Andreas Bergner, Serge P. Parel
ABSTRACT: Ligand-based virtual screening and computational hit expansion methods undoubtedly facilitate the finding of novel active chemical entities, utilizing already existing knowledge of active compounds. It as been demonstrated that the parallel execution of complementary similarity search methods enhances the performance of such virtual screening campaigns. In this article, we examine the use of virtualized emplate (query, seed) structures as an extension to common search methods, such as fingerprint and pharmacophore graph-based similarity searches. We demonstrate that template virtualization by bioisosteric enumeration and other rule-based methods, in combination with standard similarity search techniques, represents a powerful approach for hit expansion following high-throughput screening campaigns. The reliability of the methods is demonstrated by four different test data sets representing different target classes and two hit finding case studies on the epigenetic targets G9a and LSD1.
J. Chem. Inf. Model., 2013, 53(5), pp 1057–1066 (full text)