---
license: cc-by-4.0
language:
  - en
pretty_name: AutoTPD+
tags:
  - targeted-protein-degradation
  - protac
  - molecular-glue
  - drug-discovery
  - tabular
arxiv: 2605.11221
configs:
  - config_name: autotpdplus-protac
    data_files:
      - split: train
        path: AutoTPDplus-PROTAC.csv
  - config_name: autotpdplus-mg
    data_files:
      - split: train
        path: AutoTPDplus-MG.csv
  - config_name: autotpd-protac
    data_files:
      - split: train
        path: AutoTPD-PROTAC.csv
  - config_name: autotpd-mg
    data_files:
      - split: train
        path: AutoTPD-MG.csv
---

# AutoTPD+

[![Project website](https://img.shields.io/badge/Project-Website-2ea44f)](https://yaochenr.github.io/LLM-TPD-Extraction/)
[![Code](https://img.shields.io/badge/Code-GitHub-181717?logo=github)](https://github.com/yaochenr/LLM-TPD-Extraction)
[![Paper](https://img.shields.io/badge/Paper-arXiv%3A2605.11221-b31b1b?logo=arxiv)](https://arxiv.org/abs/2605.11221)
[![License](https://img.shields.io/badge/License-CC%20BY%204.0-f5c518)](https://creativecommons.org/licenses/by/4.0/)

AutoTPD+ is a living database of PROTAC and molecular glue degradation assays mined from the literature using an agentic LLM workflow.

It contains standardized compound, target, recruiter, assay, cell-line, DC50, and Dmax information. AutoTPD+ combines literature extracted records with PROTAC-DB 3.0 and MolGlueDB and is designed for monthly updates.

- **7,622** assay records
- **1,008** source publications
- **2,408** unique chemical structures
- **2 modalities:** PROTAC and molecular glue

## Files

| File | Description | Records |
|---|---|---:|
| `AutoTPDplus-PROTAC.csv` | Living PROTAC dataset | 6,766 |
| `AutoTPDplus-MG.csv` | Living molecular glue dataset | 856 |
| `AutoTPD-PROTAC.csv` | Frozen PROTAC benchmark | 714 |
| `AutoTPD-MG.csv` | Frozen molecular glue benchmark | 246 |

## Data format

Each row represents one degradation assay record.

<details>
<summary><strong>All columns</strong></summary>

| **Column** | **Description** |
| :--- | :--- |
| **IDENTIFIERS** | |
| `AutoTPD_ID` | Stable row identifier, unchanged across monthly releases |
| `Record_ID` | Record ID in the source database; empty for LLM-extracted records |
| `Compound_ID` | Compound ID in the source database; empty for LLM-extracted records |
| `DOI` | Source publication |
| **COMPOUND** | |
| `Compound_Name` | Compound name or label as reported in the paper |
| `IUPAC_Name` | IUPAC name |
| `IUPAC_Name_Source` | Where the IUPAC name came from (PAPER, PUBCHEM) |
| `SMILES` | RDKit canonical SMILES |
| `SMILES_Source` | Where the SMILES came from (PAPER, PUBCHEM, OPSIN, source database) |
| `StdInChIKey` | Standard InChIKey computed by RDKit from the standardized canonical SMILES |
| `Connectivity_Key` | First block of the InChIKey (connectivity only, ignores stereochemistry) |
| **DEGRADATION TARGET** | |
| `Degradation_Target` | Target as reported in the paper |
| `Degradation_Target_Gene` | HGNC gene symbol |
| `Degradation_Target_Uniprot` | UniProt accession |
| `Degradation_Target_Uniprot_MutationID` | UniProt variant ID when a specific mutant is targeted |
| `Degradation_Target_Sequence` | UniProt protein sequence |
| **RECRUITER** | |
| `Recruiter` | Recruiter (usually an E3 ligase) as reported |
| `Recruiter_Gene` | Gene symbol |
| `Recruiter_Uniprot` | UniProt accession |
| `Recruiter_Sequence` | UniProt protein sequence |
| **ASSAY CONTEXT** | |
| `Assay` | Assay as described in the paper |
| `Assay_STD` | Standardized assay family |
| `Cell_Line` | Cell line as reported, or cell-free |
| `Cell_Line_STD` | Cellosaurus recommended name |
| `Cell_Line_STD_Accession` | Cellosaurus accession (RRID) |
| `Cell_Line_Modification` | Engineered construct, e.g. a HiBiT knock-in |
| `Cell_Line_Species` | Species of the cell line |
| **DC50** | |
| `DC50` | Concentration giving 50% degradation |
| `DC50_units` | Units of DC50 |
| `DC50_h` | Treatment time in hours |
| `DC50_operator` | Qualifier: `<`, `>`, `~`, `<=`, `>=` |
| `DC50_range_min` | Range low end, or mean − error |
| `DC50_range_max` | Range high end, or mean + error |
| **DMAX** | |
| `Dmax` | Maximal degradation, in % |
| `Dmax_h` | Treatment time in hours |
| `Dmax_operator` | Qualifier: `<`, `>`, `~`, `<=`, `>=` |
| `Dmax_range_min` | Range low end, or mean − error |
| `Dmax_range_max` | Range high end, or mean + error |
| **DMAX CONCENTRATION** | |
| `Dmax_conc` | Concentration at which Dmax was measured |
| `Dmax_conc_units` | Units of Dmax_conc |
| `Dmax_conc_operator` | Qualifier for Dmax_conc |
| `Dmax_conc_range_min` | Range low end |
| `Dmax_conc_range_max` | Range high end |

</details>

## Usage

```python
from datasets import load_dataset

protac = load_dataset(
    "ailab-bio/AutoTPDplus",
    "autotpdplus-protac",
    token=True,
)

molecular_glue = load_dataset(
    "ailab-bio/AutoTPDplus",
    "autotpdplus-mg",
    token=True,
)
```

## Citation

```bibtex
@misc{rao2026agenticliteratureextractionaugments,
      title={Agentic Literature Extraction Augments Targeted Protein Degradation Databases}, 
      author={Yaochen Rao and Farzaneh Jalalypour and N. M. Anoop Krishnan and Rocío Mercado},
      year={2026},
      eprint={2605.11221},
      archivePrefix={arXiv},
      primaryClass={q-bio.QM},
      url={https://arxiv.org/abs/2605.11221}, 
}
```

## License

The structured datasets are released under CC BY 4.0. Original publications, PROTAC-DB, MolGlueDB, and other upstream resources remain subject to their respective licenses.