-------------------------------------------------------------------- COLLOQUIUM OF THE COMPUTATIONAL MATERIALS SCIENCE CENTER College of Science (CDS Department CSI 898-Sec 001) -------------------------------------------------------------------- Cheminformatics: Linking chemical structure and biological activity using machine learning and datamining techniques Alex Tropsha Chair, Division of Medicinal Chemistry and Natural Products, School of Pharmacy, University of North Carolina Chapel Hill, NC Cheminformatics was born due to rapid accumulation of chemical compound related databases (containing the information on chemical structures and their measured or sometimes computed properties such as bioactivities) that should be organized, accessed, and explored. Thanks to the NIH's Roadmap Program a lot of such data has become publicly available via the Molecular Libraries Initiative (MLI) and the PubChem repository of biological assays of chemical compounds. I shall describe the application of multiple machine learning and datamining approaches (e.g., k Nearest Neighbor pattern recognition, Support Vector Machines, graph mining) to the development of Quantitative Structure-Activity Relationships (QSAR) models. Our approaches afford robust and validated models capable of accurate prediction of compound properties for molecules not included in the training sets. Such models are used to screen virtual databases of existing or synthetically feasible compounds and annotate these compounds with highly reliable predicted properties. The focus on knowledge discovery and property forecasting bring cheminformatics forward as predictive, as opposed to evaluative scientific discipline. Monday , October 1, 2007 4:30 pm Room 301, Research I, Fairfax Campus Refreshments will be served at 4:15 PM. ---------------------------------------------------------------------- Find the schedule at www.cmasc.gmu.edu/seminar/schedule.html --------------------------------------------------------------------