Provides AI-ready datasets and benchmarks for therapeutic machine learning and drug discovery tasks.
PyTDC is a comprehensive open-science platform that streamlines the application of machine learning to drug discovery and development. It offers a curated collection of AI-ready datasets covering the entire therapeutics pipeline, including molecular property prediction (ADME/Tox), drug-target interactions, and molecule generation. By providing standardized evaluation metrics, scientifically rigorous data splits like scaffold and cold-start, and integrated molecular oracles, this skill enables researchers and developers to efficiently build, benchmark, and optimize models for pharmacological prediction within their development environment.
Características Principales
010 GitHub stars
02Built-in molecular oracles for property-based generation and optimization
03Standardized drug-target (DTI) and drug-drug (DDI) interaction benchmarks
04Scientific data splitting strategies including scaffold and cold-start splits
05Access to curated ADME, toxicity, and bioactivity datasets
06Molecular utility functions for SMILES, SELFIES, and graph conversions
Casos de Uso
01Generating and optimizing lead compounds using goal-directed molecular oracles
02Predicting pharmacokinetic properties like solubility, permeability, and toxicity
03Benchmarking machine learning models on standardized pharmaceutical tasks
