Finding conserved regions in protein structures using support vector machines and structure alignment

Abstract

This paper proposes a novel method for finding conserved regions in three-dimensional protein structures. The method combines support vector machines (SVMs), feature selection and protein structure alignment. For that purpose, a new feature vector is developed based on structure alignment for fragments of protein backbone structures. The results of preliminary computational experiments suggest that the proposed method is useful to find common structural fragments in similar proteins. © 2012 Springer-Verlag.