PyDompeKeyGen

A lightweight Python / RDKit implementation of DompéKeys — a set of 1,064 SMARTS-based structural keys for mapping chemical space, built for use as an interpretable molecular fingerprint.

DompéKeys were originally developed by scientists at Dompé farmaceutici as part of the EXSCALATE drug-discovery platform and published in the Journal of Cheminformatics (Manelfi, Tazzari, et al., 2024). This package provides a small, dependency-light way to generate the corresponding fingerprint for any RDKit molecule.

✨ Features

• 🧬 1,064 curated SMARTS keys spanning amino acids, acids, metal binders, toxicophores, ring systems, and generic pharmacophoric features (H-bond donors/acceptors, etc.), organized across 5 complexity levels.
• 🎯 Single-class API — one object, one method, one RDKit ExplicitBitVect out.
• ⚡ Fast startup — the SMARTS catalog is compiled once and cached to disk (smarts.pkl), and automatically rebuilt if your installed RDKit version changes.
• 🔍 Interpretable — every set bit maps back to a named, human-readable substructure.

📦 Installation

PyDompeKeyGen is not yet published to PyPI. Install it directly from GitHub:

pip install git+https://github.com/OlivierBeq/PyDompeKeyGen.git

Or clone and install locally for development:

git clone https://github.com/OlivierBeq/PyDompeKeyGen.git
cd PyDompeKeyGen
pip install -e .

Requires Python ≥ 3.10 and RDKit (installed automatically as a dependency).

🚀 Quick start

from rdkit import Chem
from pydompekeygen import PyDompeKeys

# Load the DompéKeys catalog (built once, then cached on disk)
dompekeys = PyDompeKeys()

# Generate a fingerprint for a molecule
mol = Chem.MolFromSmiles("CC(=O)Oc1ccccc1C(=O)O")  # aspirin
fingerprint = dompekeys.GetFingerprint(mol)

print(f"{fingerprint.GetNumOnBits()} / {dompekeys.num_bits} keys matched")

See the usage guide for similarity calculations, converting fingerprints to NumPy arrays, batch processing, and inspecting which named keys matched.

📚 Documentation

📖 Citation

If you use PyDompeKeyGen in published work, please cite the original DompéKeys paper:

Manelfi, C., Tazzari, V., Lunghini, F. et al. "DompeKeys": a set of novel substructure-based descriptors for efficient chemical space mapping, development and structural interpretation of machine learning models, and indexing of large databases. J Cheminform 16, 21 (2024). https://doi.org/10.1186/s13321-024-00813-4

See docs/background.md for a BibTeX entry.

🤝 Contributing

Issues and pull requests are welcome at github.com/OlivierBeq/PyDompeKeyGen.

⚖️ License

Released under the MIT License.