Standardize

molfp.standardize

Structure normalization: standardize molecules before fingerprint computation.

Provides a pipeline for cleaning, desalting, normalizing, and deduplicating molecules — critical preprocessing before fingerprint calculation to ensure consistent and reproducible results.

Pipeline (following Greg Landrum's RSC OpenScience protocol): 1. Parse SMILES → RDKit Mol 2. Cleanup: remove Hs, disconnect metals, normalize, reionize 3. Desalt: keep largest fragment (FragmentParent) 4. Neutralize: remove charges where possible (Uncharger) 5. Canonicalize tautomer: pick a single canonical tautomer form 6. Canonical SMILES: generate standardized SMILES string 7. InChIKey: compute for deduplication

standardize_mol(smiles_or_mol)

Full standardization pipeline. Returns standardized RDKit Mol or None.

Steps: parse → cleanup → desalt → neutralize → canonical tautomer.

smiles_to_mol(input_data, *, column=None, standardize=False)

Convert SMILES to RDKit Mol objects.

Parameters:

Name	Type	Description	Default
input_data		SMILES string, list of SMILES, pd.Series, or pd.DataFrame	required
column		DataFrame column name (auto-detected if None)	None
standardize		if True, apply full standardization pipeline	False

Returns:

Type	Description
	Single Mol (for string input) or list[Mol|None] (for batch input).
	Returns None for invalid SMILES.

mol_to_smiles(input_data, *, column=None, canonical=True)

Convert RDKit Mol objects (or SMILES) to canonical SMILES strings.

Parameters:

Name	Type	Description	Default
input_data		Mol, SMILES string, list, pd.Series, or pd.DataFrame	required
column		DataFrame column name (auto-detected if None)	None
canonical		if True (default), return canonical SMILES	True

Returns:

Type	Description
	Single SMILES string (for single input) or list[str|None] (for batch).

standardize_smiles(smiles)

Standardize a SMILES string. Returns canonical SMILES or None.

get_inchikey(smiles_or_mol)

Compute InChIKey for a molecule (after standardization).

desalt(smiles_or_mol)

Remove salts/counterions, keep the largest fragment.

neutralize(smiles_or_mol)

Remove charges where chemically reasonable.

canonicalize_tautomer(smiles_or_mol)

Pick the canonical tautomer form.

standardize_batch(smiles_list, return_failures=False)

Standardize a list of SMILES. Returns list of canonical SMILES.

Parameters:

Name	Type	Description	Default
smiles_list	list[str]	list of SMILES strings	required
return_failures	bool	if True, return (results, failures) tuple	False

Returns:

Type	Description
list[str | None] | tuple[list[str | None], list[tuple[int, str]]]	list of standardized SMILES (None for failures)
list[str | None] | tuple[list[str | None], list[tuple[int, str]]]	if return_failures: (results, [(idx, original_smiles), ...])

deduplicate(smiles_list, method='inchikey')

Deduplicate molecules using InChIKey or canonical SMILES.

Parameters:

Name	Type	Description	Default
smiles_list	list[str]	list of SMILES strings	required
method	str	"inchikey" (default, handles tautomers/salts) or "canonical"	'inchikey'

Returns:

Type	Description
dict[str, Any]	dict with: unique_smiles: list of unique standardized SMILES unique_indices: indices of first occurrence duplicates: list of (idx, original_smiles, duplicate_of_idx) mapping: {original_idx: unique_idx} n_input: total input n_unique: unique count n_duplicates: duplicate count n_failed: failed to parse

preprocess(smiles_list, deduplicate_method='inchikey')

Full preprocessing pipeline: standardize + deduplicate.

Convenience function combining standardize_batch and deduplicate.

Returns:

Type	Description
dict[str, Any]	dict with: clean_smiles: list of unique standardized SMILES clean_mols: list of RDKit Mol objects (ready for FP computation) dedup_result: full deduplication result dict standardization_failures: [(idx, smiles), ...]