Find activity cliffs — near-identical molecule pairs with large potency differences — and characterize whether a SAR landscape is smooth or rugged. Computes SALI (structure-activity landscape index) over fingerprint similarity, ranks cliff pairs, identifies which compounds sit on the most cliffs, and reports the fraction of similar pairs that are cliffs as a ruggedness measure. Use to locate SAR hot spots and to predict whether QSAR will generalize on a dataset. Requires curated log-scale activity. For cliffs where you also know exactly which atoms changed, prefer matched-pairs.
