What kind of data splits does PyTDC offer?

PyTDC offers specialized splitting strategies including scaffold splits for chemical diversity, cold-drug/cold-target splits for interaction tasks, and temporal splits for time-sensitive data.

How do I evaluate my models using this skill?

The skill provides a standardized Evaluator class with metrics like ROC-AUC, PR-AUC, RMSE, and MAE tailored to specific therapeutic discovery tasks.

What is the PyTDC Claude Code skill used for?

It is used to access curated, AI-ready datasets and benchmarks for drug discovery tasks such as molecular property prediction, interaction forecasting, and molecule generation.

Can I use PyTDC for clinical trial outcome prediction?

Yes, PyTDC includes datasets for multi-instance prediction tasks like TrialOutcome and drug resistance prediction.

Does this skill support molecular property optimization?

Yes, it provides access to over 17 molecular oracles that can be used to score and optimize molecules for specific biological targets or chemical properties.

Therapeutics Data Commons (PyTDC)

Name: Therapeutics Data Commons (PyTDC)
Author: ovachiever

byovachiever

•

数据科学与机器学习

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

关于

PyTDC (Therapeutics Data Commons) is a specialized skill for drug discovery researchers and ML engineers, offering streamlined access to curated biomedical datasets across the entire therapeutics pipeline. It facilitates the development and benchmarking of models for molecular property prediction, drug-target interactions, and molecule generation by providing standardized evaluation metrics, diverse data split strategies like scaffold and cold-start, and automated molecule processing utilities. This skill is essential for anyone working on ADME/Tox prediction, retrosynthesis, or optimizing molecular properties using validated oracles within the Claude Code environment.

主要功能

Integrated molecular oracles for goal-directed generation and property optimization
15 GitHub stars
Standardized benchmark groups with rigorous 5-seed evaluation protocols
Access to 100+ AI-ready drug discovery datasets including ADME, Toxicity, and HTS
Advanced data splitting techniques including scaffold, cold-drug, and temporal splits
Chemical utilities for format conversion between SMILES, SELFIES, PyG, and DGL

使用场景

Predicting binding affinities for drug-target interactions (DTI) and drug-drug interactions (DDI)
Benchmarking machine learning models on standardized pharmaceutical tasks like ADME/Tox prediction
Generating and optimizing novel molecules with specific biological or chemical properties

关于

主要功能

Integrated molecular oracles for goal-directed generation and property optimization
15 GitHub stars
Standardized benchmark groups with rigorous 5-seed evaluation protocols
Access to 100+ AI-ready drug discovery datasets including ADME, Toxicity, and HTS
Advanced data splitting techniques including scaffold, cold-drug, and temporal splits
Chemical utilities for format conversion between SMILES, SELFIES, PyG, and DGL

使用场景

Predicting binding affinities for drug-target interactions (DTI) and drug-drug interactions (DDI)
Benchmarking machine learning models on standardized pharmaceutical tasks like ADME/Tox prediction
Generating and optimizing novel molecules with specific biological or chemical properties