Update README.md

README.md

@@ -1,31 +1,12 @@
----
-pretty_name: Unified Human Gene Embeddings
-license: other
-viewer: false
-tags:
-- biology
-- bioinformatics
-- genomics
-- gene-embeddings
-- representation-learning
-- multimodal
-- human-genes
-- single-cell
-- protein-language-models
-- genomic-language-models
----
-
 # Unified Human Gene Embeddings
 
 A standardized, multimodal collection of **68 gene-level representation matrices** aligned to a common universe of **20,097 human protein-coding genes**.
 
-The resource brings together representations derived from protein, DNA, RNA and mature mRNA sequences; bulk and single-cell expression; perturbational and functional-genomics data; biomedical literature; protein-protein interaction networks; ontologies; pathways; knowledge graphs; regulatory prediction models; and single-cell foundation-model checkpoints.
+The resource brings together representations derived from protein, DNA, RNA and mature mRNA sequences; bulk and single-cell expression; perturbational and functional-genomics data; biomedical literature; protein-protein interaction networks; ontologies; pathways; knowledge graphs; regulatory prediction models and single-cell foundation-model checkpoints.
 
-> **Release status and licensing**
->
-> This repository should remain private until redistribution rights, model/API output terms, attribution requirements, and per-source licenses have been reviewed. The collection is intentionally marked `license: other`: no single blanket license covers every source-derived matrix. In particular, the AlphaGenome-derived representation requires separate review of the applicable API/output terms before public redistribution or downstream model training. The packaged files are gene-level representations and metadata, not raw patient-level data.
+Useful scripts and documentation can be found in: https://github.com/DanielTabaresLopez/Gene-Embeddings-ETH-SSRF
 
-## At a glance
+## Summary
 
 | Item | Current release |
 |---|---:|
@@ -37,11 +18,10 @@ The resource brings together representations derived from protein, DNA, RNA and
 | Embedding dimensionality range | **20–20,421** |
 | Matrix dtype | **float32** |
 | Approximate staged repository size | **4.1 GB** |
-| Release documented here | **v1.1, updated 2026-07-13** |
 
-## Why this resource exists
+## Objetive of the project
 
-Gene embeddings are fixed numerical vectors representing gene identity. Different construction methods encode different biological priors: amino-acid sequence, nucleotide sequence, transcript structure, expression, function, literature, interaction networks, pathways, regulatory activity, or knowledge learned by foundation models.
+Gene embeddings are fixed numerical vectors representing gene identity. Different construction methods encode different biological priors: amino-acid sequence, nucleotide sequence, transcript structure, expression, function, literature, interaction networks, pathways, regulatory activity or knowledge learned by foundation models.
 
 These representations are difficult to compare directly because they are distributed across many sources, use different identifier systems, cover different subsets of genes, have different dimensions, and are released in incompatible formats. This project creates a common interface so that the same genes can be compared across otherwise unrelated representation spaces.
 
@@ -55,9 +35,7 @@ The main scientific questions are:
 
 Planned analyses include centered kernel alignment, nearest-neighbor overlap, linear predictability across modalities, dimensionality reduction, clustering, and downstream benchmarking. The present release is the standardized resource required for those analyses; it does not claim that any one embedding is universally best.
 
-## What was done in this project
-
-This repository is not a simple mirror of third-party files. The work included:
+The work included:
 
 1. **Building a common gene-identifier backbone.**  
    A 20,097-row enriched human protein-coding master table was assembled and cross-checked across Ensembl, HGNC, Entrez, UniProt, RefSeq, CCDS, STRING, transcript, and protein identifiers.
@@ -91,27 +69,19 @@ This repository is not a simple mirror of third-party files. The work included:
 .
 ├── README.md
 ├── data/
-│   ├── manifest.csv
 │   ├── master_gene_table.csv
-│   ├── embedding_id_rename_map.csv
 │   └── embeddings/
 │       └── <embedding_id>/
 │           ├── embeddings.npz
 │           ├── genes.tsv
 │           ├── metadata.json
 │           └── README.md
-├── docs/
-├── examples/
 └── scripts/
 ```
 
-The public embedding IDs omit the internal processing suffix `_v1_1`; the full processing version and provenance remain recorded in `metadata.json` and the per-embedding README files.
-
 ### Root-level data files
 
-- **`data/manifest.csv`** — one row per embedding with its public ID, matrix shape, dtype, file paths, status, display name, and file size.
 - **`data/master_gene_table.csv`** — the standardized human protein-coding gene universe and identifier backbone.
-- **`data/embedding_id_rename_map.csv`** — mapping from internal processing folder names to simplified public IDs.
 
 ### Files in every embedding folder
 
@@ -132,24 +102,6 @@ REPO_ID = "embeddingsSSRF/gene_embeddings"
 
 Authentication is required while the dataset remains private. Run `hf auth login` once on the machine where the files will be accessed.
 
-### Inspect the manifest
-
-```python
-from huggingface_hub import hf_hub_download
-import pandas as pd
-
-repo_id = "embeddingsSSRF/gene_embeddings"
-
-manifest_path = hf_hub_download(
-    repo_id=repo_id,
-    filename="data/manifest.csv",
-    repo_type="dataset",
-)
-manifest = pd.read_csv(manifest_path)
-
-print(manifest.shape)
-print(manifest[["embedding_id", "n_genes", "embedding_dim", "status"]].head())
-```
 
 ### Download and load one embedding
 
@@ -254,8 +206,6 @@ The collection contains several representation types:
 - **Feature and prediction vectors** derived from expression, CRISPR dependency, localization, perturbation, splicing, or regulatory-track outputs.
 - **Network and knowledge-graph representations** regenerated or transformed from public graphs and annotations.
 
-These types are intentionally kept together because each provides one fixed vector per gene, but they should not be interpreted as identical kinds of model representation.
-
 ### 4. Final output contract
 
 Every core embedding satisfies:
@@ -287,8 +237,8 @@ Each name below links to the embedding-specific README. Coverage is relative to
 ### Biomedical literature and text
 | Public ID | Embedding | Modality | Genes | Dimensions | Coverage |
 |---|---|---|---:|---:|---:|
-| `genept_ada_conservative` | [GenePT-ADA-v1.1-conservative](./data/embeddings/genept_ada_conservative/README.md) | biomedical literature / text | 19,099 | 1,536 | 95.03% |
-| `genept_model3_gene_protein_conservative` | [GenePT-Model3-GeneProtein-v1.1-conservative](./data/embeddings/genept_model3_gene_protein_conservative/README.md) | biomedical literature + protein text | 19,103 | 3,072 | 95.05% |
+| `genept_ada` | [GenePT-ADA-v1.1-conservative](./data/embeddings/genept_ada/README.md) | biomedical literature / text | 19,099 | 1,536 | 95.03% |
+| `genept_model3_gene_protein` | [GenePT-Model3-GeneProtein-v1.1-conservativ](./data/embeddings/genept_model3_gene_protein/README.md) | biomedical literature + protein text | 19,103 | 3,072 | 95.05% |
 | `bioconceptvec_cbow` | [BioConceptVec-CBOW-v1.1](./data/embeddings/bioconceptvec_cbow/README.md) | biomedical literature / text | 17,040 | 100 | 84.79% |
 | `bioconceptvec_skipgram` | [BioConceptVec-SkipGram-v1.1](./data/embeddings/bioconceptvec_skipgram/README.md) | biomedical literature / text | 17,040 | 100 | 84.79% |
 | `bioconceptvec_glove` | [BioConceptVec-GloVe-v1.1](./data/embeddings/bioconceptvec_glove/README.md) | biomedical literature / text | 17,040 | 100 | 84.79% |
@@ -393,30 +343,6 @@ Each name below links to the embedding-specific README. Coverage is relative to
 - **Single-cell checkpoint entries are fixed gene representations.** They are extracted gene-token or position embeddings, not cell-contextual outputs from a new expression dataset.
 - **The repository is not a substitute for source documentation.** Use each folder's README and `metadata.json` before interpreting or redistributing an embedding.
 
-## Auxiliary outputs not included in the 68-core set
-
-Three processed outputs are documented for traceability but are excluded from the core geometry collection:
-
-| Output | Shape | Reason for exclusion |
-|---|---:|---|
-| NEWT-DoRothEA | 9,192 × 128 | Signed or augmented regulon tokens were collapsed back to genes, so the result is not a clean one-token-per-gene source representation. |
-| NEWT-CollecTRI | 6,780 × 128 | Same regulon-token caveat and relatively low coverage. |
-| PPI-RAW consensus adjacency | 18,356 × 20,349 | Sparse raw adjacency baseline rather than a learned dense embedding. |
-
-## Appropriate uses
-
-This resource is designed for:
-
-- representational-similarity analysis;
-- modality comparison;
-- nearest-neighbor and gene-module exploration;
-- linear probing and downstream gene-property prediction;
-- multimodal feature fusion;
-- perturbation-response modeling;
-- controlled comparisons of gene-token priors in transcriptomic models;
-- development of methods for missing-modality prediction or cross-space alignment.
-
-Users should make task-specific choices. Literature-derived embeddings may encode annotation and research-attention bias; network embeddings inherit source-network ascertainment; sequence models depend on sequence and transcript selection; expression-derived vectors are context-dependent; and checkpoint embeddings may reflect the model vocabulary and training corpus rather than universal gene biology.
 
 ## Contributing an embedding
 
@@ -445,7 +371,6 @@ New contributions should not silently impute absent genes or treat aliases as se
 
 ## Versioning and reproducibility
 
-- **Internal processing version:** v1.1.
 - **Public folder IDs:** simplified and stable, with internal version information retained in metadata.
 - **Master universe:** 20,097 human protein-coding genes.
 - **Current core release:** 68 QC-passed representations.
@@ -471,9 +396,7 @@ No project DOI is claimed in this README. A formal citation entry should be adde
 ETH Zürich Summer Research Fellowship, 2026
 
 - **Researcher:** Daniel Tabares López
-- **Supervisor:** Prof. Valentina Boeva
-- **Advisor:** Michael Bohl
-- **Project contact:** Michael Bohl — `mibohl@ethz.ch`
+- **Supervisors:** Prof. Valentina Boeva and Michael Bohl
 
 ## Acknowledgements