The goal of this research is to develop protein threading/structure refinement algorithms which, when combined with three dimensional functional site descriptors, predict both protein structure and biochemical I function. To be relevant in this post genomic era, such methods must be applicable on a proteomic scale and 3rovide additional information beyond evolutionary approaches. Since the most successful protein structure 3rediction approaches are comparative modeling/threading, key issues to be addressed are the completeness of the library of solved structures, the ability to refine modeled proteins and bring them closer to the native structure than even the best target-template structural alignment and the ability to predict loop conformations. All predictions must be at the requisite resolution to be able to accurately identify active site and binding regions in the protein. To address these issues and build upon previous accomplishments, we propose the following Specific Aims: (1). The single chain threading algorithm, PROSPECTOR_3, will be further improved to identify templates with low sequence identity to protein targets by constructing template dependent mutation matrices using analogous templates and extended to include DNA and other prosthetic groups as part of the template. (2). Threading based alignment refinement algorithms will be developed to improve upon initial threading alignments for cases where the global fold is correct, but the alignment is not. (3). Threading will be improved to better treat multiple domain proteins by constructing a representative domain-domain template library and residue-based, domain interface pair potentials. Given the predicted structure of the domains, the interfacial potential will select the orientation if it is found in the domain-domain library or predict that none of these are adopted. (4). Multimeric threading will be improved by developing more sensitive protein-protein interfacial potentials and criteria to identify high confidence predictions. The approach will be extended from dimers to higher order multimers. (5). The threading/assembly/refinement algorithm, TASSER, will be improved and extended to multimers. A less CPU intensive version suitable for web use will be developed. (6). The specificity of the active site template library will be improved and extended to include ligand binding sites. (7). The algorithms developed in Specific Aims 1-6 will be applied to over 100 proteomes including human. We will model human GPCRs, kinases, proteases and phosphatases.
Showing the most recent 10 out of 114 publications
