Abstract

Pharmaceutical companies regularly run campaigns to evolve their proprietary chemical libraries which are among their most valuable assets. Ultimate goal with those library expansions is to address novel chemical space with maximal fit to pharmaceutically relevant targets which is beyond just applying property or druglikeness filters. In this work we present a structured and highly automated process to identify putative biological targets starting from any chemistry-driven virtual or existing compound library. Multiple ligand similarity searches are performed in ChEMBL ligand space, linking library compounds to targets from ChEMBL database. The results are presented to the computational chemist in a highly intuitive and interactive manner. For a set of targets selected by a scientist, holo crystal structures are automatically retrieved and prepared for docking. The cocrystallized ligand, ChEMBL compounds and combinatorial library are then docked by an automatic procedure. The scientist finally is provided with a holistic picture of library-target fit hypotheses to draw his conclusions about relevant targets, library adjustments, library re-designs and ideas for completely new virtual libraries.