pytdc

Access AI-ready drug discovery datasets by loading curated therapeutics data assets directly through the platform. It provides standardized splits, molecular oracles, and ADME, Tox, and DTI datasets for reproducible benchmarking workflows.

A scaffold split partitions molecular data based on chemical structure to ensure realistic model assessment. You need it for generation datasets and benchmarks to evaluate generalization on novel chemical scaffolds.

Run reproducible experiments for ADME and toxicity prediction by using provided templates, example scripts, and standardized evaluation metrics. The platform loads structured guidance and reference utilities to support setup and benchmarking.

Yes, the therapeutics data commons platform works with pandas and numpy as core dependencies. It relies on these libraries to load, manipulate, and structure AI-ready datasets for drug discovery tasks.

Yes, you can use molecular oracles for end-to-end drug discovery workflows. The platform provides oracles alongside standardized splits and generation datasets to support comprehensive benchmarking from data loading to evaluation.

Available drug discovery benchmarks for interaction datasets include multi-instance interaction benchmarks and DTI datasets. These benchmarks provide standardized evaluation metrics and partition strategies for realistic model assessment.