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.
8.3
Rating
0
Installs
Machine Learning
Category
Exceptional skill for TorchDrug with comprehensive coverage of graph neural networks for drug discovery. The description clearly delineates when to use this skill vs alternatives (deepchem, pytdc). SKILL.md provides outstanding structural organization with a concise overview and systematic routing to 8 detailed reference files covering all major use cases (molecular property prediction, protein modeling, knowledge graphs, generation, retrosynthesis). The 5 workflow examples provide clear step-by-step guidance linking to appropriate references. Integration patterns with RDKit, AlphaFold/ESM, and PyTorch Lightning are practical. The quick reference cheat sheet and troubleshooting section enhance usability. Task knowledge is thorough with code examples and best practices throughout. Novelty is strong - custom GNN architectures, multi-step retrosynthesis planning, and biomedical knowledge graph reasoning require significant domain expertise and many tokens for a CLI agent alone. Minor deduction on novelty as some molecular property prediction tasks could be handled by simpler tools, but the skill's focus on custom model development and complex architectures justifies its value.
K-Dense-AI
Skill Author
Loading SKILL.md…
