designed sequences share similar\nproperties with natural promoter sequences.\n", "date": "2023-05-18", "categories": []}
{"id": "2306.08907", "title": "MCPI: Integrating Multimodal Data for Enhanced Prediction of Compound\n  Protein Interactions", "abstract": "  The identification of compound-protein interactions (CPI) plays a critical\nrole in drug screening, drug repurposing, and combination therapy studies. The\neffectiveness of CPI prediction relies heavily on the features extracted from\nboth compounds and target proteins. While various prediction methods employ\ndifferent feature combinations, both molecular-based and network-based models\nencounter the common obstacle of incomplete feature representations. Thus, a\npromising solution to this issue is to fully integrate all relevant CPI\nfeatures. This study proposed a novel model named MCPI, which is designed to\nimprove the prediction performance of CPI by integrating multiple sources of\ninformation, including the PPI network, CCI network, and structural features of\nCPI. The results of the study indicate that the MCPI model outperformed other\nexisting methods for predicting CPI on public datasets. Furthermore, the study\nhas practical implications for drug development, as the model was applied to\nsearch for potential inhibitors among FDA-approved drugs in response to the\nSARS-CoV-2 pandemic. The prediction results were then validated through the\nliterature, suggesting that the MCPI model could be a useful tool for\nidentifying potential drug candidates. Overall, this study has the potential to\nadvance our understanding of CPI and guide drug development efforts.\n", "date": "2023-06-15", "categories": []}
{"id": "2306.09187", "title": "MolCAP: Molecular Chemical reActivity pretraining and\n  prompted-finetuning enhanced molecular representation learning", "abstract": "  Molecular representation learning (MRL) is a fundamental task for drug\ndiscovery. However, previous deep-learning (DL) methods focus excessively on\nlearning robust inner-molecular representations by mask-dominated pretraining\nframework, neglecting abundant chemical reactivity molecular relationships that\nhave been demonstrated as the determining factor for various molecular property\nprediction tasks. Here, we present MolCAP to promote MRL, a graph pretraining\nTransformer based on chemical reactivity (IMR) knowledge with prompted\nfinetuning. Results show that MolCAP outperforms comparative methods based on\ntraditional molecular pretraining framework, in 13 publicly available molecular\ndatasets across a diversity of biomedical tasks. Prompted by MolCAP, even basic\ngraph neural networks are capable of achieving surprising performance that\noutperforms previous models, indicating the promising prospect of applying\nreactivity information for MRL. In addition, manual designed molecular templets\nare potential to uncover the dataset bias. All in all, we expect our MolCAP to\ngain more chemical meaningful insights for the entire process of drug\ndiscovery.\n", "date": "2023-06-13", "categories": []}
{"id": "2306.07484", "title": "Multi-objective Molecular Optimization for Opioid Use Disorder Treatment\n  Using Generative Network Complex", "abstract": "  Opioid Use Disorder (OUD) has emerged as a significant global public health\nissue, with complex multifaceted conditions. Due to the lack of effective\ntreatment options for various conditions, there is a pressing need for the\ndiscovery of new medications. In this study, we propose a deep generative model\nthat combines a stochastic differential equation (SDE)-based diffusion modeling\nwith the latent space of a pretrained autoencoder model. The molecular\ngenerator enables efficient generation of molecules that are effective on\nmultiple targets, specifically the mu, kappa, and delta opioid receptors.\nFurthermore, we assess the ADMET (absorption, distribution, metabolism,\nexcretion, and toxicity) properties of the generated molecules to identify\ndrug-like compounds. To enhance the pharmacokinetic properties of some lead\ncompounds, we employ a molecular optimization approach. We obtain a diverse set\nof drug-like molecules. We construct binding affinity predictors by integrating\nmolecular fingerprints derived from autoencoder embeddings, transformer\nembeddings, and topological Laplacians with advanced machine learning\nalgorithms. Further experimental studies are needed to evaluate the\npharmacological effects of these drug-like compounds for OUD treatment. Our\nmachine learning platform serves as a valuable tool in designing and optimizing\neffective molecules for addressing OUD.\n", "date": "2023-06-13", "categories": []}
{"id": "2306.06292", "title": "PLPCA: Persistent Laplacian Enhanced-PCA for Microarray Data Analysis", "abstract": "  Over the years, Principal Component Analysis (PCA) has served as the baseline\napproach for dimensionality reduction in gene expression data analysis. It\nprimary objective is to identify a subset of disease-causing genes from a vast\npool of thousands of genes. However, PCA possesses inherent limitations that\nhinder its interpretability, introduce classification ambiguity, and fail to\ncapture complex geometric structures in the data. Although these limitations\nhave been partially addressed in the literature by incorporating various\nregularizers such as graph Laplacian regularization, existing improved PCA\nmethods still face challenges related to multiscale analysis and capturing\nhigher-order interactions in the data. To address these challenges, we propose\na novel approach called Persistent Laplacian-enhanced Principal Component\nAnalysis (PLPCA). PLPCA amalgamates the advantages of earlier regularized PCA\nmethods with persistent spectral graph theory, specifically persistent\nLaplacians derived from algebraic topology. In contrast to graph Laplacians,\npersistent Laplacians enable multiscale analysis through filtration and\nincorporate higher-order simplicial complexes to capture higher-order\ninteractions in the data. We evaluate and validate the performance of PLPCA\nusing benchmark microarray datasets that involve normal tissue samples and four\ndifferent cancer tissues. Our extensive studies demonstrate that PLPCA\noutperforms all other state-of-the-art models for classification tasks after\ndimensionality reduction.\n", "date": "2023-06-09", "categories": []}
{"id": "2301.11765", "title": "ExplainableFold: Understanding AlphaFold Prediction with Explainable AI", "abstract": "  This paper presents ExplainableFold, an explainable AI framework for protein\nstructure prediction. Despite the success of AI-based methods such as AlphaFold\nin this field, the underlying reasons for their predictions remain unclear due\nto the black-box nature of deep learning models. To address this, we propose a\ncounterfactual learning framework inspired by biological principles to generate\ncounterfactual explanations for protein structure prediction, enabling a\ndry-lab experimentation approach. Our experimental results demonstrate the\nability of ExplainableFold to generate high-quality explanations for\nAlphaFold's predictions, providing near-experimental understanding of the\neffects of amino acids on 3D protein structure. This framework has the\npotential to facilitate a deeper understanding of protein structures.\n", "date": "2023-01-27", "categories": []}
{"id": "2306.05813", "title": "Incorporating Prior Knowledge in Deep Learning Models via Pathway\n  Activity Autoencoders", "abstract": "  Motivation: Despite advances in the computational analysis of high-throughput\nmolecular profiling assays (e.g. transcriptomics), a dichotomy exists between\nmethods that are simple and interpretable, and ones that are complex but with\nlower degree of interpretability. Furthermore, very few methods deal with\ntrying to translate interpretability in biologically relevant terms, such as\nknown pathway cascades. Biological pathways reflecting signalling events or\nmetabolic conversions are Small improvements or modifications of existing\nalgorithms will generally not be suitable, unless novel biological results have\nbeen predicted and verified. Determining which pathways are implicated in\ndisease and incorporating such pathway data as prior knowledge may enhance\npredictive modelling and personalised strategies for diagnosis, treatment and\nprevention of disease.\n  Results: We propose a novel prior-knowledge-based deep auto-encoding\nframework, PAAE, together with its accompanying generative variant, PAVAE, for\nRNA-seq data in cancer. Through comprehensive comparisons among various\nlearning models, we show that, despite having access to a smaller set of\nfeatures, our PAAE and PAVAE models achieve better out-of-set reconstruction\nresults compared to common methodologies. Furthermore, we compare our model\nwith equivalent baselines on a classification task and show that they achieve\nbetter results than models which have access to the full input gene set.\nAnother result is that using vanilla variational frameworks might negatively\nimpact both reconstruction outputs as well as classification performance.\nFinally, our work directly contributes by providing comprehensive\ninterpretability analyses on our models on top of improving prognostication for\ntranslational medicine.\n", "date": "2023-06-09", "categories": []}
{"id": "2306.04658", "title": "Mathematics-assisted directed evolution and protein engineering", "abstract": "  Directed evolution is a molecular biology technique that is transforming\nprotein engineering by creating proteins with desirable properties and\nfunctions. However, it is experimentally impossible to perform the deep\nmutational scanning of the entire protein library due to the enormous\nmutational space, which scales as $20^N$ , where N is the number of amino\nacids. This has led to the rapid growth of AI-assisted directed evolution\n(AIDE) or AI-assisted protein engineering (AIPE) as an emerging research field.\nAided with advanced natural language processing (NLP) techniques, including\nlong short-term memory, autoencoder, and transformer, sequence-based embeddings\nhave been dominant approaches in AIDE and AIPE. Persistent Laplacians, an\nemerging technique in topological data analysis (TDA), have made\nstructure-based embeddings a superb option in AIDE and AIPE. We argue that a\nclass of persistent topological Laplacians (PTLs), including persistent\nLaplacians, persistent path Laplacians, persistent sheaf Laplacians, persistent\nhypergraph Laplacians, persistent hyperdigraph Laplacians, and evolutionary de\nRham-Hodge theory, can effectively overcome the limitations of the current TDA\nand offer a new generation of more powerful TDA approaches. In the general\nframework of topological deep learning, mathematics-assisted directed evolution\n(MADE) has a great potential for future protein engineering.\n", "date": "2023-06-06", "categories": []}
{"id": "2306.03228", "title": "Discovering Novel Biological Traits From Images Using Phylogeny-Guided\n  Neural Networks", "abstract": "  Discovering evolutionary traits that are heritable across species on the tree\nof life (also referred to as a phylogenetic tree) is of great interest to\nbiologists to understand how organisms diversify and evolve. However, the\nmeasurement of traits is often a subjective and labor-intensive process, making\ntrait discovery a highly label-scarce problem. We present a novel approach for\ndiscovering evolutionary traits directly from images without relying on trait\nlabels. Our proposed approach, Phylo-NN, encodes the image of an organism into\na sequence of quantized feature vectors -- or codes -- where different segments\nof the sequence capture evolutionary signals at varying ancestry levels in the\nphylogeny. We demonstrate the effectiveness of our approach in producing\nbiologically meaningful results in a number of downstream tasks including\nspecies image generation and species-to-species image translation, using fish\nspecies as a target example.\n", "date": "2023-06-05", "categories": []}
{"id": "2306.01824", "title": "Enhancing the Protein Tertiary Structure Prediction by Multiple Sequence\n  Alignment Generation", "abstract": "  The field of protein folding research has been greatly advanced by deep\nlearning methods, with AlphaFold2 (AF2) demonstrating exceptional performance\nand atomic-level precision. As co-evolution is integral to protein structure\nprediction, AF2's accuracy is significantly influenced by the depth of multiple\nsequence alignment (MSA), which requires extensive exploration of a large\nprotein database for similar sequences. However, not all protein sequences\npossess abundant homologous families, and consequently, AF2's performance can\ndegrade on such queries, at times failing to produce meaningful results. To\naddress this, we introduce a novel generative language model, MSA-Augmenter,\nwhich leverages protein-specific attention mechanisms and large-scale MSAs to\ngenerate useful, novel protein sequences not currently found in databases.\nThese sequences supplement shallow MSAs, enhancing the accuracy of structural\nproperty predictions. Our experiments on CASP14 demonstrate that MSA-Augmenter\ncan generate de novo sequences that retain co-evolutionary information from\ninferior MSAs, thereby improving protein structure prediction quality on top of\nstrong AF2.\n", "date": "2023-06-02", "categories": []}
{"id": "2306.02532", "title": "R-Mixup: Riemannian Mixup for Biological Networks", "abstract": "  Biological networks are commonly used in biomedical and healthcare domains to\neffectively model the structure of complex biological systems with interactions\nlinking biological entities. However, due to their characteristics of high\ndimensionality and low sample size, directly applying deep learning models on\nbiological networks usually faces severe overfitting. In this work, we propose\nR-MIXUP, a Mixup-based data augmentation technique that suits the symmetric\npositive definite (SPD) property of adjacency matrices from biological networks\nwith optimized training efficiency. The interpolation process in R-MIXUP\nleverages the log-Euclidean distance metrics from the Riemannian manifold,\neffectively addressing the swelling effect and arbitrarily incorrect label\nissues of vanilla Mixup. We demonstrate the effectiveness of R-MIXUP with five\nreal-world biological network datasets on both regression and classification\ntasks. Besides, we derive a commonly ignored necessary condition for\nidentifying the SPD matrices of biological networks and empirically study its\ninfluence on the model performance. The code implementation can be found in\nAppendix E.\n", "date": "2023-06-05", "categories": []}
{"id": "2306.01005", "title": "AbODE: Ab Initio Antibody Design using Conjoined ODEs", "abstract": "  Antibodies are Y-shaped proteins that neutralize pathogens and constitute the\ncore of our adaptive immune system. De novo generation of new antibodies that\ntarget specific antigens holds the key to accelerating vaccine discovery.\nHowever, this co-design of the amino acid sequence and the 3D structure\nsubsumes and accentuates some central challenges from multiple tasks, including\nprotein folding (sequence to structure), inverse folding (structure to\nsequence), and docking (binding). We strive to surmount these challenges with a\nnew generative model AbODE that extends graph PDEs to accommodate both\ncontextual information and external interactions. Unlike existing approaches,\nAbODE uses a single round of full-shot decoding and elicits continuous\ndifferential attention that encapsulates and evolves with latent interactions\nwithin the antibody as well as those involving the antigen. We unravel\nfundamental connections between AbODE and temporal networks as well as\ngraph-matching networks. The proposed model significantly outperforms existing\nmethods on standard metrics across benchmarks.\n", "date": "2023-05-31", "categories": []}
{"id": "2210.06069", "title": "E3Bind: An End-to-End Equivariant Network for Protein-Ligand Docking", "abstract": "  In silico prediction of the ligand binding pose to a given protein target is\na crucial but challenging task in drug discovery. This work focuses on blind\nflexible selfdocking, where we aim to predict the positions, orientations and\nconformations of docked molecules. Traditional physics-based methods usually\nsuffer from inaccurate scoring functions and high inference cost. Recently,\ndata-driven methods based on deep learning techniques are attracting growing\ninterest thanks to their efficiency during inference and promising performance.\nThese methods usually either adopt a two-stage approach by first predicting the\ndistances between proteins and ligands and then generating the final\ncoordinates based on the predicted distances, or directly predicting the global\nroto-translation of ligands. In this paper, we take a different route. Inspired\nby the resounding success of AlphaFold2 for protein structure prediction, we\npropose E3Bind, an end-to-end equivariant network that iteratively updates the\nligand pose. E3Bind models the protein-ligand interaction through careful\nconsideration of the geometric constraints in docking and the local context of\nthe binding site. Experiments on standard benchmark datasets demonstrate the\nsuperior performance of our end-to-end trainable model compared to traditional\nand recently-proposed deep learning methods.\n", "date": "2022-10-12", "categories": []}
{"id": "2302.00203", "title": "End-to-End Full-Atom Antibody Design", "abstract": "  Antibody design is an essential yet challenging task in various domains like\ntherapeutics and biology. There are two major defects in current learning-based\nmethods: 1) tackling only a certain subtask of the whole antibody design\npipeline, making them suboptimal or resource-intensive. 2) omitting either the\nframework regions or side chains, thus incapable of capturing the full-atom\ngeometry. To address these pitfalls, we propose dynamic Multi-channel\nEquivariant grAph Network (dyMEAN), an end-to-end full-atom model for\nE(3)-equivariant antibody design given the epitope and the incomplete sequence\nof the antibody. Specifically, we first explore structural initialization as a\nknowledgeable guess of the antibody structure and then propose shadow paratope\nto bridge the epitope-antibody connections. Both 1D sequences and 3D structures\nare updated via an adaptive multi-channel equivariant encoder that is able to\nprocess protein residues of variable sizes when considering full atoms.\nFinally, the updated antibody is docked to the epitope via the alignment of the\nshadow paratope. Experiments on epitope-binding CDR-H3 design, complex\nstructure prediction, and affinity optimization demonstrate the superiority of\nour end-to-end framework and full-atom modeling.\n", "date": "2023-02-01", "categories": []}
{"id": "2305.15151", "title": "Knowledge-Design: Pushing the Limit of Protein Design via Knowledge\n  Refinement", "abstract": "  Recent studies have shown competitive performance in protein design that aims\nto find the amino acid sequence folding into the desired structure. However,\nmost of them disregard the importance of predictive confidence, fail to cover\nthe vast protein space, and do not incorporate common protein knowledge. After\nwitnessing the great success of pretrained models on diverse protein-related\ntasks and the fact that recovery is highly correlated with confidence, we\nwonder whether this knowledge can push the limits of protein design further. As\na solution, we propose a knowledge-aware module that refines low-quality\nresidues. We also introduce a memory-retrieval mechanism to save more than 50\\%\nof the training time. We extensively evaluate our proposed method on the CATH,\nTS50, and TS500 datasets and our results show that our Knowledge-Design method\noutperforms the previous PiFold method by approximately 9\\% on the CATH\ndataset. Specifically, Knowledge-Design is the first method that achieves 60+\\%\nrecovery on CATH, TS50 and TS500 benchmarks. We also provide additional\nanalysis to demonstrate the effectiveness of our proposed method. The code will\nbe publicly available.\n", "date": "2023-05-20", "categories": []}
{"id": "2208.01439", "title": "Unsupervised machine learning framework for discriminating major\n  variants of concern during COVID-19", "abstract": "  Due to the high mutation rate of the virus, the COVID-19 pandemic evolved\nrapidly. Certain variants of the virus, such as Delta and Omicron, emerged with\naltered viral properties leading to severe transmission and death rates. These\nvariants burdened the medical systems worldwide with a major impact to travel,\nproductivity, and the world economy. Unsupervised machine learning methods have\nthe ability to compress, characterize, and visualize unlabelled data. This\npaper presents a framework that utilizes unsupervised machine learning methods\nto discriminate and visualize the associations between major COVID-19 variants\nbased on their genome sequences. These methods comprise a combination of\nselected dimensionality reduction and clustering techniques. The framework\nprocesses the RNA sequences by performing a k-mer analysis on the data and\nfurther visualises and compares the results using selected dimensionality\nreduction methods that include principal component analysis (PCA),\nt-distributed stochastic neighbour embedding (t-SNE), and uniform manifold\napproximation projection (UMAP). Our framework also employs agglomerative\nhierarchical clustering to visualize the mutational differences among major\nvariants of concern and country-wise mutational differences for selected\nvariants (Delta and Omicron) using dendrograms. We also provide country-wise\nmutational differences for selected variants via dendrograms. We find that the\nproposed framework can effectively distinguish between the major variants and\nhas the potential to identify emerging variants in the future.\n", "date": "2022-08-01", "categories": []}
{"id": "2305.15441", "title": "Improving few-shot learning-based protein engineering with evolutionary\n  sampling", "abstract": "  Designing novel functional proteins remains a slow and expensive process due\nto a variety of protein engineering challenges; in particular, the number of\nprotein variants that can be experimentally tested in a given assay pales in\ncomparison to the vastness of the overall sequence space, resulting in low hit\nrates and expensive wet lab testing cycles. In this paper, we propose a\nfew-shot learning approach to novel protein design that aims to accelerate the\nexpensive wet lab testing cycle and is capable of leveraging a training dataset\nthat is both small and skewed ($\\approx 10^5$ datapoints, $< 1\\%$ positive\nhits). Our approach is composed of two parts: a semi-supervised transfer\nlearning approach to generate a discrete fitness landscape for a desired\nprotein function and a novel evolutionary Monte Carlo Markov Chain sampling\nalgorithm to more efficiently explore the fitness landscape. We demonstrate the\nperformance of our approach by experimentally screening predicted high fitness\ngene activators, resulting in a dramatically improved hit rate compared to\nexisting methods. Our method can be easily adapted to other protein engineering\nand design problems, particularly where the cost associated with obtaining\nlabeled data is significantly high. We have provided open source code for our\nmethod at https://\ngithub.com/SuperSecretBioTech/evolutionary_monte_carlo_search.\n", "date": "2023-05-23", "categories": []}
{"id": "2303.04691", "title": "Self-contained Beta-with-Spikes Approximation for Inference Under a\n  Wright-Fisher Model", "abstract": "  We construct a reliable estimation of evolutionary parameters within the\nWright-Fisher model, which describes changes in allele frequencies due to\nselection and genetic drift, from time-series data. Such data exists for\nbiological populations, for example via artificial evolution experiments, and\nfor the cultural evolution of behavior, such as linguistic corpora that\ndocument historical usage of different words with similar meanings. Our method\nof analysis builds on a Beta-with-Spikes approximation to the distribution of\nallele frequencies predicted by the Wright-Fisher model. We introduce a\nself-contained scheme for estimating the parameters in the approximation, and\ndemonstrate its robustness with synthetic data, especially in the\nstrong-selection and near-extinction regimes where previous approaches fail. We\nfurther apply to allele frequency data for baker's yeast (Saccharomyces\ncerevisiae), finding a significant signal of selection in cases where\nindependent evidence supports such a conclusion. We further demonstrate the\npossibility of detecting time-points at which evolutionary parameters change in\nthe context of a historical spelling reform in the Spanish language.\n", "date": "2023-03-08", "categories": []}
{"id": "2305.15156", "title": "SyNDock: N Rigid Protein Docking via Learnable Group Synchronization", "abstract": "  The regulation of various cellular processes heavily relies on the protein\ncomplexes within a living cell, necessitating a comprehensive understanding of\ntheir three-dimensional structures to elucidate the underlying mechanisms.\nWhile neural docking techniques have exhibited promising outcomes in binary\nprotein docking, the application of advanced neural architectures to multimeric\nprotein docking remains uncertain. This study introduces SyNDock, an automated\nframework that swiftly assembles precise multimeric complexes within seconds,\nshowcasing performance that can potentially surpass or be on par with recent\nadvanced approaches. SyNDock possesses several appealing advantages not present\nin previous approaches. Firstly, SyNDock formulates multimeric protein docking\nas a problem of learning global transformations to holistically depict the\nplacement of chain units of a complex, enabling a learning-centric solution.\nSecondly, SyNDock proposes a trainable two-step SE(3) algorithm, involving\ninitial pairwise transformation and confidence estimation, followed by global\ntransformation synchronization. This enables effective learning for assembling\nthe complex in a globally consistent manner. Lastly, extensive experiments\nconducted on our proposed benchmark dataset demonstrate that SyNDock\noutperforms existing docking software in crucial performance metrics, including\naccuracy and runtime. For instance, it achieves a 4.5% improvement in\nperformance and a remarkable millionfold acceleration in speed.\n", "date": "2023-05-23", "categories": []}
{"id": "2302.02277", "title": "SE(3) diffusion model with application to protein backbone generation", "abstract": "  The design of novel protein structures remains a challenge in protein\nengineering for applications across biomedicine and chemistry. In this line of\nwork, a diffusion model over rigid bodies in 3D (referred to as frames) has\nshown success in generating novel, functional protein backbones that have not\nbeen observed in nature. However, there exists no principled methodological\nframework for diffusion on SE(3), the space of orientation preserving rigid\nmotions in R3, that operates on frames and confers the group invariance. We\naddress these shortcomings by developing theoretical foundations of SE(3)\ninvariant diffusion models on multiple frames followed by a novel framework,\nFrameDiff, for learning the SE(3) equivariant score over multiple frames. We\napply FrameDiff on monomer backbone generation and find it can generate\ndesignable monomers up to 500 amino acids without relying on a pretrained\nprotein structure prediction network that has been integral to previous\nmethods. We find our samples are capable of generalizing beyond any known\nprotein structure.\n", "date": "2023-02-05", "categories": []}
{"id": "2305.12618", "title": "Atomic and Subgraph-aware Bilateral Aggregation for Molecular\n  Representation Learning", "abstract": "  Molecular representation learning is a crucial task in predicting molecular\nproperties. Molecules are often modeled as graphs where atoms and chemical\nbonds are represented as nodes and edges, respectively, and Graph Neural\nNetworks (GNNs) have been commonly utilized to predict atom-related properties,\nsuch as reactivity and solubility. However, functional groups (subgraphs) are\nclosely related to some chemical properties of molecules, such as efficacy, and\nmetabolic properties, which cannot be solely determined by individual atoms. In\nthis paper, we introduce a new model for molecular representation learning\ncalled the Atomic and Subgraph-aware Bilateral Aggregation (ASBA), which\naddresses the limitations of previous atom-wise and subgraph-wise models by\nincorporating both types of information. ASBA consists of two branches, one for\natom-wise information and the other for subgraph-wise information. Considering\nexisting atom-wise GNNs cannot properly extract invariant subgraph features, we\npropose a decomposition-polymerization GNN architecture for the subgraph-wise\nbranch. Furthermore, we propose cooperative node-level and graph-level\nself-supervised learning strategies for ASBA to improve its generalization. Our\nmethod offers a more comprehensive way to learn representations for molecular\nproperty prediction and has broad potential in drug and material discovery\napplications. Extensive experiments have demonstrated the effectiveness of our\nmethod.\n", "date": "2023-05-22", "categories": []}
{"id": "2305.11194", "title": "Vaxformer: Antigenicity-controlled Transformer for Vaccine Design\n  Against SARS-CoV-2", "abstract": "  The SARS-CoV-2 pandemic has emphasised the importance of developing a\nuniversal vaccine that can protect against current and future variants of the\nvirus. The present study proposes a novel conditional protein Language Model\narchitecture, called Vaxformer, which is designed to produce natural-looking\nantigenicity-controlled SARS-CoV-2 spike proteins. We evaluate the generated\nprotein sequences of the Vaxformer model using DDGun protein stability measure,\nnetMHCpan antigenicity score, and a structure fidelity score with AlphaFold to\ngauge its viability for vaccine development. Our results show that Vaxformer\noutperforms the existing state-of-the-art Conditional Variational Autoencoder\nmodel to generate antigenicity-controlled SARS-CoV-2 spike proteins. These\nfindings suggest promising opportunities for conditional Transformer models to\nexpand our understanding of vaccine design and their role in mitigating global\nhealth challenges. The code used in this study is available at\nhttps://github.com/aryopg/vaxformer .\n", "date": "2023-05-18", "categories": []}
{"id": "2302.02947", "title": "GPS++: Reviving the Art of Message Passing for Molecular Property\n  Prediction", "abstract": "  We present GPS++, a hybrid Message Passing Neural Network / Graph Transformer\nmodel for molecular property prediction. Our model integrates a well-tuned\nlocal message passing component and biased global attention with other key\nideas from prior literature to achieve state-of-the-art results on large-scale\nmolecular dataset PCQM4Mv2. Through a thorough ablation study we highlight the\nimpact of individual components and find that nearly all of the model's\nperformance can be maintained without any use of global self-attention, showing\nthat message passing is still a competitive approach for 3D molecular property\nprediction despite the recent dominance of graph transformers. We also find\nthat our approach is significantly more accurate than prior art when 3D\npositional information is not available.\n", "date": "2023-02-06", "categories": []}
{"id": "2305.06334", "title": "Optimizing Drug Design by Merging Generative AI With Active Learning\n  Frameworks", "abstract": "  Traditional drug discovery programs are being transformed by the advent of\nmachine learning methods. Among these, Generative AI methods (GM) have gained\nattention due to their ability to design new molecules and enhance specific\nproperties of existing ones. However, current GM methods have limitations, such\nas low affinity towards the target, unknown ADME/PK properties, or the lack of\nsynthetic tractability. To improve the applicability domain of GM methods, we\nhave developed a workflow based on a variational autoencoder coupled with\nactive learning steps. The designed GM workflow iteratively learns from\nmolecular metrics, including drug likeliness, synthesizability, similarity, and\ndocking scores. In addition, we also included a hierarchical set of criteria\nbased on advanced molecular modeling simulations during a final selection step.\nWe tested our GM workflow on two model systems, CDK2 and KRAS. In both cases,\nour model generated chemically viable molecules with a high predicted affinity\ntoward the targets. Particularly, the proportion of high-affinity molecules\ninferred by our GM workflow was significantly greater than that in the training\ndata. Notably, we also uncovered novel scaffolds significantly dissimilar to\nthose known for each target. These results highlight the potential of our GM\nworkflow to explore novel chemical space for specific targets, thereby opening\nup new possibilities for drug discovery endeavors.\n", "date": "2023-05-04", "categories": []}
{"id": "2208.11030", "title": "Link prediction with continuous-time classical and quantum walks", "abstract": "  Protein-protein interaction (PPI) networks consist of the physical and/or\nfunctional interactions between the proteins of an organism. Since the\nbiophysical and high-throughput methods used to form PPI networks are\nexpensive, time-consuming, and often contain inaccuracies, the resulting\nnetworks are usually incomplete. In order to infer missing interactions in\nthese networks, we propose a novel class of link prediction methods based on\ncontinuous-time classical and quantum random walks. In the case of quantum\nwalks, we examine the usage of both the network adjacency and Laplacian\nmatrices for controlling the walk dynamics. We define a score function based on\nthe corresponding transition probabilities and perform tests on four real-world\nPPI datasets. Our results show that continuous-time classical random walks and\nquantum walks using the network adjacency matrix can successfully predict\nmissing protein-protein interactions, with performance rivalling the state of\nthe art.\n", "date": "2022-08-23", "categories": []}
{"id": "2305.04883", "title": "Fuzzy Gene Selection and Cancer Classification Based on Deep Learning\n  Model", "abstract": "  Machine learning (ML) approaches have been used to develop highly accurate\nand efficient applications in many fields including bio-medical science.\nHowever, even with advanced ML techniques, cancer classification using gene\nexpression data is still complicated because of the high dimensionality of the\ndatasets employed. We developed a new fuzzy gene selection technique (FGS) to\nidentify informative genes to facilitate cancer classification and reduce the\ndimensionality of the available gene expression data. Three feature selection\nmethods (Mutual Information, F-ClassIf, and Chi-squared) were evaluated and\nemployed to obtain the score and rank for each gene. Then, using Fuzzification\nand Defuzzification methods to obtain the best single score for each gene,\nwhich aids in the identification of significant genes. Our study applied the\nfuzzy measures to six gene expression datasets including four Microarray and\ntwo RNA-seq datasets for evaluating the proposed algorithm. With our\nFGS-enhanced method, the cancer classification model achieved 96.5%,96.2%,96%,\nand 95.9% for accuracy, precision, recall, and f1-score respectively, which is\nsignificantly higher than 69.2% accuracy, 57.8% precision, 66% recall, and\n58.2% f1-score when the standard MLP method was used. In examining the six\ndatasets that were used, the proposed model demonstrates it's capacity to\nclassify cancer effectively.\n", "date": "2023-05-04", "categories": []}
{"id": "2305.04827", "title": "Modeling glycemia in humans by means of Grammatical Evolution", "abstract": "  Diabetes mellitus is a disease that affects to hundreds of millions of people\nworldwide. Maintaining a good control of the disease is critical to avoid\nsevere long-term complications. In recent years, several artificial pancreas\nsystems have been proposed and developed, which are increasingly advanced.\nHowever there is still a lot of research to do. One of the main problems that\narises in the (semi) automatic control of diabetes, is to get a model\nexplaining how glycemia (glucose levels in blood) varies with insulin, food\nintakes and other factors, fitting the characteristics of each individual or\npatient. This paper proposes the application of evolutionary computation\ntechniques to obtain customized models of patients, unlike most of previous\napproaches which obtain averaged models. The proposal is based on a kind of\ngenetic programming based on grammars known as Grammatical Evolution (GE). The\nproposal has been tested with in-silico patient data and results are clearly\npositive. We present also a study of four different grammars and five objective\nfunctions. In the test phase the models characterized the glucose with a mean\npercentage average error of 13.69\\%, modeling well also both hyper and\nhypoglycemic situations.\n", "date": "2023-04-27", "categories": []}
{"id": "2212.00735", "title": "xTrimoABFold: De novo Antibody Structure Prediction without MSA", "abstract": "  In the field of antibody engineering, an essential task is to design a novel\nantibody whose paratopes bind to a specific antigen with correct epitopes.\nUnderstanding antibody structure and its paratope can facilitate a mechanistic\nunderstanding of its function. Therefore, antibody structure prediction from\nits sequence alone has always been a highly valuable problem for de novo\nantibody design. AlphaFold2, a breakthrough in the field of structural biology,\nprovides a solution to predict protein structure based on protein sequences and\ncomputationally expensive coevolutionary multiple sequence alignments (MSAs).\nHowever, the computational efficiency and undesirable prediction accuracy of\nantibodies, especially on the complementarity-determining regions (CDRs) of\nantibodies limit their applications in the industrially high-throughput drug\ndesign. To learn an informative representation of antibodies, we employed a\ndeep antibody language model (ALM) on curated sequences from the observed\nantibody space database via a transformer model. We also developed a novel\nmodel named xTrimoABFold to predict antibody structure from antibody sequence\nbased on the pretrained ALM as well as efficient evoformers and structural\nmodules. The model was trained end-to-end on the antibody structures in PDB by\nminimizing the ensemble loss of domain-specific focal loss on CDR and the\nframe-aligned point loss. xTrimoABFold outperforms AlphaFold2 and other protein\nlanguage model based SOTAs, e.g., OmegaFold, HelixFold-Single, and IgFold with\na large significant margin (30+\\% improvement on RMSD) while performing 151\ntimes faster than AlphaFold2. To the best of our knowledge, xTrimoABFold\nachieved state-of-the-art antibody structure prediction. Its improvement in\nboth accuracy and efficiency makes it a valuable tool for de novo antibody\ndesign and could make further improvements in immuno-theory.\n", "date": "2022-11-30", "categories": []}
{"id": "2301.06454", "title": "PlasmoFAB: A Benchmark to Foster Machine Learning for Plasmodium\n  falciparum Protein Antigen Candidate Prediction", "abstract": "  Motivation: Machine learning methods can be used to support scientific\ndiscovery in healthcare-related research fields. However, these methods can\nonly be reliably used if they can be trained on high-quality and curated\ndatasets. Currently, no such dataset for the exploration of Plasmodium\nfalciparum protein antigen candidates exists. The parasite Plasmodium\nfalciparum causes the infectious disease malaria. Thus, identifying potential\nantigens is of utmost importance for the development of antimalarial drugs and\nvaccines. Since exploring antigen candidates experimentally is an expensive and\ntime-consuming process, applying machine learning methods to support this\nprocess has the potential to accelerate the development of drugs and vaccines,\nwhich are needed for fighting and controlling malaria. Results: We developed\nPlasmoFAB, a curated benchmark that can be used to train machine learning\nmethods for the exploration of Plasmodium falciparum protein antigen\ncandidates. We combined an extensive literature search with domain expertise to\ncreate high-quality labels for Plasmodium falciparum specific proteins that\ndistinguish between antigen candidates and intracellular proteins.\nAdditionally, we used our benchmark to compare different well-known prediction\nmodels and available protein localization prediction services on the task of\nidentifying protein antigen candidates. We show that available general-purpose\nservices are unable to provide sufficient performance on identifying protein\nantigen candidates and are outperformed by our models that were trained on this\ntailored data. Availability: PlasmoFAB is publicly available on Zenodo with DOI\n10.5281/zenodo.7433087. Furthermore, all scripts that were used in the creation\nof PlasmoFAB and the training and evaluation of machine learning models are\nopen source and publicly available on GitHub here:\nhttps://github.com/msmdev/PlasmoFAB.\n", "date": "2023-01-16", "categories": []}
{"id": "2305.01580", "title": "Molecular design method based on novel molecular representation and\n  variational auto-encoder", "abstract": "  Based on the traditional VAE, a novel neural network model is presented, with\nthe latest molecular representation, SELFIES, to improve the effect of\ngenerating new molecules. In this model, multi-layer convolutional network and\nFisher information are added to the original encoding layer to learn the data\ncharacteristics and guide the encoding process, which makes the features of the\ndata hiding layer more aggregated, and integrates the Long Short Term Memory\nneural network (LSTM) into the decoding layer for better data generation, which\neffectively solves the degradation phenomenon generated by the encoding layer\nand decoding layer of the original VAE model. Through experiments on zinc\nmolecular data sets, it is found that the similarity in the new VAE is 8.47%\nhigher than that of the original ones. SELFIES are better at generating a\nvariety of molecules than the traditional molecular representation, SELFIES.\nExperiments have shown that using SELFIES and the new VAE model presented in\nthis paper can improve the effectiveness of generating new molecules.\n", "date": "2023-02-20", "categories": []}
{"id": "2211.02504", "title": "Geometry-Complete Perceptron Networks for 3D Molecular Graphs", "abstract": "  The field of geometric deep learning has had a profound impact on the\ndevelopment of innovative and powerful graph neural network architectures.\nDisciplines such as computer vision and computational biology have benefited\nsignificantly from such methodological advances, which has led to breakthroughs\nin scientific domains such as protein structure prediction and design. In this\nwork, we introduce GCPNet, a new geometry-complete, SE(3)-equivariant graph\nneural network designed for 3D molecular graph representation learning.\nRigorous experiments across four distinct geometric tasks demonstrate that\nGCPNet's predictions (1) for protein-ligand binding affinity achieve a\nstatistically significant correlation of 0.608, more than 5% greater than\ncurrent state-of-the-art methods; (2) for protein structure ranking achieve\nstatistically significant target-local and dataset-global correlations of 0.616\nand 0.871, respectively; (3) for Newtownian many-body systems modeling achieve\na task-averaged mean squared error less than 0.01, more than 15% better than\ncurrent methods; and (4) for molecular chirality recognition achieve a\nstate-of-the-art prediction accuracy of 98.7%, better than any other machine\nlearning method to date. The source code, data, and instructions to train new\nmodels or reproduce our results are freely available at\nhttps://github.com/BioinfoMachineLearning/GCPNet.\n", "date": "2022-11-04", "categories": []}
{"id": "2304.13737", "title": "AIRIVA: A Deep Generative Model of Adaptive Immune Repertoires", "abstract": "  Recent advances in immunomics have shown that T-cell receptor (TCR)\nsignatures can accurately predict active or recent infection by leveraging the\nhigh specificity of TCR binding to disease antigens. However, the extreme\ndiversity of the adaptive immune repertoire presents challenges in reliably\nidentifying disease-specific TCRs. Population genetics and sequencing depth can\nalso have strong systematic effects on repertoires, which requires careful\nconsideration when developing diagnostic models. We present an Adaptive Immune\nRepertoire-Invariant Variational Autoencoder (AIRIVA), a generative model that\nlearns a low-dimensional, interpretable, and compositional representation of\nTCR repertoires to disentangle such systematic effects in repertoires. We apply\nAIRIVA to two infectious disease case-studies: COVID-19 (natural infection and\nvaccination) and the Herpes Simplex Virus (HSV-1 and HSV-2), and empirically\nshow that we can disentangle the individual disease signals. We further\ndemonstrate AIRIVA's capability to: learn from unlabelled samples; generate\nin-silico TCR repertoires by intervening on the latent factors; and identify\ndisease-associated TCRs validated using TCR annotations from external assay\ndata.\n", "date": "2023-04-26", "categories": []}
{"id": "2208.03325", "title": "Isoform Function Prediction Using a Deep Neural Network", "abstract": "  Isoforms are mRNAs produced from the same gene site in the phenomenon called\nAlternative Splicing. Studies have shown that more than 95% of human multi-exon\ngenes have undergone alternative splicing. Although there are few changes in\nmRNA sequence, They may have a systematic effect on cell function and\nregulation. It is widely reported that isoforms of a gene have distinct or even\ncontrasting functions. Most studies have shown that alternative splicing plays\na significant role in human health and disease. Despite the wide range of gene\nfunction studies, there is little information about isoforms' functionalities.\nRecently, some computational methods based on Multiple Instance Learning have\nbeen proposed to predict isoform function using gene function and gene\nexpression profile. However, their performance is not desirable due to the lack\nof labeled training data. In addition, probabilistic models such as Conditional\nRandom Field (CRF) have been used to model the relation between isoforms. This\nproject uses all the data and valuable information such as isoform sequences,\nexpression profiles, and gene ontology graphs and proposes a comprehensive\nmodel based on Deep Neural Networks. The UniProt Gene Ontology (GO) database is\nused as a standard reference for gene functions. The NCBI RefSeq database is\nused for extracting gene and isoform sequences, and the NCBI SRA database is\nused for expression profile data. Metrics such as Receiver Operating\nCharacteristic Area Under the Curve (ROC AUC) and Precision-Recall Under the\nCurve (PR AUC) are used to measure the prediction accuracy.\n", "date": "2022-08-05", "categories": []}
{"id": "2212.01384", "title": "KGML-xDTD: A Knowledge Graph-based Machine Learning Framework for Drug\n  Treatment Prediction and Mechanism Description", "abstract": "  Background: Computational drug repurposing is a cost- and time-efficient\napproach that aims to identify new therapeutic targets or diseases\n(indications) of existing drugs/compounds. It is especially critical for\nemerging and/or orphan diseases due to its cheaper investment and shorter\nresearch cycle compared with traditional wet-lab drug discovery approaches.\nHowever, the underlying mechanisms of action (MOAs) between repurposed drugs\nand their target diseases remain largely unknown, which is still a main\nobstacle for computational drug repurposing methods to be widely adopted in\nclinical settings.\n  Results: In this work, we propose KGML-xDTD: a Knowledge Graph-based Machine\nLearning framework for explainably predicting Drugs Treating Diseases. It is a\ntwo-module framework that not only predicts the treatment probabilities between\ndrugs/compounds and diseases but also biologically explains them via knowledge\ngraph (KG) path-based, testable mechanisms of action (MOAs). We leverage\nknowledge-and-publication based information to extract biologically meaningful\n\"demonstration paths\" as the intermediate guidance in the Graph-based\nReinforcement Learning (GRL) path-finding process. Comprehensive experiments\nand case study analyses show that the proposed framework can achieve\nstate-of-the-art performance in both predictions of drug repurposing and\nrecapitulation of human-curated drug MOA paths.\n  Conclusions: KGML-xDTD is the first model framework that can offer KG-path\nexplanations for drug repurposing predictions by leveraging the combination of\nprediction outcomes and existing biological knowledge and publications. We\nbelieve it can effectively reduce \"black-box\" concerns and increase prediction\nconfidence for drug repurposing based on predicted path-based explanations, and\nfurther accelerate the process of drug discovery for emerging diseases.\n", "date": "2022-11-30", "categories": []}
{"id": "2304.12328", "title": "Virus2Vec: Viral Sequence Classification Using Machine Learning", "abstract": "  Understanding the host-specificity of different families of viruses sheds\nlight on the origin of, e.g., SARS-CoV-2, rabies, and other such zoonotic\npathogens in humans. It enables epidemiologists, medical professionals, and\npolicymakers to curb existing epidemics and prevent future ones promptly. In\nthe family Coronaviridae (of which SARS-CoV-2 is a member), it is well-known\nthat the spike protein is the point of contact between the virus and the host\ncell membrane. On the other hand, the two traditional mammalian orders,\nCarnivora (carnivores) and Chiroptera (bats) are recognized to be responsible\nfor maintaining and spreading the Rabies Lyssavirus (RABV). We propose\nVirus2Vec, a feature-vector representation for viral (nucleotide or amino acid)\nsequences that enable vector-space-based machine learning models to identify\nviral hosts. Virus2Vec generates numerical feature vectors for unaligned\nsequences, allowing us to forego the computationally expensive sequence\nalignment step from the pipeline. Virus2Vec leverages the power of both the\n\\emph{minimizer} and position weight matrix (PWM) to generate compact feature\nvectors. Using several classifiers, we empirically evaluate Virus2Vec on\nreal-world spike sequences of Coronaviridae and rabies virus sequence data to\npredict the host (identifying the reservoirs of infection). Our results\ndemonstrate that Virus2Vec outperforms the predictive accuracies of baseline\nand state-of-the-art methods.\n", "date": "2023-04-24", "categories": []}
{"id": "2208.11126", "title": "Retrieval-based Controllable Molecule Generation", "abstract": "  Generating new molecules with specified chemical and biological properties\nvia generative models has emerged as a promising direction for drug discovery.\nHowever, existing methods require extensive training/fine-tuning with a large\ndataset, often unavailable in real-world generation tasks. In this work, we\npropose a new retrieval-based framework for controllable molecule generation.\nWe use a small set of exemplar molecules, i.e., those that (partially) satisfy\nthe design criteria, to steer the pre-trained generative model towards\nsynthesizing molecules that satisfy the given design criteria. We design a\nretrieval mechanism that retrieves and fuses the exemplar molecules with the\ninput molecule, which is trained by a new self-supervised objective that\npredicts the nearest neighbor of the input molecule. We also propose an\niterative refinement process to dynamically update the generated molecules and\nretrieval database for better generalization. Our approach is agnostic to the\nchoice of generative models and requires no task-specific fine-tuning. On\nvarious tasks ranging from simple design criteria to a challenging real-world\nscenario for designing lead compounds that bind to the SARS-CoV-2 main\nprotease, we demonstrate our approach extrapolates well beyond the retrieval\ndatabase, and achieves better performance and wider applicability than previous\nmethods. Code is available at https://github.com/NVlabs/RetMol.\n", "date": "2022-08-23", "categories": []}
{"id": "2304.10946", "title": "CancerGPT: Few-shot Drug Pair Synergy Prediction using Large Pre-trained\n  Language Models", "abstract": "  Large pre-trained language models (LLMs) have been shown to have significant\npotential in few-shot learning across various fields, even with minimal\ntraining data. However, their ability to generalize to unseen tasks in more\ncomplex fields, such as biology, has yet to be fully evaluated. LLMs can offer\na promising alternative approach for biological inference, particularly in\ncases where structured data and sample size are limited, by extracting prior\nknowledge from text corpora. Our proposed few-shot learning approach uses LLMs\nto predict the synergy of drug pairs in rare tissues that lack structured data\nand features. Our experiments, which involved seven rare tissues from different\ncancer types, demonstrated that the LLM-based prediction model achieved\nsignificant accuracy with very few or zero samples. Our proposed model, the\nCancerGPT (with $\\sim$ 124M parameters), was even comparable to the larger\nfine-tuned GPT-3 model (with $\\sim$ 175B parameters). Our research is the first\nto tackle drug pair synergy prediction in rare tissues with limited data. We\nare also the first to utilize an LLM-based prediction model for biological\nreaction prediction tasks.\n", "date": "2023-04-18", "categories": []}
{"id": "2304.10061", "title": "Scaling the leading accuracy of deep equivariant models to biomolecular\n  simulations of realistic size", "abstract": "  This work brings the leading accuracy, sample efficiency, and robustness of\ndeep equivariant neural networks to the extreme computational scale. This is\nachieved through a combination of innovative model architecture, massive\nparallelization, and models and implementations optimized for efficient GPU\nutilization. The resulting Allegro architecture bridges the accuracy-speed\ntradeoff of atomistic simulations and enables description of dynamics in\nstructures of unprecedented complexity at quantum fidelity. To illustrate the\nscalability of Allegro, we perform nanoseconds-long stable simulations of\nprotein dynamics and scale up to a 44-million atom structure of a complete,\nall-atom, explicitly solvated HIV capsid on the Perlmutter supercomputer. We\ndemonstrate excellent strong scaling up to 100 million atoms and 70% weak\nscaling to 5120 A100 GPUs.\n", "date": "2023-04-20", "categories": []}
{"id": "2206.12747", "title": "HyGNN: Drug-Drug Interaction Prediction via Hypergraph Neural Network", "abstract": "  Drug-Drug Interactions (DDIs) may hamper the functionalities of drugs, and in\nthe worst scenario, they may lead to adverse drug reactions (ADRs). Predicting\nall DDIs is a challenging and critical problem. Most existing computational\nmodels integrate drug-centric information from different sources and leverage\nthem as features in machine learning classifiers to predict DDIs. However,\nthese models have a high chance of failure, especially for the new drugs when\nall the information is not available. This paper proposes a novel Hypergraph\nNeural Network (HyGNN) model based on only the SMILES string of drugs,\navailable for any drug, for the DDI prediction problem. To capture the drug\nsimilarities, we create a hypergraph from drugs' chemical substructures\nextracted from the SMILES strings. Then, we develop HyGNN consisting of a novel\nattention-based hypergraph edge encoder to get the representation of drugs as\nhyperedges and a decoder to predict the interactions between drug pairs.\nFurthermore, we conduct extensive experiments to evaluate our model and compare\nit with several state-of-the-art methods. Experimental results demonstrate that\nour proposed HyGNN model effectively predicts DDIs and impressively outperforms\nthe baselines with a maximum ROC-AUC and PR-AUC of 97.9% and 98.1%,\nrespectively.\n", "date": "2022-06-25", "categories": []}
{"id": "2304.06731", "title": "PCD2Vec: A Poisson Correction Distance-Based Approach for Viral Host\n  Classification", "abstract": "  Coronaviruses are membrane-enveloped, non-segmented positive-strand RNA\nviruses belonging to the Coronaviridae family. Various animal species, mainly\nmammalian and avian, are severely infected by various coronaviruses, causing\nserious concerns like the recent pandemic (COVID-19). Therefore, building a\ndeeper understanding of these viruses is essential to devise prevention and\nmitigation mechanisms. In the Coronavirus genome, an essential structural\nregion is the spike region, and it's responsible for attaching the virus to the\nhost cell membrane. Therefore, the usage of only the spike protein, instead of\nthe full genome, provides most of the essential information for performing\nanalyses such as host classification. In this paper, we propose a novel method\nfor predicting the host specificity of coronaviruses by analyzing spike protein\nsequences from different viral subgenera and species. Our method involves using\nthe Poisson correction distance to generate a distance matrix, followed by\nusing a radial basis function (RBF) kernel and kernel principal component\nanalysis (PCA) to generate a low-dimensional embedding. Finally, we apply\nclassification algorithms to the low-dimensional embedding to generate the\nresulting predictions of the host specificity of coronaviruses. We provide\ntheoretical proofs for the non-negativity, symmetry, and triangle inequality\nproperties of the Poisson correction distance metric, which are important\nproperties in a machine-learning setting. By encoding the spike protein\nstructure and sequences using this comprehensive approach, we aim to uncover\nhidden patterns in the biological sequences to make accurate predictions about\nhost specificity. Finally, our classification results illustrate that our\nmethod can achieve higher predictive accuracy and improve performance over\nexisting baselines.\n", "date": "2023-04-13", "categories": []}
{"id": "2209.12643", "title": "PiFold: Toward effective and efficient protein inverse folding", "abstract": "  How can we design protein sequences folding into the desired structures\neffectively and efficiently? AI methods for structure-based protein design have\nattracted increasing attention in recent years; however, few methods can\nsimultaneously improve the accuracy and efficiency due to the lack of\nexpressive features and autoregressive sequence decoder. To address these\nissues, we propose PiFold, which contains a novel residue featurizer and PiGNN\nlayers to generate protein sequences in a one-shot way with improved recovery.\nExperiments show that PiFold could achieve 51.66\\% recovery on CATH 4.2, while\nthe inference speed is 70 times faster than the autoregressive competitors. In\naddition, PiFold achieves 58.72\\% and 60.42\\% recovery scores on TS50 and\nTS500, respectively. We conduct comprehensive ablation studies to reveal the\nrole of different types of protein features and model designs, inspiring\nfurther simplification and improvement. The PyTorch code is available at\n\\href{https://github.com/A4Bio/PiFold}{GitHub}.\n", "date": "2022-09-22", "categories": []}
{"id": "2304.05542", "title": "CLCLSA: Cross-omics Linked embedding with Contrastive Learning and Self\n  Attention for multi-omics integration with incomplete multi-omics data", "abstract": "  Integration of heterogeneous and high-dimensional multi-omics data is\nbecoming increasingly important in understanding genetic data. Each omics\ntechnique only provides a limited view of the underlying biological process and\nintegrating heterogeneous omics layers simultaneously would lead to a more\ncomprehensive and detailed understanding of diseases and phenotypes. However,\none obstacle faced when performing multi-omics data integration is the\nexistence of unpaired multi-omics data due to instrument sensitivity and cost.\nStudies may fail if certain aspects of the subjects are missing or incomplete.\nIn this paper, we propose a deep learning method for multi-omics integration\nwith incomplete data by Cross-omics Linked unified embedding with Contrastive\nLearning and Self Attention (CLCLSA). Utilizing complete multi-omics data as\nsupervision, the model employs cross-omics autoencoders to learn the feature\nrepresentation across different types of biological data. The multi-omics\ncontrastive learning, which is used to maximize the mutual information between\ndifferent types of omics, is employed before latent feature concatenation. In\naddition, the feature-level self-attention and omics-level self-attention are\nemployed to dynamically identify the most informative features for multi-omics\ndata integration. Extensive experiments were conducted on four public\nmulti-omics datasets. The experimental results indicated that the proposed\nCLCLSA outperformed the state-of-the-art approaches for multi-omics data\nclassification using incomplete multi-omics data.\n", "date": "2023-04-12", "categories": []}
{"id": "2304.05823", "title": "DiscoGen: Learning to Discover Gene Regulatory Networks", "abstract": "  Accurately inferring Gene Regulatory Networks (GRNs) is a critical and\nchallenging task in biology. GRNs model the activatory and inhibitory\ninteractions between genes and are inherently causal in nature. To accurately\nidentify GRNs, perturbational data is required. However, most GRN discovery\nmethods only operate on observational data. Recent advances in neural\nnetwork-based causal discovery methods have significantly improved causal\ndiscovery, including handling interventional data, improvements in performance\nand scalability. However, applying state-of-the-art (SOTA) causal discovery\nmethods in biology poses challenges, such as noisy data and a large number of\nsamples. Thus, adapting the causal discovery methods is necessary to handle\nthese challenges. In this paper, we introduce DiscoGen, a neural network-based\nGRN discovery method that can denoise gene expression measurements and handle\ninterventional data. We demonstrate that our model outperforms SOTA neural\nnetwork-based causal discovery methods.\n", "date": "2023-04-12", "categories": []}
{"id": "2304.04982", "title": "Biological Factor Regulatory Neural Network", "abstract": "  Genes are fundamental for analyzing biological systems and many recent works\nproposed to utilize gene expression for various biological tasks by deep\nlearning models. Despite their promising performance, it is hard for deep\nneural networks to provide biological insights for humans due to their\nblack-box nature. Recently, some works integrated biological knowledge with\nneural networks to improve the transparency and performance of their models.\nHowever, these methods can only incorporate partial biological knowledge,\nleading to suboptimal performance. In this paper, we propose the Biological\nFactor Regulatory Neural Network (BFReg-NN), a generic framework to model\nrelations among biological factors in cell systems. BFReg-NN starts from gene\nexpression data and is capable of merging most existing biological knowledge\ninto the model, including the regulatory relations among genes or proteins\n(e.g., gene regulatory networks (GRN), protein-protein interaction networks\n(PPI)) and the hierarchical relations among genes, proteins and pathways (e.g.,\nseveral genes/proteins are contained in a pathway). Moreover, BFReg-NN also has\nthe ability to provide new biologically meaningful insights because of its\nwhite-box characteristics. Experimental results on different gene\nexpression-based tasks verify the superiority of BFReg-NN compared with\nbaselines. Our case studies also show that the key insights found by BFReg-NN\nare consistent with the biological literature.\n", "date": "2023-04-11", "categories": []}
{"id": "2304.03775", "title": "Biological Sequence Kernels with Guaranteed Flexibility", "abstract": "  Applying machine learning to biological sequences - DNA, RNA and protein -\nhas enormous potential to advance human health, environmental sustainability,\nand fundamental biological understanding. However, many existing machine\nlearning methods are ineffective or unreliable in this problem domain. We study\nthese challenges theoretically, through the lens of kernels. Methods based on\nkernels are ubiquitous: they are used to predict molecular phenotypes, design\nnovel proteins, compare sequence distributions, and more. Many methods that do\nnot use kernels explicitly still rely on them implicitly, including a wide\nvariety of both deep learning and physics-based techniques. While kernels for\nother types of data are well-studied theoretically, the structure of biological\nsequence space (discrete, variable length sequences), as well as biological\nnotions of sequence similarity, present unique mathematical challenges. We\nformally analyze how well kernels for biological sequences can approximate\narbitrary functions on sequence space and how well they can distinguish\ndifferent sequence distributions. In particular, we establish conditions under\nwhich biological sequence kernels are universal, characteristic and metrize the\nspace of distributions. We show that a large number of existing kernel-based\nmachine learning methods for biological sequences fail to meet our conditions\nand can as a consequence fail severely. We develop straightforward and\ncomputationally tractable ways of modifying existing kernels to satisfy our\nconditions, imbuing them with strong guarantees on accuracy and reliability.\nOur proof techniques build on and extend the theory of kernels with discrete\nmasses. We illustrate our theoretical results in simulation and on real\nbiological data sets.\n", "date": "2023-04-06", "categories": []}
{"id": "2304.02891", "title": "ViralVectors: Compact and Scalable Alignment-free Virome Feature\n  Generation", "abstract": "  The amount of sequencing data for SARS-CoV-2 is several orders of magnitude\nlarger than any virus. This will continue to grow geometrically for SARS-CoV-2,\nand other viruses, as many countries heavily finance genomic surveillance\nefforts. Hence, we need methods for processing large amounts of sequence data\nto allow for effective yet timely decision-making. Such data will come from\nheterogeneous sources: aligned, unaligned, or even unassembled raw nucleotide\nor amino acid sequencing reads pertaining to the whole genome or regions (e.g.,\nspike) of interest. In this work, we propose \\emph{ViralVectors}, a compact\nfeature vector generation from virome sequencing data that allows effective\ndownstream analysis. Such generation is based on \\emph{minimizers}, a type of\nlightweight \"signature\" of a sequence, used traditionally in assembly and read\nmapping -- to our knowledge, the first use minimizers in this way. We validate\nour approach on different types of sequencing data: (a) 2.5M SARS-CoV-2 spike\nsequences (to show scalability); (b) 3K Coronaviridae spike sequences (to show\nrobustness to more genomic variability); and (c) 4K raw WGS reads sets taken\nfrom nasal-swab PCR tests (to show the ability to process unassembled reads).\nOur results show that ViralVectors outperforms current benchmarks in most\nclassification and clustering tasks.\n", "date": "2023-04-06", "categories": []}
{"id": "2304.02682", "title": "nD-PDPA: nDimensional Probability Density Profile Analysis", "abstract": "  Despite the recent advances in various Structural Genomics Projects, a large\ngap remains between the number of sequenced and structurally characterized\nproteins. Some reasons for this discrepancy include technical difficulties,\nlabor, and the cost related to determining a structure by experimental methods\nsuch as NMR spectroscopy. Several computational methods have been developed to\nexpand the applicability of NMR spectroscopy by addressing temporal and\neconomical problems more efficiently. While these methods demonstrate\nsuccessful outcomes to solve more challenging and structurally novel proteins,\nthe cost has not been reduced significantly. Probability Density Profile\nAnalysis (PDPA) has been previously introduced by our lab to directly address\nthe economics of structure determination of routine proteins and the\nidentification of novel structures from a minimal set of unassigned NMR data.\n2D-PDPA (in which 2D denotes incorporation of data from two alignment media)\nhas been successful in identifying the structural homolog of an unknown protein\nwithin a library of ~1000 decoy structures. In order to further expand the\nselectivity and sensitivity of PDPA, the incorporation of additional data was\nnecessary. However, the expansion of the original PDPA approach was limited by\nits computational requirements where the inclusion of additional data would\nrender it computationally intractable. Here we present the most recent\ndevelopments of PDPA method (nD-PDPA: n Dimensional Probability Density Profile\nAnalysis) that eliminate 2D-PDPA's computational limitations, and allows\ninclusion of RDC data from multiple vector types in multiple alignment media.\n", "date": "2023-04-05", "categories": []}
{"id": "2304.02697", "title": "Revolutionizing Single Cell Analysis: The Power of Large Language Models\n  for Cell Type Annotation", "abstract": "  In recent years, single cell RNA sequencing has become a widely used\ntechnique to study cellular diversity and function. However, accurately\nannotating cell types from single cell data has been a challenging task, as it\nrequires extensive knowledge of cell biology and gene function. The emergence\nof large language models such as ChatGPT and New Bing in 2023 has\nrevolutionized this process by integrating the scientific literature and\nproviding accurate annotations of cell types. This breakthrough enables\nresearchers to conduct literature reviews more efficiently and accurately, and\ncan potentially uncover new insights into cell type annotation. By using\nChatGPT to annotate single cell data, we can relate rare cell type to their\nfunction and reveal specific differentiation trajectories of cell subtypes that\nwere previously overlooked. This can have important applications in\nunderstanding cancer progression, mammalian development, and stem cell\ndifferentiation, and can potentially lead to the discovery of key cells that\ninterrupt the differentiation pathway and solve key problems in the life\nsciences. Overall, the future of cell type annotation in single cell data looks\npromising and the Large Language model will be an important milestone in the\nhistory of single cell analysis.\n", "date": "2023-04-05", "categories": []}
{"id": "2304.02729", "title": "Constructing Phylogenetic Networks via Cherry Picking and Machine\n  Learning", "abstract": "  Combining a set of phylogenetic trees into a single phylogenetic network that\nexplains all of them is a fundamental challenge in evolutionary studies.\nExisting methods are computationally expensive and can either handle only small\nnumbers of phylogenetic trees or are limited to severely restricted classes of\nnetworks. In this paper, we apply the recently-introduced theoretical framework\nof cherry picking to design a class of efficient heuristics that are guaranteed\nto produce a network containing each of the input trees, for datasets\nconsisting of binary trees. Some of the heuristics in this framework are based\non the design and training of a machine learning model that captures essential\ninformation on the structure of the input trees and guides the algorithms\ntowards better solutions. We also propose simple and fast randomised heuristics\nthat prove to be very effective when run multiple times.\n  Unlike the existing exact methods, our heuristics are applicable to datasets\nof practical size, and the experimental study we conducted on both simulated\nand real data shows that these solutions are qualitatively good, always within\nsome small constant factor from the optimum. Moreover, our machine-learned\nheuristics are one of the first applications of machine learning to\nphylogenetics and show its promise.\n", "date": "2023-03-31", "categories": []}
{"id": "2304.03239", "title": "Assessing the Reproducibility of Machine-learning-based Biomarker\n  Discovery in Parkinson's Disease", "abstract": "  Genome-Wide Association Studies (GWAS) help identify genetic variations in\npeople with diseases such as Parkinson's disease (PD), which are less common in\nthose without the disease. Thus, GWAS data can be used to identify genetic\nvariations associated with the disease. Feature selection and machine learning\napproaches can be used to analyze GWAS data and identify potential disease\nbiomarkers. However, GWAS studies have technical variations that affect the\nreproducibility of identified biomarkers, such as differences in genotyping\nplatforms and selection criteria for individuals to be genotyped. To address\nthis issue, we collected five GWAS datasets from the database of Genotypes and\nPhenotypes (dbGaP) and explored several data integration strategies. We\nevaluated the agreement among different strategies in terms of the Single\nNucleotide Polymorphisms (SNPs) that were identified as potential PD\nbiomarkers. Our results showed a low concordance of biomarkers discovered using\ndifferent datasets or integration strategies. However, we identified fifty SNPs\nthat were identified at least twice, which could potentially serve as novel PD\nbiomarkers. These SNPs are indirectly linked to PD in the literature but have\nnot been directly associated with PD before. These findings open up new\npotential avenues of investigation.\n", "date": "2023-04-06", "categories": []}
{"id": "2304.02198", "title": "EigenFold: Generative Protein Structure Prediction with Diffusion Models", "abstract": "  Protein structure prediction has reached revolutionary levels of accuracy on\nsingle structures, yet distributional modeling paradigms are needed to capture\nthe conformational ensembles and flexibility that underlie biological function.\nTowards this goal, we develop EigenFold, a diffusion generative modeling\nframework for sampling a distribution of structures from a given protein\nsequence. We define a diffusion process that models the structure as a system\nof harmonic oscillators and which naturally induces a cascading-resolution\ngenerative process along the eigenmodes of the system. On recent CAMEO targets,\nEigenFold achieves a median TMScore of 0.84, while providing a more\ncomprehensive picture of model uncertainty via the ensemble of sampled\nstructures relative to existing methods. We then assess EigenFold's ability to\nmodel and predict conformational heterogeneity for fold-switching proteins and\nligand-induced conformational change. Code is available at\nhttps://github.com/bjing2016/EigenFold.\n", "date": "2023-04-05", "categories": []}
{"id": "2304.01565", "title": "A Survey on Graph Diffusion Models: Generative AI in Science for\n  Molecule, Protein and Material", "abstract": "  Diffusion models have become a new SOTA generative modeling method in various\nfields, for which there are multiple survey works that provide an overall\nsurvey. With the number of articles on diffusion models increasing\nexponentially in the past few years, there is an increasing need for surveys of\ndiffusion models on specific fields. In this work, we are committed to\nconducting a survey on the graph diffusion models. Even though our focus is to\ncover the progress of diffusion models in graphs, we first briefly summarize\nhow other generative modeling methods are used for graphs. After that, we\nintroduce the mechanism of diffusion models in various forms, which facilitates\nthe discussion on the graph diffusion models. The applications of graph\ndiffusion models mainly fall into the category of AI-generated content (AIGC)\nin science, for which we mainly focus on how graph diffusion models are\nutilized for generating molecules and proteins but also cover other cases,\nincluding materials design. Moreover, we discuss the issue of evaluating\ndiffusion models in the graph domain and the existing challenges.\n", "date": "2023-04-04", "categories": []}
{"id": "2209.08171", "title": "Deep learning for reconstructing protein structures from cryo-EM density\n  maps: recent advances and future directions", "abstract": "  Cryo-Electron Microscopy (cryo-EM) has emerged as a key technology to\ndetermine the structure of proteins, particularly large protein complexes and\nassemblies in recent years. A key challenge in cryo-EM data analysis is to\nautomatically reconstruct accurate protein structures from cryo-EM density\nmaps. In this review, we briefly overview various deep learning methods for\nbuilding protein structures from cryo-EM density maps, analyze their impact,\nand discuss the challenges of preparing high-quality data sets for training\ndeep learning models. Looking into the future, more advanced deep learning\nmodels of effectively integrating cryo-EM data with other sources of\ncomplementary data such as protein sequences and AlphaFold-predicted structures\nneed to be developed to further advance the field.\n", "date": "2022-09-16", "categories": []}
{"id": "2303.18017", "title": "Rapid prediction of lab-grown tissue properties using deep learning", "abstract": "  The interactions between cells and the extracellular matrix are vital for the\nself-organisation of tissues. In this paper we present proof-of-concept to use\nmachine learning tools to predict the role of this mechanobiology in the\nself-organisation of cell-laden hydrogels grown in tethered moulds. We develop\na process for the automated generation of mould designs with and without key\nsymmetries. We create a large training set with $N=6500$ cases by running\ndetailed biophysical simulations of cell-matrix interactions using the\ncontractile network dipole orientation (CONDOR) model for the self-organisation\nof cellular hydrogels within these moulds. These are used to train an\nimplementation of the \\texttt{pix2pix} deep learning model, reserving $740$\ncases that were unseen in the training of the neural network for training and\nvalidation. Comparison between the predictions of the machine learning\ntechnique and the reserved predictions from the biophysical algorithm show that\nthe machine learning algorithm makes excellent predictions. The machine\nlearning algorithm is significantly faster than the biophysical method, opening\nthe possibility of very high throughput rational design of moulds for\npharmaceutical testing, regenerative medicine and fundamental studies of\nbiology. Future extensions for scaffolds and 3D bioprinting will open\nadditional applications.\n", "date": "2023-03-31", "categories": []}
{"id": "2208.06073", "title": "Conditional Antibody Design as 3D Equivariant Graph Translation", "abstract": "  Antibody design is valuable for therapeutic usage and biological research.\nExisting deep-learning-based methods encounter several key issues: 1)\nincomplete context for Complementarity-Determining Regions (CDRs) generation;\n2) incapability of capturing the entire 3D geometry of the input structure; 3)\ninefficient prediction of the CDR sequences in an autoregressive manner. In\nthis paper, we propose Multi-channel Equivariant Attention Network (MEAN) to\nco-design 1D sequences and 3D structures of CDRs. To be specific, MEAN\nformulates antibody design as a conditional graph translation problem by\nimporting extra components including the target antigen and the light chain of\nthe antibody. Then, MEAN resorts to E(3)-equivariant message passing along with\na proposed attention mechanism to better capture the geometrical correlation\nbetween different components. Finally, it outputs both the 1D sequences and 3D\nstructure via a multi-round progressive full-shot scheme, which enjoys more\nefficiency and precision against previous autoregressive approaches. Our method\nsignificantly surpasses state-of-the-art models in sequence and structure\nmodeling, antigen-binding CDR design, and binding affinity optimization.\nSpecifically, the relative improvement to baselines is about 23% in\nantigen-binding CDR design and 34% for affinity optimization.\n", "date": "2022-08-12", "categories": []}
{"id": "2208.09016", "title": "Improving Small Molecule Generation using Mutual Information Machine", "abstract": "  We address the task of controlled generation of small molecules, which\nentails finding novel molecules with desired properties under certain\nconstraints (e.g., similarity to a reference molecule). Here we introduce\nMolMIM, a probabilistic auto-encoder for small molecule drug discovery that\nlearns an informative and clustered latent space. MolMIM is trained with Mutual\nInformation Machine (MIM) learning, and provides a fixed length representation\nof variable length SMILES strings. Since encoder-decoder models can learn\nrepresentations with ``holes'' of invalid samples, here we propose a novel\nextension to the training procedure which promotes a dense latent space, and\nallows the model to sample valid molecules from random perturbations of latent\ncodes. We provide a thorough comparison of MolMIM to several variable-size and\nfixed-size encoder-decoder models, demonstrating MolMIM's superior generation\nas measured in terms of validity, uniqueness, and novelty. We then utilize\nCMA-ES, a naive black-box and gradient free search algorithm, over MolMIM's\nlatent space for the task of property guided molecule optimization. We achieve\nstate-of-the-art results in several constrained single property optimization\ntasks as well as in the challenging task of multi-objective optimization,\nimproving over previous success rate SOTA by more than 5\\% . We attribute the\nstrong results to MolMIM's latent representation which clusters similar\nmolecules in the latent space, whereas CMA-ES is often used as a baseline\noptimization method. We also demonstrate MolMIM to be favourable in a compute\nlimited regime, making it an attractive model for such cases.\n", "date": "2022-08-18", "categories": []}
{"id": "2303.16209", "title": "AmorProt: Amino Acid Molecular Fingerprints Repurposing based Protein\n  Fingerprint", "abstract": "  As protein therapeutics play an important role in almost all medical fields,\nnumerous studies have been conducted on proteins using artificial intelligence.\nArtificial intelligence has enabled data driven predictions without the need\nfor expensive experiments. Nevertheless, unlike the various molecular\nfingerprint algorithms that have been developed, protein fingerprint algorithms\nhave rarely been studied. In this study, we proposed the amino acid molecular\nfingerprints repurposing based protein (AmorProt) fingerprint, a protein\nsequence representation method that effectively uses the molecular fingerprints\ncorresponding to 20 amino acids. Subsequently, the performances of the tree\nbased machine learning and artificial neural network models were compared using\n(1) amyloid classification and (2) isoelectric point regression. Finally, the\napplicability and advantages of the developed platform were demonstrated\nthrough a case study and the following experiments: (3) comparison of dataset\ndependence with feature based methods; (4) feature importance analysis; and (5)\nprotein space analysis. Consequently, the significantly improved model\nperformance and data set independent versatility of the AmorProt fingerprint\nwere verified. The results revealed that the current protein representation\nmethod can be applied to various fields related to proteins, such as predicting\ntheir fundamental properties or interaction with ligands.\n", "date": "2023-03-27", "categories": []}
{"id": "2303.15520", "title": "Learning Harmonic Molecular Representations on Riemannian Manifold", "abstract": "  Molecular representation learning plays a crucial role in AI-assisted drug\ndiscovery research. Encoding 3D molecular structures through Euclidean neural\nnetworks has become the prevailing method in the geometric deep learning\ncommunity. However, the equivariance constraints and message passing in\nEuclidean space may limit the network expressive power. In this work, we\npropose a Harmonic Molecular Representation learning (HMR) framework, which\nrepresents a molecule using the Laplace-Beltrami eigenfunctions of its\nmolecular surface. HMR offers a multi-resolution representation of molecular\ngeometric and chemical features on 2D Riemannian manifold. We also introduce a\nharmonic message passing method to realize efficient spectral message passing\nover the surface manifold for better molecular encoding. Our proposed method\nshows comparable predictive power to current models in small molecule property\nprediction, and outperforms the state-of-the-art deep learning models for\nligand-binding protein pocket classification and the rigid protein docking\nchallenge, demonstrating its versatility in molecular representation learning.\n", "date": "2023-03-27", "categories": []}
{"id": "2302.07061", "title": "Do Deep Learning Methods Really Perform Better in Molecular Conformation\n  Generation?", "abstract": "  Molecular conformation generation (MCG) is a fundamental and important\nproblem in drug discovery. Many traditional methods have been developed to\nsolve the MCG problem, such as systematic searching, model-building, random\nsearching, distance geometry, molecular dynamics, Monte Carlo methods, etc.\nHowever, they have some limitations depending on the molecular structures.\nRecently, there are plenty of deep learning based MCG methods, which claim they\nlargely outperform the traditional methods. However, to our surprise, we design\na simple and cheap algorithm (parameter-free) based on the traditional methods\nand find it is comparable to or even outperforms deep learning based MCG\nmethods in the widely used GEOM-QM9 and GEOM-Drugs benchmarks. In particular,\nour design algorithm is simply the clustering of the RDKIT-generated\nconformations. We hope our findings can help the community to revise the deep\nlearning methods for MCG. The code of the proposed algorithm could be found at\nhttps://gist.github.com/ZhouGengmo/5b565f51adafcd911c0bc115b2ef027c.\n", "date": "2023-02-14", "categories": []}
{"id": "2208.01918", "title": "DeepProphet2 -- A Deep Learning Gene Recommendation Engine", "abstract": "  New powerful tools for tackling life science problems have been created by\nrecent advances in machine learning. The purpose of the paper is to discuss the\npotential advantages of gene recommendation performed by artificial\nintelligence (AI). Indeed, gene recommendation engines try to solve this\nproblem: if the user is interested in a set of genes, which other genes are\nlikely to be related to the starting set and should be investigated? This task\nwas solved with a custom deep learning recommendation engine, DeepProphet2\n(DP2), which is freely available to researchers worldwide via\nhttps://www.generecommender.com?utm_source=DeepProphet2_paper&utm_medium=pdf.\nHereafter, insights behind the algorithm and its practical applications are\nillustrated.\n  The gene recommendation problem can be addressed by mapping the genes to a\nmetric space where a distance can be defined to represent the real semantic\ndistance between them. To achieve this objective a transformer-based model has\nbeen trained on a well-curated freely available paper corpus, PubMed. The paper\ndescribes multiple optimization procedures that were employed to obtain the\nbest bias-variance trade-off, focusing on embedding size and network depth. In\nthis context, the model's ability to discover sets of genes implicated in\ndiseases and pathways was assessed through cross-validation. A simple\nassumption guided the procedure: the network had no direct knowledge of\npathways and diseases but learned genes' similarities and the interactions\namong them. Moreover, to further investigate the space where the neural network\nrepresents genes, the dimensionality of the embedding was reduced, and the\nresults were projected onto a human-comprehensible space. In conclusion, a set\nof use cases illustrates the algorithm's potential applications in a real word\nsetting.\n", "date": "2022-08-03", "categories": []}
{"id": "2212.00543", "title": "Fine-Grained Selective Similarity Integration for Drug-Target\n  Interaction Prediction", "abstract": "  The discovery of drug-target interactions (DTIs) is a pivotal process in\npharmaceutical development. Computational approaches are a promising and\nefficient alternative to tedious and costly wet-lab experiments for predicting\nnovel DTIs from numerous candidates. Recently, with the availability of\nabundant heterogeneous biological information from diverse data sources,\ncomputational methods have been able to leverage multiple drug and target\nsimilarities to boost the performance of DTI prediction. Similarity integration\nis an effective and flexible strategy to extract crucial information across\ncomplementary similarity views, providing a compressed input for any\nsimilarity-based DTI prediction model. However, existing similarity integration\nmethods filter and fuse similarities from a global perspective, neglecting the\nutility of similarity views for each drug and target. In this study, we propose\na Fine-Grained Selective similarity integration approach, called FGS, which\nemploys a local interaction consistency-based weight matrix to capture and\nexploit the importance of similarities at a finer granularity in both\nsimilarity selection and combination steps. We evaluate FGS on five DTI\nprediction datasets under various prediction settings. Experimental results\nshow that our method not only outperforms similarity integration competitors\nwith comparable computational costs, but also achieves better prediction\nperformance than state-of-the-art DTI prediction approaches by collaborating\nwith conventional base models. Furthermore, case studies on the analysis of\nsimilarity weights and on the verification of novel predictions confirm the\npractical ability of FGS.\n", "date": "2022-12-01", "categories": []}
{"id": "2303.11434", "title": "ResDTA: Predicting Drug-Target Binding Affinity Using Residual Skip\n  Connections", "abstract": "  The discovery of novel drug target (DT) interactions is an important step in\nthe drug development process. The majority of computer techniques for\npredicting DT interactions have focused on binary classification, with the goal\nof determining whether or not a DT pair interacts. Protein ligand interactions,\non the other hand, assume a continuous range of binding strength values, also\nknown as binding affinity, and forecasting this value remains a difficulty. As\nthe amount of affinity data in DT knowledge-bases grows, advanced learning\ntechniques such as deep learning architectures can be used to predict binding\naffinities. In this paper, we present a deep-learning-based methodology for\npredicting DT binding affinities using just sequencing information from both\ntargets and drugs. The results show that the proposed deep learning-based model\nthat uses the 1D representations of targets and drugs is an effective approach\nfor drug target binding affinity prediction and it does not require additional\nchemical domain knowledge to work with. The model in which high-level\nrepresentations of a drug and a target are constructed via CNNs that uses\nresidual skip connections and also with an additional stream to create a\nhigh-level combined representation of the drug-target pair achieved the best\nConcordance Index (CI) performance in one of the largest benchmark datasets,\noutperforming the recent state-of-the-art method AttentionDTA and many other\nmachine-learning and deep-learning based baseline methods for DT binding\naffinity prediction that uses the 1D representations of targets and drugs.\n", "date": "2023-03-20", "categories": []}
{"id": "2303.11494", "title": "FlexVDW: A machine learning approach to account for protein flexibility\n  in ligand docking", "abstract": "  Most widely used ligand docking methods assume a rigid protein structure.\nThis leads to problems when the structure of the target protein deforms upon\nligand binding. In particular, the ligand's true binding pose is often scored\nvery unfavorably due to apparent clashes between ligand and protein atoms,\nwhich lead to extremely high values of the calculated van der Waals energy\nterm. Traditionally, this problem has been addressed by explicitly searching\nfor receptor conformations to account for the flexibility of the receptor in\nligand binding. Here we present a deep learning model trained to take receptor\nflexibility into account implicitly when predicting van der Waals energy. We\nshow that incorporating this machine-learned energy term into a\nstate-of-the-art physics-based scoring function improves small molecule ligand\npose prediction results in cases with substantial protein deformation, without\ndegrading performance in cases with minimal protein deformation. This work\ndemonstrates the feasibility of learning effects of protein flexibility on\nligand binding without explicitly modeling changes in protein structure.\n", "date": "2023-03-20", "categories": []}
{"id": "2303.11695", "title": "Transcriptomics-based matching of drugs to diseases with deep learning", "abstract": "  In this work we present a deep learning approach to conduct hypothesis-free,\ntranscriptomics-based matching of drugs for diseases. Our proposed neural\nnetwork architecture is trained on approved drug-disease indications, taking as\ninput the relevant disease and drug differential gene expression profiles, and\nlearns to identify novel indications. We assemble an evaluation dataset of\ndisease-drug indications spanning 68 diseases and evaluate in silico our\napproach against the most widely used transcriptomics-based matching baselines,\nCMap and the Characteristic Direction. Our results show a more than 200%\nimprovement over both baselines in terms of standard retrieval metrics. We\nfurther showcase our model's ability to capture different genes' expressions\ninteractions among drugs and diseases. We provide our trained models, data and\ncode to predict with them at https://github.com/healx/dgem-nn-public.\n", "date": "2023-03-21", "categories": []}
{"id": "2303.10429", "title": "Protein Sequence Design with Batch Bayesian Optimisation", "abstract": "  Protein sequence design is a challenging problem in protein engineering,\nwhich aims to discover novel proteins with useful biological functions.\nDirected evolution is a widely-used approach for protein sequence design, which\nmimics the evolution cycle in a laboratory environment and conducts an\niterative protocol. However, the burden of laboratory experiments can be\nreduced by using machine learning approaches to build a surrogate model of the\nprotein landscape and conducting in-silico population selection through\nmodel-based fitness prediction. In this paper, we propose a new method based on\nBatch Bayesian Optimization (Batch BO), a well-established optimization method,\nfor protein sequence design. By incorporating Batch BO into the directed\nevolution process, our method is able to make more informed decisions about\nwhich sequences to select for artificial evolution, leading to improved\nperformance and faster convergence. We evaluate our method on a suite of\nin-silico protein sequence design tasks and demonstrate substantial improvement\nover baseline algorithms.\n", "date": "2023-03-18", "categories": []}
{"id": "2302.09335", "title": "Knowledge Graph Completion based on Tensor Decomposition for Disease\n  Gene Prediction", "abstract": "  Accurate identification of disease genes has consistently been one of the\nkeys to decoding a disease's molecular mechanism. Most current approaches focus\non constructing biological networks and utilizing machine learning, especially,\ndeep learning to identify disease genes, but ignore the complex relations\nbetween entities in the biological knowledge graph. In this paper, we construct\na biological knowledge graph centered on diseases and genes, and develop an\nend-to-end Knowledge graph completion model for Disease Gene Prediction using\ninteractional tensor decomposition (called KDGene). KDGene introduces an\ninteraction module between the embeddings of entities and relations to tensor\ndecomposition, which can effectively enhance the information interaction in\nbiological knowledge. Experimental results show that KDGene significantly\noutperforms state-of-the-art algorithms. Furthermore, the comprehensive\nbiological analysis of the case of diabetes mellitus confirms KDGene's ability\nfor identifying new and accurate candidate genes. This work proposes a scalable\nknowledge graph completion framework to identify disease candidate genes, from\nwhich the results are promising to provide valuable references for further wet\nexperiments.\n", "date": "2023-02-18", "categories": []}
{"id": "2303.08915", "title": "LRDB: LSTM Raw data DNA Base-caller based on long-short term models in\n  an active learning environment", "abstract": "  The first important step in extracting DNA characters is using the output\ndata of MinION devices in the form of electrical current signals. Various\ncutting-edge base callers use this data to detect the DNA characters based on\nthe input. In this paper, we discuss several shortcomings of prior base callers\nin the case of time-critical applications, privacy-aware design, and the\nproblem of catastrophic forgetting. Next, we propose the LRDB model, a\nlightweight open-source model for private developments with a better\nread-identity (0.35% increase) for the target bacterial samples in the paper.\nWe have limited the extent of training data and benefited from the transfer\nlearning algorithm to make the active usage of the LRDB viable in critical\napplications. Henceforth, less training time for adapting to new DNA samples\n(in our case, Bacterial samples) is needed. Furthermore, LRDB can be modified\nconcerning the user constraints as the results show a negligible accuracy loss\nin case of using fewer parameters. We have also assessed the noise-tolerance\nproperty, which offers about a 1.439% decline in accuracy for a 15dB noise\ninjection, and the performance metrics show that the model executes in a medium\nspeed range compared with current cutting-edge models.\n", "date": "2023-03-15", "categories": []}
{"id": "2211.14939", "title": "Applying Deep Reinforcement Learning to the HP Model for Protein\n  Structure Prediction", "abstract": "  A central problem in computational biophysics is protein structure\nprediction, i.e., finding the optimal folding of a given amino acid sequence.\nThis problem has been studied in a classical abstract model, the HP model,\nwhere the protein is modeled as a sequence of H (hydrophobic) and P (polar)\namino acids on a lattice. The objective is to find conformations maximizing H-H\ncontacts. It is known that even in this reduced setting, the problem is\nintractable (NP-hard). In this work, we apply deep reinforcement learning (DRL)\nto the two-dimensional HP model. We can obtain the conformations of best known\nenergies for benchmark HP sequences with lengths from 20 to 50. Our DRL is\nbased on a deep Q-network (DQN). We find that a DQN based on long short-term\nmemory (LSTM) architecture greatly enhances the RL learning ability and\nsignificantly improves the search process. DRL can sample the state space\nefficiently, without the need of manual heuristics. Experimentally we show that\nit can find multiple distinct best-known solutions per trial. This study\ndemonstrates the effectiveness of deep reinforcement learning in the HP model\nfor protein folding.\n", "date": "2022-11-27", "categories": []}
{"id": "2303.06154", "title": "Resource saving taxonomy classification with k-mer distributions and\n  machine learning", "abstract": "  Modern high throughput sequencing technologies like metagenomic sequencing\ngenerate millions of sequences which have to be classified based on their\ntaxonomic rank. Modern approaches either apply local alignment and comparison\nto existing data sets like MMseqs2 or use deep neural networks as it is done in\nDeepMicrobes and BERTax. Alignment-based approaches are costly in terms of\nruntime, especially since databases get larger and larger. For the deep\nlearning-based approaches, specialized hardware is necessary for a computation,\nwhich consumes large amounts of energy. In this paper, we propose to use\n$k$-mer distributions obtained from DNA as features to classify its taxonomic\norigin using machine learning approaches like the subspace $k$-nearest\nneighbors algorithm, neural networks or bagged decision trees. In addition, we\npropose a feature space data set balancing approach, which allows reducing the\ndata set for training and improves the performance of the classifiers. By\ncomparing performance, time, and memory consumption of our approach to those of\nstate-of-the-art algorithms (BERTax and MMseqs2) using several datasets, we\nshow that our approach improves the classification on the genus level and\nachieves comparable results for the superkingdom and phylum level.\n  Link:\nhttps://es-cloud.cs.uni-tuebingen.de/d/8e2ab8c3fdd444e1a135/?p=%2FTaxonomyClassification&mode=list\n", "date": "2023-03-10", "categories": []}
{"id": "2303.06857", "title": "An automated pipeline to create an atlas of in situ hybridization gene\n  expression data in the adult marmoset brain", "abstract": "  We present the first automated pipeline to create an atlas of in situ\nhybridization gene expression in the adult marmoset brain in the same\nstereotaxic space. The pipeline consists of segmentation of gene expression\nfrom microscopy images and registration of images to a standard space.\nAutomation of this pipeline is necessary to analyze the large volume of data in\nthe genome-wide whole-brain dataset, and to process images that have varying\nintensity profiles and expression patterns with minimal human bias. To reduce\nthe number of labelled images required for training, we develop a\nsemi-supervised segmentation model. We further develop an iterative algorithm\nto register images to a standard space, enabling comparative analysis between\ngenes and concurrent visualization with other datasets, thereby facilitating a\nmore holistic understanding of primate brain structure and function.\n", "date": "2023-03-13", "categories": []}
{"id": "2303.02162", "title": "T-Cell Receptor Optimization with Reinforcement Learning and Mutation\n  Policies for Precesion Immunotherapy", "abstract": "  T cells monitor the health status of cells by identifying foreign peptides\ndisplayed on their surface. T-cell receptors (TCRs), which are protein\ncomplexes found on the surface of T cells, are able to bind to these peptides.\nThis process is known as TCR recognition and constitutes a key step for immune\nresponse. Optimizing TCR sequences for TCR recognition represents a fundamental\nstep towards the development of personalized treatments to trigger immune\nresponses killing cancerous or virus-infected cells. In this paper, we\nformulated the search for these optimized TCRs as a reinforcement learning (RL)\nproblem, and presented a framework TCRPPO with a mutation policy using proximal\npolicy optimization. TCRPPO mutates TCRs into effective ones that can recognize\ngiven peptides. TCRPPO leverages a reward function that combines the\nlikelihoods of mutated sequences being valid TCRs measured by a new scoring\nfunction based on deep autoencoders, with the probabilities of mutated\nsequences recognizing peptides from a peptide-TCR interaction predictor. We\ncompared TCRPPO with multiple baseline methods and demonstrated that TCRPPO\nsignificantly outperforms all the baseline methods to generate positive binding\nand valid TCRs. These results demonstrate the potential of TCRPPO for both\nprecision immunotherapy and peptide-recognizing TCR motif discovery.\n", "date": "2023-03-02", "categories": []}
{"id": "2303.00176", "title": "Zyxin is all you need: machine learning adherent cell mechanics", "abstract": "  Cellular form and function emerge from complex mechanochemical systems within\nthe cytoplasm. No systematic strategy currently exists to infer large-scale\nphysical properties of a cell from its many molecular components. This is a\nsignificant obstacle to understanding biophysical processes such as cell\nadhesion and migration. Here, we develop a data-driven biophysical modeling\napproach to learn the mechanical behavior of adherent cells. We first train\nneural networks to predict forces generated by adherent cells from images of\ncytoskeletal proteins. Strikingly, experimental images of a single focal\nadhesion protein, such as zyxin, are sufficient to predict forces and\ngeneralize to unseen biological regimes. This protein field alone contains\nenough information to yield accurate predictions even if forces themselves are\ngenerated by many interacting proteins. We next develop two approaches - one\nexplicitly constrained by physics, the other more agnostic - that help\nconstruct data-driven continuum models of cellular forces using this single\nfocal adhesion field. Both strategies consistently reveal that cellular forces\nare encoded by two different length scales in adhesion protein distributions.\nBeyond adherent cell mechanics, our work serves as a case study for how to\nintegrate neural networks in the construction of predictive phenomenological\nmodels in cell biology, even when little knowledge of the underlying\nmicroscopic mechanisms exist.\n", "date": "2023-03-01", "categories": []}
{"id": "2303.01569", "title": "Chemically Transferable Generative Backmapping of Coarse-Grained\n  Proteins", "abstract": "  Coarse-graining (CG) accelerates molecular simulations of protein dynamics by\nsimulating sets of atoms as singular beads. Backmapping is the opposite\noperation of bringing lost atomistic details back from the CG representation.\nWhile machine learning (ML) has produced accurate and efficient CG simulations\nof proteins, fast and reliable backmapping remains a challenge. Rule-based\nmethods produce poor all-atom geometries, needing computationally costly\nrefinement through additional simulations. Recently proposed ML approaches\noutperform traditional baselines but are not transferable between proteins and\nsometimes generate unphysical atom placements with steric clashes and\nimplausible torsion angles. This work addresses both issues to build a fast,\ntransferable, and reliable generative backmapping tool for CG protein\nrepresentations. We achieve generalization and reliability through a combined\nset of innovations: representation based on internal coordinates; an\nequivariant encoder/prior; a custom loss function that helps ensure local\nstructure, global structure, and physical constraints; and expert curation of\nhigh-quality out-of-equilibrium protein data for training. Our results pave the\nway for out-of-the-box backmapping of coarse-grained simulations for arbitrary\nproteins.\n", "date": "2023-03-02", "categories": []}
{"id": "2301.12112", "title": "On Pre-trained Language Models for Antibody", "abstract": "  Antibodies are vital proteins offering robust protection for the human body\nfrom pathogens. The development of general protein and antibody-specific\npre-trained language models both facilitate antibody prediction tasks. However,\nthere have been limited studies that comprehensively explore the representation\ncapability of distinct pre-trained language models on different antibody tasks.\nTo investigate the problem, we aim to answer several key questions in this\npaper, such as how pre-trained language models perform in antibody tasks with\ndifferent specificity and how introducing specific biological mechanisms to the\npre-training process can benefit the model. Additionally, we evaluate if the\nlearned antibody pre-trained representations can be applied to real-world\nantibody problems, like drug discovery and immune process understanding.\nPreviously, no benchmark available largely hindered the study to answer these\nquestions. To aid in our investigation, we provide an AnTibody Understanding\nEvaluation (ATUE) benchmark. We comprehensively evaluate the performance of\nprotein pre-trained language models by empirical study along with conclusions\nand new insights. Our ATUE and code are released at\nhttps://github.com/dqwang122/EATLM.\n", "date": "2023-01-28", "categories": []}
{"id": "2302.07134", "title": "Do Deep Learning Models Really Outperform Traditional Approaches in\n  Molecular Docking?", "abstract": "  Molecular docking, given a ligand molecule and a ligand binding site (called\n``pocket'') on a protein, predicting the binding mode of the protein-ligand\ncomplex, is a widely used technique in drug design. Many deep learning models\nhave been developed for molecular docking, while most existing deep learning\nmodels perform docking on the whole protein, rather than on a given pocket as\nthe traditional molecular docking approaches, which does not match common\nneeds. What's more, they claim to perform better than traditional molecular\ndocking, but the approach of comparison is not fair, since traditional methods\nare not designed for docking on the whole protein without a given pocket. In\nthis paper, we design a series of experiments to examine the actual performance\nof these deep learning models and traditional methods. For a fair comparison,\nwe decompose the docking on the whole protein into two steps, pocket searching\nand docking on a given pocket, and build pipelines to evaluate traditional\nmethods and deep learning methods respectively. We find that deep learning\nmodels are actually good at pocket searching, but traditional methods are\nbetter than deep learning models at docking on given pockets. Overall, our work\nexplicitly reveals some potential problems in current deep learning models for\nmolecular docking and provides several suggestions for future works.\n", "date": "2023-02-14", "categories": []}
{"id": "2302.10888", "title": "Data-Efficient Protein 3D Geometric Pretraining via Refinement of\n  Diffused Protein Structure Decoy", "abstract": "  Learning meaningful protein representation is important for a variety of\nbiological downstream tasks such as structure-based drug design. Having\nwitnessed the success of protein sequence pretraining, pretraining for\nstructural data which is more informative has become a promising research\ntopic. However, there are three major challenges facing protein structure\npretraining: insufficient sample diversity, physically unrealistic modeling,\nand the lack of protein-specific pretext tasks. To try to address these\nchallenges, we present the 3D Geometric Pretraining. In this paper, we propose\na unified framework for protein pretraining and a 3D geometric-based,\ndata-efficient, and protein-specific pretext task: RefineDiff (Refine the\nDiffused Protein Structure Decoy). After pretraining our geometric-aware model\nwith this task on limited data(less than 1% of SOTA models), we obtained\ninformative protein representations that can achieve comparable performance for\nvarious downstream tasks.\n", "date": "2023-02-05", "categories": []}
{"id": "2302.10907", "title": "Diffusion Models in Bioinformatics: A New Wave of Deep Learning\n  Revolution in Action", "abstract": "  Denoising diffusion models have emerged as one of the most powerful\ngenerative models in recent years. They have achieved remarkable success in\nmany fields, such as computer vision, natural language processing (NLP), and\nbioinformatics. Although there are a few excellent reviews on diffusion models\nand their applications in computer vision and NLP, there is a lack of an\noverview of their applications in bioinformatics. This review aims to provide a\nrather thorough overview of the applications of diffusion models in\nbioinformatics to aid their further development in bioinformatics and\ncomputational biology. We start with an introduction of the key concepts and\ntheoretical foundations of three cornerstone diffusion modeling frameworks\n(denoising diffusion probabilistic models, noise-conditioned scoring networks,\nand stochastic differential equations), followed by a comprehensive description\nof diffusion models employed in the different domains of bioinformatics,\nincluding cryo-EM data enhancement, single-cell data analysis, protein design\nand generation, drug and small molecule design, and protein-ligand interaction.\nThe review is concluded with a summary of the potential new development and\napplications of diffusion models in bioinformatics.\n", "date": "2023-02-13", "categories": []}
{"id": "2302.10923", "title": "PointFISH -- learning point cloud representations for RNA localization\n  patterns", "abstract": "  Subcellular RNA localization is a critical mechanism for the spatial control\nof gene expression. Its mechanism and precise functional role is not yet very\nwell understood. Single Molecule Fluorescence in Situ Hybridization (smFISH)\nimages allow for the detection of individual RNA molecules with subcellular\naccuracy. In return, smFISH requires robust methods to quantify and classify\nRNA spatial distribution. Here, we present PointFISH, a novel computational\napproach for the recognition of RNA localization patterns. PointFISH is an\nattention-based network for computing continuous vector representations of RNA\npoint clouds. Trained on simulations only, it can directly process extracted\ncoordinates from experimental smFISH images. The resulting embedding allows\nscalable and flexible spatial transcriptomics analysis and matches performance\nof hand-crafted pipelines.\n", "date": "2023-02-21", "categories": []}
{"id": "2302.10348", "title": "Binding-and-folding recognition of an intrinsically disordered protein\n  using online learning molecular dynamics", "abstract": "  Intrinsically disordered proteins participate in many biological processes by\nfolding upon binding with other proteins. However, coupled folding and binding\nprocesses are not well understood from an atomistic point of view. One of the\nmain questions is whether folding occurs prior to or after binding. Here we use\na novel unbiased high-throughput adaptive sampling approach to reconstruct the\nbinding and folding between the disordered transactivation domain of\n\\mbox{c-Myb} and the KIX domain of the CREB-binding protein. The reconstructed\nlong-term dynamical process highlights the binding of a short stretch of amino\nacids on \\mbox{c-Myb} as a folded $\\alpha$-helix. Leucine residues, specially\nLeu298 to Leu302, establish initial native contacts that prime the binding and\nfolding of the rest of the peptide, with a mixture of conformational selection\non the N-terminal region with an induced fit of the C-terminal.\n", "date": "2023-02-20", "categories": []}
{"id": "2302.08680", "title": "Modeling Polypharmacy and Predicting Drug-Drug Interactions using Deep\n  Generative Models on Multimodal Graphs", "abstract": "  Latent representations of drugs and their targets produced by contemporary\ngraph autoencoder models have proved useful in predicting many types of\nnode-pair interactions on large networks, including drug-drug, drug-target, and\ntarget-target interactions. However, most existing approaches model either the\nnode's latent spaces in which node distributions are rigid or do not\neffectively capture the interrelations between drugs; these limitations hinder\nthe methods from accurately predicting drug-pair interactions. In this paper,\nwe present the effectiveness of variational graph autoencoders (VGAE) in\nmodeling latent node representations on multimodal networks. Our approach can\nproduce flexible latent spaces for each node type of the multimodal graph; the\nembeddings are used later for predicting links among node pairs under different\nedge types. To further enhance the models' performance, we suggest a new method\nthat concatenates Morgan fingerprints, which capture the molecular structures\nof each drug, with their latent embeddings before preceding them to the\ndecoding stage for link prediction. Our proposed model shows competitive\nresults on three multimodal networks: (1) a multimodal graph consisting of drug\nand protein nodes, (2) a multimodal graph constructed from a subset of the\nDrugBank database involving drug nodes under different interaction types, and\n(3) a multimodal graph consisting of drug and cell line nodes. Our source code\nis publicly available at https://github.com/HySonLab/drug-interactions.\n", "date": "2023-02-17", "categories": []}
{"id": "2302.06860", "title": "BLIAM: Literature-based Data Synthesis for Synergistic Drug Combination\n  Prediction", "abstract": "  Language models pre-trained on scientific literature corpora have\nsubstantially advanced scientific discovery by offering high-quality feature\nrepresentations for downstream applications. However, these features are often\nnot interpretable, and thus can reveal limited insights to domain experts.\nInstead of obtaining features from language models, we propose BLIAM, a\nliterature-based data synthesis approach to directly generate training data\npoints that are interpretable and model-agnostic to downstream applications.\nThe key idea of BLIAM is to create prompts using existing training data and\nthen use these prompts to synthesize new data points. BLIAM performs these two\nsteps iteratively as new data points will define more informative prompts and\nnew prompts will in turn synthesize more accurate data points. Notably,\nliterature-based data augmentation might introduce data leakage since labels of\ntest data points in downstream applications might have already been mentioned\nin the language model corpus. To prevent such leakage, we introduce GDSC-combo,\na large-scale drug combination discovery dataset that was published after the\nbiomedical language model was trained. We found that BLIAM substantially\noutperforms a non-augmented approach and manual prompting in this rigorous data\nsplit setting. BLIAM can be further used to synthesize data points for novel\ndrugs and cell lines that were not even measured in biomedical experiments. In\naddition to the promising prediction performance, the data points synthesized\nby BLIAM are interpretable and model-agnostic, enabling in silico augmentation\nfor in vitro experiments.\n", "date": "2023-02-14", "categories": []}
{"id": "2302.07120", "title": "PrefixMol: Target- and Chemistry-aware Molecule Design via Prefix\n  Embedding", "abstract": "  Is there a unified model for generating molecules considering different\nconditions, such as binding pockets and chemical properties? Although\ntarget-aware generative models have made significant advances in drug design,\nthey do not consider chemistry conditions and cannot guarantee the desired\nchemical properties. Unfortunately, merging the target-aware and chemical-aware\nmodels into a unified model to meet customized requirements may lead to the\nproblem of negative transfer. Inspired by the success of multi-task learning in\nthe NLP area, we use prefix embeddings to provide a novel generative model that\nconsiders both the targeted pocket's circumstances and a variety of chemical\nproperties. All conditional information is represented as learnable features,\nwhich the generative model subsequently employs as a contextual prompt.\nExperiments show that our model exhibits good controllability in both single\nand multi-conditional molecular generation. The controllability enables us to\noutperform previous structure-based drug design methods. More interestingly, we\nopen up the attention mechanism and reveal coupling relationships between\nconditions, providing guidance for multi-conditional molecule generation.\n", "date": "2023-02-14", "categories": []}
{"id": "2210.00006", "title": "A Graph Neural Network Approach to Automated Model Building in Cryo-EM\n  Maps", "abstract": "  Electron cryo-microscopy (cryo-EM) produces three-dimensional (3D) maps of\nthe electrostatic potential of biological macromolecules, including proteins.\nAlong with knowledge about the imaged molecules, cryo-EM maps allow de novo\natomic modelling, which is typically done through a laborious manual process.\nTaking inspiration from recent advances in machine learning applications to\nprotein structure prediction, we propose a graph neural network (GNN) approach\nfor automated model building of proteins in cryo-EM maps. The GNN acts on a\ngraph with nodes assigned to individual amino acids and edges representing the\nprotein chain. Combining information from the voxel-based cryo-EM data, the\namino acid sequence data and prior knowledge about protein geometries, the GNN\nrefines the geometry of the protein chain and classifies the amino acids for\neach of its nodes. Application to 28 test cases shows that our approach\noutperforms the state-of-the-art and approximates manual building for cryo-EM\nmaps with resolutions better than 3.5 \\r{A}.\n", "date": "2022-09-30", "categories": []}
{"id": "2302.03907", "title": "DDeMON: Ontology-based function prediction by Deep Learning from Dynamic\n  Multiplex Networks", "abstract": "  Biological systems can be studied at multiple levels of information,\nincluding gene, protein, RNA and different interaction networks levels. The\ngoal of this work is to explore how the fusion of systems' level information\nwith temporal dynamics of gene expression can be used in combination with\nnon-linear approximation power of deep neural networks to predict novel gene\nfunctions in a non-model organism potato \\emph{Solanum tuberosum}. We propose\nDDeMON (Dynamic Deep learning from temporal Multiplex Ontology-annotated\nNetworks), an approach for scalable, systems-level inference of function\nannotation using time-dependent multiscale biological information. The proposed\nmethod, which is capable of considering billions of potential links between the\ngenes of interest, was applied on experimental gene expression data and the\nbackground knowledge network to reliably classify genes with unknown function\ninto five different functional ontology categories, linked to the experimental\ndata set. Predicted novel functions of genes were validated using extensive\nprotein domain search approach.\n", "date": "2023-02-08", "categories": []}
{"id": "2302.03991", "title": "Prediction approaches for partly missing multi-omics covariate data: A\n  literature review and an empirical comparison study", "abstract": "  As the availability of omics data has increased in the last few years, more\nmulti-omics data have been generated, that is, high-dimensional molecular data\nconsisting of several types such as genomic, transcriptomic, or proteomic data,\nall obtained from the same patients. Such data lend themselves to being used as\ncovariates in automatic outcome prediction because each omics type may\ncontribute unique information, possibly improving predictions compared to using\nonly one omics data type. Frequently, however, in the training data and the\ndata to which automatic prediction rules should be applied, the test data, the\ndifferent omics data types are not available for all patients. We refer to this\ntype of data as block-wise missing multi-omics data. First, we provide a\nliterature review on existing prediction methods applicable to such data.\nSubsequently, using a collection of 13 publicly available multi-omics data\nsets, we compare the predictive performances of several of these approaches for\ndifferent block-wise missingness patterns. Finally, we discuss the results of\nthis empirical comparison study and draw some tentative conclusions.\n", "date": "2023-02-08", "categories": []}
{"id": "2302.00981", "title": "Predicting Molecule-Target Interaction by Learning Biomedical Network\n  and Molecule Representations", "abstract": "  The study of molecule-target interaction is quite important for drug\ndiscovery in terms of target identification, hit identification, pathway study,\ndrug-drug interaction, etc. Most existing methodologies utilize either\nbiomedical network information or molecule structural features to predict\npotential interaction link. However, the biomedical network information based\nmethods usually suffer from cold start problem, while structure based methods\noften give limited performance due to the structure/interaction assumption and\ndata quality. To address these issues, we propose a pseudo-siamese Graph Neural\nNetwork method, namely MTINet+, which learns both biomedical network\ntopological and molecule structural/chemical information as representations to\npredict potential interaction of given molecule and target pair. In MTINet+,\n1-hop subgraphs of given molecule and target pair are extracted from known\ninteraction of biomedical network as topological information, meanwhile the\nmolecule structural and chemical attributes are processed as molecule\ninformation. MTINet+ learns these two types of information as embedding\nfeatures for predicting the pair link. In the experiments of different\nmolecule-target interaction tasks, MTINet+ significantly outperforms over the\nstate-of-the-art baselines. In addition, in our designed network sparsity\nexperiments , MTINet+ shows strong robustness against different sparse\nbiomedical networks.\n", "date": "2023-02-02", "categories": []}
{"id": "2302.01435", "title": "Target specific peptide design using latent space approximate trajectory\n  collector", "abstract": "  Despite the prevalence and many successes of deep learning applications in de\nnovo molecular design, the problem of peptide generation targeting specific\nproteins remains unsolved. A main barrier for this is the scarcity of the\nhigh-quality training data. To tackle the issue, we propose a novel machine\nlearning based peptide design architecture, called Latent Space Approximate\nTrajectory Collector (LSATC). It consists of a series of samplers on an\noptimization trajectory on a highly non-convex energy landscape that\napproximates the distributions of peptides with desired properties in a latent\nspace. The process involves little human intervention and can be implemented in\nan end-to-end manner. We demonstrate the model by the design of peptide\nextensions targeting Beta-catenin, a key nuclear effector protein involved in\ncanonical Wnt signalling. When compared with a random sampler, LSATC can sample\npeptides with $36\\%$ lower binding scores in a $16$ times smaller interquartile\nrange (IQR) and $284\\%$ less hydrophobicity with a $1.4$ times smaller IQR.\nLSATC also largely outperforms other common generative models. Finally, we\nutilized a clustering algorithm to select 4 peptides from the 100 LSATC\ndesigned peptides for experimental validation. The result confirms that all the\nfour peptides extended by LSATC show improved Beta-catenin binding by at least\n$20.0\\%$, and two of the peptides show a $3$ fold increase in binding affinity\nas compared to the base peptide.\n", "date": "2023-02-02", "categories": []}
{"id": "2302.01596", "title": "A Novel Fuzzy Bi-Clustering Algorithm with AFS for Identification of\n  Co-Regulated Genes", "abstract": "  The identification of co-regulated genes and their transcription-factor\nbinding sites (TFBS) are the key steps toward understanding transcription\nregulation. In addition to effective laboratory assays, various bi-clustering\nalgorithms for detection of the co-expressed genes have been developed.\nBi-clustering methods are used to discover subgroups of genes with similar\nexpression patterns under to-be-identified subsets of experimental conditions\nwhen applied to gene expression data. By building two fuzzy partition matrices\nof the gene expression data with the Axiomatic Fuzzy Set (AFS) theory, this\npaper proposes a novel fuzzy bi-clustering algorithm for identification of\nco-regulated genes. Specifically, the gene expression data is transformed into\ntwo fuzzy partition matrices via sub-preference relations theory of AFS at\nfirst. One of the matrices is considering the genes as the universe and the\nconditions as the concept, the other one is considering the genes as the\nconcept and the conditions as the universe. The identification of the\nco-regulated genes (bi-clusters) is carried out on the two partition matrices\nat the same time. Then, a novel fuzzy-based similarity criterion is defined\nbased on the partition matrixes, and a cyclic optimization algorithm is\ndesigned to discover the significant bi-clusters at expression level. The above\nprocedures guarantee that the generated bi-clusters have more significant\nexpression values than that of extracted by the traditional bi-clustering\nmethods. Finally, the performance of the proposed method is evaluated with the\nperformance of the three well-known bi-clustering algorithms on publicly\navailable real microarray datasets. The experimental results are in agreement\nwith the theoretical analysis and show that the proposed algorithm can\neffectively detect the co-regulated genes without any prior knowledge of the\ngene expression data.\n", "date": "2023-02-03", "categories": []}
{"id": "2212.08379", "title": "GeneFormer: Learned Gene Compression using Transformer-based Context\n  Modeling", "abstract": "  With the development of gene sequencing technology, an explosive growth of\ngene data has been witnessed. And the storage of gene data has become an\nimportant issue. Traditional gene data compression methods rely on general\nsoftware like G-zip, which fails to utilize the interrelation of nucleotide\nsequence. Recently, many researchers begin to investigate deep learning based\ngene data compression method. In this paper, we propose a transformer-based\ngene compression method named GeneFormer. Specifically, we first introduce a\nmodified transformer structure to fully explore the nucleotide sequence\ndependency. Then, we propose fixed-length parallel grouping to accelerate the\ndecoding speed of our autoregressive model. Experimental results on real-world\ndatasets show that our method saves 29.7% bit rate compared with the\nstate-of-the-art method, and the decoding speed is significantly faster than\nall existing learning-based gene compression methods.\n", "date": "2022-12-16", "categories": []}
{"id": "2301.12222", "title": "Machine Learning Methods for Cancer Classification Using Gene Expression\n  Data: A Review", "abstract": "  Cancer is a term that denotes a group of diseases caused by abnormal growth\nof cells that can spread in different parts of the body. According to the World\nHealth Organization (WHO), cancer is the second major cause of death after\ncardiovascular diseases. Gene expression can play a fundamental role in the\nearly detection of cancer, as it is indicative of the biochemical processes in\ntissue and cells, as well as the genetic characteristics of an organism.\nDeoxyribonucleic Acid (DNA) microarrays and Ribonucleic Acid (RNA)- sequencing\nmethods for gene expression data allow quantifying the expression levels of\ngenes and produce valuable data for computational analysis. This study reviews\nrecent progress in gene expression analysis for cancer classification using\nmachine learning methods. Both conventional and deep learning-based approaches\nare reviewed, with an emphasis on the ap-plication of deep learning models due\nto their comparative advantages for identifying gene patterns that are\ndistinctive for various types of cancers. Relevant works that employ the most\ncommonly used deep neural network architectures are covered, including\nmulti-layer perceptrons, convolutional, recurrent, graph, and transformer\nnetworks. This survey also presents an overview of the data collection methods\nfor gene expression analysis and lists important datasets that are commonly\nused for supervised machine learning for this task. Furthermore, reviewed are\npertinent techniques for feature engineering and data preprocessing that are\ntypically used to handle the high dimensionality of gene expression data,\ncaused by a large number of genes present in data samples. The paper concludes\nwith a discussion of future research directions for machine learning-based gene\nexpression analysis for cancer classification.\n", "date": "2023-01-28", "categories": []}
{"id": "2206.08336", "title": "Constrained Submodular Optimization for Vaccine Design", "abstract": "  Advances in machine learning have enabled the prediction of immune system\nresponses to prophylactic and therapeutic vaccines. However, the engineering\ntask of designing vaccines remains a challenge. In particular, the genetic\nvariability of the human immune system makes it difficult to design peptide\nvaccines that provide widespread immunity in vaccinated populations. We\nintroduce a framework for evaluating and designing peptide vaccines that uses\nprobabilistic machine learning models, and demonstrate its ability to produce\ndesigns for a SARS-CoV-2 vaccine that outperform previous designs. We provide a\ntheoretical analysis of the approximability, scalability, and complexity of our\nframework.\n", "date": "2022-06-16", "categories": []}
{"id": "2301.11763", "title": "Gene Teams are on the Field: Evaluation of Variants in Gene-Networks\n  Using High Dimensional Modelling", "abstract": "  In medical genetics, each genetic variant is evaluated as an independent\nentity regarding its clinical importance. However, in most complex diseases,\nvariant combinations in specific gene networks, rather than the presence of a\nparticular single variant, predominates. In the case of complex diseases,\ndisease status can be evaluated by considering the success level of a team of\nspecific variants. We propose a high dimensional modelling based method to\nanalyse all the variants in a gene network together. To evaluate our method, we\nselected two gene networks, mTOR and TGF-Beta. For each pathway, we generated\n400 control and 400 patient group samples. mTOR and TGF-? pathways contain 31\nand 93 genes of varying sizes, respectively. We produced Chaos Game\nRepresentation images for each gene sequence to obtain 2-D binary patterns.\nThese patterns were arranged in succession, and a 3-D tensor structure was\nachieved for each gene network. Features for each data sample were acquired by\nexploiting Enhanced Multivariance Products Representation to 3-D data. Features\nwere split as training and testing vectors. Training vectors were employed to\ntrain a Support Vector Machines classification model. We achieved more than 96%\nand 99% classification accuracies for mTOR and TGF-Beta networks, respectively,\nusing a limited amount of training samples.\n", "date": "2023-01-27", "categories": []}
{"id": "2301.10772", "title": "Gene-SGAN: a method for discovering disease subtypes with imaging and\n  genetic signatures via multi-view weakly-supervised deep clustering", "abstract": "  Disease heterogeneity has been a critical challenge for precision diagnosis\nand treatment, especially in neurologic and neuropsychiatric diseases. Many\ndiseases can display multiple distinct brain phenotypes across individuals,\npotentially reflecting disease subtypes that can be captured using MRI and\nmachine learning methods. However, biological interpretability and treatment\nrelevance are limited if the derived subtypes are not associated with genetic\ndrivers or susceptibility factors. Herein, we describe Gene-SGAN - a\nmulti-view, weakly-supervised deep clustering method - which dissects disease\nheterogeneity by jointly considering phenotypic and genetic data, thereby\nconferring genetic correlations to the disease subtypes and associated\nendophenotypic signatures. We first validate the generalizability,\ninterpretability, and robustness of Gene-SGAN in semi-synthetic experiments. We\nthen demonstrate its application to real multi-site datasets from 28,858\nindividuals, deriving subtypes of Alzheimer's disease and brain endophenotypes\nassociated with hypertension, from MRI and SNP data. Derived brain phenotypes\ndisplayed significant differences in neuroanatomical patterns, genetic\ndeterminants, biological and clinical biomarkers, indicating potentially\ndistinct underlying neuropathologic processes, genetic drivers, and\nsusceptibility factors. Overall, Gene-SGAN is broadly applicable to disease\nsubtyping and endophenotype discovery, and is herein tested on disease-related,\ngenetically-driven neuroimaging phenotypes.\n", "date": "2023-01-25", "categories": []}
{"id": "2301.09533", "title": "Solving the HP model with Nested Monte Carlo Search", "abstract": "  In this paper we present a new Monte Carlo Search (MCS) algorithm for finding\nthe ground state energy of proteins in the HP-model. We also compare it briefly\nto other MCS algorithms not usually used on the HP-model and provide an\noverview of the algorithms used on HP-model. The algorithm presented in this\npaper does not beat state of the art algorithms, see PERM (Hsu and Grassberger\n2011), REMC (Thachuk, Shmygelska, and Hoos 2007) or WLRE (W\\\"ust and Landau\n2012) for better results.\n  Hsu, H.-P.; and Grassberger, P. 2011. A review of Monte Carlo simulations of\npolymers with PERM. Journal of Statistical Physics, 144 (3): 597 to 637.\n  Thachuk, C.; Shmygelska, A.; and Hoos, H. H. 2007. A replica exchange Monte\nCarlo algorithm for protein folding in the HP model. BMC Bioinformatics, 8(1):\n342.\n  W\\\"ust, T.; and Landau, D. P. 2012. Optimized Wang-Landau sampling of lattice\npolymers: Ground state search and folding thermodynamics of HP model proteins.\nThe Journal of Chemical Physics, 137(6): 064903.\n", "date": "2023-01-23", "categories": []}
{"id": "2301.09642", "title": "DiffSDS: A language diffusion model for protein backbone inpainting\n  under geometric conditions and constraints", "abstract": "  Have you ever been troubled by the complexity and computational cost of SE(3)\nprotein structure modeling and been amazed by the simplicity and power of\nlanguage modeling? Recent work has shown promise in simplifying protein\nstructures as sequences of protein angles; therefore, language models could be\nused for unconstrained protein backbone generation. Unfortunately, such\nsimplification is unsuitable for the constrained protein inpainting problem,\nwhere the model needs to recover masked structures conditioned on unmasked\nones, as it dramatically increases the computing cost of geometric constraints.\nTo overcome this dilemma, we suggest inserting a hidden \\textbf{a}tomic\n\\textbf{d}irection \\textbf{s}pace (\\textbf{ADS}) upon the language model,\nconverting invariant backbone angles into equivalent direction vectors and\npreserving the simplicity, called Seq2Direct encoder ($\\text{Enc}_{s2d}$).\nGeometric constraints could be efficiently imposed on the newly introduced\ndirection space. A Direct2Seq decoder ($\\text{Dec}_{d2s}$) with mathematical\nguarantees is also introduced to develop a \\textbf{SDS}\n($\\text{Enc}_{s2d}$+$\\text{Dec}_{d2s}$) model. We apply the SDS model as the\ndenoising neural network during the conditional diffusion process, resulting in\na constrained generative model--\\textbf{DiffSDS}. Extensive experiments show\nthat the plug-and-play ADS could transform the language model into a strong\nstructural model without loss of simplicity. More importantly, the proposed\nDiffSDS outperforms previous strong baselines by a large margin on the task of\nprotein inpainting.\n", "date": "2023-01-22", "categories": []}
{"id": "2301.07568", "title": "Beating the Best: Improving on AlphaFold2 at Protein Structure\n  Prediction", "abstract": "  The goal of Protein Structure Prediction (PSP) problem is to predict a\nprotein's 3D structure (confirmation) from its amino acid sequence. The problem\nhas been a 'holy grail' of science since the Noble prize-winning work of\nAnfinsen demonstrated that protein conformation was determined by sequence. A\nrecent and important step towards this goal was the development of AlphaFold2,\ncurrently the best PSP method. AlphaFold2 is probably the highest profile\napplication of AI to science. Both AlphaFold2 and RoseTTAFold (another\nimpressive PSP method) have been published and placed in the public domain\n(code & models). Stacking is a form of ensemble machine learning ML in which\nmultiple baseline models are first learnt, then a meta-model is learnt using\nthe outputs of the baseline level model to form a model that outperforms the\nbase models. Stacking has been successful in many applications. We developed\nthe ARStack PSP method by stacking AlphaFold2 and RoseTTAFold. ARStack\nsignificantly outperforms AlphaFold2. We rigorously demonstrate this using two\nsets of non-homologous proteins, and a test set of protein structures published\nafter that of AlphaFold2 and RoseTTAFold. As more high quality prediction\nmethods are published it is likely that ensemble methods will increasingly\noutperform any single method.\n", "date": "2023-01-18", "categories": []}
{"id": "2208.02194", "title": "Interpretable bilinear attention network with domain adaptation improves\n  drug-target prediction", "abstract": "  Predicting drug-target interaction is key for drug discovery. Recent deep\nlearning-based methods show promising performance but two challenges remain:\n(i) how to explicitly model and learn local interactions between drugs and\ntargets for better prediction and interpretation; (ii) how to generalize\nprediction performance on novel drug-target pairs from different distribution.\nIn this work, we propose DrugBAN, a deep bilinear attention network (BAN)\nframework with domain adaptation to explicitly learn pair-wise local\ninteractions between drugs and targets, and adapt on out-of-distribution data.\nDrugBAN works on drug molecular graphs and target protein sequences to perform\nprediction, with conditional domain adversarial learning to align learned\ninteraction representations across different distributions for better\ngeneralization on novel drug-target pairs. Experiments on three benchmark\ndatasets under both in-domain and cross-domain settings show that DrugBAN\nachieves the best overall performance against five state-of-the-art baselines.\nMoreover, visualizing the learned bilinear attention map provides interpretable\ninsights from prediction results.\n", "date": "2022-08-03", "categories": []}
{"id": "2301.05931", "title": "Drug Synergistic Combinations Predictions via Large-Scale Pre-Training\n  and Graph Structure Learning", "abstract": "  Drug combination therapy is a well-established strategy for disease treatment\nwith better effectiveness and less safety degradation. However, identifying\nnovel drug combinations through wet-lab experiments is resource intensive due\nto the vast combinatorial search space. Recently, computational approaches,\nspecifically deep learning models have emerged as an efficient way to discover\nsynergistic combinations. While previous methods reported fair performance,\ntheir models usually do not take advantage of multi-modal data and they are\nunable to handle new drugs or cell lines. In this study, we collected data from\nvarious datasets covering various drug-related aspects. Then, we take advantage\nof large-scale pre-training models to generate informative representations and\nfeatures for drugs, proteins, and diseases. Based on that, a message-passing\ngraph is built on top to propagate information together with graph structure\nlearning flexibility. This is first introduced in the biological networks and\nenables us to generate pseudo-relations in the graph. Our framework achieves\nstate-of-the-art results in comparison with other deep learning-based methods\non synergistic prediction benchmark datasets. We are also capable of\ninferencing new drug combination data in a test on an independent set released\nby AstraZeneca, where 10% of improvement over previous methods is observed. In\naddition, we're robust against unseen drugs and surpass almost 15% AU ROC\ncompared to the second-best model. We believe our framework contributes to both\nthe future wet-lab discovery of novel drugs and the building of promising\nguidance for precise combination medicine.\n", "date": "2023-01-14", "categories": []}
{"id": "2301.06194", "title": "Geometric Graph Learning with Extended Atom-Types Features for\n  Protein-Ligand Binding Affinity Prediction", "abstract": "  Understanding and accurately predicting protein-ligand binding affinity are\nessential in the drug design and discovery process. At present, machine\nlearning-based methodologies are gaining popularity as a means of predicting\nbinding affinity due to their efficiency and accuracy, as well as the\nincreasing availability of structural and binding affinity data for\nprotein-ligand complexes. In biomolecular studies, graph theory has been widely\napplied since graphs can be used to model molecules or molecular complexes in a\nnatural manner. In the present work, we upgrade the graph-based learners for\nthe study of protein-ligand interactions by integrating extensive atom types\nsuch as SYBYL and extended connectivity interactive features (ECIF) into\nmultiscale weighted colored graphs (MWCG). By pairing with the gradient\nboosting decision tree (GBDT) machine learning algorithm, our approach results\nin two different methods, namely $^\\text{sybyl}\\text{GGL}$-Score and\n$^\\text{ecif}\\text{GGL}$-Score. Both of our models are extensively validated in\ntheir scoring power using three commonly used benchmark datasets in the drug\ndesign area, namely CASF-2007, CASF-2013, and CASF-2016. The performance of our\nbest model $^\\text{sybyl}\\text{GGL}$-Score is compared with other\nstate-of-the-art models in the binding affinity prediction for each benchmark.\nWhile both of our models achieve state-of-the-art results, the SYBYL atom-type\nmodel $^\\text{sybyl}\\text{GGL}$-Score outperforms other methods by a wide\nmargin in all benchmarks.\n", "date": "2023-01-15", "categories": []}
{"id": "2301.04093", "title": "On the Robustness of AlphaFold: A COVID-19 Case Study", "abstract": "  Protein folding neural networks (PFNNs) such as AlphaFold predict remarkably\naccurate structures of proteins compared to other approaches. However, the\nrobustness of such networks has heretofore not been explored. This is\nparticularly relevant given the broad social implications of such technologies\nand the fact that biologically small perturbations in the protein sequence do\nnot generally lead to drastic changes in the protein structure. In this paper,\nwe demonstrate that AlphaFold does not exhibit such robustness despite its high\naccuracy. This raises the challenge of detecting and quantifying the extent to\nwhich these predicted protein structures can be trusted. To measure the\nrobustness of the predicted structures, we utilize (i) the root-mean-square\ndeviation (RMSD) and (ii) the Global Distance Test (GDT) similarity measure\nbetween the predicted structure of the original sequence and the structure of\nits adversarially perturbed version. We prove that the problem of minimally\nperturbing protein sequences to fool protein folding neural networks is\nNP-complete. Based on the well-established BLOSUM62 sequence alignment scoring\nmatrix, we generate adversarial protein sequences and show that the RMSD\nbetween the predicted protein structure and the structure of the original\nsequence are very large when the adversarial changes are bounded by (i) 20\nunits in the BLOSUM62 distance, and (ii) five residues (out of hundreds or\nthousands of residues) in the given protein sequence. In our experimental\nevaluation, we consider 111 COVID-19 proteins in the Universal Protein resource\n(UniProt), a central resource for protein data managed by the European\nBioinformatics Institute, Swiss Institute of Bioinformatics, and the US Protein\nInformation Resource. These result in an overall GDT similarity test score\naverage of around 34%, demonstrating a substantial drop in the performance of\nAlphaFold.\n", "date": "2023-01-10", "categories": []}
{"id": "2301.04653", "title": "Optirank: classification for RNA-Seq data with optimal ranking reference\n  genes", "abstract": "  Classification algorithms using RNA-Sequencing (RNA-Seq) data as input are\nused in a variety of biological applications. By nature, RNA-Seq data is\nsubject to uncontrolled fluctuations both within and especially across\ndatasets, which presents a major difficulty for a trained classifier to\ngeneralize to an external dataset. Replacing raw gene counts with the rank of\ngene counts inside an observation has proven effective to mitigate this\nproblem. However, the rank of a feature is by definition relative to all other\nfeatures, including highly variable features that introduce noise in the\nranking. To address this problem and obtain more robust ranks, we propose a\nlogistic regression model, optirank, which learns simultaneously the parameters\nof the model and the genes to use as a reference set in the ranking. We show\nthe effectiveness of this method on simulated data. We also consider real\nclassification tasks, which present different kinds of distribution shifts\nbetween train and test data. Those tasks concern a variety of applications,\nsuch as cancer of unknown primary classification, identification of specific\ngene signatures, and determination of cell type in single-cell RNA-Seq\ndatasets. On those real tasks, optirank performs at least as well as the\nvanilla logistic regression on classical ranks, while producing sparser\nsolutions. In addition, to increase the robustness against dataset shifts, we\npropose a multi-source learning scheme and demonstrate its effectiveness when\nused in combination with rank-based classifiers.\n", "date": "2023-01-11", "categories": []}
{"id": "2301.04225", "title": "Inferring Gene Regulatory Neural Networks for Bacterial Decision Making\n  in Biofilms", "abstract": "  Bacterial cells are sensitive to a range of external signals used to learn\nthe environment. These incoming external signals are then processed using a\nGene Regulatory Network (GRN), exhibiting similarities to modern computing\nalgorithms. An in-depth analysis of gene expression dynamics suggests an\ninherited Gene Regulatory Neural Network (GRNN) behavior within the GRN that\nenables the cellular decision-making based on received signals from the\nenvironment and neighbor cells. In this study, we extract a sub-network of\n\\textit{Pseudomonas aeruginosa} GRN that is associated with one virulence\nfactor: pyocyanin production as a use case to investigate the GRNN behaviors.\nFurther, using Graph Neural Network (GNN) architecture, we model a single\nspecies biofilm to reveal the role of GRNN dynamics on ecosystem-wide\ndecision-making. Varying environmental conditions, we prove that the extracted\nGRNN computes input signals similar to natural decision-making process of the\ncell. Identifying of neural network behaviors in GRNs may lead to more accurate\nbacterial cell activity predictive models for many applications, including\nhuman health-related problems and agricultural applications. Further, this\nmodel can produce data on causal relationships throughout the network, enabling\nthe possibility of designing tailor-made infection-controlling mechanisms. More\ninterestingly, these GRNNs can perform computational tasks for bio-hybrid\ncomputing systems.\n", "date": "2023-01-10", "categories": []}
{"id": "2301.02916", "title": "Unsupervised ensemble-based phenotyping helps enhance the\n  discoverability of genes related to heart morphology", "abstract": "  Recent genome-wide association studies (GWAS) have been successful in\nidentifying associations between genetic variants and simple cardiac parameters\nderived from cardiac magnetic resonance (CMR) images. However, the emergence of\nbig databases including genetic data linked to CMR, facilitates investigation\nof more nuanced patterns of shape variability. Here, we propose a new framework\nfor gene discovery entitled Unsupervised Phenotype Ensembles (UPE). UPE builds\na redundant yet highly expressive representation by pooling a set of phenotypes\nlearned in an unsupervised manner, using deep learning models trained with\ndifferent hyperparameters. These phenotypes are then analyzed via (GWAS),\nretaining only highly confident and stable associations across the ensemble. We\napply our approach to the UK Biobank database to extract left-ventricular (LV)\ngeometric features from image-derived three-dimensional meshes. We demonstrate\nthat our approach greatly improves the discoverability of genes influencing LV\nshape, identifying 11 loci with study-wide significance and 8 with suggestive\nsignificance. We argue that our approach would enable more extensive discovery\nof gene associations with image-derived phenotypes for other organs or image\nmodalities.\n", "date": "2023-01-07", "categories": []}
{"id": "2301.01769", "title": "Comprehensive analysis of gene expression profiles to radiation exposure\n  reveals molecular signatures of low-dose radiation response", "abstract": "  There are various sources of ionizing radiation exposure, where medical\nexposure for radiation therapy or diagnosis is the most common human-made\nsource. Understanding how gene expression is modulated after ionizing radiation\nexposure and investigating the presence of any dose-dependent gene expression\npatterns have broad implications for health risks from radiotherapy, medical\nradiation diagnostic procedures, as well as other environmental exposure. In\nthis paper, we perform a comprehensive pathway-based analysis of gene\nexpression profiles in response to low-dose radiation exposure, in order to\nexamine the potential mechanism of gene regulation underlying such responses.\nTo accomplish this goal, we employ a statistical framework to determine whether\na specific group of genes belonging to a known pathway display coordinated\nexpression patterns that are modulated in a manner consistent with the\nradiation level. Findings in our study suggest that there exist complex yet\nconsistent signatures that reflect the molecular response to radiation\nexposure, which differ between low-dose and high-dose radiation.\n", "date": "2023-01-03", "categories": []}
{"id": "2212.14121", "title": "CellTranspose: Few-shot Domain Adaptation for Cellular Instance\n  Segmentation", "abstract": "  Automated cellular instance segmentation is a process utilized for\naccelerating biological research for the past two decades, and recent\nadvancements have produced higher quality results with less effort from the\nbiologist. Most current endeavors focus on completely cutting the researcher\nout of the picture by generating highly generalized models. However, these\nmodels invariably fail when faced with novel data, distributed differently than\nthe ones used for training. Rather than approaching the problem with methods\nthat presume the availability of large amounts of target data and computing\npower for retraining, in this work we address the even greater challenge of\ndesigning an approach that requires minimal amounts of new annotated data as\nwell as training time. We do so by designing specialized contrastive losses\nthat leverage the few annotated samples very efficiently. A large set of\nresults show that 3 to 5 annotations lead to models with accuracy that: 1)\nsignificantly mitigate the covariate shift effects; 2) matches or surpasses\nother adaptation methods; 3) even approaches methods that have been fully\nretrained on the target distribution. The adaptation training is only a few\nminutes, paving a path towards a balance between model performance, computing\nrequirements and expert-level annotation needs.\n", "date": "2022-12-28", "categories": []}
{"id": "2212.13295", "title": "Structure-based drug discovery with deep learning", "abstract": "  Artificial intelligence (AI) in the form of deep learning bears promise for\ndrug discovery and chemical biology, $\\textit{e.g.}$, to predict protein\nstructure and molecular bioactivity, plan organic synthesis, and design\nmolecules $\\textit{de novo}$. While most of the deep learning efforts in drug\ndiscovery have focused on ligand-based approaches, structure-based drug\ndiscovery has the potential to tackle unsolved challenges, such as affinity\nprediction for unexplored protein targets, binding-mechanism elucidation, and\nthe rationalization of related chemical kinetic properties. Advances in deep\nlearning methodologies and the availability of accurate predictions for protein\ntertiary structure advocate for a $\\textit{renaissance}$ in structure-based\napproaches for drug discovery guided by AI. This review summarizes the most\nprominent algorithmic concepts in structure-based deep learning for drug\ndiscovery, and forecasts opportunities, applications, and challenges ahead.\n", "date": "2022-12-26", "categories": []}
{"id": "2212.12542", "title": "Neural Networks beyond explainability: Selective inference for sequence\n  motifs", "abstract": "  Over the past decade, neural networks have been successful at making\npredictions from biological sequences, especially in the context of regulatory\ngenomics. As in other fields of deep learning, tools have been devised to\nextract features such as sequence motifs that can explain the predictions made\nby a trained network. Here we intend to go beyond explainable machine learning\nand introduce SEISM, a selective inference procedure to test the association\nbetween these extracted features and the predicted phenotype. In particular, we\ndiscuss how training a one-layer convolutional network is formally equivalent\nto selecting motifs maximizing some association score. We adapt existing\nsampling-based selective inference procedures by quantizing this selection over\nan infinite set to a large but finite grid. Finally, we show that sampling\nunder a specific choice of parameters is sufficient to characterize the\ncomposite null hypothesis typically used for selective inference-a result that\ngoes well beyond our particular framework. We illustrate the behavior of our\nmethod in terms of calibration, power and speed and discuss its power/speed\ntrade-off with a simpler data-split strategy. SEISM paves the way to an easier\nanalysis of neural networks used in regulatory genomics, and to more powerful\nmethods for genome wide association studies (GWAS).\n", "date": "2022-12-23", "categories": []}
{"id": "2212.13171", "title": "Higher order organizational features can distinguish protein interaction\n  networks of disease classes: a case study of neoplasms and neurological\n  diseases", "abstract": "  Neoplasms (NPs) and neurological diseases and disorders (NDDs) are amongst\nthe major classes of diseases underlying deaths of a disproportionate number of\npeople worldwide. To determine if there exist some distinctive features in the\nlocal wiring patterns of protein interactions emerging at the onset of a\ndisease belonging to either of these two classes, we examined 112 and 175\nprotein interaction networks belonging to NPs and NDDs, respectively. Orbit\nusage profiles (OUPs) for each of these networks were enumerated by\ninvestigating the networks' local topology. 56 non-redundant OUPs (nrOUPs) were\nderived and used as network features for classification between these two\ndisease classes. Four machine learning classifiers, namely, k-nearest neighbour\n(KNN), support vector machine (SVM), deep neural network (DNN), random forest\n(RF) were trained on these data. DNN obtained the greatest average AUPRC\n(0.988) among these classifiers. DNNs developed on node2vec and the proposed\nnrOUPs embeddings were compared using 5-fold cross validation on the basis of\naverage values of the six of performance measures, viz., AUPRC, Accuracy,\nSensitivity, Specificity, Precision and MCC. It was found that nrOUPs based\nclassifier performed better in all of these six performance measures.\n", "date": "2022-12-26", "categories": []}
{"id": "2212.10567", "title": "Anticancer Peptides Classification using Kernel Sparse Representation\n  Classifier", "abstract": "  Cancer is one of the most challenging diseases because of its complexity,\nvariability, and diversity of causes. It has been one of the major research\ntopics over the past decades, yet it is still poorly understood. To this end,\nmultifaceted therapeutic frameworks are indispensable. \\emph{Anticancer\npeptides} (ACPs) are the most promising treatment option, but their large-scale\nidentification and synthesis require reliable prediction methods, which is\nstill a problem. In this paper, we present an intuitive classification strategy\nthat differs from the traditional \\emph{black box} method and is based on the\nwell-known statistical theory of \\emph{sparse-representation classification}\n(SRC). Specifically, we create over-complete dictionary matrices by embedding\nthe \\emph{composition of the K-spaced amino acid pairs} (CKSAAP). Unlike the\ntraditional SRC frameworks, we use an efficient \\emph{matching pursuit} solver\ninstead of the computationally expensive \\emph{basis pursuit} solver in this\nstrategy. Furthermore, the \\emph{kernel principal component analysis} (KPCA) is\nemployed to cope with non-linearity and dimension reduction of the feature\nspace whereas the \\emph{synthetic minority oversampling technique} (SMOTE) is\nused to balance the dictionary. The proposed method is evaluated on two\nbenchmark datasets for well-known statistical parameters and is found to\noutperform the existing methods. The results show the highest sensitivity with\nthe most balanced accuracy, which might be beneficial in understanding\nstructural and chemical aspects and developing new ACPs. The Google-Colab\nimplementation of the proposed method is available at the author's GitHub page\n(\\href{https://github.com/ehtisham-Fazal/ACP-Kernel-SRC}{https://github.com/ehtisham-fazal/ACP-Kernel-SRC}).\n", "date": "2022-12-19", "categories": []}
{"id": "2212.10675", "title": "There's Plenty of Room Right Here: Biological Systems as Evolved,\n  Overloaded, Multi-scale Machines", "abstract": "  The applicability of computational models to the biological world is an\nactive topic of debate. We argue that a useful path forward results from\nabandoning hard boundaries between categories and adopting an\nobserver-dependent, pragmatic view. Such a view dissolves the contingent\ndichotomies driven by human cognitive biases (e.g., tendency to oversimplify)\nand prior technological limitations in favor of a more continuous, gradualist\nview necessitated by the study of evolution, developmental biology, and\nintelligent machines. Efforts to re-shape living systems for biomedical or\nbioengineering purposes require prediction and control of their function at\nmultiple scales. This is challenging for many reasons, one of which is that\nliving systems perform multiple functions in the same place at the same time.\nWe refer to this as \"polycomputing\" - the ability of the same substrate to\nsimultaneously compute different things. This ability is an important way in\nwhich living things are a kind of computer, but not the familiar, linear,\ndeterministic kind; rather, living things are computers in the broad sense of\ncomputational materials as reported in the rapidly-growing physical computing\nliterature. We argue that an observer-centered framework for the computations\nperformed by evolved and designed systems will improve the understanding of\nmeso-scale events, as it has already done at quantum and relativistic scales.\nHere, we review examples of biological and technological polycomputing, and\ndevelop the idea that overloading of different functions on the same hardware\nis an important design principle that helps understand and build both evolved\nand designed systems. Learning to hack existing polycomputing substrates, as\nwell as evolve and design new ones, will have massive impacts on regenerative\nmedicine, robotics, and computer engineering.\n", "date": "2022-12-20", "categories": []}
{"id": "2212.09991", "title": "Dynamic Molecular Graph-based Implementation for Biophysical Properties\n  Prediction", "abstract": "  Neural Networks (GNNs) have revolutionized the molecular discovery to\nunderstand patterns and identify unknown features that can aid in predicting\nbiophysical properties and protein-ligand interactions. However, current models\ntypically rely on 2-dimensional molecular representations as input, and while\nutilization of 2\\3- dimensional structural data has gained deserved traction in\nrecent years as many of these models are still limited to static graph\nrepresentations. We propose a novel approach based on the transformer model\nutilizing GNNs for characterizing dynamic features of protein-ligand\ninteractions. Our message passing transformer pre-trains on a set of molecular\ndynamic data based off of physics-based simulations to learn coordinate\nconstruction and make binding probability and affinity predictions as a\ndownstream task. Through extensive testing we compare our results with the\nexisting models, our MDA-PLI model was able to outperform the molecular\ninteraction prediction models with an RMSE of 1.2958. The geometric encodings\nenabled by our transformer architecture and the addition of time series data\nadd a new dimensionality to this form of research.\n", "date": "2022-12-20", "categories": []}
{"id": "2212.00136", "title": "DEL-Dock: Molecular Docking-Enabled Modeling of DNA-Encoded Libraries", "abstract": "  DNA-Encoded Library (DEL) technology has enabled significant advances in hit\nidentification by enabling efficient testing of combinatorially-generated\nmolecular libraries. DEL screens measure protein binding affinity though\nsequencing reads of molecules tagged with unique DNA-barcodes that survive a\nseries of selection experiments. Computational models have been deployed to\nlearn the latent binding affinities that are correlated to the sequenced count\ndata; however, this correlation is often obfuscated by various sources of noise\nintroduced in its complicated data-generation process. In order to denoise DEL\ncount data and screen for molecules with good binding affinity, computational\nmodels require the correct assumptions in their modeling structure to capture\nthe correct signals underlying the data. Recent advances in DEL models have\nfocused on probabilistic formulations of count data, but existing approaches\nhave thus far been limited to only utilizing 2-D molecule-level\nrepresentations. We introduce a new paradigm, DEL-Dock, that combines\nligand-based descriptors with 3-D spatial information from docked\nprotein-ligand complexes. 3-D spatial information allows our model to learn\nover the actual binding modality rather than using only structured-based\ninformation of the ligand. We show that our model is capable of effectively\ndenoising DEL count data to predict molecule enrichment scores that are better\ncorrelated with experimental binding affinity measurements compared to prior\nworks. Moreover, by learning over a collection of docked poses we demonstrate\nthat our model, trained only on DEL data, implicitly learns to perform good\ndocking pose selection without requiring external supervision from\nexpensive-to-source protein crystal structures.\n", "date": "2022-11-30", "categories": []}
{"id": "2212.05184", "title": "Information retrieval in single cell chromatin analysis using TF-IDF\n  transformation methods", "abstract": "  Single-cell sequencing assay for transposase-accessible chromatin\n(scATAC-seq) assesses genome-wide chromatin accessibility in thousands of cells\nto reveal regulatory landscapes in high resolutions. However, the analysis\npresents challenges due to the high dimensionality and sparsity of the data.\nSeveral methods have been developed, including transformation techniques of\nterm-frequency inverse-document frequency (TF-IDF), dimension reduction methods\nsuch as singular value decomposition (SVD), factor analysis, and autoencoders.\nYet, a comprehensive study on the mentioned methods has not been fully\nperformed. It is not clear what is the best practice when analyzing scATAC-seq\ndata. We compared several scenarios for transformation and dimension reduction\nas well as the SVD-based feature analysis to investigate potential enhancements\nin scATAC-seq information retrieval. Additionally, we investigate if\nautoencoders benefit from the TF-IDF transformation. Our results reveal that\nthe TF-IDF transformation generally leads to improved clustering and\nbiologically relevant feature extraction.\n", "date": "2022-12-10", "categories": []}
{"id": "2212.05991", "title": "Graph algorithms for predicting subcellular localization at the pathway\n  level", "abstract": "  Protein subcellular localization is an important factor in normal cellular\nprocesses and disease. While many protein localization resources treat it as\nstatic, protein localization is dynamic and heavily influenced by biological\ncontext. Biological pathways are graphs that represent a specific biological\ncontext and can be inferred from large-scale data. We develop graph algorithms\nto predict the localization of all interactions in a biological pathway as an\nedge-labeling task. We compare a variety of models including graph neural\nnetworks, probabilistic graphical models, and discriminative classifiers for\npredicting localization annotations from curated pathway databases. We also\nperform a case study where we construct biological pathways and predict\nlocalizations of human fibroblasts undergoing viral infection. Pathway\nlocalization prediction is a promising approach for integrating publicly\navailable localization data into the analysis of large-scale biological data.\n", "date": "2022-12-12", "categories": []}
{"id": "2212.03456", "title": "Dock2D: Synthetic data for the molecular recognition problem", "abstract": "  Predicting the physical interaction of proteins is a cornerstone problem in\ncomputational biology. New classes of learning-based algorithms are actively\nbeing developed, and are typically trained end-to-end on protein complex\nstructures extracted from the Protein Data Bank. These training datasets tend\nto be large and difficult to use for prototyping and, unlike image or natural\nlanguage datasets, they are not easily interpretable by non-experts. We present\nDock2D-IP and Dock2D-IF, two \"toy\" datasets that can be used to select\nalgorithms predicting protein-protein interactions$\\unicode{x2014}$or any other\ntype of molecular interactions. Using two-dimensional shapes as input, each\nexample from Dock2D-IP (\"interaction pose\") describes the interaction pose of\ntwo shapes known to interact and each example from Dock2D-IF (\"interaction\nfact\") describes whether two shapes form a stable complex or not. We propose a\nnumber of baseline solutions to the problem and show that the same underlying\nenergy function can be learned either by solving the interaction pose task\n(formulated as an energy-minimization \"docking\" problem) or the\nfact-of-interaction task (formulated as a binding free energy estimation\nproblem).\n", "date": "2022-12-07", "categories": []}
{"id": "2212.02750", "title": "Improving Molecule Properties Through 2-Stage VAE", "abstract": "  Variational autoencoder (VAE) is a popular method for drug discovery and\nthere had been a great deal of architectures and pipelines proposed to improve\nits performance. But the VAE model itself suffers from deficiencies such as\npoor manifold recovery when data lie on low-dimensional manifold embedded in\nhigher dimensional ambient space and they manifest themselves in each\napplications differently. The consequences of it in drug discovery is somewhat\nunder-explored. In this paper, we study how to improve the similarity of the\ndata generated via VAE and the training dataset by improving manifold recovery\nvia a 2-stage VAE where the second stage VAE is trained on the latent space of\nthe first one. We experimentally evaluated our approach using the ChEMBL\ndataset as well as a polymer datasets. In both dataset, the 2-stage VAE method\nis able to improve the property statistics significantly from a pre-existing\nmethod.\n", "date": "2022-12-06", "categories": []}
{"id": "2212.01575", "title": "Multi-view deep learning based molecule design and structural\n  optimization accelerates the SARS-CoV-2 inhibitor discovery", "abstract": "  In this work, we propose MEDICO, a Multi-viEw Deep generative model for\nmolecule generation, structural optimization, and the SARS-CoV-2 Inhibitor\ndisCOvery. To the best of our knowledge, MEDICO is the first-of-this-kind graph\ngenerative model that can generate molecular graphs similar to the structure of\ntargeted molecules, with a multi-view representation learning framework to\nsufficiently and adaptively learn comprehensive structural semantics from\ntargeted molecular topology and geometry. We show that our MEDICO significantly\noutperforms the state-of-the-art methods in generating valid, unique, and novel\nmolecules under benchmarking comparisons. In particular, we showcase the\nmulti-view deep learning model enables us to generate not only the molecules\nstructurally similar to the targeted molecules but also the molecules with\ndesired chemical properties, demonstrating the strong capability of our model\nin exploring the chemical space deeply. Moreover, case study results on\ntargeted molecule generation for the SARS-CoV-2 main protease (Mpro) show that\nby integrating molecule docking into our model as chemical priori, we\nsuccessfully generate new small molecules with desired drug-like properties for\nthe Mpro, potentially accelerating the de novo design of Covid-19 drugs.\nFurther, we apply MEDICO to the structural optimization of three well-known\nMpro inhibitors (N3, 11a, and GC376) and achieve ~88% improvement in their\nbinding affinity to Mpro, demonstrating the application value of our model for\nthe development of therapeutics for SARS-CoV-2 infection.\n", "date": "2022-12-03", "categories": []}
{"id": "2212.02404", "title": "Energy-based Generative Models for Target-specific Drug Discovery", "abstract": "  Drug targets are the main focus of drug discovery due to their key role in\ndisease pathogenesis. Computational approaches are widely applied to drug\ndevelopment because of the increasing availability of biological molecular\ndatasets. Popular generative approaches can create new drug molecules by\nlearning the given molecule distributions. However, these approaches are mostly\nnot for target-specific drug discovery. We developed an energy-based\nprobabilistic model for computational target-specific drug discovery. Results\nshow that our proposed TagMol can generate molecules with similar binding\naffinity scores as real molecules. GAT-based models showed faster and better\nlearning relative to GCN baseline models.\n", "date": "2022-12-05", "categories": []}
{"id": "2212.00015", "title": "Scalable Pathogen Detection from Next Generation DNA Sequencing with\n  Deep Learning", "abstract": "  Next-generation sequencing technologies have enhanced the scope of\nInternet-of-Things (IoT) to include genomics for personalized medicine through\nthe increased availability of an abundance of genome data collected from\nheterogeneous sources at a reduced cost. Given the sheer magnitude of the\ncollected data and the significant challenges offered by the presence of highly\nsimilar genomic structure across species, there is a need for robust, scalable\nanalysis platforms to extract actionable knowledge such as the presence of\npotentially zoonotic pathogens. The emergence of zoonotic diseases from novel\npathogens, such as the influenza virus in 1918 and SARS-CoV-2 in 2019 that can\njump species barriers and lead to pandemic underscores the need for scalable\nmetagenome analysis. In this work, we propose MG2Vec, a deep learning-based\nsolution that uses the transformer network as its backbone, to learn robust\nfeatures from raw metagenome sequences for downstream biomedical tasks such as\ntargeted and generalized pathogen detection. Extensive experiments on four\nincreasingly challenging, yet realistic diagnostic settings, show that the\nproposed approach can help detect pathogens from uncurated, real-world clinical\nsamples with minimal human supervision in the form of labels. Further, we\ndemonstrate that the learned representations can generalize to completely\nunrelated pathogens across diseases and species for large-scale metagenome\nanalysis. We provide a comprehensive evaluation of a novel representation\nlearning framework for metagenome-based disease diagnostics with deep learning\nand provide a way forward for extracting and using robust vector\nrepresentations from low-cost next generation sequencing to develop\ngeneralizable diagnostic tools.\n", "date": "2022-11-30", "categories": []}
{"id": "2212.00502", "title": "Cellular Automata Model for Non-Structural Proteins Comparing\n  Transmissibility and Pathogenesis of SARS Covid (CoV-2, CoV) and MERS Covid", "abstract": "  Significantly higher transmissibility of SARS CoV-2 (2019) compared to SARS\nCoV (2003) can be attributed to mutations of structural proteins (Spike S,\nNucleocapsid N, Membrane M, and Envelope E) and the role played by\nnon-structural proteins (nsps) and accessory proteins (ORFs) for viral\nreplication, assembly and shedding. The non-structural proteins (nsps) avail\nhost protein synthesis machinery to initiate viral replication, along with\nneutralization of host immune defense. The key protein out of the 16 nsps, is\nthe non-structural protein nsp1, also known as the leader protein. Nsp1 leads\nthe process of hijacking host resources by blocking host translation. This\npaper concentrates on the analysis of nsps of SARS covid (CoV-2, CoV) and MERS\ncovid based on Cellular Automata enhanced Machine Learning (CAML) model\ndeveloped for study of biological strings. This computational model compares\ndeviation of structure - function of CoV-2 from that of CoV employing CAML\nmodel parameters derived out of CA evolution of amino acid chains of nsps. This\ncomparative analysis points to - (i) higher transmissibility of CoV-2 compared\nto CoV for major nsps, and (ii) deviation of MERS covid from SARS CoV in\nrespect of virulence and pathogenesis. A Machine Learning (ML) framework has\nbeen designed to map the CAML model parameters to the physical domain features\nreported in in-vitro/in-vivo/in-silico experimental studies. The ML framework\nenables us to learn the permissible range of model parameters derived out of\nmutational study of sixteen nsps of three viruses.\n", "date": "2022-11-25", "categories": []}
{"id": "2211.16508", "title": "Reinforced Genetic Algorithm for Structure-based Drug Design", "abstract": "  Structure-based drug design (SBDD) aims to discover drug candidates by\nfinding molecules (ligands) that bind tightly to a disease-related protein\n(targets), which is the primary approach to computer-aided drug discovery.\nRecently, applying deep generative models for three-dimensional (3D) molecular\ndesign conditioned on protein pockets to solve SBDD has attracted much\nattention, but their formulation as probabilistic modeling often leads to\nunsatisfactory optimization performance. On the other hand, traditional\ncombinatorial optimization methods such as genetic algorithms (GA) have\ndemonstrated state-of-the-art performance in various molecular optimization\ntasks. However, they do not utilize protein target structure to inform design\nsteps but rely on a random-walk-like exploration, which leads to unstable\nperformance and no knowledge transfer between different tasks despite the\nsimilar binding physics. To achieve a more stable and efficient SBDD, we\npropose Reinforced Genetic Algorithm (RGA) that uses neural models to\nprioritize the profitable design steps and suppress random-walk behavior. The\nneural models take the 3D structure of the targets and ligands as inputs and\nare pre-trained using native complex structures to utilize the knowledge of the\nshared binding physics from different targets and then fine-tuned during\noptimization. We conduct thorough empirical studies on optimizing binding\naffinity to various disease targets and show that RGA outperforms the baselines\nin terms of docking scores and is more robust to random initializations. The\nablation study also indicates that the training on different targets helps\nimprove performance by leveraging the shared underlying physics of the binding\nprocesses. The code is available at\nhttps://github.com/futianfan/reinforced-genetic-algorithm.\n", "date": "2022-11-28", "categories": []}
{"id": "2211.15720", "title": "Predicting pathways for old and new metabolites through clustering", "abstract": "  The diverse metabolic pathways are fundamental to all living organisms, as\nthey harvest energy, synthesize biomass components, produce molecules to\ninteract with the microenvironment, and neutralize toxins. While discovery of\nnew metabolites and pathways continues, the prediction of pathways for new\nmetabolites can be challenging. It can take vast amounts of time to elucidate\npathways for new metabolites; thus, according to HMDB only 60% of metabolites\nget assigned to pathways. Here, we present an approach to identify pathways\nbased on metabolite structure. We extracted 201 features from SMILES\nannotations, and identified new metabolites from PubMed abstracts and HMDB.\nAfter applying clustering algorithms to both groups of features, we quantified\ncorrelations between metabolites, and found the clusters accurately linked 92%\nof known metabolites to their respective pathways. Thus, this approach could be\nvaluable for predicting metabolic pathways for new metabolites.\n", "date": "2022-11-28", "categories": []}
{"id": "2211.10546", "title": "Evaluating COVID-19 Sequence Data Using Nearest-Neighbors Based Network\n  Model", "abstract": "  The SARS-CoV-2 coronavirus is the cause of the COVID-19 disease in humans.\nLike many coronaviruses, it can adapt to different hosts and evolve into\ndifferent lineages. It is well-known that the major SARS-CoV-2 lineages are\ncharacterized by mutations that happen predominantly in the spike protein.\nUnderstanding the spike protein structure and how it can be perturbed is vital\nfor understanding and determining if a lineage is of concern. These are crucial\nto identifying and controlling current outbreaks and preventing future\npandemics. Machine learning (ML) methods are a viable solution to this effort,\ngiven the volume of available sequencing data, much of which is unaligned or\neven unassembled. However, such ML methods require fixed-length numerical\nfeature vectors in Euclidean space to be applicable. Similarly, euclidean space\nis not considered the best choice when working with the classification and\nclustering tasks for biological sequences. For this purpose, we design a method\nthat converts the protein (spike) sequences into the sequence similarity\nnetwork (SSN). We can then use SSN as an input for the classical algorithms\nfrom the graph mining domain for the typical tasks such as classification and\nclustering to understand the data. We show that the proposed alignment-free\nmethod is able to outperform the current SOTA method in terms of clustering\nresults. Similarly, we are able to achieve higher classification accuracy using\nwell-known Node2Vec-based embedding compared to other baseline embedding\napproaches.\n", "date": "2022-11-19", "categories": []}
{"id": "2208.04944", "title": "Bridging the gap between target-based and cell-based drug discovery with\n  a graph generative multi-task model", "abstract": "  Drug discovery is vitally important for protecting human against disease.\nTarget-based screening is one of the most popular methods to develop new drugs\nin the past several decades. This method efficiently screens candidate drugs\ninhibiting target protein in vitro, but it often fails due to inadequate\nactivity of the selected drugs in vivo. Accurate computational methods are\nneeded to bridge this gap. Here, we propose a novel graph multi task deep\nlearning model to identify compounds carrying both target inhibitory and cell\nactive (MATIC) properties. On a carefully curated SARS-CoV-2 dataset, the\nproposed MATIC model shows advantages comparing with traditional method in\nscreening effective compounds in vivo. Next, we explored the model\ninterpretability and found that the learned features for target inhibition (in\nvitro) or cell active (in vivo) tasks are different with molecular property\ncorrelations and atom functional attentions. Based on these findings, we\nutilized a monte carlo based reinforcement learning generative model to\ngenerate novel multi-property compounds with both in vitro and in vivo\nefficacy, thus bridging the gap between target-based and cell-based drug\ndiscovery.\n", "date": "2022-08-09", "categories": []}
{"id": "2211.10442", "title": "Deep learning methods for drug response prediction in cancer:\n  predominant and emerging trends", "abstract": "  Cancer claims millions of lives yearly worldwide. While many therapies have\nbeen made available in recent years, by in large cancer remains unsolved.\nExploiting computational predictive models to study and treat cancer holds\ngreat promise in improving drug development and personalized design of\ntreatment plans, ultimately suppressing tumors, alleviating suffering, and\nprolonging lives of patients. A wave of recent papers demonstrates promising\nresults in predicting cancer response to drug treatments while utilizing deep\nlearning methods. These papers investigate diverse data representations, neural\nnetwork architectures, learning methodologies, and evaluations schemes.\nHowever, deciphering promising predominant and emerging trends is difficult due\nto the variety of explored methods and lack of standardized framework for\ncomparing drug response prediction models. To obtain a comprehensive landscape\nof deep learning methods, we conducted an extensive search and analysis of deep\nlearning models that predict the response to single drug treatments. A total of\n60 deep learning-based models have been curated and summary plots were\ngenerated. Based on the analysis, observable patterns and prevalence of methods\nhave been revealed. This review allows to better understand the current state\nof the field and identify major challenges and promising solution paths.\n", "date": "2022-11-18", "categories": []}
{"id": "2211.09862", "title": "Knowledge distillation for fast and accurate DNA sequence correction", "abstract": "  Accurate genome sequencing can improve our understanding of biology and the\ngenetic basis of disease. The standard approach for generating DNA sequences\nfrom PacBio instruments relies on HMM-based models. Here, we introduce\nDistilled DeepConsensus - a distilled transformer-encoder model for sequence\ncorrection, which improves upon the HMM-based methods with runtime constraints\nin mind. Distilled DeepConsensus is 1.3x faster and 1.5x smaller than its\nlarger counterpart while improving the yield of high quality reads (Q30) over\nthe HMM-based method by 1.69x (vs. 1.73x for larger model). With improved\naccuracy of genomic sequences, Distilled DeepConsensus improves downstream\napplications of genomic sequence analysis such as reducing variant calling\nerrors by 39% (34% for larger model) and improving genome assembly quality by\n3.8% (4.2% for larger model). We show that the representations learned by\nDistilled DeepConsensus are similar between faster and slower models.\n", "date": "2022-11-17", "categories": []}
{"id": "2211.08406", "title": "Incorporating Pre-training Paradigm for Antibody Sequence-Structure\n  Co-design", "abstract": "  Antibodies are versatile proteins that can bind to pathogens and provide\neffective protection for human body. Recently, deep learning-based\ncomputational antibody design has attracted popular attention since it\nautomatically mines the antibody patterns from data that could be complementary\nto human experiences. However, the computational methods heavily rely on\nhigh-quality antibody structure data, which is quite limited. Besides, the\ncomplementarity-determining region (CDR), which is the key component of an\nantibody that determines the specificity and binding affinity, is highly\nvariable and hard to predict. Therefore, the data limitation issue further\nraises the difficulty of CDR generation for antibodies. Fortunately, there\nexists a large amount of sequence data of antibodies that can help model the\nCDR and alleviate the reliance on structure data. By witnessing the success of\npre-training models for protein modeling, in this paper, we develop the\nantibody pre-training language model and incorporate it into the\n(antigen-specific) antibody design model in a systemic way. Specifically, we\nfirst pre-train an antibody language model based on the sequence data, then\npropose a one-shot way for sequence and structure generation of CDR to avoid\nthe heavy cost and error propagation from an autoregressive manner, and finally\nleverage the pre-trained antibody model for the antigen-specific antibody\ngeneration model with some carefully designed modules. Through various\nexperiments, we show that our method achieves superior performances over\nprevious baselines on different tasks, such as sequence and structure\ngeneration and antigen-binding CDR-H3 design.\n", "date": "2022-10-26", "categories": []}
{"id": "2211.08267", "title": "Reads2Vec: Efficient Embedding of Raw High-Throughput Sequencing Reads\n  Data", "abstract": "  The massive amount of genomic data appearing for SARS-CoV-2 since the\nbeginning of the COVID-19 pandemic has challenged traditional methods for\nstudying its dynamics. As a result, new methods such as Pangolin, which can\nscale to the millions of samples of SARS-CoV-2 currently available, have\nappeared. Such a tool is tailored to take as input assembled, aligned and\ncurated full-length sequences, such as those found in the GISAID database. As\nhigh-throughput sequencing technologies continue to advance, such assembly,\nalignment and curation may become a bottleneck, creating a need for methods\nwhich can process raw sequencing reads directly.\n  In this paper, we propose Reads2Vec, an alignment-free embedding approach\nthat can generate a fixed-length feature vector representation directly from\nthe raw sequencing reads without requiring assembly. Furthermore, since such an\nembedding is a numerical representation, it may be applied to highly optimized\nclassification and clustering algorithms. Experiments on simulated data show\nthat our proposed embedding obtains better classification results and better\nclustering properties contrary to existing alignment-free baselines. In a study\non real data, we show that alignment-free embeddings have better clustering\nproperties than the Pangolin tool and that the spike region of the SARS-CoV-2\ngenome heavily informs the alignment-free clusterings, which is consistent with\ncurrent biological knowledge of SARS-CoV-2.\n", "date": "2022-11-15", "categories": []}
{"id": "2211.06430", "title": "Efficient HLA imputation from sequential SNPs data by Transformer", "abstract": "  Human leukocyte antigen (HLA) genes are associated with a variety of\ndiseases, however direct typing of HLA is time and cost consuming. Thus various\nimputation methods using sequential SNPs data have been proposed based on\nstatistical or deep learning models, e.g. CNN-based model, named DEEP*HLA.\nHowever, imputation efficiency is not sufficient for in frequent alleles and a\nlarge size of reference panel is required. Here, we developed a\nTransformer-based model to impute HLA alleles, named \"HLA Reliable IMputatioN\nby Transformer (HLARIMNT)\" to take advantage of sequential nature of SNPs data.\nWe validated the performance of HLARIMNT using two different reference panels;\nPan-Asian reference panel (n = 530) and Type 1 Diabetes Genetics Consortium\n(T1DGC) reference panel (n = 5,225), as well as the mixture of those two panels\n(n = 1,060). HLARIMNT achieved higher accuracy than DEEP*HLA by several\nindices, especially for infrequent alleles. We also varied the size of data\nused for training, and HLARIMNT imputed more accurately among any size of\ntraining data. These results suggest that Transformer-based model may impute\nefficiently not only HLA types but also any other gene types from sequential\nSNPs data.\n", "date": "2022-11-11", "categories": []}
{"id": "2211.06792", "title": "Drug-target affinity prediction method based on consistent expression of\n  heterogeneous data", "abstract": "  The first step in drug discovery is finding drug molecule moieties with\nmedicinal activity against specific targets. Therefore, it is crucial to\ninvestigate the interaction between drug-target proteins and small chemical\nmolecules. However, traditional experimental methods for discovering potential\nsmall drug molecules are labor-intensive and time-consuming. There is currently\na lot of interest in building computational models to screen small drug\nmolecules using drug molecule-related databases. In this paper, we propose a\nmethod for predicting drug-target binding affinity using deep learning models.\nThis method uses a modified GRU and GNN to extract features from the\ndrug-target protein sequences and the drug molecule map, respectively, to\nobtain their feature vectors. The combined vectors are used as vector\nrepresentations of drug-target molecule pairs and then fed into a fully\nconnected network to predict drug-target binding affinity. This proposed model\ndemonstrates its accuracy and effectiveness in predicting drug-target binding\naffinity on the DAVIS and KIBA datasets.\n", "date": "2022-11-13", "categories": []}
{"id": "2211.07105", "title": "Bayesian Reconstruction and Differential Testing of Excised mRNA", "abstract": "  Characterizing the differential excision of mRNA is critical for\nunderstanding the functional complexity of a cell or tissue, from normal\ndevelopmental processes to disease pathogenesis. Most transcript reconstruction\nmethods infer full-length transcripts from high-throughput sequencing data.\nHowever, this is a challenging task due to incomplete annotations and the\ndifferential expression of transcripts across cell-types, tissues, and\nexperimental conditions. Several recent methods circumvent these difficulties\nby considering local splicing events, but these methods lose transcript-level\nsplicing information and may conflate transcripts. We develop the first\nprobabilistic model that reconciles the transcript and local splicing\nperspectives. First, we formalize the sequence of mRNA excisions (SME)\nreconstruction problem, which aims to assemble variable-length sequences of\nmRNA excisions from RNA-sequencing data. We then present a novel hierarchical\nBayesian admixture model for the Reconstruction of Excised mRNA (BREM). BREM\ninterpolates between local splicing events and full-length transcripts and thus\nfocuses only on SMEs that have high posterior probability. We develop posterior\ninference algorithms based on Gibbs sampling and local search of independent\nsets and characterize differential SME usage using generalized linear models\nbased on converged BREM model parameters. We show that BREM achieves higher F1\nscore for reconstruction tasks and improved accuracy and sensitivity in\ndifferential splicing when compared with four state-of-the-art transcript and\nlocal splicing methods on simulated data. Lastly, we evaluate BREM on both bulk\nand scRNA sequencing data based on transcript reconstruction, novelty of\ntranscripts produced, model sensitivity to hyperparameters, and a functional\nanalysis of differentially expressed SMEs, demonstrating that BREM captures\nrelevant biological signal.\n", "date": "2022-11-14", "categories": []}
{"id": "2211.05935", "title": "Inferring probabilistic Boolean networks from steady-state gene data\n  samples", "abstract": "  Probabilistic Boolean Networks have been proposed for estimating the\nbehaviour of dynamical systems as they combine rule-based modelling with\nuncertainty principles. Inferring PBNs directly from gene data is challenging\nhowever, especially when data is costly to collect and/or noisy, e.g., in the\ncase of gene expression profile data. In this paper, we present a reproducible\nmethod for inferring PBNs directly from real gene expression data measurements\ntaken when the system was at a steady state. The steady-state dynamics of PBNs\nis of special interest in the analysis of biological machinery. The proposed\napproach does not rely on reconstructing the state evolution of the network,\nwhich is computationally intractable for larger networks. We demonstrate the\nmethod on samples of real gene expression profiling data from a well-known\nstudy on metastatic melanoma. The pipeline is implemented using Python and we\nmake it publicly available.\n", "date": "2022-11-11", "categories": []}
{"id": "2206.01092", "title": "Innovations in Integrating Machine Learning and Agent-Based Modeling of\n  Biomedical Systems", "abstract": "  Agent-based modeling (ABM) is a well-established paradigm for simulating\ncomplex systems via interactions between constituent entities. Machine learning\n(ML) refers to approaches whereby statistical algorithms 'learn' from data on\ntheir own, without imposing a priori theories of system behavior. Biological\nsystems -- from molecules, to cells, to entire organisms -- consist of vast\nnumbers of entities, governed by complex webs of interactions that span many\nspatiotemporal scales and exhibit nonlinearity, stochasticity and intricate\ncoupling between entities. The macroscopic properties and collective dynamics\nof such systems are difficult to capture via continuum modelling and mean-field\nformalisms. ABM takes a 'bottom-up' approach that obviates these difficulties\nby enabling one to easily propose and test a set of well-defined 'rules' to be\napplied to the individual entities (agents) in a system. Evaluating a system\nand propagating its state over discrete time-steps effectively simulates the\nsystem, allowing observables to be computed and system properties to be\nanalyzed. Because the rules that govern an ABM can be difficult to abstract and\nformulate from experimental data, there is an opportunity to use ML to help\ninfer optimal, system-specific ABM rules. Once such rule-sets are devised, ABM\ncalculations can generate a wealth of data, and ML can be applied there too --\ne.g., to probe statistical measures that meaningfully describe a system's\nstochastic properties. As an example of synergy in the other direction (from\nABM to ML), ABM simulations can generate realistic datasets for training ML\nalgorithms (e.g., for regularization, to mitigate overfitting). In these ways,\none can envision various synergistic ABM$\\rightleftharpoons$ML loops. This\nreview summarizes how ABM and ML have been integrated in contexts that span\nspatiotemporal scales, from cellular to population-level epidemiology.\n", "date": "2022-06-02", "categories": []}
{"id": "2211.03793", "title": "Uncertainty Quantification for Atlas-Level Cell Type Transfer", "abstract": "  Single-cell reference atlases are large-scale, cell-level maps that capture\ncellular heterogeneity within an organ using single cell genomics. Given their\nsize and cellular diversity, these atlases serve as high-quality training data\nfor the transfer of cell type labels to new datasets. Such label transfer,\nhowever, must be robust to domain shifts in gene expression due to measurement\ntechnique, lab specifics and more general batch effects. This requires methods\nthat provide uncertainty estimates on the cell type predictions to ensure\ncorrect interpretation. Here, for the first time, we introduce uncertainty\nquantification methods for cell type classification on single-cell reference\natlases. We benchmark four model classes and show that currently used models\nlack calibration, robustness, and actionable uncertainty scores. Furthermore,\nwe demonstrate how models that quantify uncertainty are better suited to detect\nunseen cell types in the setting of atlas-level cell type transfer.\n", "date": "2022-11-07", "categories": []}
{"id": "2209.06716", "title": "Modelling Technical and Biological Effects in scRNA-seq data with\n  Scalable GPLVMs", "abstract": "  Single-cell RNA-seq datasets are growing in size and complexity, enabling the\nstudy of cellular composition changes in various biological/clinical contexts.\nScalable dimensionality reduction techniques are in need to disentangle\nbiological variation in them, while accounting for technical and biological\nconfounders. In this work, we extend a popular approach for probabilistic\nnon-linear dimensionality reduction, the Gaussian process latent variable\nmodel, to scale to massive single-cell datasets while explicitly accounting for\ntechnical and biological confounders. The key idea is to use an augmented\nkernel which preserves the factorisability of the lower bound allowing for fast\nstochastic variational inference. We demonstrate its ability to reconstruct\nlatent signatures of innate immunity recovered in Kumasaka et al. (2021) with\n9x lower training time. We further analyze a COVID dataset and demonstrate\nacross a cohort of 130 individuals, that this framework enables data\nintegration while capturing interpretable signatures of infection.\nSpecifically, we explore COVID severity as a latent dimension to refine patient\nstratification and capture disease-specific gene expression.\n", "date": "2022-09-14", "categories": []}
{"id": "2210.12385", "title": "Deep Learning in Single-Cell Analysis", "abstract": "  Single-cell technologies are revolutionizing the entire field of biology. The\nlarge volumes of data generated by single-cell technologies are\nhigh-dimensional, sparse, heterogeneous, and have complicated dependency\nstructures, making analyses using conventional machine learning approaches\nchallenging and impractical. In tackling these challenges, deep learning often\ndemonstrates superior performance compared to traditional machine learning\nmethods. In this work, we give a comprehensive survey on deep learning in\nsingle-cell analysis. We first introduce background on single-cell technologies\nand their development, as well as fundamental concepts of deep learning\nincluding the most popular deep architectures. We present an overview of the\nsingle-cell analytic pipeline pursued in research applications while noting\ndivergences due to data sources or specific applications. We then review seven\npopular tasks spanning through different stages of the single-cell analysis\npipeline, including multimodal integration, imputation, clustering, spatial\ndomain identification, cell-type deconvolution, cell segmentation, and\ncell-type annotation. Under each task, we describe the most recent developments\nin classical and deep learning methods and discuss their advantages and\ndisadvantages. Deep learning tools and benchmark datasets are also summarized\nfor each task. Finally, we discuss the future directions and the most recent\nchallenges. This survey will serve as a reference for biologists and computer\nscientists, encouraging collaborations.\n", "date": "2022-10-22", "categories": []}
{"id": "2211.02657", "title": "MolE: a molecular foundation model for drug discovery", "abstract": "  Models that accurately predict properties based on chemical structure are\nvaluable tools in drug discovery. However, for many properties, public and\nprivate training sets are typically small, and it is difficult for the models\nto generalize well outside of the training data. Recently, large language\nmodels have addressed this problem by using self-supervised pretraining on\nlarge unlabeled datasets, followed by fine-tuning on smaller, labeled datasets.\nIn this paper, we report MolE, a molecular foundation model that adapts the\nDeBERTa architecture to be used on molecular graphs together with a two-step\npretraining strategy. The first step of pretraining is a self-supervised\napproach focused on learning chemical structures, and the second step is a\nmassive multi-task approach to learn biological information. We show that\nfine-tuning pretrained MolE achieves state-of-the-art results on 9 of the 22\nADMET tasks included in the Therapeutic Data Commons.\n", "date": "2022-11-03", "categories": []}
{"id": "2211.02660", "title": "De novo PROTAC design using graph-based deep generative models", "abstract": "  PROteolysis TArgeting Chimeras (PROTACs) are an emerging therapeutic modality\nfor degrading a protein of interest (POI) by marking it for degradation by the\nproteasome. Recent developments in artificial intelligence (AI) suggest that\ndeep generative models can assist with the de novo design of molecules with\ndesired properties, and their application to PROTAC design remains largely\nunexplored. We show that a graph-based generative model can be used to propose\nnovel PROTAC-like structures from empty graphs. Our model can be guided towards\nthe generation of large molecules (30--140 heavy atoms) predicted to degrade a\nPOI through policy-gradient reinforcement learning (RL). Rewards during RL are\napplied using a boosted tree surrogate model that predicts a molecule's\ndegradation potential for each POI. Using this approach, we steer the\ngenerative model towards compounds with higher likelihoods of predicted\ndegradation activity. Despite being trained on sparse public data, the\ngenerative model proposes molecules with substructures found in known\ndegraders. After fine-tuning, predicted activity against a challenging POI\nincreases from 50% to >80% with near-perfect chemical validity for sampled\ncompounds, suggesting this is a promising approach for the optimization of\nlarge, PROTAC-like molecules for targeted protein degradation.\n", "date": "2022-11-04", "categories": []}
{"id": "2211.02935", "title": "Efficient Cavity Searching for Gene Network of Influenza A Virus", "abstract": "  High order structures (cavities and cliques) of the gene network of influenza\nA virus reveal tight associations among viruses during evolution and are key\nsignals that indicate viral cross-species infection and cause pandemics. As\nindicators for sensing the dynamic changes of viral genes, these higher order\nstructures have been the focus of attention in the field of virology. However,\nthe size of the viral gene network is usually huge, and searching these\nstructures in the networks introduces unacceptable delay. To mitigate this\nissue, in this paper, we propose a simple-yet-effective model named HyperSearch\nbased on deep learning to search cavities in a computable complex network for\ninfluenza virus genetics. Extensive experiments conducted on a public influenza\nvirus dataset demonstrate the effectiveness of HyperSearch over other advanced\ndeep-learning methods without any elaborated model crafting. Moreover,\nHyperSearch can finish the search works in minutes while 0-1 programming takes\ndays. Since the proposed method is simple and easy to be transferred to other\ncomplex networks, HyperSearch has the potential to facilitate the monitoring of\ndynamic changes in viral genes and help humans keep up with the pace of virus\nmutations.\n", "date": "2022-11-05", "categories": []}
{"id": "2210.01123", "title": "Towards Learned Simulators for Cell Migration", "abstract": "  Simulators driven by deep learning are gaining popularity as a tool for\nefficiently emulating accurate but expensive numerical simulators. Successful\napplications of such neural simulators can be found in the domains of physics,\nchemistry, and structural biology, amongst others. Likewise, a neural simulator\nfor cellular dynamics can augment lab experiments and traditional computational\nmethods to enhance our understanding of a cell's interaction with its physical\nenvironment. In this work, we propose an autoregressive probabilistic model\nthat can reproduce spatiotemporal dynamics of single cell migration,\ntraditionally simulated with the Cellular Potts model. We observe that standard\nsingle-step training methods do not only lead to inconsistent rollout\nstability, but also fail to accurately capture the stochastic aspects of the\ndynamics, and we propose training strategies to mitigate these issues. Our\nevaluation on two proof-of-concept experimental scenarios shows that neural\nmethods have the potential to faithfully simulate stochastic cellular dynamics\nat least an order of magnitude faster than a state-of-the-art implementation of\nthe Cellular Potts model.\n", "date": "2022-10-02", "categories": []}
{"id": "2211.00625", "title": "Machine learning can guide experimental approaches for protein\n  digestibility estimations", "abstract": "  Food protein digestibility and bioavailability are critical aspects in\naddressing human nutritional demands, particularly when seeking sustainable\nalternatives to animal-based proteins. In this study, we propose a machine\nlearning approach to predict the true ileal digestibility coefficient of food\nitems. The model makes use of a unique curated dataset that combines\nnutritional information from different foods with FASTA sequences of some of\ntheir protein families. We extracted the biochemical properties of the proteins\nand combined these properties with embeddings from a Transformer-based protein\nLanguage Model (pLM). In addition, we used SHAP to identify features that\ncontribute most to the model prediction and provide interpretability. This\nfirst AI-based model for predicting food protein digestibility has an accuracy\nof 90% compared to existing experimental techniques. With this accuracy, our\nmodel can eliminate the need for lengthy in-vivo or in-vitro experiments,\nmaking the process of creating new foods faster, cheaper, and more ethical.\n", "date": "2022-11-01", "categories": []}
{"id": "2210.16721", "title": "Exemplar Guided Deep Neural Network for Spatial Transcriptomics Analysis\n  of Gene Expression Prediction", "abstract": "  Spatial transcriptomics (ST) is essential for understanding diseases and\ndeveloping novel treatments. It measures gene expression of each fine-grained\narea (i.e., different windows) in the tissue slide with low throughput. This\npaper proposes an Exemplar Guided Network (EGN) to accurately and efficiently\npredict gene expression directly from each window of a tissue slide image. We\napply exemplar learning to dynamically boost gene expression prediction from\nnearest/similar exemplars of a given tissue slide image window. Our EGN\nframework composes of three main components: 1) an extractor to structure a\nrepresentation space for unsupervised exemplar retrievals; 2) a vision\ntransformer (ViT) backbone to progressively extract representations of the\ninput window; and 3) an Exemplar Bridging (EB) block to adaptively revise the\nintermediate ViT representations by using the nearest exemplars. Finally, we\ncomplete the gene expression prediction task with a simple attention-based\nprediction block. Experiments on standard benchmark datasets indicate the\nsuperiority of our approach when comparing with the past state-of-the-art\n(SOTA) methods.\n", "date": "2022-10-30", "categories": []}
{"id": "2209.07527", "title": "Improved proteasomal cleavage prediction with positive-unlabeled\n  learning", "abstract": "  Accurate in silico modeling of the antigen processing pathway is crucial to\nenable personalized epitope vaccine design for cancer. An important step of\nsuch pathway is the degradation of the vaccine into smaller peptides by the\nproteasome, some of which are going to be presented to T cells by the MHC\ncomplex. While predicting MHC-peptide presentation has received a lot of\nattention recently, proteasomal cleavage prediction remains a relatively\nunexplored area in light of recent advancesin high-throughput mass\nspectrometry-based MHC ligandomics. Moreover, as such experimental techniques\ndo not allow to identify regions that cannot be cleaved, the latest predictors\ngenerate decoy negative samples and treat them as true negatives when training,\neven though some of them could actually be positives. In this work, we thus\npresent a new predictor trained with an expanded dataset and the solid\ntheoretical underpinning of positive-unlabeled learning, achieving a new\nstate-of-the-art in proteasomal cleavage prediction. The improved predictive\ncapabilities will in turn enable more precise vaccine development improving the\nefficacy of epitope-based vaccines. Pretrained models are available on GitHub\n", "date": "2022-09-14", "categories": []}
{"id": "2210.16098", "title": "Predicting Protein-Ligand Binding Affinity with Equivariant Line Graph\n  Network", "abstract": "  Binding affinity prediction of three-dimensional (3D) protein ligand\ncomplexes is critical for drug repositioning and virtual drug screening.\nExisting approaches transform a 3D protein-ligand complex to a two-dimensional\n(2D) graph, and then use graph neural networks (GNNs) to predict its binding\naffinity. However, the node and edge features of the 2D graph are extracted\nbased on invariant local coordinate systems of the 3D complex. As a result, the\nmethod can not fully learn the global information of the complex, such as, the\nphysical symmetry and the topological information of bonds. To address these\nissues, we propose a novel Equivariant Line Graph Network (ELGN) for affinity\nprediction of 3D protein ligand complexes. The proposed ELGN firstly adds a\nsuper node to the 3D complex, and then builds a line graph based on the 3D\ncomplex. After that, ELGN uses a new E(3)-equivariant network layer to pass the\nmessages between nodes and edges based on the global coordinate system of the\n3D complex. Experimental results on two real datasets demonstrate the\neffectiveness of ELGN over several state-of-the-art baselines.\n", "date": "2022-10-27", "categories": []}
{"id": "2208.05004", "title": "CoViT: Real-time phylogenetics for the SARS-CoV-2 pandemic using Vision\n  Transformers", "abstract": "  Real-time viral genome detection, taxonomic classification and phylogenetic\nanalysis are critical for efficient tracking and control of viral pandemics\nsuch as Covid-19. However, the unprecedented and still growing amounts of viral\ngenome data create a computational bottleneck, which effectively prevents the\nreal-time pandemic tracking. For genomic tracing to work effectively, each new\nviral genome sequence must be placed in its pangenomic context. Re-inferring\nthe full phylogeny of SARS-CoV-2, with datasets containing millions of samples,\nis prohibitively slow even using powerful computational resources. We are\nattempting to alleviate the computational bottleneck by modifying and applying\nVision Transformer, a recently developed neural network model for image\nrecognition, to taxonomic classification and placement of viral genomes, such\nas SARS-CoV-2. Our solution, CoViT, places SARS-CoV-2 genome accessions onto\nSARS-CoV-2 phylogenetic tree with the accuracy of 94.2%. Since CoViT is a\nclassification neural network, it provides more than one likely placement.\nSpecifically, one of the two most likely placements suggested by CoViT is\ncorrect with the probability of 97.9%. The probability of the correct placement\nto be found among the five most likely placements generated by CoViT is 99.8%.\nThe placement time is 0.055s per individual genome running on NVIDIAs GeForce\nRTX 2080 Ti GPU. We make CoViT available to research community through GitHub:\nhttps://github.com/zuherJahshan/covit.\n", "date": "2022-08-09", "categories": []}
{"id": "2210.12229", "title": "Deep Reinforcement Learning for Stabilization of Large-scale\n  Probabilistic Boolean Networks", "abstract": "  The ability to direct a Probabilistic Boolean Network (PBN) to a desired\nstate is important to applications such as targeted therapeutics in cancer\nbiology. Reinforcement Learning (RL) has been proposed as a framework that\nsolves a discrete-time optimal control problem cast as a Markov Decision\nProcess. We focus on an integrative framework powered by a model-free deep RL\nmethod that can address different flavours of the control problem (e.g., with\nor without control inputs; attractor state or a subset of the state space as\nthe target domain). The method is agnostic to the distribution of probabilities\nfor the next state, hence it does not use the probability transition matrix.\nThe time complexity is linear on the time steps, or interactions between the\nagent (deep RL) and the environment (PBN), during training. Indeed, we explore\nthe scalability of the deep RL approach to (set) stabilization of large-scale\nPBNs and demonstrate successful control on large networks, including a\nmetastatic melanoma PBN with 200 nodes.\n", "date": "2022-10-21", "categories": []}
{"id": "2210.13709", "title": "InForecaster: Forecasting Influenza Hemagglutinin Mutations Through the\n  Lens of Anomaly Detection", "abstract": "  The influenza virus hemagglutinin is an important part of the virus\nattachment to the host cells. The hemagglutinin proteins are one of the genetic\nregions of the virus with a high potential for mutations. Due to the importance\nof predicting mutations in producing effective and low-cost vaccines, solutions\nthat attempt to approach this problem have recently gained a significant\nattention. A historical record of mutations have been used to train predictive\nmodels in such solutions. However, the imbalance between mutations and the\npreserved proteins is a big challenge for the development of such models that\nneeds to be addressed. Here, we propose to tackle this challenge through\nanomaly detection (AD). AD is a well-established field in Machine Learning (ML)\nthat tries to distinguish unseen anomalies from the normal patterns using only\nnormal training samples. By considering mutations as the anomalous behavior, we\ncould benefit existing rich solutions in this field that have emerged recently.\nSuch methods also fit the problem setup of extreme imbalance between the number\nof unmutated vs. mutated training samples. Motivated by this formulation, our\nmethod tries to find a compact representation for unmutated samples while\nforcing anomalies to be separated from the normal ones. This helps the model to\nlearn a shared unique representation between normal training samples as much as\npossible, which improves the discernibility and detectability of mutated\nsamples from the unmutated ones at the test time. We conduct a large number of\nexperiments on four publicly available datasets, consisting of 3 different\nhemagglutinin protein datasets, and one SARS-CoV-2 dataset, and show the\neffectiveness of our method through different standard criteria.\n", "date": "2022-10-25", "categories": []}
{"id": "2210.12158", "title": "Graph Coloring via Neural Networks for Haplotype Assembly and Viral\n  Quasispecies Reconstruction", "abstract": "  Understanding genetic variation, e.g., through mutations, in organisms is\ncrucial to unravel their effects on the environment and human health. A\nfundamental characterization can be obtained by solving the haplotype assembly\nproblem, which yields the variation across multiple copies of chromosomes.\nVariations among fast evolving viruses that lead to different strains (called\nquasispecies) are also deciphered with similar approaches. In both these cases,\nhigh-throughput sequencing technologies that provide oversampled mixtures of\nlarge noisy fragments (reads) of genomes, are used to infer constituent\ncomponents (haplotypes or quasispecies). The problem is harder for polyploid\nspecies where there are more than two copies of chromosomes. State-of-the-art\nneural approaches to solve this NP-hard problem do not adequately model\nrelations among the reads that are important for deconvolving the input signal.\nWe address this problem by developing a new method, called NeurHap, that\ncombines graph representation learning with combinatorial optimization. Our\nexperiments demonstrate substantially better performance of NeurHap in real and\nsynthetic datasets compared to competing approaches.\n", "date": "2022-10-21", "categories": []}
{"id": "2210.11482", "title": "A Methodology for the Prediction of Drug Target Interaction using CDK\n  Descriptors", "abstract": "  Detecting probable Drug Target Interaction (DTI) is a critical task in drug\ndiscovery. Conventional DTI studies are expensive, labor-intensive, and take a\nlot of time, hence there are significant reasons to construct useful\ncomputational techniques that may successfully anticipate possible DTIs.\nAlthough certain methods have been developed for this cause, numerous\ninteractions are yet to be discovered, and prediction accuracy is still low. To\nmeet these challenges, we propose a DTI prediction model built on molecular\nstructure of drugs and sequence of target proteins. In the proposed model, we\nuse Simplified Molecular Input Line Entry System (SMILES) to create CDK\ndescriptors, Molecular ACCess System (MACCS) fingerprints, Electrotopological\nstate (Estate) fingerprints and amino acid sequences of targets to get Pseudo\nAmino Acid Composition (PseAAC). We target to evaluate performance of DTI\nprediction models using CDK descriptors. For comparison, we use benchmark data\nand evaluate models performance on two widely used fingerprints, MACCS\nfingerprints and Estate fingerprints. The evaluation of performances shows that\nCDK descriptors are superior at predicting DTIs. The proposed method also\noutperforms other previously published techniques significantly.\n", "date": "2022-10-20", "categories": []}
{"id": "2210.10784", "title": "Graph Regularized Probabilistic Matrix Factorization for Drug-Drug\n  Interactions Prediction", "abstract": "  Co-administration of two or more drugs simultaneously can result in adverse\ndrug reactions. Identifying drug-drug interactions (DDIs) is necessary,\nespecially for drug development and for repurposing old drugs. DDI prediction\ncan be viewed as a matrix completion task, for which matrix factorization (MF)\nappears as a suitable solution. This paper presents a novel Graph Regularized\nProbabilistic Matrix Factorization (GRPMF) method, which incorporates expert\nknowledge through a novel graph-based regularization strategy within an MF\nframework. An efficient and sounded optimization algorithm is proposed to solve\nthe resulting non-convex problem in an alternating fashion. The performance of\nthe proposed method is evaluated through the DrugBank dataset, and comparisons\nare provided against state-of-the-art techniques. The results demonstrate the\nsuperior performance of GRPMF when compared to its counterparts.\n", "date": "2022-10-19", "categories": []}
{"id": "2206.06583", "title": "Exploring evolution-aware & -free protein language models as protein\n  function predictors", "abstract": "  Large-scale Protein Language Models (PLMs) have improved performance in\nprotein prediction tasks, ranging from 3D structure prediction to various\nfunction predictions. In particular, AlphaFold, a ground-breaking AI system,\ncould potentially reshape structural biology. However, the utility of the PLM\nmodule in AlphaFold, Evoformer, has not been explored beyond structure\nprediction. In this paper, we investigate the representation ability of three\npopular PLMs: ESM-1b (single sequence), MSA-Transformer (multiple sequence\nalignment) and Evoformer (structural), with a special focus on Evoformer.\nSpecifically, we aim to answer the following key questions: (i) Does the\nEvoformer trained as part of AlphaFold produce representations amenable to\npredicting protein function? (ii) If yes, can Evoformer replace ESM-1b and\nMSA-Transformer? (ii) How much do these PLMs rely on evolution-related protein\ndata? In this regard, are they complementary to each other? We compare these\nmodels by empirical study along with new insights and conclusions. All code and\ndatasets for reproducibility are available at\nhttps://github.com/elttaes/Revisiting-PLMs.\n", "date": "2022-06-14", "categories": []}
{"id": "2210.05677", "title": "Application of Deep Learning on Single-Cell RNA-sequencing Data\n  Analysis: A Review", "abstract": "  Single-cell RNA-sequencing (scRNA-seq) has become a routinely used technique\nto quantify the gene expression profile of thousands of single cells\nsimultaneously. Analysis of scRNA-seq data plays an important role in the study\nof cell states and phenotypes, and has helped elucidate biological processes,\nsuch as those occurring during development of complex organisms and improved\nour understanding of disease states, such as cancer, diabetes, and COVID, among\nothers. Deep learning, a recent advance of artificial intelligence that has\nbeen used to address many problems involving large datasets, has also emerged\nas a promising tool for scRNA-seq data analysis, as it has a capacity to\nextract informative, compact features from noisy, heterogeneous, and\nhigh-dimensional scRNA-seq data to improve downstream analysis. The present\nreview aims at surveying recently developed deep learning techniques in\nscRNA-seq data analysis, identifying key steps within the scRNA-seq data\nanalysis pipeline that have been advanced by deep learning, and explaining the\nbenefits of deep learning over more conventional analysis tools. Finally, we\nsummarize the challenges in current deep learning approaches faced within\nscRNA-seq data and discuss potential directions for improvements in deep\nalgorithms for scRNA-seq data analysis.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.05274", "title": "Equivariant 3D-Conditional Diffusion Models for Molecular Linker Design", "abstract": "  Fragment-based drug discovery has been an effective paradigm in early-stage\ndrug development. An open challenge in this area is designing linkers between\ndisconnected molecular fragments of interest to obtain chemically-relevant\ncandidate drug molecules. In this work, we propose DiffLinker, an\nE(3)-equivariant 3D-conditional diffusion model for molecular linker design.\nGiven a set of disconnected fragments, our model places missing atoms in\nbetween and designs a molecule incorporating all the initial fragments. Unlike\nprevious approaches that are only able to connect pairs of molecular fragments,\nour method can link an arbitrary number of fragments. Additionally, the model\nautomatically determines the number of atoms in the linker and its attachment\npoints to the input fragments. We demonstrate that DiffLinker outperforms other\nmethods on the standard datasets generating more diverse and\nsynthetically-accessible molecules. Besides, we experimentally test our method\nin real-world applications, showing that it can successfully generate valid\nlinkers conditioned on target protein pockets.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.04096", "title": "PropertyDAG: Multi-objective Bayesian optimization of partially ordered,\n  mixed-variable properties for biological sequence design", "abstract": "  Bayesian optimization offers a sample-efficient framework for navigating the\nexploration-exploitation trade-off in the vast design space of biological\nsequences. Whereas it is possible to optimize the various properties of\ninterest jointly using a multi-objective acquisition function, such as the\nexpected hypervolume improvement (EHVI), this approach does not account for\nobjectives with a hierarchical dependency structure. We consider a common use\ncase where some regions of the Pareto frontier are prioritized over others\naccording to a specified $\\textit{partial ordering}$ in the objectives. For\ninstance, when designing antibodies, we would like to maximize the binding\naffinity to a target antigen only if it can be expressed in live cell culture\n-- modeling the experimental dependency in which affinity can only be measured\nfor antibodies that can be expressed and thus produced in viable quantities. In\ngeneral, we may want to confer a partial ordering to the properties such that\neach property is optimized conditioned on its parent properties satisfying some\nfeasibility condition. To this end, we present PropertyDAG, a framework that\noperates on top of the traditional multi-objective BO to impose this desired\nordering on the objectives, e.g. expression $\\rightarrow$ affinity. We\ndemonstrate its performance over multiple simulated active learning iterations\non a penicillin production task, toy numerical problem, and a real-world\nantibody design task.\n", "date": "2022-10-08", "categories": []}
{"id": "2210.02881", "title": "Antibody Representation Learning for Drug Discovery", "abstract": "  Therapeutic antibody development has become an increasingly popular approach\nfor drug development. To date, antibody therapeutics are largely developed\nusing large scale experimental screens of antibody libraries containing\nhundreds of millions of antibody sequences. The high cost and difficulty of\ndeveloping therapeutic antibodies create a pressing need for computational\nmethods to predict antibody properties and create bespoke designs. However, the\nrelationship between antibody sequence and activity is a complex physical\nprocess and traditional iterative design approaches rely on large scale assays\nand random mutagenesis. Deep learning methods have emerged as a promising way\nto learn antibody property predictors, but predicting antibody properties and\ntarget-specific activities depends critically on the choice of antibody\nrepresentations and data linking sequences to properties is often limited.\nExisting works have not yet investigated the value, limitations and\nopportunities of these methods in application to antibody-based drug discovery.\nIn this paper, we present results on a novel SARS-CoV-2 antibody binding\ndataset and an additional benchmark dataset. We compare three classes of\nmodels: conventional statistical sequence models, supervised learning on each\ndataset independently, and fine-tuning an antibody specific pre-trained\nlanguage model. Experimental results suggest that self-supervised pretraining\nof feature representation consistently offers significant improvement in over\nprevious approaches. We also investigate the impact of data size on the model\nperformance, and discuss challenges and opportunities that the machine learning\ncommunity can address to advance in silico engineering and design of\ntherapeutic antibodies.\n", "date": "2022-10-05", "categories": []}
{"id": "2208.08910", "title": "Learned Indexing in Proteins: Extended Work on Substituting Complex\n  Distance Calculations with Embedding and Clustering Techniques", "abstract": "  Despite the constant evolution of similarity searching research, it continues\nto face the same challenges stemming from the complexity of the data, such as\nthe curse of dimensionality and computationally expensive distance functions.\nVarious machine learning techniques have proven capable of replacing elaborate\nmathematical models with combinations of simple linear functions, often gaining\nspeed and simplicity at the cost of formal guarantees of accuracy and\ncorrectness of querying.\n  The authors explore the potential of this research trend by presenting a\nlightweight solution for the complex problem of 3D protein structure search.\nThe solution consists of three steps -- (i) transformation of 3D protein\nstructural information into very compact vectors, (ii) use of a probabilistic\nmodel to group these vectors and respond to queries by returning a given number\nof similar objects, and (iii) a final filtering step which applies basic vector\ndistance functions to refine the result.\n", "date": "2022-08-18", "categories": []}
{"id": "2210.00345", "title": "Identifying Selections Operating on HIV-1 Reverse Transcriptase via\n  Uniform Manifold Approximation and Projection", "abstract": "  We analyze 14,651 HIV1 reverse transcriptase (HIV RT) sequences from the\nStanford HIV Drug Resistance Database labeled with treatment regimen in order\nto study the evolution this enzyme under drug selection in the clinic. Our goal\nis to identify distinct sectors of HIV RT's sequence space that are undergoing\nevolution as a way to identify individual selections and/or evolutionary\nsolutions. We utilize Uniform Manifold Approximation and Projection (UMAP), a\ngraph-based dimensionality reduction technique uniquely suited for the\ndetection of non-linear dependencies and visualize the results using an\nunsupervised clustering algorithm based on density analysis. Our analysis\nproduced 21 distinct clusters of sequences. Supporting the biological\nsignificance of these clusters, they tend to represent phylogenetically related\nsequences with strong correspondence to distinct treatment regimens. Thus, this\nmethod for visualization of areas of HIV RT undergoing evolution can help infer\ninformation about selective pressures, although it is correlative. The mutation\nsignatures associated with each cluster may represent the higher-order\nepistatic context facilitating these evolutionary pathways, information that is\ngenerally not accessible by other types of mutational co-dependence analyses.\n", "date": "2022-10-01", "categories": []}
{"id": "2209.15181", "title": "RL-MD: A Novel Reinforcement Learning Approach for DNA Motif Discovery", "abstract": "  The extraction of sequence patterns from a collection of functionally linked\nunlabeled DNA sequences is known as DNA motif discovery, and it is a key task\nin computational biology. Several deep learning-based techniques have recently\nbeen introduced to address this issue. However, these algorithms can not be\nused in real-world situations because of the need for labeled data. Here, we\npresented RL-MD, a novel reinforcement learning based approach for DNA motif\ndiscovery task. RL-MD takes unlabelled data as input, employs a relative\ninformation-based method to evaluate each proposed motif, and utilizes these\ncontinuous evaluation results as the reward. The experiments show that RL-MD\ncan identify high-quality motifs in real-world data.\n", "date": "2022-09-30", "categories": []}
{"id": "2209.15399", "title": "Parea: multi-view ensemble clustering for cancer subtype discovery", "abstract": "  Multi-view clustering methods are essential for the stratification of\npatients into sub-groups of similar molecular characteristics. In recent years,\na wide range of methods has been developed for this purpose. However, due to\nthe high diversity of cancer-related data, a single method may not perform\nsufficiently well in all cases. We present Parea, a multi-view hierarchical\nensemble clustering approach for disease subtype discovery. We demonstrate its\nperformance on several machine learning benchmark datasets. We apply and\nvalidate our methodology on real-world multi-view cancer patient data. Parea\noutperforms the current state-of-the-art on six out of seven analysed cancer\ntypes. We have integrated the Parea method into our developed Python package\nPyrea (https://github.com/mdbloice/Pyrea), which enables the effortless and\nflexible design of ensemble workflows while incorporating a wide range of\nfusion and clustering algorithms.\n", "date": "2022-09-30", "categories": []}
{"id": "2209.13591", "title": "Deep learning forward and reverse primer design to detect SARS-CoV-2\n  emerging variants", "abstract": "  Surges that have been observed at different periods in the number of COVID-19\ncases are associated with the emergence of multiple SARS-CoV-2 (Severe Acute\nRespiratory Virus) variants. The design of methods to support laboratory\ndetection are crucial in the monitoring of these variants. Hence, in this\npaper, we develop a semi-automated method to design both forward and reverse\nprimer sets to detect SARS-CoV-2 variants. To proceed, we train deep\nConvolution Neural Networks (CNNs) to classify labelled SARS-CoV-2 variants and\nidentify partial genomic features needed for the forward and reverse Polymerase\nChain Reaction (PCR) primer design. Our proposed approach supplements existing\nones while promoting the emerging concept of neural network assisted primer\ndesign for PCR. Our CNN model was trained using a database of SARS-CoV-2\nfull-length genomes from GISAID and tested on a separate dataset from NCBI,\nwith 98\\% accuracy for the classification of variants. This result is based on\nthe development of three different methods of feature extraction, and the\nselected primer sequences for each SARS-CoV-2 variant detection (except\nOmicron) were present in more than 95 \\% of sequences in an independent set of\n5000 same variant sequences, and below 5 \\% in other independent datasets with\n5000 sequences of each variant. In total, we obtain 22 forward and reverse\nprimer pairs with flexible length sizes (18-25 base pairs) with an expected\namplicon length ranging between 42 and 3322 nucleotides. Besides the feature\nappearance, in-silico primer checks confirmed that the identified primer pairs\nare suitable for accurate SARS-CoV-2 variant detection by means of PCR tests.\n", "date": "2022-09-25", "categories": []}
{"id": "2209.13014", "title": "Predicting Protein-Ligand Binding Affinity via Joint Global-Local\n  Interaction Modeling", "abstract": "  The prediction of protein-ligand binding affinity is of great significance\nfor discovering lead compounds in drug research. Facing this challenging task,\nmost existing prediction methods rely on the topological and/or spatial\nstructure of molecules and the local interactions while ignoring the\nmulti-level inter-molecular interactions between proteins and ligands, which\noften lead to sub-optimal performance. To solve this issue, we propose a novel\nglobal-local interaction (GLI) framework to predict protein-ligand binding\naffinity. In particular, our GLI framework considers the inter-molecular\ninteractions between proteins and ligands, which involve not only the\nhigh-energy short-range interactions between closed atoms but also the\nlow-energy long-range interactions between non-bonded atoms. For each pair of\nprotein and ligand, our GLI embeds the long-range interactions globally and\naggregates local short-range interactions, respectively. Such a joint\nglobal-local interaction modeling strategy helps to improve prediction\naccuracy, and the whole framework is compatible with various neural\nnetwork-based modules. Experiments demonstrate that our GLI framework\noutperforms state-of-the-art methods with simple neural network architectures\nand moderate computational costs.\n", "date": "2022-09-18", "categories": []}
{"id": "2209.11923", "title": "DeepChrome 2.0: Investigating and Improving Architectures,\n  Visualizations, & Experiments", "abstract": "  Histone modifications play a critical role in gene regulation. Consequently,\npredicting gene expression from histone modification signals is a highly\nmotivated problem in epigenetics. We build upon the work of DeepChrome by Singh\net al. (2016), who trained classifiers that map histone modification signals to\ngene expression. We present a novel visualization technique for providing\ninsight into combinatorial relationships among histone modifications for gene\nregulation that uses a generative adversarial network to generate histone\nmodification signals. We also explore and compare various architectural\nchanges, with results suggesting that the 645k-parameter convolutional neural\nnetwork from DeepChrome has the same predictive power as a 12-parameter linear\nnetwork. Results from cross-cell prediction experiments, where the model is\ntrained and tested on datasets of varying sizes, cell-types, and correlations,\nsuggest the relationship between histone modification signals and gene\nexpression is independent of cell type. We release our PyTorch\nre-implementation of DeepChrome on GitHub\n\\footnote{\\url{github.com/ssss1029/gene_expression_294}}.\\parfillskip=0pt\n", "date": "2022-09-24", "categories": []}
{"id": "2208.12811", "title": "PRIME: Uncovering Circadian Oscillation Patterns and Associations with\n  AD in Untimed Genome-wide Gene Expression across Multiple Brain Regions", "abstract": "  The disruption of circadian rhythm is a cardinal symptom for Alzheimer's\ndisease (AD) patients. The full circadian rhythm orchestration of gene\nexpression in the human brain and its inherent associations with AD remain\nlargely unknown. We present a novel comprehensive approach, PRIME, to detect\nand analyze rhythmic oscillation patterns in untimed high-dimensional gene\nexpression data across multiple datasets. To demonstrate the utility of PRIME,\nfirstly, we validate it by a time course expression dataset from mouse liver as\na cross-species and cross-organ validation. Then, we apply it to study\noscillation patterns in untimed genome-wide gene expression from 19 human brain\nregions of controls and AD patients. Our findings reveal clear, synchronized\noscillation patterns in 15 pairs of brain regions of control, while these\noscillation patterns either disappear or dim for AD. It is worth noting that\nPRIME discovers the circadian rhythmic patterns without requiring the sample's\ntimestamps. The codes for PRIME, along with codes to reproduce the figures in\nthis paper, are available at https://github.com/xinxingwu-uk/PRIME.\n", "date": "2022-08-25", "categories": []}
{"id": "2209.07423", "title": "Can Pre-trained Models Really Learn Better Molecular Representations for\n  AI-aided Drug Discovery?", "abstract": "  Self-supervised pre-training is gaining increasingly more popularity in\nAI-aided drug discovery, leading to more and more pre-trained models with the\npromise that they can extract better feature representations for molecules.\nYet, the quality of learned representations have not been fully explored. In\nthis work, inspired by the two phenomena of Activity Cliffs (ACs) and Scaffold\nHopping (SH) in traditional Quantitative Structure-Activity Relationship (QSAR)\nanalysis, we propose a method named Representation-Property Relationship\nAnalysis (RePRA) to evaluate the quality of the representations extracted by\nthe pre-trained model and visualize the relationship between the\nrepresentations and properties. The concepts of ACs and SH are generalized from\nthe structure-activity context to the representation-property context, and the\nunderlying principles of RePRA are analyzed theoretically. Two scores are\ndesigned to measure the generalized ACs and SH detected by RePRA, and therefore\nthe quality of representations can be evaluated. In experiments,\nrepresentations of molecules from 10 target tasks generated by 7 pre-trained\nmodels are analyzed. The results indicate that the state-of-the-art pre-trained\nmodels can overcome some shortcomings of canonical Extended-Connectivity\nFingerPrints (ECFP), while the correlation between the basis of the\nrepresentation space and specific molecular substructures are not explicit.\nThus, some representations could be even worse than the canonical fingerprints.\nOur method enables researchers to evaluate the quality of molecular\nrepresentations generated by their proposed self-supervised pre-trained models.\nAnd our findings can guide the community to develop better pre-training\ntechniques to regularize the occurrence of ACs and SH.\n", "date": "2022-08-21", "categories": []}
{"id": "2209.04698", "title": "Structured Q-learning For Antibody Design", "abstract": "  Optimizing combinatorial structures is core to many real-world problems, such\nas those encountered in life sciences. For example, one of the crucial steps\ninvolved in antibody design is to find an arrangement of amino acids in a\nprotein sequence that improves its binding with a pathogen. Combinatorial\noptimization of antibodies is difficult due to extremely large search spaces\nand non-linear objectives. Even for modest antibody design problems, where\nproteins have a sequence length of eleven, we are faced with searching over\n2.05 x 10^14 structures. Applying traditional Reinforcement Learning algorithms\nsuch as Q-learning to combinatorial optimization results in poor performance.\nWe propose Structured Q-learning (SQL), an extension of Q-learning that\nincorporates structural priors for combinatorial optimization. Using a\nmolecular docking simulator, we demonstrate that SQL finds high binding energy\nsequences and performs favourably against baselines on eight challenging\nantibody design tasks, including designing antibodies for SARS-COV.\n", "date": "2022-09-10", "categories": []}
{"id": "2209.06259", "title": "Designing Biological Sequences via Meta-Reinforcement Learning and\n  Bayesian Optimization", "abstract": "  The ability to accelerate the design of biological sequences can have a\nsubstantial impact on the progress of the medical field. The problem can be\nframed as a global optimization problem where the objective is an expensive\nblack-box function such that we can query large batches restricted with a\nlimitation of a low number of rounds. Bayesian Optimization is a principled\nmethod for tackling this problem. However, the astronomically large state space\nof biological sequences renders brute-force iterating over all possible\nsequences infeasible. In this paper, we propose MetaRLBO where we train an\nautoregressive generative model via Meta-Reinforcement Learning to propose\npromising sequences for selection via Bayesian Optimization. We pose this\nproblem as that of finding an optimal policy over a distribution of MDPs\ninduced by sampling subsets of the data acquired in the previous rounds. Our\nin-silico experiments show that meta-learning over such ensembles provides\nrobustness against reward misspecification and achieves competitive results\ncompared to existing strong baselines.\n", "date": "2022-09-13", "categories": []}
{"id": "2209.06158", "title": "Tailoring Molecules for Protein Pockets: a Transformer-based Generative\n  Solution for Structured-based Drug Design", "abstract": "  Structure-based drug design is drawing growing attentions in computer-aided\ndrug discovery. Compared with the virtual screening approach where a\npre-defined library of compounds are computationally screened, de novo drug\ndesign based on the structure of a target protein can provide novel drug\ncandidates. In this paper, we present a generative solution named TamGent\n(Target-aware molecule generator with Transformer) that can directly generate\ncandidate drugs from scratch for a given target, overcoming the limits imposed\nby existing compound libraries. Following the Transformer framework (a\nstate-of-the-art framework in deep learning), we design a variant of\nTransformer encoder to process 3D geometric information of targets and\npre-train the Transformer decoder on 10 million compounds from PubChem for\ncandidate drug generation. Systematical evaluation on candidate compounds\ngenerated for targets from DrugBank shows that both binding affinity and\ndrugability are largely improved. TamGent outperforms previous baselines in\nterms of both effectiveness and efficiency. The method is further verified by\ngenerating candidate compounds for the SARS-CoV-2 main protease and the\noncogenic mutant KRAS G12C. The results show that our method not only\nre-discovers previously verified drug molecules , but also generates novel\nmolecules with better docking scores, expanding the compound pool and\npotentially leading to the discovery of novel drugs.\n", "date": "2022-08-30", "categories": []}
{"id": "2209.04952", "title": "Efficient Approximate Kernel Based Spike Sequence Classification", "abstract": "  Machine learning (ML) models, such as SVM, for tasks like classification and\nclustering of sequences, require a definition of distance/similarity between\npairs of sequences. Several methods have been proposed to compute the\nsimilarity between sequences, such as the exact approach that counts the number\nof matches between $k$-mers (sub-sequences of length $k$) and an approximate\napproach that estimates pairwise similarity scores. Although exact methods\nyield better classification performance, they pose high computational costs,\nlimiting their applicability to a small number of sequences. The approximate\nalgorithms are proven to be more scalable and perform comparably to (sometimes\nbetter than) the exact methods -- they are designed in a \"general\" way to deal\nwith different types of sequences (e.g., music, protein, etc.). Although\ngeneral applicability is a desired property of an algorithm, it is not the case\nin all scenarios. For example, in the current COVID-19 (coronavirus) pandemic,\nthere is a need for an approach that can deal specifically with the\ncoronavirus. To this end, we propose a series of ways to improve the\nperformance of the approximate kernel (using minimizers and information gain)\nin order to enhance its predictive performance pm coronavirus sequences. More\nspecifically, we improve the quality of the approximate kernel using domain\nknowledge (computed using information gain) and efficient preprocessing (using\nminimizers computation) to classify coronavirus spike protein sequences\ncorresponding to different variants (e.g., Alpha, Beta, Gamma). We report\nresults using different classification and clustering algorithms and evaluate\ntheir performance using multiple evaluation metrics. Using two datasets, we\nshow that our proposed method helps improve the kernel's performance compared\nto the baseline and state-of-the-art approaches in the healthcare domain.\n", "date": "2022-09-11", "categories": []}
{"id": "2209.00981", "title": "Exploiting Pretrained Biochemical Language Models for Targeted Drug\n  Design", "abstract": "  Motivation: The development of novel compounds targeting proteins of interest\nis one of the most important tasks in the pharmaceutical industry. Deep\ngenerative models have been applied to targeted molecular design and have shown\npromising results. Recently, target-specific molecule generation has been\nviewed as a translation between the protein language and the chemical language.\nHowever, such a model is limited by the availability of interacting\nprotein-ligand pairs. On the other hand, large amounts of unlabeled protein\nsequences and chemical compounds are available and have been used to train\nlanguage models that learn useful representations. In this study, we propose\nexploiting pretrained biochemical language models to initialize (i.e. warm\nstart) targeted molecule generation models. We investigate two warm start\nstrategies: (i) a one-stage strategy where the initialized model is trained on\ntargeted molecule generation (ii) a two-stage strategy containing a\npre-finetuning on molecular generation followed by target specific training. We\nalso compare two decoding strategies to generate compounds: beam search and\nsampling.\n  Results: The results show that the warm-started models perform better than a\nbaseline model trained from scratch. The two proposed warm-start strategies\nachieve similar results to each other with respect to widely used metrics from\nbenchmarks. However, docking evaluation of the generated compounds for a number\nof novel proteins suggests that the one-stage strategy generalizes better than\nthe two-stage strategy. Additionally, we observe that beam search outperforms\nsampling in both docking evaluation and benchmark metrics for assessing\ncompound quality.\n  Availability and implementation: The source code is available at\nhttps://github.com/boun-tabi/biochemical-lms-for-drug-design and the materials\nare archived in Zenodo at https://doi.org/10.5281/zenodo.6832145\n", "date": "2022-09-02", "categories": []}
{"id": "2208.11248", "title": "Secondary Protein Structure Prediction Using Neural Networks", "abstract": "  In this paper we experiment with using neural network structures to predict a\nprotein's secondary structure ({\\alpha} helix positions) from only its primary\nstructure (amino acid sequence). We implement a fully connected neural network\n(FCNN) and preform three experiments using that FCNN. Firstly, we do a\ncross-species comparison of models trained and tested on mouse and human\ndatasets. Secondly, we test the impact of varying the length of protein\nsequence we input into the model. Thirdly, we compare custom error functions\ndesigned to focus on the center of the input window. At the end of paper we\npropose a alternative, recurrent neural network model which can be applied to\nthe problem.\n", "date": "2022-08-24", "categories": []}
{"id": "2208.11467", "title": "Tracking by weakly-supervised learning and graph optimization for\n  whole-embryo C. elegans lineages", "abstract": "  Tracking all nuclei of an embryo in noisy and dense fluorescence microscopy\ndata is a challenging task. We build upon a recent method for nuclei tracking\nthat combines weakly-supervised learning from a small set of nuclei center\npoint annotations with an integer linear program (ILP) for optimal cell lineage\nextraction. Our work specifically addresses the following challenging\nproperties of C. elegans embryo recordings: (1) Many cell divisions as compared\nto benchmark recordings of other organisms, and (2) the presence of polar\nbodies that are easily mistaken as cell nuclei. To cope with (1), we devise and\nincorporate a learnt cell division detector. To cope with (2), we employ a\nlearnt polar body detector. We further propose automated ILP weights tuning via\na structured SVM, alleviating the need for tedious manual set-up of a\nrespective grid search. Our method outperforms the previous leader of the cell\ntracking challenge on the Fluo-N3DH-CE embryo dataset. We report a further\nextensive quantitative evaluation on two more C. elegans datasets. We will make\nthese datasets public to serve as an extended benchmark for future method\ndevelopment. Our results suggest considerable improvements yielded by our\nmethod, especially in terms of the correctness of division event detection and\nthe number and length of fully correct track segments. Code:\nhttps://github.com/funkelab/linajea\n", "date": "2022-08-24", "categories": []}
{"id": "2208.09559", "title": "Neural network facilitated ab initio derivation of linear formula: A\n  case study on formulating the relationship between DNA motifs and gene\n  expression", "abstract": "  Developing models with high interpretability and even deriving formulas to\nquantify relationships between biological data is an emerging need. We propose\nhere a framework for ab initio derivation of sequence motifs and linear formula\nusing a new approach based on the interpretable neural network model called\ncontextual regression model. We showed that this linear model could predict\ngene expression levels using promoter sequences with a performance comparable\nto deep neural network models. We uncovered a list of 300 motifs with important\nregulatory roles on gene expression and showed that they also had significant\ncontributions to cell-type specific gene expression in 154 diverse cell types.\nThis work illustrates the possibility of deriving formulas to represent biology\nlaws that may not be easily elucidated.\n(https://github.com/Wang-lab-UCSD/Motif_Finding_Contextual_Regression)\n", "date": "2022-08-19", "categories": []}
{"id": "2208.07997", "title": "Deep Learning Enabled Time-Lapse 3D Cell Analysis", "abstract": "  This paper presents a method for time-lapse 3D cell analysis. Specifically,\nwe consider the problem of accurately localizing and quantitatively analyzing\nsub-cellular features, and for tracking individual cells from time-lapse 3D\nconfocal cell image stacks. The heterogeneity of cells and the volume of\nmulti-dimensional images presents a major challenge for fully automated\nanalysis of morphogenesis and development of cells. This paper is motivated by\nthe pavement cell growth process, and building a quantitative morphogenesis\nmodel. We propose a deep feature based segmentation method to accurately detect\nand label each cell region. An adjacency graph based method is used to extract\nsub-cellular features of the segmented cells. Finally, the robust graph based\ntracking algorithm using multiple cell features is proposed for associating\ncells at different time instances. Extensive experiment results are provided\nand demonstrate the robustness of the proposed method. The code is available on\nGithub and the method is available as a service through the BisQue portal.\n", "date": "2022-08-17", "categories": []}
{"id": "2208.06361", "title": "Hyperbolic Molecular Representation Learning for Drug Repositioning", "abstract": "  Learning accurate drug representations is essential for task such as\ncomputational drug repositioning. A drug hierarchy is a valuable source that\nencodes knowledge of relations among drugs in a tree-like structure where drugs\nthat act on the same organs, treat the same disease, or bind to the same\nbiological target are grouped together. However, its utility in learning drug\nrepresentations has not yet been explored, and currently described drug\nrepresentations cannot place novel molecules in a drug hierarchy. Here, we\ndevelop a semi-supervised drug embedding that incorporates two sources of\ninformation: (1) underlying chemical grammar that is inferred from chemical\nstructures of drugs and drug-like molecules (unsupervised), and (2)\nhierarchical relations that are encoded in an expert-crafted hierarchy of\napproved drugs (supervised). We use the Variational Auto-Encoder (VAE)\nframework to encode the chemical structures of molecules and use the drug-drug\nsimilarity information obtained from the hierarchy to induce the clustering of\ndrugs in hyperbolic space. The hyperbolic space is amenable for encoding\nhierarchical relations. Our qualitative results support that the learned drug\nembedding can induce the hierarchical relations among drugs. We demonstrate\nthat the learned drug embedding can be used for drug repositioning.\n", "date": "2022-07-06", "categories": []}
{"id": "2208.05341", "title": "Diversifying Design of Nucleic Acid Aptamers Using Unsupervised Machine\n  Learning", "abstract": "  Inverse design of short single-stranded RNA and DNA sequences (aptamers) is\nthe task of finding sequences that satisfy a set of desired criteria. Relevant\ncriteria may be, for example, the presence of specific folding motifs, binding\nto molecular ligands, sensing properties, etc. Most practical approaches to\naptamer design identify a small set of promising candidate sequences using\nhigh-throughput experiments (e.g. SELEX), and then optimize performance by\nintroducing only minor modifications to the empirically found candidates.\nSequences that possess the desired properties but differ drastically in\nchemical composition will add diversity to the search space and facilitate the\ndiscovery of useful nucleic acid aptamers. Systematic diversification protocols\nare needed. Here we propose to use an unsupervised machine learning model known\nas the Potts model to discover new, useful sequences with controllable sequence\ndiversity. We start by training a Potts model using the maximum entropy\nprinciple on a small set of empirically identified sequences unified by a\ncommon feature. To generate new candidate sequences with a controllable degree\nof diversity, we take advantage of the model's spectral feature: an energy\nbandgap separating sequences that are similar to the training set from those\nthat are distinct. By controlling the Potts energy range that is sampled, we\ngenerate sequences that are distinct from the training set yet still likely to\nhave the encoded features. To demonstrate performance, we apply our approach to\ndesign diverse pools of sequences with specified secondary structure motifs in\n30-mer RNA and DNA aptamers.\n", "date": "2022-08-10", "categories": []}
{"id": "2208.04314", "title": "TripHLApan: predicting HLA molecules binding peptides based on triple\n  coding matrix and transfer learning", "abstract": "  Human leukocyte antigen (HLA) is an important molecule family in the field of\nhuman immunity, which recognizes foreign threats and triggers immune responses\nby presenting peptides to T cells. In recent years, the synthesis of tumor\nvaccines to induce specific immune responses has become the forefront of cancer\ntreatment. Computationally modeling the binding patterns between peptide and\nHLA can greatly accelerate the development of tumor vaccines. However, most of\nthe prediction methods performance is very limited and they cannot fully take\nadvantage of the analysis of existing biological knowledge as the basis of\nmodeling. In this paper, we propose TripHLApan, a novel pan-specific prediction\nmodel, for HLA molecular peptide binding prediction. TripHLApan exhibits\npowerful prediction ability by integrating triple coding matrix, BiGRU +\nAttention models, and transfer learning strategy. The comprehensive evaluations\ndemonstrate the effectiveness of TripHLApan in predicting HLA-I and HLA-II\npeptide binding in different test environments. The predictive power of HLA-I\nis further demonstrated in the latest data set. In addition, we show that\nTripHLApan has strong binding reconstitution ability in the samples of a\nmelanoma patient. In conclusion, TripHLApan is a powerful tool for predicting\nthe binding of HLA-I and HLA-II molecular peptides for the synthesis of tumor\nvaccines.\n", "date": "2022-08-06", "categories": []}
{"id": "2208.04524", "title": "Multiple Instance Neural Networks Based on Sparse Attention for Cancer\n  Detection using T-cell Receptor Sequences", "abstract": "  Early detection of cancers has been much explored due to its paramount\nimportance in biomedical fields. Among different types of data used to answer\nthis biological question, studies based on T cell receptors (TCRs) are under\nrecent spotlight due to the growing appreciation of the roles of the host\nimmunity system in tumor biology. However, the one-to-many correspondence\nbetween a patient and multiple TCR sequences hinders researchers from simply\nadopting classical statistical/machine learning methods. There were recent\nattempts to model this type of data in the context of multiple instance\nlearning (MIL).\n  Despite the novel application of MIL to cancer detection using TCR sequences\nand the demonstrated adequate performance in several tumor types, there is\nstill room for improvement, especially for certain cancer types. Furthermore,\nexplainable neural network models are not fully investigated for this\napplication.\n  In this article, we propose multiple instance neural networks based on sparse\nattention (MINN-SA) to enhance the performance in cancer detection and\nexplainability. The sparse attention structure drops out uninformative\ninstances in each bag, achieving both interpretability and better predictive\nperformance in combination with the skip connection.\n  Our experiments show that MINN-SA yields the highest area under the ROC curve\n(AUC) scores on average measured across 10 different types of cancers, compared\nto existing MIL approaches. Moreover, we observe from the estimated attentions\nthat MINN-SA can identify the TCRs that are specific for tumor antigens in the\nsame T cell repertoire.\n", "date": "2022-08-09", "categories": []}
{"id": "2207.13611", "title": "A Semi-automatic Cell Tracking Process Towards Completing the 4D Atlas\n  of C. elegans Development", "abstract": "  The nematode Caenorhabditis elegans (C. elegans) is used as a model organism\nto better understand developmental biology and neurobiology. C. elegans\nfeatures an invariant cell lineage, which has been catalogued and observed\nusing fluorescence microscopy images. However, established methods to track\ncells in late-stage development fail to generalize once sporadic muscular\ntwitching has begun. We build upon methodology which uses skin cells as\nfiducial markers to carry out cell tracking despite random twitching. In\nparticular, we present a cell nucleus segmentation and tracking procedure which\nwas integrated into a 3D rendering GUI to improve efficiency in tracking cells\nacross late-stage development. Results on images depicting aforementioned\nmuscle cell nuclei across three test embryos suggest the fiducial markers in\nconjunction with a classic tracking paradigm overcome sporadic twitching.\n", "date": "2022-07-27", "categories": []}
{"id": "2206.03823", "title": "Multi-channel neural networks for predicting influenza A virus hosts and\n  antigenic types", "abstract": "  Influenza occurs every season and occasionally causes pandemics. Despite its\nlow mortality rate, influenza is a major public health concern, as it can be\ncomplicated by severe diseases like pneumonia. A fast, accurate and low-cost\nmethod to predict the origin host and subtype of influenza viruses could help\nreduce virus transmission and benefit resource-poor areas. In this work, we\npropose multi-channel neural networks to predict antigenic types and hosts of\ninfluenza A viruses with hemagglutinin and neuraminidase protein sequences. An\nintegrated data set containing complete protein sequences were used to produce\na pre-trained model, and two other data sets were used for testing the model's\nperformance. One test set contained complete protein sequences, and another\ntest set contained incomplete protein sequences. The results suggest that\nmulti-channel neural networks are applicable and promising for predicting\ninfluenza A virus hosts and antigenic subtypes with complete and partial\nprotein sequences.\n", "date": "2022-06-08", "categories": []}
{"id": "2207.14639", "title": "Subtype-Former: a deep learning approach for cancer subtype discovery\n  with multi-omics data", "abstract": "  Motivation: Cancer is heterogeneous, affecting the precise approach to\npersonalized treatment. Accurate subtyping can lead to better survival rates\nfor cancer patients. High-throughput technologies provide multiple omics data\nfor cancer subtyping. However, precise cancer subtyping remains challenging due\nto the large amount and high dimensionality of omics data. Results: This study\nproposed Subtype-Former, a deep learning method based on MLP and Transformer\nBlock, to extract the low-dimensional representation of the multi-omics data.\nK-means and Consensus Clustering are also used to achieve accurate subtyping\nresults. We compared Subtype-Former with the other state-of-the-art subtyping\nmethods across the TCGA 10 cancer types. We found that Subtype-Former can\nperform better on the benchmark datasets of more than 5000 tumors based on the\nsurvival analysis. In addition, Subtype-Former also achieved outstanding\nresults in pan-cancer subtyping, which can help analyze the commonalities and\ndifferences across various cancer types at the molecular level. Finally, we\napplied Subtype-Former to the TCGA 10 types of cancers. We identified 50\nessential biomarkers, which can be used to study targeted cancer drugs and\npromote the development of cancer treatments in the era of precision medicine.\n", "date": "2022-07-28", "categories": []}
{"id": "2207.14106", "title": "MarkerMap: nonlinear marker selection for single-cell studies", "abstract": "  Single-cell RNA-seq data allow the quantification of cell type differences\nacross a growing set of biological contexts. However, pinpointing a small\nsubset of genomic features explaining this variability can be ill-defined and\ncomputationally intractable. Here we introduce MarkerMap, a generative model\nfor selecting minimal gene sets which are maximally informative of cell type\norigin and enable whole transcriptome reconstruction. MarkerMap provides a\nscalable framework for both supervised marker selection, aimed at identifying\nspecific cell type populations, and unsupervised marker selection, aimed at\ngene expression imputation and reconstruction. We benchmark MarkerMap's\ncompetitive performance against previously published approaches on real single\ncell gene expression data sets. MarkerMap is available as a pip installable\npackage, as a community resource aimed at developing explainable machine\nlearning techniques for enhancing interpretability in single-cell studies.\n", "date": "2022-07-28", "categories": []}
{"id": "2207.07734", "title": "COEM: Cross-Modal Embedding for MetaCell Identification", "abstract": "  Metacells are disjoint and homogeneous groups of single-cell profiles,\nrepresenting discrete and highly granular cell states. Existing metacell\nalgorithms tend to use only one modality to infer metacells, even though\nsingle-cell multi-omics datasets profile multiple molecular modalities within\nthe same cell. Here, we present \\textbf{C}ross-M\\textbf{O}dal\n\\textbf{E}mbedding for \\textbf{M}etaCell Identification (COEM), which utilizes\nan embedded space leveraging the information of both scATAC-seq and scRNA-seq\nto perform aggregation, balancing the trade-off between fine resolution and\nsufficient sequencing coverage. COEM outperforms the state-of-the-art method\nSEACells by efficiently identifying accurate and well-separated metacells\nacross datasets with continuous and discrete cell types. Furthermore, COEM\nsignificantly improves peak-to-gene association analyses, and facilitates\ncomplex gene regulatory inference tasks.\n", "date": "2022-07-15", "categories": []}
{"id": "2207.09783", "title": "Cancer Subtyping by Improved Transcriptomic Features Using Vector\n  Quantized Variational Autoencoder", "abstract": "  Defining and separating cancer subtypes is essential for facilitating\npersonalized therapy modality and prognosis of patients. The definition of\nsubtypes has been constantly recalibrated as a result of our deepened\nunderstanding. During this recalibration, researchers often rely on clustering\nof cancer data to provide an intuitive visual reference that could reveal the\nintrinsic characteristics of subtypes. The data being clustered are often omics\ndata such as transcriptomics that have strong correlations to the underlying\nbiological mechanism. However, while existing studies have shown promising\nresults, they suffer from issues associated with omics data: sample scarcity\nand high dimensionality. As such, existing methods often impose unrealistic\nassumptions to extract useful features from the data while avoiding overfitting\nto spurious correlations. In this paper, we propose to leverage a recent strong\ngenerative model, Vector Quantized Variational AutoEncoder (VQ-VAE), to tackle\nthe data issues and extract informative latent features that are crucial to the\nquality of subsequent clustering by retaining only information relevant to\nreconstructing the input. VQ-VAE does not impose strict assumptions and hence\nits latent features are better representations of the input, capable of\nyielding superior clustering performance with any mainstream clustering method.\nExtensive experiments and medical analysis on multiple datasets comprising 10\ndistinct cancers demonstrate the VQ-VAE clustering results can significantly\nand robustly improve prognosis over prevalent subtyping systems.\n", "date": "2022-07-20", "categories": []}
{"id": "2207.08898", "title": "Benchmarking Machine Learning Robustness in Covid-19 Genome Sequence\n  Classification", "abstract": "  The rapid spread of the COVID-19 pandemic has resulted in an unprecedented\namount of sequence data of the SARS-CoV-2 genome -- millions of sequences and\ncounting. This amount of data, while being orders of magnitude beyond the\ncapacity of traditional approaches to understanding the diversity, dynamics,\nand evolution of viruses is nonetheless a rich resource for machine learning\n(ML) approaches as alternatives for extracting such important information from\nthese data. It is of hence utmost importance to design a framework for testing\nand benchmarking the robustness of these ML models.\n  This paper makes the first effort (to our knowledge) to benchmark the\nrobustness of ML models by simulating biological sequences with errors. In this\npaper, we introduce several ways to perturb SARS-CoV-2 genome sequences to\nmimic the error profiles of common sequencing platforms such as Illumina and\nPacBio. We show from experiments on a wide array of ML models that some\nsimulation-based approaches are more robust (and accurate) than others for\nspecific embedding methods to certain adversarial attacks to the input\nsequences. Our benchmarking framework may assist researchers in properly\nassessing different ML models and help them understand the behavior of the\nSARS-CoV-2 virus or avoid possible future pandemics.\n", "date": "2022-07-18", "categories": []}
{"id": "2207.06616", "title": "Antibody-Antigen Docking and Design via Hierarchical Equivariant\n  Refinement", "abstract": "  Computational antibody design seeks to automatically create an antibody that\nbinds to an antigen. The binding affinity is governed by the 3D binding\ninterface where antibody residues (paratope) closely interact with antigen\nresidues (epitope). Thus, predicting 3D paratope-epitope complex (docking) is\nthe key to finding the best paratope. In this paper, we propose a new model\ncalled Hierarchical Equivariant Refinement Network (HERN) for paratope docking\nand design. During docking, HERN employs a hierarchical message passing network\nto predict atomic forces and use them to refine a binding complex in an\niterative, equivariant manner. During generation, its autoregressive decoder\nprogressively docks generated paratopes and builds a geometric representation\nof the binding interface to guide the next residue choice. Our results show\nthat HERN significantly outperforms prior state-of-the-art on paratope docking\nand design benchmarks.\n", "date": "2022-07-14", "categories": []}
{"id": "2207.06630", "title": "Identifying Orientation-specific Lipid-protein Fingerprints using Deep\n  Learning", "abstract": "  Improved understanding of the relation between the behavior of RAS and RAF\nproteins and the local lipid environment in the cell membrane is critical for\ngetting insights into the mechanisms underlying cancer formation. In this work,\nwe employ deep learning (DL) to learn this relationship by predicting protein\norientational states of RAS and RAS-RAF protein complexes with respect to the\nlipid membrane based on the lipid densities around the protein domains from\ncoarse-grained (CG) molecular dynamics (MD) simulations. Our DL model can\npredict six protein states with an overall accuracy of over 80%. The findings\nof this work offer new insights into how the proteins modulate the lipid\nenvironment, which in turn may assist designing novel therapies to regulate\nsuch interactions in the mechanisms associated with cancer development.\n", "date": "2022-07-14", "categories": []}
{"id": "2207.06678", "title": "Deep Learning Methods for Protein Family Classification on PDB\n  Sequencing Data", "abstract": "  Composed of amino acid chains that influence how they fold and thus dictating\ntheir function and features, proteins are a class of macromolecules that play a\ncentral role in major biological processes and are required for the structure,\nfunction, and regulation of the body's tissues. Understanding protein functions\nis vital to the development of therapeutics and precision medicine, and hence\nthe ability to classify proteins and their functions based on measurable\nfeatures is crucial; indeed, the automatic inference of a protein's properties\nfrom its sequence of amino acids, known as its primary structure, remains an\nimportant open problem within the field of bioinformatics, especially given the\nrecent advancements in sequencing technologies and the extensive number of\nknown but uncategorized proteins with unknown properties. In this work, we\ndemonstrate and compare the performance of several deep learning frameworks,\nincluding novel bi-directional LSTM and convolutional models, on widely\navailable sequencing data from the Protein Data Bank (PDB) of the Research\nCollaboratory for Structural Bioinformatics (RCSB), as well as benchmark this\nperformance against classical machine learning approaches, including k-nearest\nneighbors and multinomial regression classifiers, trained on experimental data.\nOur results show that our deep learning models deliver superior performance to\nclassical machine learning methods, with the convolutional architecture\nproviding the most impressive inference performance.\n", "date": "2022-07-14", "categories": []}
{"id": "2207.05477", "title": "HelixFold: An Efficient Implementation of AlphaFold2 using PaddlePaddle", "abstract": "  Accurate protein structure prediction can significantly accelerate the\ndevelopment of life science. The accuracy of AlphaFold2, a frontier end-to-end\nstructure prediction system, is already close to that of the experimental\ndetermination techniques. Due to the complex model architecture and large\nmemory consumption, it requires lots of computational resources and time to\nimplement the training and inference of AlphaFold2 from scratch. The cost of\nrunning the original AlphaFold2 is expensive for most individuals and\ninstitutions. Therefore, reducing this cost could accelerate the development of\nlife science. We implement AlphaFold2 using PaddlePaddle, namely HelixFold, to\nimprove training and inference speed and reduce memory consumption. The\nperformance is improved by operator fusion, tensor fusion, and hybrid\nparallelism computation, while the memory is optimized through Recompute,\nBFloat16, and memory read/write in-place. Compared with the original AlphaFold2\n(implemented with Jax) and OpenFold (implemented with PyTorch), HelixFold needs\nonly 7.5 days to complete the full end-to-end training and only 5.3 days when\nusing hybrid parallelism, while both AlphaFold2 and OpenFold take about 11\ndays. HelixFold saves 1x training time. We verified that HelixFold's accuracy\ncould be on par with AlphaFold2 on the CASP14 and CAMEO datasets. HelixFold's\ncode is available on GitHub for free download:\nhttps://github.com/PaddlePaddle/PaddleHelix/tree/dev/apps/protein_folding/helixfold,\nand we also provide stable web services on\nhttps://paddlehelix.baidu.com/app/drug/protein/forecast.\n", "date": "2022-07-12", "categories": []}
{"id": "2207.06237", "title": "Hierarchy exploitation to detect missing annotations on hierarchical\n  multi-label classification", "abstract": "  The availability of genomic data has grown exponentially in the last decade,\nmainly due to the development of new sequencing technologies. Based on the\ninteractions between genes (and gene products) extracted from the increasing\ngenomic data, numerous studies have focused on the identification of\nassociations between genes and functions. While these studies have shown great\npromise, the problem of annotating genes with functions remains an open\nchallenge. In this work, we present a method to detect missing annotations in\nhierarchical multi-label classification datasets. We propose a method that\nexploits the class hierarchy by computing aggregated probabilities to the paths\nof classes from the leaves to the root for each instance. The proposed method\nis presented in the context of predicting missing gene function annotations,\nwhere these aggregated probabilities are further used to select a set of\nannotations to be verified through in vivo experiments. The experiments on\nOriza sativa Japonica, a variety of rice, showcase that incorporating the\nhierarchy of classes into the method often improves the predictive performance\nand our proposed method yields superior results when compared to competitor\nmethods from the literature.\n", "date": "2022-07-13", "categories": []}
{"id": "2207.05682", "title": "Long Short-Term Memory to predict 3D Amino acids Positions in GPCR\n  Molecular Dynamics", "abstract": "  G-Protein Coupled Receptors (GPCRs) are a big family of eukaryotic cell\ntransmembrane proteins, responsible for numerous biological processes. From a\npractical viewpoint around 34\\% of the drugs approved by the US Food and Drug\nAdministration target these receptors. They can be analyzed from their\nsimulated molecular dynamics, including the prediction of their behavior in the\npresence of drugs. In this paper, the capability of Long Short-Term Memory\nNetworks (LSTMs) are evaluated to learn and predict the molecular dynamic\ntrajectories of a receptor. Several models were trained with the 3D position of\nthe amino acids of the receptor considering different transformations on the\nposition of the amino acid, such as their centers of mass, the geometric\ncenters and the position of the $\\alpha$--carbon for each amino acid. The error\nof the prediction of the position was evaluated by the mean average error (MAE)\nand root-mean-square deviation (RMSD). The LSTM models show a robust\nperformance, with results comparable to the state-of-the-art in non-dynamic 3D\npredictions. The best MAE and RMSD values were found for the mass center of the\namino acids with 0.078 {\\AA} and 0.156 {\\AA} respectively. This work shows the\npotential of LSTM to predict the molecular dynamics of GPRCs.\n", "date": "2022-06-02", "categories": []}
{"id": "2207.04166", "title": "Variational Mixtures of ODEs for Inferring Cellular Gene Expression\n  Dynamics", "abstract": "  A key problem in computational biology is discovering the gene expression\nchanges that regulate cell fate transitions, in which one cell type turns into\nanother. However, each individual cell cannot be tracked longitudinally, and\ncells at the same point in real time may be at different stages of the\ntransition process. This can be viewed as a problem of learning the behavior of\na dynamical system from observations whose times are unknown. Additionally, a\nsingle progenitor cell type often bifurcates into multiple child cell types,\nfurther complicating the problem of modeling the dynamics. To address this\nproblem, we developed an approach called variational mixtures of ordinary\ndifferential equations. By using a simple family of ODEs informed by the\nbiochemistry of gene expression to constrain the likelihood of a deep\ngenerative model, we can simultaneously infer the latent time and latent state\nof each cell and predict its future gene expression state. The model can be\ninterpreted as a mixture of ODEs whose parameters vary continuously across a\nlatent space of cell states. Our approach dramatically improves data fit,\nlatent time inference, and future cell state estimation of single-cell gene\nexpression data compared to previous approaches.\n", "date": "2022-07-09", "categories": []}
{"id": "2207.04333", "title": "Emerging Patterns in the Continuum Representation of Protein-Lipid\n  Fingerprints", "abstract": "  Capturing intricate biological phenomena often requires multiscale modeling\nwhere coarse and inexpensive models are developed using limited components of\nexpensive and high-fidelity models. Here, we consider such a multiscale\nframework in the context of cancer biology and address the challenge of\nevaluating the descriptive capabilities of a continuum model developed using\n1-dimensional statistics from a molecular dynamics model. Using deep learning,\nwe develop a highly predictive classification model that identifies complex and\nemergent behavior from the continuum model. With over 99.9% accuracy\ndemonstrated for two simulations, our approach confirms the existence of\nprotein-specific \"lipid fingerprints\", i.e. spatial rearrangements of lipids in\nresponse to proteins of interest. Through this demonstration, our model also\nprovides external validation of the continuum model, affirms the value of such\nmultiscale modeling, and can foster new insights through further analysis of\nthese fingerprints.\n", "date": "2022-07-09", "categories": []}
{"id": "2207.04878", "title": "Stacked Autoencoder Based Multi-Omics Data Integration for Cancer\n  Survival Prediction", "abstract": "  Cancer survival prediction is important for developing personalized\ntreatments and inducing disease-causing mechanisms. Multi-omics data\nintegration is attracting widespread interest in cancer research for providing\ninformation for understanding cancer progression at multiple genetic levels.\nMany works, however, are limited because of the high dimensionality and\nheterogeneity of multi-omics data. In this paper, we propose a novel method to\nintegrate multi-omics data for cancer survival prediction, called Stacked\nAutoEncoder-based Survival Prediction Neural Network (SAEsurv-net). In the\ncancer survival prediction for TCGA cases, SAEsurv-net addresses the curse of\ndimensionality with a two-stage dimensionality reduction strategy and handles\nmulti-omics heterogeneity with a stacked autoencoder model. The two-stage\ndimensionality reduction strategy achieves a balance between computation\ncomplexity and information exploiting. The stacked autoencoder model removes\nmost heterogeneities such as data's type and size in the first group of\nautoencoders, and integrates multiple omics data in the second autoencoder. The\nexperiments show that SAEsurv-net outperforms models based on a single type of\ndata as well as other state-of-the-art methods.\n", "date": "2022-07-08", "categories": []}
{"id": "2207.00584", "title": "Distribution-based Sketching of Single-Cell Samples", "abstract": "  Modern high-throughput single-cell immune profiling technologies, such as\nflow and mass cytometry and single-cell RNA sequencing can readily measure the\nexpression of a large number of protein or gene features across the millions of\ncells in a multi-patient cohort. While bioinformatics approaches can be used to\nlink immune cell heterogeneity to external variables of interest, such as,\nclinical outcome or experimental label, they often struggle to accommodate such\na large number of profiled cells. To ease this computational burden, a limited\nnumber of cells are typically \\emph{sketched} or subsampled from each patient.\nHowever, existing sketching approaches fail to adequately subsample rare cells\nfrom rare cell-populations, or fail to preserve the true frequencies of\nparticular immune cell-types. Here, we propose a novel sketching approach based\non Kernel Herding that selects a limited subsample of all cells while\npreserving the underlying frequencies of immune cell-types. We tested our\napproach on three flow and mass cytometry datasets and on one single-cell RNA\nsequencing dataset and demonstrate that the sketched cells (1) more accurately\nrepresent the overall cellular landscape and (2) facilitate increased\nperformance in downstream analysis tasks, such as classifying patients\naccording to their clinical outcome. An implementation of sketching with Kernel\nHerding is publicly available at\n\\url{https://github.com/vishalathreya/Set-Summarization}.\n", "date": "2022-06-30", "categories": []}
{"id": "2206.05311", "title": "Graph-in-Graph Network for Automatic Gene Ontology Description\n  Generation", "abstract": "  Gene Ontology (GO) is the primary gene function knowledge base that enables\ncomputational tasks in biomedicine. The basic element of GO is a term, which\nincludes a set of genes with the same function. Existing research efforts of GO\nmainly focus on predicting gene term associations. Other tasks, such as\ngenerating descriptions of new terms, are rarely pursued. In this paper, we\npropose a novel task: GO term description generation. This task aims to\nautomatically generate a sentence that describes the function of a GO term\nbelonging to one of the three categories, i.e., molecular function, biological\nprocess, and cellular component. To address this task, we propose a\nGraph-in-Graph network that can efficiently leverage the structural information\nof GO. The proposed network introduces a two-layer graph: the first layer is a\ngraph of GO terms where each node is also a graph (gene graph). Such a\nGraph-in-Graph network can derive the biological functions of GO terms and\ngenerate proper descriptions. To validate the effectiveness of the proposed\nnetwork, we build three large-scale benchmark datasets. By incorporating the\nproposed Graph-in-Graph network, the performances of seven different\nsequence-to-sequence models can be substantially boosted across all evaluation\nmetrics, with up to 34.7%, 14.5%, and 39.1% relative improvements in BLEU,\nROUGE-L, and METEOR, respectively.\n", "date": "2022-06-10", "categories": []}
{"id": "2206.12240", "title": "PSP: Million-level Protein Sequence Dataset for Protein Structure\n  Prediction", "abstract": "  Proteins are essential component of human life and their structures are\nimportant for function and mechanism analysis. Recent work has shown the\npotential of AI-driven methods for protein structure prediction. However, the\ndevelopment of new models is restricted by the lack of dataset and benchmark\ntraining procedure. To the best of our knowledge, the existing open source\ndatasets are far less to satisfy the needs of modern protein sequence-structure\nrelated research. To solve this problem, we present the first million-level\nprotein structure prediction dataset with high coverage and diversity, named as\nPSP. This dataset consists of 570k true structure sequences (10TB) and 745k\ncomplementary distillation sequences (15TB). We provide in addition the\nbenchmark training procedure for SOTA protein structure prediction model on\nthis dataset. We validate the utility of this dataset for training by\nparticipating CAMEO contest in which our model won the first place. We hope our\nPSP dataset together with the training benchmark can enable a broader community\nof AI/biology researchers for AI-driven protein related research.\n", "date": "2022-06-24", "categories": []}
{"id": "2206.12241", "title": "CoSP: Co-supervised pretraining of pocket and ligand", "abstract": "  Can we inject the pocket-ligand interaction knowledge into the pre-trained\nmodel and jointly learn their chemical space? Pretraining molecules and\nproteins has attracted considerable attention in recent years, while most of\nthese approaches focus on learning one of the chemical spaces and lack the\ninjection of biological knowledge. We propose a co-supervised pretraining\n(CoSP) framework to simultaneously learn 3D pocket and ligand representations.\nWe use a gated geometric message passing layer to model both 3D pockets and\nligands, where each node's chemical features, geometric position and\norientation are considered. To learn biological meaningful embeddings, we\ninject the pocket-ligand interaction knowledge into the pretraining model via\ncontrastive loss. Considering the specificity of molecules, we further propose\na chemical similarity-enhanced negative sampling strategy to improve the\ncontrastive learning performance. Through extensive experiments, we conclude\nthat CoSP can achieve competitive results in pocket matching, molecule property\npredictions, and virtual screening.\n", "date": "2022-06-23", "categories": []}
{"id": "2206.02290", "title": "A knowledge graph representation learning approach to predict novel\n  kinase-substrate interactions", "abstract": "  The human proteome contains a vast network of interacting kinases and\nsubstrates. Even though some kinases have proven to be immensely useful as\ntherapeutic targets, a majority are still understudied. In this work, we\npresent a novel knowledge graph representation learning approach to predict\nnovel interaction partners for understudied kinases. Our approach uses a\nphosphoproteomic knowledge graph constructed by integrating data from iPTMnet,\nProtein Ontology, Gene Ontology and BioKG. The representation of kinases and\nsubstrates in this knowledge graph are learned by performing directed random\nwalks on triples coupled with a modified SkipGram or CBOW model. These\nrepresentations are then used as an input to a supervised classification model\nto predict novel interactions for understudied kinases. We also present a\npost-predictive analysis of the predicted interactions and an ablation study of\nthe phosphoproteomic knowledge graph to gain an insight into the biology of the\nunderstudied kinases.\n", "date": "2022-06-05", "categories": []}
{"id": "2206.04683", "title": "Molecular dynamics without molecules: searching the conformational space\n  of proteins with generative neural networks", "abstract": "  All-atom and coarse-grained molecular dynamics are two widely used\ncomputational tools to study the conformational states of proteins. Yet, these\ntwo simulation methods suffer from the fact that without access to\nsupercomputing resources, the time and length scales at which these states\nbecome detectable are difficult to achieve. One alternative to such methods is\nbased on encoding the atomistic trajectory of molecular dynamics as a shorthand\nversion devoid of physical particles, and then learning to propagate the\nencoded trajectory through the use of artificial intelligence. Here we show\nthat a simple textual representation of the frames of molecular dynamics\ntrajectories as vectors of Ramachandran basin classes retains most of the\nstructural information of the full atomistic representation of a protein in\neach frame, and can be used to generate equivalent atom-less trajectories\nsuitable to train different types of generative neural networks. In turn, the\ntrained generative models can be used to extend indefinitely the atom-less\ndynamics or to sample the conformational space of proteins from their\nrepresentation in the models latent space. We define intuitively this\nmethodology as molecular dynamics without molecules, and show that it enables\nto cover physically relevant states of proteins that are difficult to access\nwith traditional molecular dynamics.\n", "date": "2022-06-09", "categories": []}
{"id": "2206.02092", "title": "Bandit Theory and Thompson Sampling-Guided Directed Evolution for\n  Sequence Optimization", "abstract": "  Directed Evolution (DE), a landmark wet-lab method originated in 1960s,\nenables discovery of novel protein designs via evolving a population of\ncandidate sequences. Recent advances in biotechnology has made it possible to\ncollect high-throughput data, allowing the use of machine learning to map out a\nprotein's sequence-to-function relation. There is a growing interest in machine\nlearning-assisted DE for accelerating protein optimization. Yet the theoretical\nunderstanding of DE, as well as the use of machine learning in DE, remains\nlimited. In this paper, we connect DE with the bandit learning theory and make\na first attempt to study regret minimization in DE. We propose a Thompson\nSampling-guided Directed Evolution (TS-DE) framework for sequence optimization,\nwhere the sequence-to-function mapping is unknown and querying a single value\nis subject to costly and noisy measurements. TS-DE updates a posterior of the\nfunction based on collected measurements. It uses a posterior-sampled function\nestimate to guide the crossover recombination and mutation steps in DE. In the\ncase of a linear model, we show that TS-DE enjoys a Bayesian regret of order\n$\\tilde O(d^{2}\\sqrt{MT})$, where $d$ is feature dimension, $M$ is population\nsize and $T$ is number of rounds. This regret bound is nearly optimal,\nconfirming that bandit learning can provably accelerate DE. It may have\nimplications for more general sequence optimization and evolutionary\nalgorithms.\n", "date": "2022-06-05", "categories": []}
{"id": "2206.00668", "title": "Learning to Untangle Genome Assembly with Graph Convolutional Networks", "abstract": "  A quest to determine the complete sequence of a human DNA from telomere to\ntelomere started three decades ago and was finally completed in 2021. This\naccomplishment was a result of a tremendous effort of numerous experts who\nengineered various tools and performed laborious manual inspection to achieve\nthe first gapless genome sequence. However, such method can hardly be used as a\ngeneral approach to assemble different genomes, especially when the assembly\nspeed is critical given the large amount of data. In this work, we explore a\ndifferent approach to the central part of the genome assembly task that\nconsists of untangling a large assembly graph from which a genomic sequence\nneeds to be reconstructed. Our main motivation is to reduce human-engineered\nheuristics and use deep learning to develop more generalizable reconstruction\ntechniques. Precisely, we introduce a new learning framework to train a graph\nconvolutional network to resolve assembly graphs by finding a correct path\nthrough them. The training is supervised with a dataset generated from the\nresolved CHM13 human sequence and tested on assembly graphs built using real\nhuman PacBio HiFi reads. Experimental results show that a model, trained on\nsimulated graphs generated solely from a single chromosome, is able to\nremarkably resolve all other chromosomes. Moreover, the model outperforms\nhand-crafted heuristics from a state-of-the-art \\textit{de novo} assembler on\nthe same graphs. Reconstructed chromosomes with graph networks are more\naccurate on nucleotide level, report lower number of contigs, higher genome\nreconstructed fraction and NG50/NGA50 assessment metrics.\n", "date": "2022-06-01", "categories": []}
{"id": "2206.00262", "title": "Self-supervised Learning for Label Sparsity in Computational Drug\n  Repositioning", "abstract": "  The computational drug repositioning aims to discover new uses for marketed\ndrugs, which can accelerate the drug development process and play an important\nrole in the existing drug discovery system. However, the number of validated\ndrug-disease associations is scarce compared to the number of drugs and\ndiseases in the real world. Too few labeled samples will make the\nclassification model unable to learn effective latent factors of drugs,\nresulting in poor generalization performance. In this work, we propose a\nmulti-task self-supervised learning framework for computational drug\nrepositioning. The framework tackles label sparsity by learning a better drug\nrepresentation. Specifically, we take the drug-disease association prediction\nproblem as the main task, and the auxiliary task is to use data augmentation\nstrategies and contrast learning to mine the internal relationships of the\noriginal drug features, so as to automatically learn a better drug\nrepresentation without supervised labels. And through joint training, it is\nensured that the auxiliary task can improve the prediction accuracy of the main\ntask. More precisely, the auxiliary task improves drug representation and\nserving as additional regularization to improve generalization. Furthermore, we\ndesign a multi-input decoding network to improve the reconstruction ability of\nthe autoencoder model. We evaluate our model using three real-world datasets.\nThe experimental results demonstrate the effectiveness of the multi-task\nself-supervised learning framework, and its predictive ability is superior to\nthe state-of-the-art model.\n", "date": "2022-06-01", "categories": []}
{"id": "2606.21109", "title": "Chem2Gen-Bench: Benchmarking Chemical-to-Genetic Translation in Perturbation Response Space", "abstract": "Virtual-cell and perturbation models are increasingly used to predict cellular responses for biomedical discovery, but chemical and genetic perturbations are not automatically interchangeable. Existing evaluations often study chemical response prediction or genetic perturbation prediction separately, leaving target-matched chemical-to-genetic translation under-tested. We introduce Chem2Gen-Bench, a benchmark comprising 260,084 chemical and 1,099,045 genetic perturbation profiles organized into cell-target contexts, and evaluate pairwise alignment, retrieval, protocol covariate associations, feature spaces, and foundation-model embeddings. Across matched contexts, translation fidelity is measurable but heterogeneous; background adjustment increases the association between pairwise similarity and retrieval success, while paired tests show lower mean retrieval success after adjustment under the evaluated settings. In a target-matched K562 audit, the evaluated foundation-model embeddings did not consistently improve over gene-delta baselines. Chem2Gen-Bench provides an auditable framework for testing when chemical and genetic perturbations align around shared targets and when representation gains are supported by matched perturbation evidence.", "date": "2026-06-19", "categories": []}
{"id": "2606.21116", "title": "ConnectomeBench2: A Unified Benchmark for Automated Connectomic Proofreading", "abstract": "Proofreading--correcting segmentation errors in 3D brain reconstructions--is the rate-limiting step in synapse-resolution connectomics. We release ConnectomeBench2, a unified multi-species dataset of over 716,485 expert-labeled proofreading decisions with >4,500,000 associated images spanning four major open connectomes (mouse, human, zebrafish, fly), spanning both split and merge error correction. Trained on this dataset, a single Vision Transformer with shared encoders for mesh geometry and electron microscopy reaches human-level accuracy across species for split error correction and merge error identification, with performance scaling with data size and modality. Beyond accuracy, we show that the model is well-calibrated within distribution, that measures of distribution distance predict where calibration and accuracy will degrade on unseen data, and that connectomics-specific pretraining and active learning-based sample selection show potential to substantially reduce the labeling effort needed to extend to new species and brain regions. The benchmark provides the infrastructure to train and evaluate increasingly capable vision models for connectomic proofreading.\n  Data and code availability. The ConnectomeBench2 dataset is released on Hugging Face at https://huggingface.co/datasets/jeffbbrown2/ConnectomeBench2. The accompanying codebase is available on GitHub at https://github.com/timfarkas/ConnectomeBench2.", "date": "2026-06-19", "categories": []}
{"id": "2606.21156", "title": "Contrastive and Adaptive Multi-modal Masked Autoencoder for Spatial Transcriptomics", "abstract": "The high cost of spatial transcriptomics (ST) has driven extensive studies into predicting gene expression directly from H&E histology images. However, this prediction task faces an inherent limitation, as tissue morphology alone provides insufficient information to fully resolve underlying gene expression. To address this limitation, a recent study leverages partial gene expression to guide the prediction process alongside histology images. Building on this paradigm, we approach the prediction task as a spatial imputation problem, employing a Masked Autoencoder (MAE) to utilize a small fraction of gene expression as genetic anchors for inferring whole-slide gene expression profiles. Specifically, we propose a bio-saliency score and a learning-to-rank strategy to adaptively identify the most informative spots within the tissue. Based on these identified spots, our framework selects contiguous regions as genetic anchors to ensure suitability for real-world ST profiling hardware. To effectively leverage these anchors, we design a cross-modal joint encoder that integrates visual and genetic modalities. By aligning the selected anchors with their corresponding visual features via contrastive learning, the encoder generates robust joint representations to accurately predict gene expression across the whole slide. Notably, our framework consistently surpasses existing methods in both histology-only prediction and spatial imputation, achieving superior accuracy even without genetic anchors and further excelling with as little as 10% transcriptomic coverage. Our code is available at https://github.com/Kyyle2114/CAMMST.", "date": "2026-06-19", "categories": []}
{"id": "2606.21876", "title": "Protein contacts are already in the attention: a single-forward-pass alternative to the Categorical Jacobian", "abstract": "The Categorical Jacobian of Zhang et al. (2024) reads protein contacts from a language model by perturbing every residue with every alternative amino acid, about $19L$ forward passes. We show the signal it reconstructs is already concentrated in a small subset of attention heads: averaging the top-$K$ contact-relevant heads -- selected on as few as 10 labeled proteins, with no fitted per-pair or per-head weights -- recovers contacts in a single forward pass and matches or beats the Categorical Jacobian for every bidirectional model where it is defined (bar the smallest, 8M). Our primary test is leakage-clean: on a CAMEO split where neither selection nor evaluation touches data the models have plausibly memorized, the head readout beats the Categorical Jacobian on ESM-2-650M by +9pp ($N = 29$, $p < 0.001$), with the within-model margin reproducing across architectures. Ablations localize the gain to labeled head selection, not to averaging: at a matched label budget the unweighted mean ties a supervised $L_1$ logistic regression on the same heads. Both methods fall 30-36pp from their in-distribution Zhang numbers to the leakage-clean split, which we read as an upper bound on how much prior numbers reflect pretraining overlap. We additionally introduce representation-CJ, a hidden-state generalization of the Jacobian to architectures without a masked-LM head (the output-head-independent analogue of logit-CJ), agreeing with the Categorical Jacobian where both are defined (per-protein Pearson $r \\approx 0.95$); show that the optimal $K$ tracks how diffusely a model spreads its contact heads; and find both methods lose the signal on the two causal LMs we test, suggesting attention-encoded pair structure may depend on bidirectional pretraining.", "date": "2026-06-20", "categories": []}
{"id": "2606.22077", "title": "Morphology-Aware Multimodal Representation Learning for Insect Phylogenetic Reconstruction", "abstract": "Morphological traits provide important evidence for phylogenetic reconstruction and evolutionary relationship analysis. Recent image-based approaches have introduced deep learning, particularly convolutional models, to derive morphological features from specimen images, but these methods generally rely on single-modality visual representations and do not explicitly incorporate morphological semantics. This study proposes a morphology-aware multimodal alignment framework for insect phylogenetic reconstruction. The framework combines specimen images with curated morphological descriptions by adapting a vision transformer through parameter-efficient fine-tuning and supervised contrastive learning, followed by image-text alignment in a shared latent space. The learned image embeddings are then used as continuous traits for Bayesian phylogenetic reconstruction. On the public Rove-Tree-11 dataset, comparative and ablation experiments across multiple visual backbones and feature adaptation strategies demonstrate that multimodal alignment improves topological agreement with the reference phylogeny. The results indicate that the proposed framework can derive morphology-aware visual traits for computational phylogenetic reconstruction.", "date": "2026-06-20", "categories": []}
{"id": "2606.22138", "title": "BioMatrix: Towards a Comprehensive Biological Foundation Model Spanning the Modality Matrix of Sequences, Structures, and Language", "abstract": "We present BioMatrix, the first multimodal foundation model that natively integrates sequences, structures, and natural language for both molecules and proteins within a single decoder-only architecture. Existing biological foundation models pursue native multimodality and broad entity coverage separately: those that fuse multiple modalities under a shared objective remain confined to a single entity type, while those spanning multiple entity types either omit explicit structural modeling or rely on adapter-based designs in which the model cannot natively generate the very modalities it can read. BioMatrix closes this gap by mapping molecular sequences (supporting both SMILES and SELFIES notations), molecular structures, protein sequences, protein structures, and natural language into a shared discrete token space through a unified tokenization scheme, so that all modalities are consumed and produced uniformly under a single next-token prediction objective -- without external encoders, projection adapters, or modality-specific output heads. Built upon the Qwen3 language model (1.7B and 4B), BioMatrix is continually pretrained on 304.4 billion tokens spanning general and domain-specific text, sequence and structure views of molecules and proteins, and cross-modal corpora that interleave biomolecular entities with scientific text and link distinct entities through molecule-protein and protein-protein interaction data. After tuning on a comprehensive suite of downstream applications covering 80 tasks across 6 categories -- encompassing single-entity and multi-entity understanding and generation tasks across and within modalities -- BioMatrix achieves state-of-the-art or competitive performance on 77 out of 80 tasks, demonstrating that a single, natively multimodal generalist model can effectively match or surpass specialized approaches across a wide range of biological tasks.", "date": "2026-06-20", "categories": []}
{"id": "2606.22181", "title": "Residue-Level Attributions in Protein Language Models Do Not Recover Allergen Epitopes", "abstract": "Deep allergenicity classifiers are increasingly used in safety screening of novel foods, and recent protein language models have substantially improved protein-level allergenicity prediction. However, whether their explanations capture biologically meaningful information remains unclear. We introduce an epitope-grounded residue-level benchmark for quantitatively evaluating attribution faithfulness in protein allergenicity models. Across frozen ESM-2, multi-task ESM-2, and DeepPlantAllergy, protein-level classification was robust, yet classification-head explanation signals did not significantly exceed random in their residue-level alignment with annotated epitopes across AUROC, AUPRC, and Precision@k. Integrated Gradients identified residues that were functionally important to the model, but not overlapping annotated epitopes. Saturation mutagenesis further suggested classifiers may rely on physicochemical and compositional sequence features rather than epitope-specific mechanisms. Residue-level importance signals should therefore not be interpreted as immunological explanations for safety screening or hypoallergen design without quantitative validation. Code available: https://github.com/Jeffateth/XAllergen2.0-paper", "date": "2026-06-20", "categories": []}
{"id": "2606.22731", "title": "Closed-loop Auto Research for Molecular Property Prediction: Discovering and Certifying Generalizable Improvements", "abstract": "Closed-loop Auto Research extends automated machine learning from fixed-dataset fitting to changing the research workflow, with language-model agents editing representations and model code and acquiring external evidence. Molecular property prediction spans many small endpoints. We ask whether this action space yields improvements generalizing beyond the validation signal selecting them.\n  We isolate three Auto Research axes, features, models, and external evidence, under a file-level ablation lock attributing each gain to one axis over a strong baseline. Across 36 endpoints in three benchmark suites we score each selected configuration once on a held-out test whose labels the search never read. A routed pipeline taking each endpoint's best validation axis reaches positive held-out gains of 0.013, 0.011, and 0.042, the transferable axis differing by suite, data on TDC, model on Polaris, feature and model on MoleculeNet. The largest model-search gain falls from 0.041 on validation to 0.003 on test, while curated data reaches 0.022 but negative 0.019 on test, two non-transfer signatures. Curated external data raises held-out CYP2C9-substrate performance by 0.17 and half-life by 0.08, admitted through a contamination filter rejecting same-source files overlapping 64 to 89 percent of test structures, necessary but not sufficient for transfer. A matched-trial automated machine learning control did not reproduce the agent's code-level model intervention, reaching 0.006 against 0.042, and the pipeline stays competitive with an 84M-parameter pretrained 3D model on the shared training split.\n  The experiments stay within molecular property prediction, but separating discovery from held-out certification is a domain-agnostic lesson for any closed-loop system optimising a proxy for a held-out quantity.", "date": "2026-06-22", "categories": []}
{"id": "2606.22823", "title": "Retrieval-Augmented Multimodal Learning for Enzyme-Substrate Interaction Prediction Under Low-Homology Shift", "abstract": "Enzyme substrate interaction (ESI) prediction is a fundamental computational task for biocatalyst discovery and reaction screening in large biochemical spaces. In practical settings, ESI prediction is challenged by sparse positive supervision and low-homology distribution shift, where test enzymes share limited sequence identity with those observed during training. To address these challenges, we propose RAMMESI, a retrieval-augmented multimodal framework for robust ESI prediction. RAMMESI learns explicit pairwise enzyme-substrate representations through directional cross-modal interaction modeling and adaptive fusion. To enhance robustness, RAMMESI retrieves neighboring enzymes at inference time, recombines them with the query substrate, and aggregates the resulting pairwise predictions as contextual evidence. To improve learning under sparse positive supervision, we further adopt an imbalance-aware weighted-BCE objective. Experiments on two ESI benchmarks under sequence-identity-aware splits demonstrate that RAMMESI achieves consistently strong performance, with particular advantages in more challenging low-identity regimes. In addition, the retrieval module improves multiple ESI backbones in a plug-and-play manner, suggesting that retrieval provides a general mechanism for improving robustness under homology shift.", "date": "2026-06-22", "categories": []}
{"id": "2606.23830", "title": "Deciphering Fingerprints of 3D Molecular Surfaces for Accurate Epitope Prediction", "abstract": "Molecular surfaces encode the geometric and physicochemical patterns that determine antibody-antigen recognition, central to epitope prediction. However, existing methods rely on sequences or backbone structures and struggle to capture discontinuous, surface-driven epitopes. This study presents SurfBind, a surface-centric learning framework for epitope prediction that operates directly on molecular surface representations. SurfBind integrates geometric and physicochemical cues through a Transformer-based architecture with patch-level surface modeling, binder-aware cross-attention, and a hierarchical coarse-to-fine prediction paradigm. Experiments on challenging epitope identification benchmarks, including SAbDab and DB5.5, demonstrate that SurfBind achieves state-of-the-art performance and strong generalization across unseen antibodies and conformational states, highlighting the value of interaction-aware surface modeling for understanding the crucial mechanisms of protein-protein interactions.", "date": "2026-06-22", "categories": []}
{"id": "2606.23856", "title": "Sesame: Structure-Aware Molecular Generation via Spatial Density-Map Conditioning", "abstract": "Generative molecular models for drug design are a promising direction with much active research. In the next phase of computational drug design, such models will need to understand small molecule structure and protein-ligand interactions, and they will need to possess the machinery to generate molecules de novo. Incorporating each feature poses a critical challenge. Equally important, yet often treated as secondary, is the ability to grow a molecule from a partial starting point -- a scaffold or fragment supplied by a chemist -- which is the central operation of lead optimization. We present Sesame (Spatial Evoformer for a Structure-Aware Molecular Engine), a diffusion-based molecular generation model that leverages a novel spatial pairformer module to condition on partial molecular structure and the surrounding protein pocket, both expressed as continuous spatial density maps. This single conditioning mechanism supports both de novo generation and fragment-conditioned lead optimization, letting a medicinal chemist prune a hit to a scaffold and have Sesame grow it in productive ways. In addition to this module, we also introduce a diffusion framework for joint denoising of atom types, bond types, and positions, along with a trajectory finetuning scheme that trains on the model's own sampling rollouts to improve generation quality. Sesame is trained on a large corpus of ligand-only and protein-ligand datasets.", "date": "2026-06-22", "categories": []}
{"id": "2606.24940", "title": "Stable-Shift: Biologically Structured Prediction of Transcriptional Responses to Unseen Gene Perturbations", "abstract": "Predicting transcriptional responses to genetic perturbations could reduce the experimental burden of functional genomics, but extrapolation to genes that were never perturbed during training remains difficult. We present Stable-Shift, a structured method for estimating unseen-gene responses. Stable-Shift aggregates single-cell measurements into perturbation-level expression shifts, fits a low-rank response basis using training perturbations only, and predicts an unseen gene's coordinates in that basis from biological context. The context combines STRING interactions, network structure, control-cell expression statistics, and Gene Ontology annotations; the evaluated implementation uses graph convolution to integrate these inputs. On the supplied K562 Perturb-seq benchmark, Stable-Shift obtained 0.592 cosine similarity, compared with 0.569 for GEARS, together with higher Spearman correlation and top-gene precision among the evaluated methods. Its mean cosine similarity over five unseen-gene splits was 0.589 +/- 0.008. The same ordering was observed in the supplied graph-aware, residualized, gene-space, and Norman-dataset comparisons. These results support further study of biologically structured latent-response prediction, while the lower gene-space accuracy and sensitivity to sparse graph neighborhoods limit the scope of the present conclusions.", "date": "2026-06-22", "categories": []}
{"id": "2606.24235", "title": "SP-Mind: An Autonomous Reasoning Agent for Spatial Proteomics Analysis", "abstract": "Spatial proteomics enables single-cell-resolution characterization of protein expression within tissue architecture, playing a critical role in understanding tumor microenvironments and guiding precision medicine. However, current analysis workflows remain fragmented, requiring expert manual orchestration of heterogeneous tools and limiting research scalability and reproducibility. We present SP-Mind, the first autonomous AI agent designed to unify the spatial proteomics analysis pipeline, from raw multiplexed tissue imaging to downstream phenotype discovery. Equipped with expert-curated biological analysis skills and specialized computational tools, SP-Mind converts natural-language queries into end-to-end analytical workflows without task-specific fine-tuning. To rigorously evaluate its capabilities, we introduce SP-Bench, a comprehensive benchmark spanning diverse tissue types, comprising 102 tasks across 18 distinct categories. Through extensive evaluation on SP-Bench and established downstream tasks, SP-Mind achieves state-of-the-art performance compared to existing open-source biomedical agent baselines.", "date": "2026-06-23", "categories": []}
{"id": "2606.24779", "title": "DeepBD: A Grounded Agentic Workflow for Variant Prioritization and Diagnosis of Genetic Birth Defects", "abstract": "Birth defects are a major cause of fetal loss, neonatal morbidity and long-term disability. In the subset with suspected genetic etiologies, exome and genome sequencing have moved many cases from variant detection to post-sequencing interpretation: clinicians must rank patient-specific candidate variants under incomplete fetal or infant phenotypes and heterogeneous evidence from population genetics, variant-effect prediction, gene-disease validity, phenotype ontologies, cellular and pathway context, protein structure and clinical literature. We present DeepBD, a grounded agentic workflow for variant prioritization and diagnostic interpretation of genetic birth defects. DeepBD organizes the workflow into LLM-assisted case structuring, a pretrained evidence engine, specialist evidence modules and a grounded diagnostic review layer. The evidence engine learns patient-specific variant scores from structured rule evidence, sequence and variant-effect representations and phenotype-conditioned biological context, whereas specialist modules and the agentic layer provide tool-based refinement, candidate-pool review and diagnosis-oriented synthesis from ranked candidates. Developed using an in-house fetal and infant cohort comprising 18,622 cases, DeepBD achieved Recall@1/3/5/10 of 0.658/0.882/0.912/0.929 on an internal held-out solved-case benchmark, outperforming standalone Exomiser, DeepRare and prompted LLM reranking baselines evaluated on Exomiser-derived top-20 candidate variants. Ablation and overlap analyses show that rule evidence, mechanistic context, and specialist refinement provide complementary signals. These findings support a grounded agentic workflow that separates evidence integration, tool-based refinement, and LLM-assisted diagnostic review for retrospective variant prioritization in genetic birth defects.", "date": "2026-06-23", "categories": []}
{"id": "2606.26168", "title": "Implementation of reinforcement learning in chemical reaction networks: application to phototaxis as curiosity-driven exploration", "abstract": "Living systems navigate environments using noisy and incomplete sensory signals. In unicellular algae, phototaxis is often modeled as a mechanistic run--tumble process driven by stimulus--response rules. However, such descriptions overlook how organisms actively sample their environment to reduce sensory ambiguity. From a minimal cognition perspective, we reframe this navigation as a subjective, information-driven sensorimotor process. To this end, we propose a framework linking a Partially Observable Markov Decision Process (POMDP) with biochemical reaction dynamics. Environmental variables are hidden, while the cell updates a minimal internal state from each observation through a memoryless Bayesian step. These internal dynamics balance orienting toward light with exploratory reorientation and can be implemented through Chemical-Reaction-Network Ordinary Differential Equations (CRN--ODEs). Our model includes a biophysical observation process for photoreception and a chemically computable polynomial bound on information gain. Using Inverse Reinforcement Learning (IRL) on 30 experimentally recorded Chlamydomonas trajectories, we infer the behavioral objective consistent with observed phototactic motion and benchmark the resulting dynamics with standard Stochastic Simulation Algorithm (SSA) baselines. Our model reproduces the empirical alignment-to-light distribution, comparable to objective SSA baselines on this dataset. Within this framework, run--tumble alternation emerges as an information-acquisition strategy: tumbling reorients the cell to sample new sensory configurations and resolve sensor ambiguity, demonstrating how intracellular biochemical networks can support adaptive information-seeking behavior in cellular navigation.", "date": "2026-06-24", "categories": []}
{"id": "2606.26179", "title": "KG-TRACE: A Neuro-Symbolic Framework for Mechanistic Grounding in Antimicrobial Resistance Prediction", "abstract": "While WGS-based AMR prediction has reached high accuracy, existing models lack a mechanism to ground neural attributions in established biological pathways. We present KG-TRACE, a novel neuro-symbolic framework that integrates the WHO mutation knowledge graph (KG) as a structured biological constraint on a neural genomic model. Unlike existing methods that learn statistical patterns in isolation, KG-TRACE fuses genomic features and RotatE-based KG embeddings through a learned epistemic trust gate, dynamically weighting neural evidence against symbolic biological knowledge.\n  Evaluated on the CRyPTIC M. tuberculosis cohort, KG-TRACE achieves an AUROC of 0.9760 for isoniazid, achieving competitive accuracy while its primary value lies in symbolic grounding, not predictive uplift. More importantly, we introduce the Biological Grounding Ratio (BGR), a dataset-level metric that quantifies alignment between neural attributions and established biology. Our framework achieves a 92.5% symbolic coverage of isoniazid-resistant predictions and effectively identifies MDR co-occurrence artifacts by issuing laboratory follow-up flags for 'UNCERTAIN' cases. We demonstrate that neuro-symbolic grounding provides a verifiable audit trail for clinicians, bridging the gap between predictive accuracy and clinical trust.", "date": "2026-06-24", "categories": []}
{"id": "2606.26563", "title": "scBench-Long: Verifiable Benchmarking of Long-Horizon Single-Cell Biology", "abstract": "Single-cell studies require analysts to convert raw measurements into specific biological claims through multi-step workflows and integration of metadata, assay context, and auxiliary evidence. Existing AI-biology benchmarks largely measure broad knowledge, executable workflows, or local analysis steps. We introduce scBench-Long, a benchmark for long-horizon single-cell biology in which agents must recover scientific conclusions from raw or near-raw data without prescribed methods. The benchmark contains 21 evaluations spanning melanoma CD8 T-cell reactivity, CD8 RNA+ATAC regulatory inference, human--monkey chimera development, KRAS-driven lung tumor aging, and lethal COVID-19 lung pathology. Tasks cover paired scRNA/TCR sequencing, RNA and chromatin profiling, cross-species transcriptomics, combinatorial scRNA-seq, single-nucleus RNA-seq, immune repertoires, ortholog maps, ligand--receptor resources, and validation evidence. Candidate claims are reproduced, reviewed, and converted into controlled answer vocabularies with deterministic grading and trajectory rubrics. Across 1,068 completed trajectories, the strongest model--harness pair passes 16/63 runs (25.4\\%). scBench-Long evaluates whether agents can move beyond local analysis steps and make complex scientific claims that are supported by single-cell data.", "date": "2026-06-25", "categories": []}
{"id": "2606.27096", "title": "Transformer-Based Classification of Bacterial Raman Spectra with LOOCV", "abstract": "Transformer-based models have recently attracted increasing attention for Raman spectral classification. In this study, a transformer-based approach was systematically evaluated using a nested leave-one-replicate-out cross-validation framework and compared with conventional machine-learning pipelines combining PCA or ICA with LDA, SVM, and Random Forest classifiers. A bacterial Raman dataset comprising 5,417 single-cell spectra from six bacterial species and nine independent measurement replicates was used. The transformer consistently achieved the highest classification performance across independent test replicates and significantly outperformed all conventional approaches. Analysis of the learned latent feature space revealed improved class separation compared with PCA- and ICA-based representations. Furthermore, the transformer maintained superior performance when applied directly to raw Raman spectra without preprocessing, demonstrating robust behavior across measurement replicates. These findings highlight the potential of transformer-based models for robust Raman spectral classification and emphasize the importance of replicate-aware validation for realistic model evaluation.", "date": "2026-06-25", "categories": []}
{"id": "2606.27413", "title": "GRAFT: Biological Graph and Hypergraph Benchmarks for Linked Gene Expression and Phenotypic Trait Prediction in Arabidopsis thaliana", "abstract": "Understanding which genes control which traits in an organism remains one of the central challenges in biology. Despite significant advances in data collection technology, our ability to map genes to traits is still limited. This genome-to-phenome (G2P) challenge spans several problem domains, including plant breeding, and requires methods capable of reasoning over high-dimensional, heterogeneous, and biologically structured data. Current datasets and data repositories, however, are not well-equipped for this task. Current studies do not link gene expression and trait data, and most focus on very specific traits, limiting the breadth of possible correlations. To address this gap, we present the novel Gene-Graph Regression for Arabidopsis Functional Traits (GRAFT) dataset, a curated multi-modal dataset linking gene expression profiles with phenotypic trait measurements in Arabidopsis thaliana, a model organism in plant biology. GRAFT supports tasks such as phenotype prediction and interpretable graph learning. In addition, we benchmark conventional regression and explanatory baselines, including a biologically-informed hypergraph baseline, to validate gene-trait associations. To the best of our knowledge, this is the first dataset to provide multimodal gene information and heterogeneous trait or phenotype data for the same Arabidopsis thaliana specimens. With GRAFT, we aim to foster research to accurately understand the relationship between genotypes and phenotypes using gene information, higher-order gene pairings, and trait data from multiple sources.", "date": "2026-06-25", "categories": []}
{"id": "2606.27440", "title": "PairSAE: Mechanistic Interpretability from Pair Representations in Protein Co-Folding", "abstract": "Foundation models for structural biology have achieved remarkable performance in predicting biomolecular structure and show promise for the design of proteins and small molecules. Yet understanding which internal features drive their outputs remains challenging. Standard sparse autoencoders (SAEs), effective on transformer-style sequence embeddings, do not transfer cleanly to pairformer-like architectures: naively operating on pairwise representations yields a quadratic blow-up of features and obscures concepts distributed jointly across sequence and pair representations.\n  We introduce PairSAE, which summarizes pairwise tensors via an N-mode SVD into token-wise interaction roles, then uses a sparse autoencoder to learn a shared set of token-level features that decode into both sequence and pair representations. Evaluated on Boltz-2 activations for PLINDER protein-ligand complexes, PairSAE yields interpretable features that align with UniProt annotations and predict Boltz-2 affinity values. These results indicate that PairSAE links the latent space of foundation models for structural biology to interpretable structural concepts, clarifying what the model \"knows\" while avoiding pairformer-induced pitfalls that limit conventional SAEs.", "date": "2026-06-25", "categories": []}
{"id": "2606.27657", "title": "Reconstructing the Developmental Trajectory of Adipocytes in Human Adipose Tissue Using Single-Cell RNA Sequencing", "abstract": "Obesity is a global health crisis associated with metabolic disorders such as type 2 diabetes and cardiovascular disease. This study employed single-cell RNA sequencing to reconstruct the developmental trajectory of human adipocytes from adipose tissue samples. Our analysis identified 15 transcriptionally distinct cell clusters, including 7 transitional states, revealing the dynamic process of adipocyte differentiation. We detected 16 functionally active signaling pathways mediating cellular communication between adipocytes and their progenitors. Among these, insulin-like growth factor (IGF) and fibroblast growth factor (FGF) pathways emerged as the most prominent networks, showing consistent activity across differentiation stages (p<0.05). The study revealed depot-specific differences, with visceral adipocytes undergoing additional extracellular matrix remodeling absent in subcutaneous differentiation. Spatial analysis further showed that IGF signaling was particularly active in perivascular niches, while FGF activity dominated in mature adipocyte zones. These results provide the first comprehensive map of human adipocyte development, highlighting IGF and FGF pathways as potential therapeutic targets. The identified signaling networks offer new insights for developing interventions to promote healthy adipose expansion or inhibit pathological fat accumulation. This work advances our fundamental understanding of adipose tissue biology while providing clinically relevant data for metabolic disorder treatments.", "date": "2026-06-26", "categories": []}
{"id": "2606.27824", "title": "Pepti-drift: Toxicity-Repulsive Drifting for Antigen-Conditioned Discrete Peptide Generation", "abstract": "Peptides are a promising therapeutic modality that combine the chemical tunability of small molecules with the target specificity of macromolecular therapeutics. However, designing antigen-specific binding peptides while avoiding toxicity remains a major challenge for therapeutic peptide discovery. Here, we present Pepti-drift, a toxicity-aware latent refinement framework that generates peptide candidates through a single antigen-conditioned drift step. In a peptide embedding space, Pepti-drift learns to attract generated peptide latents toward antigen-matched binding peptides while repelling them from toxicity-associated regions. This is challenging because binding-promoting physicochemical features often overlap with toxicity-associated features in peptide representation space. To address this, we introduce a warm-up strategy to stabilize this competing objective by first learning binding-oriented attraction and then increasing toxicity repulsion. Pepti-drift achieves highly efficient generation, running 16.2-fold faster than PepMLM and 1,092.0-fold faster than PepTune. Generated peptides show 100% validity, 98.1% uniqueness, the highest sequence diversity, and near-zero cross-antigen reuse. Further evaluation indicates consistently reduced toxicity and hemolysis risk across most peptide-length ranges while retaining target-related predictive binding signal. Pepti-drift thus provides a fast, scalable, and controllable framework for antigen-specific peptide design that directly encodes safe-and-active properties.", "date": "2026-06-26", "categories": []}
{"id": "2606.27939", "title": "Two-Stage Fine-Tuning for Protein Sequence Generation with Targeted Amino-Acid Composition", "abstract": "Protein language models are standard priors for biological sequence generation, but steering them toward explicit distributional design targets remains largely unexplored. We study a constrained protein generation problem in which sequences must match a desired amino-acid (AA) composition profile while preserving plausible sequence statistics and diversity. The motivating application is synthetic feed protein design, where the AA composition of dietary proteins directly determines their nutritional value. We propose a two-stage pipeline in which domain-adaptive fine-tuning (FT) on an in-domain protein dataset is followed by iterative reward-weighted FT via reinforcement learning (RL) anchored against the FT model as a frozen reference. We evaluate the pipeline on two AA compositions and find that FT brings the average composition close to the target, while the subsequent RL enforces specific sequence constraints that FT alone cannot satisfy. We additionally evaluate the design choices of the proposed composition reward term against two baselines and an ablated variant, isolate the contribution of each training stage, and verify that AA composition alignment is achieved without degrading sequence quality.", "date": "2026-06-26", "categories": []}
{"id": "2606.28459", "title": "scKDGM: KAN-guided Dynamic Graph Masked Learning for Single-Cell RNA-seq Clustering", "abstract": "Single-cell RNA sequencing (scRNA-seq) clustering is essential for identifying cell types, but high dimensionality, sparsity, dropout, and technical noise hinder robust expression representation and cell graph construction. Existing masked autoencoders mainly use expression recovery for feature reconstruction, while graph clustering methods usually depend on fixed KNN graphs and do not feed recovered expression back into graph optimization. We propose scKDGM, a KAN-guided dynamic graph masked learning framework for scRNA-seq clustering. scKDGM uses graph-aware distribution preserving gene masking (GDP-Mask) to perturb cell identity, a KAN-based TAKGCN encoder to learn masked-view representations, mask-guided expression recovery to construct a dynamic graph, and cross-view contrastive learning to transfer recovery signals into topology updates. A ZINB loss models overdispersion and zero inflation. Experiments on 12 real scRNA-seq datasets show that scKDGM outperforms 10 baselines in average NMI and ARI.", "date": "2026-06-26", "categories": []}
{"id": "2606.28659", "title": "Transformer-Based Active Learning for Data-Efficient Vaccine Epitope Selection in PRRS", "abstract": "High-fidelity molecular docking simulations can produce biologically relevant estimates of epitope-receptor binding affinity but are computationally expensive and therefore limit the number of candidates that can be screened for vaccine design. In this work, we evaluate machine learning (ML) approaches where variants of active learning are used to classify instances of high binding affinity between 9-mer epitopes and a well-conserved swine leukocyte antigen (SLA) receptor in the context of Porcine Reproductive and Respiratory Syndrome (PRRS). We use an internally generated dataset of 80 epitope-SLA docking affinities, each requiring more than 48 hours of high-performance computing (HPC). Multiple model families (linear, MLP, CNN, and a small transformer) are trained under strict low-data conditions within a pool-based active learning loop. In each case, optimal model configurations are identified by conducting large-scale hyperparameter optimization over the combined space of model architecture, training configuration, acquisition policy, and ensemble decision rules. To mitigate the effects of data subsample selection, each candidate configuration is evaluated by averaging performance over many randomized and balanced training and validation data subsets. Across experiments, transformer-based sequence models consistently emerged as the best-performing architecture, with active incremental learning yielding significant improvement over a baseline random sample acquisition strategy. Under moderate training data availability (N=30), the optimized ML-model configuration outperforms a standard baseline trained on twice the amount of data. Under higher training data availability (N=60), the same configuration achieves a peak accuracy of 86.8%, consistent with an upper bound of 85% classification accuracy based on two independent estimates of conformational noise.", "date": "2026-06-27", "categories": []}
{"id": "2606.29883", "title": "Building artificial intelligence virtual tissue (AIVT) for tissue state representation, feature prediction, and dynamic simulation", "abstract": "Modeling tissue states and their transitions is essential for understanding tissue homeostasis in health and pathological remodeling in disease. However, conventional computational modeling approaches are inadequate to capture the complexity of tissues as spatially organized, multiscale biological systems. Artificial intelligence (AI) has shown a remarkable ability for representing intricate systems, creating new opportunities to characterize tissue states and their transitions. Here, we propose the concept of AI virtual tissue (AIVT), an AI framework grounded in spatial multimodal data for modeling tissues in health and disease. AIVT is designed to learn unified, spatially resolved, and dynamically manipulatable representations of tissue state, enabling tissue state representation and analysis, molecular and morphological feature prediction, and simulation of spatiotemporal tissue dynamics. We outline the fundamental assumptions, core capabilities, architectural components, as well as data and algorithm foundations of AIVT as a framework for AI-driven tissue modeling.", "date": "2026-06-29", "categories": []}
{"id": "2606.30695", "title": "Modeling Cell-Cycle-Aware Single-Cell Drug Perturbation Responses", "abstract": "Single-cell drug perturbation models should predict not only transcriptional response magnitude, but also whether a treatment alters the proliferative state of a cell. This is challenging because cell-cycle variation is often treated as nuisance variation, and benchmark pipelines rarely treat drug-induced phase changes as a primary prediction target. We introduce scCycleMol, a cell-cycle-aware perturbation prediction framework built on a curated 24-hour SciPlex3 benchmark with standardized molecule identities, dose and cell-line metadata, and gene expression with cell-cycle supervision derived from treated states. Instead of using cell-cycle state as an input covariate, scCycleMol derives supervision from predicted treated expression and propagates it through a learnable full-expression cell-cycle head with circular G1/S/G2M phase targets. We evaluate marker-based supervision, molecular representations, and pretraining strategies to isolate sources of improvement. Across a SciPlex3 benchmark with over 600k cells, 186 perturbation conditions, multiple cancer cell lines, and thousands of genes, scCycleMol improves out-of-distribution expression prediction compared with conditional perturbation baselines. The best LINCS-pretrained circular model achieves 0.9093 expected all-gene r squared and 0.6843 expected differentially expressed gene r squared, compared with 0.6800 and 0.5400 for LINCS-pretrained ChemCPA. Closed-loop cell-cycle supervision improves phase accuracy by about 0.5 to 0.6 points while maintaining nearly unchanged expression prediction. A Tahoe-pretrained variant reaches 0.9609 phase accuracy, highlighting the benefit of explicit cell-cycle-aware supervision in perturbation modeling.", "date": "2026-06-29", "categories": []}
{"id": "2606.30902", "title": "Structure-Regularized Interpretable TCR-Epitope Prediction", "abstract": "T cell receptor (TCR)-epitope binding prediction is essential for understanding adaptive immunity and developing immunotherapies. Existing sequence- and structure-based models often generalize poorly to unseen epitopes and provide limited interpretability. Furthermore, the impact of generated structures on model learning remains unclear. We present TCR-SRIM, a structure-regularized interpretable-by-design model that combines protein language model embeddings with interpretable contact prototypes to capture residue-level TCR-epitope interactions. TCR-SRIM achieves state-of-the-art predictive performance and improved interpretation quality on the TCR-XAI benchmark. Using its inherent interpretability, we further evaluate the effect of generated structures on model learning. While structures predicted by AlphaFold3, TCRModel2, and tFold-TCR yield competitive performance, they lead to less accurate interaction patterns and reduced binding-site diversity than experimentally-resolved structures. Our results highlight limitations of current structure prediction models for TCR-epitope learning and demonstrate the value of interpretable-by-design models for studying generated biological structures.", "date": "2026-06-29", "categories": []}
{"id": "2606.31332", "title": "CryoACE: An Atom-centric Framework for Accurate and Automated Model Building in Cryo-EM", "abstract": "Protein automodeling from cryo-EM density maps faces unique challenges in enforcing physicochemical validity and managing conformational heterogeneity. Current solvers are often limited to static predictions or require computationally intensive heuristic searches. We present CryoACE, an end-to-end framework that reconstructs precise atomic graphs for both homogeneous and heterogeneous structures. Our method features two key innovations: an atom-centric reconstruction paradigm, where density features are sampled directly at atomic coordinates and iteratively recycled to refine structures, replacing expensive voxel convolutions for efficient multimodal fusion; and a training-free guidance mechanism that leverages predicted local resolution priors to resolve dynamic ambiguity. Validated on a newly constructed high-quality dataset, CryoACE significantly outperforms existing baselines on static benchmarks and, for the first time, unveils atomic-level dynamic conformations on complex real-world datasets like EMPIAR-10345 without relying on pre-built static structures.", "date": "2026-06-30", "categories": []}