---
dataset_info:
  features:
  - name: smiles
    dtype: string
  - name: labels
    sequence: int64
  - name: mask
    sequence: int64
  - name: MW
    dtype: float64
  - name: LogP
    dtype: float64
  - name: TPSA
    dtype: float64
  - name: Complexity
    dtype: float64
  splits:
  - name: train
    num_bytes: 1605736737.3156543
    num_examples: 350183
  - name: test
    num_bytes: 401435330.6843458
    num_examples: 87546
  download_size: 31528613
  dataset_size: 2007172068
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
tags:
- chemistry
size_categories:
- 100K<n<1M
task_categories:
- tabular-classification
---

## ChEMBL Drug-Target Activity Dataset

This dataset was extracted from [`ChEMBL34` database](https://doi.org/10.6019/CHEMBL.database.34). It is designed for multitask classification of drug-target activities. It links compound structures with activity data for multiple assays, enabling multitask learning experiments in drug discovery. Key features of the dataset include:

### Multitask Format
- Each assay ID is treated as a separate binary classification task.
- Binary labels (0 for inactive, 1 for active) and masks (indicating valid labels) are provided for each task.

### ChEMBL Assay IDs
This dataset includes 281 ChEMBL assays, corresponding to 281 binary-classification tasks. You can find the full list of assay_ids (same sequence as in labels and mask) in the [assay_ids.json](assay_ids.json) file.

Detail of the assays can be found from [ChEMBL Assays](https://www.ebi.ac.uk/chembl/explore/assays/).

### Column Descriptions



| **Column Name** | **Description**                                                                                           |
|------------------|----------------------------------------------------------------------------------------------------------|
| **smiles**       | Simplified Molecular Input Line Entry System (SMILES) representation of chemical compounds.              |
| **labels**       | A list of length = number of tasks. Could be 0 for inactive, 1 for active, and -1 for missing labels     |
| **mask**         | A list of length = number of tasks. Could be 0 for non-present and 1 for present and valid label         |
| **chembl_id**    | Unique identifier for the compound in the ChEMBL database.                                               |
| **MW**           | Molecular weight of the compound, representing the sum of atomic masses in Daltons (g/mol).              |
| **LogP**         | Partition coefficient (LogP) of the compound, indicating its lipophilicity or hydrophobicity.            |
| **TPSA**         | Topological polar surface area of the compound, reflecting the molecule's polarity and transportability. |
| **Complexity**   | Molecular complexity, measured as the fraction of sp³ hybridized carbons in the molecule (0–1 range).    |