rdkit-molecular-descriptors— agent skill

Purpose

Turn a SMILES string into a compact descriptor summary suitable for quick triage, featurization, and metadata enrichment.

When to use

• You need a local RDKit-backed molecular summary from a SMILES string.
• You want canonical SMILES, formula, weight, and simple physicochemical descriptors.

When not to use

• You need conformer generation, docking, or reaction modeling.
• You need batch-scale descriptor computation for millions of compounds.

Inputs

• One SMILES string
• Optional molecule name
• Optional JSON output path

Outputs

• JSON summary with canonical SMILES, formula, exact mass, logP, TPSA, HBA/HBD, rotatable bonds, ring count, and heavy-atom count

Requirements

• slurm/envs/chem-tools with RDKit available

Procedure

1. Run slurm/envs/chem-tools/bin/python skills/drug-discovery-and-cheminformatics/rdkit-molecular-descriptors/scripts/compute_rdkit_descriptors.py --smiles "CC(=O)OC1=CC=CC=C1C(=O)O" --name aspirin.
2. Inspect canonical_smiles, formula, molecular_weight, and tpsa.
3. Reuse the JSON payload for downstream filtering or reporting.

Validation

• The command exits successfully.
• canonical_smiles is non-empty for a known valid molecule.
• Descriptor fields are numeric where expected.

Failure modes and fixes

• Invalid SMILES: verify the input string and quoting.
• Missing RDKit environment: run the script with slurm/envs/chem-tools/bin/python.

Safety and limits

• Descriptor calculation only.
• No medicinal-chemistry or safety interpretation is implied.

Provenance

• RDKit docs: https://www.rdkit.org/docs/index.html
• RDKit repository: https://github.com/rdkit/rdkit

Related skills

• chembl-molecule-search