TorchDrug is a PyTorch-based machine learning framework designed specifically for drug discovery and molecular science tasks like property prediction and generation.

Does this skill support RDKit integration?

Yes, TorchDrug provides seamless conversion between its own internal molecular representations and RDKit objects for advanced cheminformatics tasks.

Can I use TorchDrug with AlphaFold?

Yes, TorchDrug integrates with AlphaFold and ESM to work with predicted protein structures and sequences for function or stability prediction.

Is it suitable for chemical synthesis planning?

Yes, TorchDrug contains specialized modules and datasets like USPTO-50k for building and evaluating retrosynthesis pipelines.

Which GNN models are available in this skill?

It includes popular architectures like GIN, GAT, and SchNet for general graphs, as well as protein-specific models like GearNet.

TorchDrug for Molecular Science

Name: TorchDrug for Molecular Science
Author: pur3v4d3r

bypur3v4d3r

•

データサイエンスとML

Accelerates drug discovery and molecular modeling using graph neural networks and curated biological datasets within PyTorch.

概要

TorchDrug is a specialized Claude Code skill designed for researchers and developers working on drug discovery, chemical informatics, and protein modeling. It provides a comprehensive suite of tools for predicting molecular properties like toxicity and solubility, generating novel chemical structures, and performing knowledge graph reasoning for drug repurposing. By integrating with PyTorch and RDKit, this skill streamlines the implementation of complex Graph Neural Networks (GNNs) and protein structure analysis, making it an essential resource for biomedical research and machine learning applications in the life sciences.

主な機能

Protein sequence and 3D structure modeling with AlphaFold and ESM integration
Biomedical knowledge graph reasoning for drug repurposing and link prediction
1 GitHub stars
Built-in access to 40+ curated datasets and 20+ GNN architectures like GIN and SchNet
De novo molecular generation and multi-step retrosynthesis planning
Molecular property prediction for ADMET, toxicity, and binding affinity

ユースケース

Discovering potential disease treatments through biological knowledge graph completion
Designing synthetic routes for molecules via automated retrosynthesis models
Predicting ADMET properties and solubility for novel drug candidates

概要

主な機能

Protein sequence and 3D structure modeling with AlphaFold and ESM integration
Biomedical knowledge graph reasoning for drug repurposing and link prediction
1 GitHub stars
Built-in access to 40+ curated datasets and 20+ GNN architectures like GIN and SchNet
De novo molecular generation and multi-step retrosynthesis planning
Molecular property prediction for ADMET, toxicity, and binding affinity

ユースケース

Discovering potential disease treatments through biological knowledge graph completion
Designing synthetic routes for molecules via automated retrosynthesis models
Predicting ADMET properties and solubility for novel drug candidates