What is the TorchDrug skill for Claude?

It is a specialized capability that helps Claude guide users through implementing drug discovery and molecular science tasks using the TorchDrug PyTorch-based library.

How does it handle chemical data?

It integrates seamlessly with SMILES strings and RDKit, allowing for easy conversion between standard cheminformatics formats and TorchDrug graph objects.

Can I use this skill for protein structure prediction?

Yes, it supports protein modeling using sequence models like ESM and structure-aware models like GearNet, integrating with PDB files and AlphaFold data.

Which GNN architectures are supported?

The skill provides guidance for 20+ architectures including GIN, GAT, SchNet, and specialized models for proteins and knowledge graphs.

Does it support molecular generation?

Absolutely. The skill provides patterns for de novo design using autoregressive models and reinforcement learning techniques like GCPN for property-guided generation.

TorchDrug Discovery ML

Name: TorchDrug Discovery ML
Author: K-Dense-AI

byK-Dense-AI

•

2,188

Data Science & ML

Simplifies graph-based machine learning for drug discovery, protein modeling, and molecular science using the TorchDrug framework.

About

This skill equips Claude with the domain-specific knowledge and implementation patterns needed to leverage the TorchDrug library for advanced biomedical research and machine learning. It provides expert guidance on training graph neural networks (GNNs) for molecular property prediction, performing protein sequence and structure analysis, executing knowledge graph reasoning for drug repurposing, and implementing molecular generation or retrosynthesis pipelines. It is an essential tool for researchers and developers working at the intersection of deep learning and chemistry, offering ready-to-use workflows for over 40 curated datasets and 20+ specialized model architectures.

Key Features

Knowledge graph reasoning for link prediction and drug discovery
Protein modeling using both sequence (ESM) and 3D structure (GearNet) data
Retrosynthesis planning for chemical synthesis route prediction
Molecular property prediction for ADMET and toxicity screening
Generative modeling for de novo molecular design and lead optimization
2,188 GitHub stars

Use Cases

Predicting drug-likeness and biological activity of novel chemical compounds
Identifying new therapeutic uses for existing drugs through biomedical knowledge graph analysis
Analyzing protein functions and interactions using 3D-aware graph neural networks

About

Key Features

Knowledge graph reasoning for link prediction and drug discovery
Protein modeling using both sequence (ESM) and 3D structure (GearNet) data
Retrosynthesis planning for chemical synthesis route prediction
Molecular property prediction for ADMET and toxicity screening
Generative modeling for de novo molecular design and lead optimization
2,188 GitHub stars

Use Cases

Predicting drug-likeness and biological activity of novel chemical compounds
Identifying new therapeutic uses for existing drugs through biomedical knowledge graph analysis
Analyzing protein functions and interactions using 3D-aware graph neural networks