Activity Cliffs

molfp.cliffs

Activity cliff detection: find molecule pairs with high structural similarity but large activity differences.

activity_cliffs(smiles_list, activities, fp_type='ecfp4', sim_threshold=0.8, activity_threshold=2.0, metric='tanimoto')

Find activity cliff pairs in a dataset.

An activity cliff is a pair of molecules with high structural similarity (Tanimoto >= sim_threshold) but large activity difference (|activity_a - activity_b| >= activity_threshold).

Parameters:

Name	Type	Description	Default
smiles_list	list[str]	list of SMILES strings	required
activities	list[float] | ndarray	list/array of activity values (e.g., pIC50)	required
fp_type	str	fingerprint to use for similarity	'ecfp4'
sim_threshold	float	minimum similarity to be considered structurally similar	0.8
activity_threshold	float	minimum activity difference for a cliff	2.0
metric	str	similarity metric ("tanimoto" or "dice")	'tanimoto'

Returns:

Type	Description
list[dict[str, Any]]	list of dicts, each with: idx_a, idx_b, smiles_a, smiles_b, similarity, activity_a, activity_b, delta_activity
list[dict[str, Any]]	Sorted by delta_activity descending.

cliff_summary(smiles_list, activities, fp_types=None, sim_threshold=0.8, activity_threshold=2.0)

Compare activity cliff counts across fingerprint types.

Returns list of {fp_type, n_cliffs, max_delta, mean_delta}.