chemistry

RDKit-backed molecular helpers (SMILES, SELFIES, fingerprints, properties).

exception drugs.chemistry.RDKitNotAvailable

Bases: RuntimeError

Raised when RDKit-dependent functionality is requested but unavailable.

exception drugs.chemistry.SELFIESNotAvailable

Bases: RuntimeError

Raised when SELFIES dependency is missing.

drugs.chemistry.mol_from_smiles(smiles: str, *, sanitize: bool = True) → Any | None

Parse a SMILES string into an RDKit molecule, handling errors gracefully.

When RDKit is unavailable, returns the SMILES string so downstream helpers can fall back to hashed fingerprints and lightweight heuristics.

drugs.chemistry.to_selfies(smiles: str) → str | None

Convert SMILES to SELFIES, returning None when conversion fails.

drugs.chemistry.molecular_fingerprint(mol: Any, *, method: str = 'morgan', n_bits: int = 2048, radius: int = 2, use_features: bool = False) → ndarray

Generate a molecular fingerprint as a numpy array of bits.

If RDKit is unavailable, a deterministic hashed fingerprint derived from SMILES is returned instead.

drugs.chemistry.fingerprint_similarity(fp_a: ndarray, fp_b: ndarray, *, metric: str = 'tanimoto') → float

Compute similarity between two bit vectors using common coefficients.

drugs.chemistry.rdkit_properties(mol: Any) → Dict[str, Any]

Compute common RDKit-derived molecular properties.

When RDKit is missing, returns a placeholder structure with None values and Lipinski checks skipped. This allows downstream code to continue running while reporting the absence of cheminformatics features.

drugs.chemistry.lipinski_violations(mol: Any) → List[str]

Return a list of violated Lipinski rules for the molecule.

drugs.chemistry.synthetic_accessibility(mol: Any) → float | None

Estimate synthetic accessibility score (lower is easier).

drugs.chemistry.require_rdkit() → None