Medchem

If you're triaging compound libraries in drug discovery, this gives you the standard filters in one place. Run Lipinski/Veber/CNS rules, flag PAINS and structural alerts, or build composite queries like "CNS-like AND no PAINS AND TPSA under 90" with the built-in query language. It wraps datamol and RDKit with a functional API that returns DataFrames or boolean masks, so you can chain filters over thousands of molecules without writing boilerplate. The NIBR filters and complexity percentiles are handy for screening deck curation. One catch: the Lilly demerit filter needs separate conda binaries. Best used as a first-pass filter before you apply target-specific knowledge, not as a final decision layer.

Install to Claude Code

npx -y skills add k-dense-ai/scientific-agent-skills --skill medchem --agent claude-code