pytdc

Name: pytdc
Availability: InStock
Author: K-Dense

Load AI-ready drug discovery datasets and benchmarks from Therapeutics Data Commons.

33.0k|3.2k|Updated Oct 19, 2025

Search Tags:#benchmark #datasets #drug-discovery #tdc #toxicity #adme #dti

Ready to run

npx skills add https://github.com/K-Dense-AI/scientific-agent-skills --skill pytdc-k-dense-ai

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: pytdc
Source: https://github.com/K-Dense-AI/scientific-agent-skills/tree/main/scientific-skills/pytdc
Command: npx skills add https://github.com/K-Dense-AI/scientific-agent-skills --skill pytdc-k-dense-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires tdc, numpy, pandas, and includes scripts (resource) components.

What problem does it solve?

PyTDC provides AI-ready datasets and benchmarks for drug discovery, enabling researchers to evaluate models across ADME, toxicity, DTI, and generation tasks using standardized splits and evaluation protocols.

Core Features & Use Cases

Open-science datasets and benchmarks spanning single-instance, multi-instance, and generation tasks.
Access curated datasets with scaffold, random, and cold splits; benchmark groups; and evaluation utilities via the Therapeutics Data Commons library.
Use cases: benchmarking ML models for molecular property prediction; comparing drug-target interaction predictions; prototyping drug discovery workflows.

Quick Start

Install the TDC package and run the provided scripts to load a dataset and obtain train/valid/test splits for benchmarking.

Frequently Asked Questions about pytdc

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I access AI-ready drug discovery datasets for molecular property prediction?▼

You can access AI-ready drug discovery datasets for molecular property prediction by loading curated ADME, toxicity, and DTI data through the Therapeutics Data Commons library, which provides standardized formats ready for machine learning workflows.

What data splits are available for benchmarking drug-target interaction predictions?▼

Drug-target interaction benchmarking supports scaffold, random, and cold splits. These standardized partitioning methods enable consistent model evaluation across single-instance and multi-instance drug discovery tasks using the provided benchmark groups.

Can I use pandas and numpy with Therapeutics Data Commons datasets?▼

Yes, Therapeutics Data Commons datasets integrate directly with pandas and numpy. Loaded data and generated splits output in compatible formats, allowing immediate manipulation and processing within standard Python data analysis pipelines.

How do I evaluate ML models on ADME and toxicity benchmarks?▼

You evaluate ML models on ADME and toxicity benchmarks by loading the specific task dataset, obtaining the train, valid, and test splits, and applying the standardized evaluation utilities provided by the Therapeutics Data Commons library.

What is the best way to load generation tasks for drug discovery workflows?▼

The best way to load generation tasks for drug discovery workflows is using the Therapeutics Data Commons library, which supplies open-science datasets and evaluation protocols specifically tailored for molecular generation model prototyping.

Does this Skill provide evaluation tooling for comparing drug-target interaction models?▼

Yes, the Skill provides evaluation tooling for comparing drug-target interaction models. It supplies benchmark groups and standardized evaluation protocols to assess prediction accuracy across different machine learning approaches.