torchdrug
PyTorch-native graph neural networks for molecules and proteins. Use when building custom GNN architectures for drug discovery, protein modeling, or knowledge graph reasoning. Best for custom model development, protein property prediction, retrosynthesis. For pre-trained models and diverse featurizers use deepchem; for benchmark datasets use pytdc.
Security Assessment
About torchdrug
TorchDrug is a powerful machine learning toolbox that leverages PyTorch to facilitate drug discovery and molecular science through the implementation of graph neural networks (GNNs). It addresses the challenges faced in predicting molecular properties, modeling proteins, and reasoning through biological knowledge graphs by providing a framework that incorporates various models and datasets tailored for these tasks. With TorchDrug, researchers and developers can build custom GNN architectures specifically designed for applications in drug discovery, protein modeling, and retrosynthesis planning, thus advancing their capabilities in scientific research.
FAQ
How do I install TorchDrug?
You can install TorchDrug using pip with the command `pip install torchdrug`, or with optional dependencies using `pip install torchdrug[full]`.
What types of data can I use with TorchDrug?
TorchDrug supports various data types including SMILES strings, protein sequences, PDB files, chemical reactions, and biomedical knowledge graphs.
Is TorchDrug compatible with other libraries?
Yes, TorchDrug is compatible with PyTorch, PyTorch Lightning, and can be used in conjunction with RDKit, AlphaFold, and ESM.
What are the main tasks I can perform with TorchDrug?
TorchDrug allows for molecular property prediction, protein modeling, drug-target binding prediction, molecular generation, and retrosynthesis planning.
Are there any pre-trained models available in TorchDrug?
TorchDrug includes multiple model architectures and datasets, but for pre-trained models and diverse featurizers, you may want to consider using deepchem.
Install torchdrug
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
K-Dense-AI/claude-scientific-skills