Can I use PyTDC for scientific benchmarking?

Yes, PyTDC includes standardized benchmark groups with specific data splits and evaluation metrics to ensure reproducible and comparable results in pharmaceutical ML research.

What kind of datasets does PyTDC provide?

PyTDC provides curated, AI-ready datasets for various therapeutic tasks including ADME (Absorption, Distribution, Metabolism, Excretion), Toxicity (Tox), Drug-Target Interactions (DTI), and Molecule Generation.

What are molecular oracles in PyTDC?

Oracles are scoring functions used to evaluate generated molecules for specific properties like drug-likeness, synthetic accessibility, or binding affinity to specific targets like GSK3B or DRD2.

Does it support chemical informatics utilities?

Yes, it includes utilities for SMILES/SELFIES conversion, molecular filtering (PAINS, drug-likeness), and scaffold-based data splitting to maintain chemical diversity.

PyTDC Drug Discovery ML

Name: PyTDC Drug Discovery ML
Author: lifangda

bylifangda

•

データサイエンスとML

Provides AI-ready datasets, benchmarks, and molecular oracles for drug discovery and therapeutics machine learning.

PyTDC (Therapeutics Data Commons) is a specialized skill for drug discovery researchers and ML engineers that streamlines access to curated, AI-ready datasets spanning the entire therapeutics pipeline. It offers standardized evaluation metrics, advanced data splits like scaffold and cold-start, and molecular oracles for property optimization. By providing a unified interface for tasks ranging from ADMET prediction to drug-target interactions and molecular generation, this skill enables users to build, train, and benchmark pharmaceutical ML models with high scientific rigor and minimal boilerplate code.

主な機能

0116 GitHub stars

02Access to curated ADME, toxicity, and bioactivity datasets

03Molecular generation oracles for goal-directed property optimization

04Standardized benchmark groups for systematic model evaluation

05Chemical utility functions for format conversion and molecule filtering

06Advanced data splitting strategies including scaffold and cold-target splits

ユースケース

01Benchmarking new ML architectures on standardized pharmaceutical datasets

02Generating and optimizing novel molecules with specific pharmacokinetic profiles

03Predicting drug-target binding affinity (DTI) and drug-drug interactions (DDI)

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lifangda/claude-plugins pytdc

For use in Claude.ai and ChatGPT

Download Skill