The Pharmacophore Network: A Computational Method for Exploring Structure–Activity Relationships from a Large Chemical Data Set

Historically, structure-activity relationship (SAR) analysis has focused on small sets of molecules, but in recent years, there has been increasing efforts to analyze the growing amount of data stored in public databases like ChEMBL. The pharmacophore network introduced herein is dedicated to the organization of a set of pharmacophores automatically discovered from a large data set of molecules. The network navigation allows to derive essential tasks of a drug discovery process, including the study of the relations between different chemical series, the analysis of the influence of additional chemical features on the compounds' activity, and the identification of diverse binding modes. This paper describes the method used to construct the pharmacophore network, and a case study dealing with BCR-ABL exemplifies its usage for large-scale SAR analysis. Thanks to a benchmarking study, we also demonstrate that the selection of a subset of representative pharmacophores can be used to conduct classification tasks.