Protein domain prediction for protein tertiary structure prediction

Abstract

Protein structure prediction results provide important biological information such as structural, functional analysis of proteins and protein design. Protein structure prediction methods can be divided into two categories. First, ab initio modeling methods predict structure based on physical principles. Second, Template-based modeling methods predict structures by selecting template structures from the database of known experimental structures. This can lead to more accurate predictions. Protein domain information further improves the predictive power of template-based modeling. This is because protein domain information improves the fold recognition step, one of the main parts of template-based modeling, which in turn determines the quality of final predicted protein structure. However, they do not account for the fact that some multi-domain proteins have related proteins with experimental structures of multiple domains resolved together in the structure database. The goal of GalaxyDom is to assign modeling units such that the maximal information can be extracted from the structure database for structure prediction of each unit, allowing prediction of domain-domain interactions as well as individual domain structures when reliable templates are available in the database. Our method uses HHsearch as the fold recognition tool. We use CASP domain definition in order to evaluate the true protein domain boundary. The CASP domain definition is determined by the CASP assessor and the true domain boundary is ± 10 residues. Also, the protein domain predictions of GalaxyDom find better templates than other domain prediction tools results. The better the method is at predicting reliable modeling unit regions, the better the ability of template-based modeling to find better template proteins.