torchdrug

Name: torchdrug
Availability: InStock
Author: estrella-231

Trains PyTorch-native graph neural networks for molecules, proteins, and biomedical knowledge graphs.

Updated May 24, 2026

Search Tags:#drug discovery #graph neural networks #retrosynthesis #protein modeling #molecular property prediction #knowledge graph reasoning #molecule generation

Ready to run

npx skills add https://github.com/Estrella-231/Mathematical_modeling_tongmeng --skill torchdrug-estrella-231

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: torchdrug
Source: https://github.com/Estrella-231/Mathematical_modeling_tongmeng/tree/main/.agents/skills/torchdrug
Command: npx skills add https://github.com/Estrella-231/Mathematical_modeling_tongmeng --skill torchdrug-estrella-231

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

TorchDrug eliminates the heavy engineering burden of building and training graph neural networks for drug discovery, molecular science, and biomedical knowledge reasoning.

Core Features & Use Cases

Molecule & Protein Graph Learning: Train GNNs for molecular property prediction, protein modeling, and protein property tasks from SMILES and PDB/sequence data.
Knowledge Graph Reasoning: Perform link prediction on biomedical knowledge graphs like Hetionet using embedding and reasoning tasks.
Generative Modeling & Retrosynthesis: Generate candidate molecules and plan retrosynthetic routes using trained models and task decompositions.

Quick Start

Use TorchDrug to train a molecular property predictor on a benchmark dataset and evaluate AUROC/AUPRC for classification.

Frequently Asked Questions about torchdrug

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

FAQPage Schema

How do I train graph neural networks for molecular property prediction?▼

To train graph neural networks for molecular property prediction, you can use this PyTorch-native interface to load benchmark datasets, build GNN models, and evaluate AUROC/AUPRC metrics for classification tasks.

Can I build PyTorch models for protein modeling and biomedical knowledge graphs?▼

Yes, you can build PyTorch models for protein modeling and knowledge graph reasoning by processing PDB/sequence data and performing link prediction on biomedical graphs like Hetionet.

What is the best way to generate molecules and plan retrosynthesis routes?▼

The best way to generate candidate molecules and plan retrosynthetic routes is by using trained generative models and task decompositions to automate molecular generation and synthesis planning workflows.

Do I need a specific data pipeline to train models for drug discovery tasks?▼

Yes, you need a TorchDrug-compatible data pipeline to process SMILES strings and graph structures before training models for drug discovery, molecular property prediction, and knowledge graph completion.

How does knowledge graph reasoning work for biomedical link prediction?▼

Knowledge graph reasoning for biomedical link prediction works by applying embedding and reasoning tasks to graph structures, enabling the prediction of missing relationships between entities in complex biomedical networks.