Structural alignment using network properties

Abstract

Understanding the structural means of protein function via structural comparisons have wide range of applications such as protein fold classification, protein structure modelling and design. In this thesis, a novel structural alignment algorithm based on a amino acid network model is presented. The method we present models proteins as an amino acid network, derived from contact map representation of proteins. By using this model, we obtain fast tertiary structure comparisons, and combine them with primary and secondary structure comparisons to develop an overall similarity function. The similarity function drives a dynamic programming based alignment algorithm to obtain fast and accurate structural alignments. The structural alignments obtained are used to discover functional structural subunits called domains and to discover overall structural similarity of two proteins. We compared our domain prediction results with existing domain recognition methods and saw that our method correlates well with existing methods. Our global structural alignment results are compared with CE alignments.