Sampling Technique and Scoring Function for Protein Loop Structure Refinement

Abstract

In the absence of experimentally determined protein structures, comparative modeling can facilitate structure-related researches by providing high-accuracy model structures. However, comparative modeling methods usually show limitations in local regions with sequential diversities, so protein structure refinement, or loop modeling, is required for those regions. In this thesis, two different modeling approaches are covered to enhance the performance of loop structure refinement. First, GalaxyGPCRloop, specialized in modeling of the second extracellular loop of G-protein-coupled receptors, is described. It was developed to efficiently find near-native conformations by various sampling techniques such as utilization of database information, prediction of disulfide bridge, and sampling of hairpin structure. Second, a new scoring function that implicitly considers side chain ensemble is described. The scoring function enables initial discrimination of near-native conformations from non-native ones. With the sampling technique and scoring function, several loop modeling tests show potential of performance improvement. Further development of novel algorithms integrating the sampling technique with the scoring function would be a next task for solving more challenging loop structure refinement problems.