Selective potentiators of the metabotropic glutamate receptor subtype mGluR5 have exciting potential for development of novel treatment strategies for schizophrenia and other disorders that disrupt cognitive function. The latest generation of selective mGluR5 potentiators is based on the lead compound CDPPB and features systemically active compounds with long half-lives that cross the blood-brain barrier. A high-throughput screen (HTS) for mGluR5 potentiators at Vanderbilts NIH-funded molecular libraries screening center network facility revealed a large and diverse set of about 1400 substances whose activity was validated in independent experiments. The present ChemInformatics proposal targets utilizing the power of recent machine learning techniques such as Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) to model the complex relationship between chemical structure and biological activity of mGluR5 potentiators reflected in the HTS results. An innovative encoding scheme is developed that allows mapping of the diverse chemical space into a single mathematical model. The resulting Quantitative Structure Activity Relation (QSAR) models will serve a three-fold purpose: (a) a comprehensive binding site pharmacophore will be obtained to facilitate understanding of the SAR and rationalize further experiments;(b) the models will be used to virtually screen libraries of millions of compounds which are known but not physically available for HTS at Vanderbilt to gain a priority list for acquisition or synthesis;and (c) in combination with an existing Genetic Algorithm (GA) structure generator existing lead compounds will be optimized and new structures will be designed to identify potential new targets for synthesis. Overall we hope to not only identify novel allosteric potentiators of mGluR5 and understand their activity as potential treatment of schizophrenia and other disorders that disrupt cognitive function, but also to build an innovative ChemInformatics software and database tool which can be adopted for research in other NIH molecular libraries screening centers. The developed applications will be made freely and readily accessible for academic research using a WWW interface deeply integrated in the drug development pipeline. The employed QSAR models require no crystal structure of the target protein. Hence the method can be readily applied to membrane proteins-such as GPCRs-which are target of 40-50% of modern medicinal drugs. The PI of the proposal has extensive experience in the usage of ANNs and SVMs to predict properties of organic molecules and proteins (1-9), solve protein structures (10-15), and predict activity of therapeutics (16). He implemented GAs for the design and optimization of chemical structures (17,18). For the tasks at hand he teams up with Jeff Conn, a renowned expert for researching mGluRs (19) and potential therapeutics targeting these systems (20-22).

National Institute of Health (NIH)
National Institute of Mental Health (NIMH)
Exploratory/Developmental Grants (R21)
Project #
Application #
Study Section
Special Emphasis Panel (ZRG1-MDCN-C (91))
Program Officer
Winsky, Lois M
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Vanderbilt University Medical Center
Schools of Medicine
United States
Zip Code