Application of privileged substructure identification to drug discovery

Christophe Cleva ¹, Eric Sebille ¹, Cedric Merlot ¹, Jean Bunn ¹, Wolfgang Sauer ², Daniel Domine ¹ and Dennis J. Church ¹.

1 Scientific Computing and 2 Chemistry Groups, Serono Pharmaceutical Research Institute, Geneva, Switzerland.

Fast and efficient in silico prediction tools have become essential to reduce time and costs in the drug discovery process. The nature and amount of data generated at each stage of this process are drastically different but all require statistically sound approaches using descriptors that can easily be computed and interpreted in terms of chemical structures to rationalize the succeeding experiments. This presentation describes a recently developed approach aimed at identifying the fragments statistically related to biological outcomes. Applications ranging from focused set design to selectivity prediction and using data collected from both in-house and external sources are presented.

Key words: Structure-Activity Relationships, Substructure Analysis, Virtual Screening