mims-harvard/TDC

Provides curated, production-ready datasets and benchmarks across drug discovery stages—from target identification to manufacturing—with integrated data split strategies, evaluation metrics, and molecule generation oracles. Supports diverse therapeutic modalities (small molecules, antibodies, vaccines) through a minimal-dependency Python API designed for rapid prototyping, enabling researchers to build ML predictors with just three lines of code. Includes leaderboards for standardized model comparison and incorporates multimodal biological context to bridge AI and biochemical domain expertise.

1,215 stars. No commits in the last 6 months.