alterlab-pytdc

What problem does it solve?

PyTDC provides open, AI-ready datasets and standardized benchmarks for drug discovery and development, enabling researchers to quickly load, split, and evaluate models across ADME, toxicity, DTI, and generation tasks.

Core Features & Use Cases

• Open science platform with AI-ready datasets and standardized benchmarks across the therapeutics pipeline.
• Supports single-instance, multi-instance, and generation tasks (ADME, Tox, DTI, DDI, PPI, MolGen, RetroSyn, PairMolGen) with conventional splits (scaffold, random, cold, temporal) and benchmark groups for reproducible evaluation.
• Includes practical templates and references under scripts/ and references/ to facilitate loading, splitting, evaluation, and generation workflows.

Quick Start

Load a dataset with PyTDC and obtain a scaffold split to begin benchmarking models.