Cross-Domain Toxicity Bioactivity Data From ChEMBL, Tox21, and ToxCast

897.3 KB of bioactivity data curated from the ChEMBL database, focusing on Molecular Initiating Events and Key Events, with additional datasets from Tox21 and ToxCast. Muhammad Jawahir assembled this dataset to enable reproducible machine learning research in computational toxicology. It was last updated in April 2026.

Use Cases

• Train semi-supervised domain adaptation models using bioactivity data from ChEMBL assays to predict toxicity outcomes in Tox21 or ToxCast domains.
• Benchmark binary toxicity classifiers on the `dl_toxicity_binary_final` label across multiple assay identifiers like CHEMBL218 and CHEMBL253.
• Analyze relationships between specific Molecular Initiating Events (MIEs) or Key Events (KEs) and compound toxicity using the curated assay data.

Strengths

• Data is sourced from authoritative public repositories including ChEMBL, Tox21, and ToxCast.
• Explicitly lists 10 specific ChEMBL assay/target identifiers for reproducibility.

Limitations

• The dataset is small at 897.3 KB, indicating limited scope and likely a modest number of records.
• Row count and specific column names are unknown, complicating initial assessment of model suitability.
• Potential for class imbalance or label noise inherent in bioactivity and toxicity data is not quantified.

Provenance

Source

ChEMBL database, Tox21, and ToxCast programs.

Collection Method

Curated using the ChEMBL Webresource Client to retrieve bioactivity data for specific assays.

Time Range

null

Freshness

Last updated in April 2026.

Geography

null