---

name: descriptors-fingerprints description: "Calculate RDKit molecular descriptors, fingerprints, bit-vector similarities, clustering inputs, and ML-ready feature tables." disable-model-invocation: true metadata: disco-role: operating license: BSD 3-Clause

RDKit Descriptors and Fingerprints

Use this sub-skill when a request involves molecular properties, feature engineering, fingerprint generation, similarity search, bit vectors, distance matrices, or Butina clustering.

Route Here

• Calculate scalar descriptors with rdkit.Chem.Descriptors, rdkit.Chem.rdMolDescriptors, rdkit.Chem.Lipinski, rdkit.Chem.Crippen, rdkit.Chem.MolSurf, or rdkit.Chem.QED.
• Build feature dictionaries or pandas-ready rows from molecules and descriptor functions.
• Generate Morgan, RDKit, atom-pair, or topological-torsion fingerprints with rdkit.Chem.rdFingerprintGenerator.
• Compare fingerprints with rdkit.DataStructs metrics such as Tanimoto, Dice, cosine, and bulk similarity helpers.
• Convert similarities to distances for nearest-neighbor search, clustering, or ML inputs.
• Cluster molecules with rdkit.ML.Cluster.Butina.ClusterData from condensed distance lists or feature vectors.

Route Elsewhere

• Use ../molecule-io-core/ for SMILES/SDF parsing, sanitization, molecule validation, and canonicalization before descriptor work.
• Use ../conformers-drawing/ for 3D conformer generation, shape descriptors, alignment, RMSD, or drawing similarity maps.
• Use ../contrib-utilities/ for optional contributed scorers such as SA Score, NP Score, Fraggle, MMPA, or NIBR filters.
• Use ../data-cli-integration/ for RDKit data-file locations, feature-definition files, database helpers, or non-feature-engineering pandas integration.

Start With These References

• references/descriptors.md for descriptor families, drug-like properties, and feature-table recipes.
• references/fingerprints-and-similarity.md for fingerprint generator APIs, vector types, similarity, top-k search, and Butina clustering.
• references/troubleshooting.md for deprecated Morgan helpers, invalid molecules, vector mismatches, sparse/count/vector issues, and pandas export pitfalls.
• scripts/fingerprint_similarity.py for a small, self-contained CLI that parses SMILES, builds Morgan generator fingerprints, and reports Tanimoto similarities while surfacing invalid inputs.

Common Patterns

• Prefer the current generator API: rdFingerprintGenerator.GetMorganGenerator(radius=2, fpSize=2048).GetFingerprint(mol).
• Use Descriptors.CalcMolDescriptors(mol) when a broad scalar descriptor dictionary is more useful than calling individual descriptor functions.
• Use rdMolDescriptors.CalcExactMolWt(mol), Crippen.MolLogP(mol), MolSurf.TPSA(mol), Lipinski.NumHDonors(mol), and QED.qed(mol) for focused medicinal-chemistry properties.
• Check every parsed molecule for None before descriptor or fingerprint calculation; most downstream APIs expect a valid Chem.Mol.
• For Butina from fingerprint similarities, pass distances as 1.0 - similarity in condensed lower-triangle order with isDistData=True.

Quick Smoke

Run the bundled helper on a few molecules:

python scripts/fingerprint_similarity.py --smiles "CCO" "CCCO" "c1ccccc1" --top-k 2

For a query-vs-library search:

python scripts/fingerprint_similarity.py --query "CCO" --smiles "CCO" "CCN" "c1ccccc1" --top-k 3