Bioactivity Dataset Curator & Splitter

Turn assay exports into model-ready SAR and QSAR datasets: normalize IC50, EC50, Ki, and Kd values to a common p-scale, resolve censored and replicate measurements by an explicit policy, and split compounds into leakage-resistant scaffold, random, or time sets with a reason for every rejected row.
Estimated minimum cost: $—

Examples

Activity table

One row per measurement: a structure, a value, and optionally its unit, relation (=, <, >), endpoint (IC50, EC50, Ki, Kd), assay, compound ID, and date. ChEMBL activity exports work as-is. At most 50,000 rows.

Curation and splits

Choose how units, censored values, and replicates resolve, then how compounds split into train, validation, and test sets.