Exploring structure-promiscuity relationships using dual-site promiscuity cliffs and corresponding single-site analogs

Currently available volumes of compounds and biological activity data enable large-scale analyses of compound promiscuity (multi-target activity). To aid in the exploration of structure-promiscuity relationships, promiscuity cliffs (PCs) were introduced previously. In analogy to activity cliffs, PCs were defined as pairs of structurally analogous compounds with large differences in the number of targets they are active against. Hence, PCs reveal small chemical modifications that are implicated in promiscuity. As introduced originally, PCs were identified by applying the matched molecular pair formalism and were thus confined to changes at a single substitution site. Herein, PCs with multiple substitution sites are introduced and a pilot study on a large collection of protein kinase inhibitors is reported, which provide excellent test cases for promiscuity analysis. For dual-site PCs (dsPCs), which dominated the distribution of multi-site PCs, an extended data structure was generated comprising a dsPC and two single-site analogs accounting for individual substitutions. Using a canonical representation, extended dsPCs are intuitive and easy to interpret from a chemical perspective. The analog quartet representing an extended dsPC is rich in structure-promiscuity relationship information and makes it possible to evaluate the potential interplay of chemical modifications implicated in promiscuity. Furthermore, extended dsPCs provide insights into possible experimental causes of apparent differences in analog promiscuity such as varying test frequencies. Hence, the newly introduced PC format should be of interest for exploring origins of compound promiscuity in medicinal chemistry and for formulating experimentally testable target hypotheses for analogs.