This is a solid workflow for analyzing protein interaction networks without getting lost in the biology. It walks through identifier mapping, network retrieval from STRING and BioGRID, enrichment analysis, and optional structural data. The standout feature is the domain reasoning guidance: it actually explains that a high STRING combined_score doesn't mean two proteins physically bind, and tells you to check the experimental score instead. That kind of clarity matters when you're building tools for researchers who need accurate interpretations, not just data dumps. Requires a BioGRID API key for the fallback database. Built by the Harvard medical informatics group, which shows in the thoughtful workflow design.
npx skills add https://github.com/mims-harvard/tooluniverse --skill protein-interaction-network-analysis