torchdrug

Build, train, and deploy graph neural network workflows for chemistry and biology.

1|Updated Mar 11, 2026
One-click install
npx skills add https://github.com/SciMate-AI/scicli --skill torchdrug-scimate-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: torchdrug
Source: https://github.com/SciMate-AI/scicli/tree/main/internal/skills/bundled/claude-scientific-skills/skills/torchdrug
Command: npx skills add https://github.com/SciMate-AI/scicli --skill torchdrug-scimate-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

TorchDrug provides a comprehensive, modular platform for building and evaluating graph neural network models for chemistry and biology, enabling researchers to design, train, and deploy models for molecule property prediction, protein modeling, and knowledge-graph reasoning.

Core Features & Use Cases

  • Modular model definitions and task wrappers for molecular property prediction, protein modeling, and KG reasoning
  • Extensive datasets, reference materials, and pre-trained architectures to accelerate research
  • End-to-end workflows from data loading to evaluation, including retrosynthesis planning

Quick Start

Install TorchDrug and run a basic property-prediction workflow on a sample dataset.

Frequently Asked Questions about torchdrug

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

FAQPage Schema
How do I use graph neural networks for molecular property prediction?

Graph neural networks for molecular property prediction are built by defining modular models and task wrappers to train and evaluate molecular structures. TorchDrug enables this through end-to-end workflows spanning data loading to evaluation.

Can I perform retrosynthesis planning using graph neural networks?

Retrosynthesis planning can be performed using graph neural network workflows designed for chemistry. TorchDrug provides end-to-end workflows that evaluate and plan retrosynthesis across diverse chemical reaction datasets.

Does TorchDrug support protein modeling and knowledge-graph reasoning?

TorchDrug supports both protein modeling and knowledge-graph reasoning by providing modular model definitions and task wrappers. Researchers can build, train, and deploy graph neural network architectures for these specific biology tasks.

What is the best way to build a graph neural network model for drug discovery?

The best way to build a graph neural network model for drug discovery is using a modular architecture with separate components for models, tasks, and data. TorchDrug provides these configurable components along with extensive pre-trained architectures.

Do I need specific datasets to start training models for chemistry and biology?

You need datasets formatted for graph structures to start training models for chemistry and biology. TorchDrug includes extensive reference datasets and pre-trained architectures to accelerate research and enable quick starts.