Graph-based drug discovery toolkit. Molecular property prediction (ADMET), protein modeling, knowledge graph reasoning, molecular generation, retrosynthesis, GNNs (GIN, GAT, SchNet), 40+ datasets, for PyTorch-based ML on molecules, proteins, and biomedical graphs.
8.7
Rating
0
Installs
Machine Learning
Category
Excellent skill for drug discovery ML tasks. The description is comprehensive and clearly delineates when to use TorchDrug (molecular properties, proteins, knowledge graphs, generation, retrosynthesis). SKILL.md provides a strong overview with quick-start code, then systematically covers 7 core capabilities with clear pointers to detailed reference files. Structure is exemplary: concise main file with logical sectioning, troubleshooting guide, and well-organized references for deep dives. Task knowledge is outstanding - covers complete workflows from dataset selection through model choice to evaluation. Novelty is strong: TorchDrug's specialized GNN architectures, 40+ curated datasets, and multi-modal capabilities (molecules + proteins + KGs) would require extensive manual setup and domain expertise. A CLI agent would struggle with the domain-specific model selection, proper graph construction for different molecular representations, and integration of cheminformatics tools. Minor improvement possible: could add more concrete code snippets in the workflows section, though the referenced files likely contain these details.
davila7
Skill Author
Loading SKILL.md…
