Extraction of SAR information from activity cliff clusters via matching molecular series

The vast majority of activity cliffs that occur is sets of bioactive compounds are formed in a coordinated manner. This means that multiple and overlapping cliffs are formed by groups of structural analogs with varying activity. In network representations, coordinated activity cliffs emerge as clusters of varying size and topology. Activity cliff clusters are typically rich in structure-activity relationship (SAR) information but often difficult to analyze from a medicinal chemistry viewpoint. A key question is how to best access SAR information contained in activity cliff clusters without the need to evaluate many different clusters individually. Herein, we introduce a methodology for the systematic extraction of SAR information from activity cliff clusters that utilizes the concept of matching molecular series (MMS). Sequences of activity cliff-forming compounds are isolated from clusters that follow a activity gradient and series spanning large activity differences are preferentially selected. In addition to its systematic nature, an attractive feature of the approach is that SAR information associated with extracted series is readily interpretable. We show that MMS are abundant in activity cliff clusters from the current spectrum of bioactive compounds and that many MMS share compounds. The resulting pairs of connected MMS contain compounds with closely related structural cores and alternative substitution sites that reveal SAR determinants and preferred substituents.