Build molecular features without losing label alignment.
Convert small-molecule structures into features for molecular machine learning and similarity search.
When to use it
Use when converting SMILES or RDKit molecules into molecular representations or comparing representations on the same split and assay endpoint.
It changes how the agent implements molecular featurization, preserving record alignment, keeping learned preprocessing inside training folds, and saving reproducibility information.
What you provide
Requires Python 3.11–3.14; macOS Intel requires Python 3.11–3.12 and platform-specific dependency constraints.
Requires molfeat 1.0.0. Core fingerprints run offline; pretrained models need optional extras.
This skill
https://fs.molfeat.datamol.io/artifacts/
Reads public model metadata
https://arxiv.org/abs/2609.00065
Fetches the authors' citation record