Datamol Cheminformatics Skill

What it does

This skill teaches Claude the idiomatic way to use datamol, a lightweight Pythonic layer over RDKit that returns native rdkit.Chem.Mol objects. It documents ten workflow areas: format conversion (SMILES/InChI/SELFIES), sanitization and full standardization, molecular descriptors, fingerprints (ECFP, MACCS, atom pairs) with Tanimoto distances, Butina clustering and diversity picking, Bemis-Murcko scaffold analysis and scaffold splits, BRICS/RECAP fragmentation, 3D conformer generation with RMSD clustering and SASA, visualization (grids, publication SVG, substructure highlighting), and reaction SMARTS application.

It also covers built-in parallelism via n_jobs, progress bars, cloud/HTTPS file I/O through fsspec, scikit-learn integration for QSAR modeling, defensive error handling patterns, and a troubleshooting section for parsing failures, clustering memory blowups and slow conformer generation.

Who it's for

• Drug discovery and medicinal chemistry researchers handling compound libraries
• Data scientists who find raw RDKit verbose or fiddly
• Anyone building ML datasets from molecules, including scaffold-based splits
• Teams running virtual screening or SAR analysis pipelines repeatedly

Examples

1. "Standardize the 50k SMILES in this CSV, drop invalid structures, and produce a descriptor table" → generates dm.to_mol + dm.standardize_mol + batch_compute_many_descriptors(n_jobs=-1) code.
2. "Pick the 500 most diverse compounds from this SDF using ECFP4" → dm.read_sdf → dm.to_fp → dm.pick_diverse pipeline.
3. "Split train/test by scaffold and fit a Random Forest activity model" → Bemis-Murcko scaffold grouping plus scikit-learn training code.

Install the skill I found on Claude Skill Mart.
Copy the skills/datamol folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/datamol/.
When it's done, tell me in one line what this skill can do.