Optimization of the MAD algorithm for virtual screening

Abstract

The approach termed Determination and Mapping of Activity-Specific Descriptor Value Ranges (MAD) is a conceptually novel molecular similarity method for the identification of active compounds. MAD is based on mapping of compounds to different (multiple) activity class-selective descriptor value ranges. It was recently developed in our laboratory and successfully applied in initial virtual screening trials. We have been able to show that selected molecular property descriptors can display a strong tendency to respond to unique features of compounds having similar biological activity, thus providing a basis for the approach. Accordingly, a crucial step of the MAD algorithm is the identification of activity-sensitive descriptor settings. The second critical step of MAD is the subsequent mapping of database compounds to activity-sensitive descriptor value ranges in order to identify novel active molecules. This chapter describes the optimization of the MAD approach and evaluates the second-generation algorithm on a number of different compound activity classes with a particular focus on the recognition of remote molecular similarity relationships. © 2008 Humana Press, a part of Springer Science+Business Media, LLC.