Predicting 3D Structure of Protein-Protein Complexes
Aug 7, Tue 2007
11:00am, MRB 200
Dr. Petras Kundrotas
One of the most important tasks in the post-genomic era is developing of methods that can predict 3D structures of protein-protein complexes on genome-scale with accuracy sufficient to be used for computing meaningful biophysical quantities. Successful accomplishment of such a goal requires that both the overall structure of the complex and the interfacial contacts be predicted with reasonable quality. However, since the strongest interactions occur across the interface of the complex, the accurate modeling of the interface is crucial. In this talk I present the results of a homology approach that utilizes profile-to-profile alignment to detect templates and to generate the corresponding alignment. The distinctive feature of the method is that the interfacial residues (and their immediate surrounding) are given larger weights for the matching of such residues and the gap penalties for interfacial residues are also increased. Our profile-to-profile Homology Based Complex Prediction "HOMBACOP" method was benchmarked against the ProtCom database of protein-protein complexes and tested against Boston University benchmarking set.