Assistant Professor Hui Lu
Department of Bioengineering
University of Illinois at Chicago
Chicago, IL

Wednesday, April 21, 2004
1:00 am (CT)
3269 Beckman Institute

Abstract

Many proteins perform their functions through binding to other proteins and nucleic acids. We are developing three approaches to predict the binding sites and partners of proteins. First, a structure-based modeling framework is being designed for genome scale binding prediction. Two applications within the current framework will be presented: whole genome protein-protein interaction predictions and transcription factor binding site predictions based on protein-DNA interactions. Specifically, two components are needed for the current approach to be useful on a genome scale: a successful fold recognition protocol and good binding potentials. We are currently developing, for this purpose, statistical potentials for various types of binding based on experimentally solved structures of protein complexes, as well as, a new approach to fold recognition based on kernel-based machine learning. Second, we performed a dynamics-based analysis of protein complexes, which has shown a significant correlation between the protein binding sites and their dynamic vibrations. This correlation can be used in selecting the native structure complexes from docking generated decoys. Third, we systematically analyzed the relation between protein-protein interactions and gene expression. We have found that the gene expression correlation among the interacting proteins is different for different species. Moreover, within one genome, the correlation increases with the degrees of interaction among interacting protein complexes.