Chemical Space Mapping and Structure–Activity Analysis of the ChEMBL Antiviral Compound Set

Curation, standardization and data fusion of the antiviral information present in the ChEMBL public database led to the definition of a robust data set, providing an association of antiviral compounds to seven broadly defined antiviral activity classes. Generative topographic mapping (GTM) subjected to evolutionary tuning was then used to produce maps of the antiviral chemical space, providing an optimal separation of compound families associated with the different antiviral classes. The ability to pinpoint the specific spots occupied (responsibility patterns) on a map by various classes of antiviral compounds opened the way for a GTM-supported search for privileged structural motifs, typical for each antiviral class. The privileged locations of antiviral classes were analyzed in order to highlight underlying privileged common structural motifs. Unlike in classical medicinal chemistry, where privileged structures are, almost always, predefined scaffolds, privileged structural motif detection based on GTM responsibility patterns has the decisive advantage of being able to automatically capture the nature ("resolution detail"-scaffold, detailed substructure, pharmacophore pattern, etc.) of the relevant structural motifs. Responsibility patterns were found to represent underlying structural motifs of various natures-from very fuzzy (groups of various "interchangeable" similar scaffolds), to the classical scenario in medicinal chemistry (underlying motif actually being the scaffold), to very precisely defined motifs (specifically substituted scaffolds).