Chemometric approaches to virtual screening

CINF 81

Alexander Tropsha, Scott Oloff, Shuxing Zhang, and Min Shen. Laboratory for Molecular Modeling, School of Pharmacy, University of North Carolina at Chapel Hill, CB # 7360, Beard Hall, School of Pharmacy, Chapel Hill, NC 27599-7360
We discuss novel chemometric approaches to both ligand and receptor based virtual screening, which characterize both ligands and receptors (if available) in multidimensional space of chemical descriptors. In ligand based screening, we employ rigorously validated QSAR models to mine chemical databases for compounds with high predicted activity. We demonstrate that this approach yields an exceptionally high experimental hit rate in identifying anticonvulsant compounds from a set of 250,000 molecules. We also report on a novel approach to identifying Complementary Ligands Based on Receptor Information (CoLiBRI). CoLIBRI transforms chemical structure of both ligands and their complimentary active sites into the high-dimensional descriptor space and uses specially developed chemical similarity metrics to mine targetís complementary ligands from large databases. The results illustrate that CoLiBRI is capable of identifying all known ligands of 260 test binding sites within the top 1% of the database of ca. 60,000 compounds in 95% of all cases.