Descriptor based protein structure representation using triangular spatial relationships in 3-D

Abstract

Pairwise protein structure comparison has taken significant scientific research effort in last two decades. Even though it all started with alignment-based comparison methods, recently there are several non-alignment based methods that have shown good potential. One such approach is based on shape descriptors. These methods use histograms or vectors to represent the molecular shapes. They have shown to improve comparison speed but require reference frame transformations, are limited to sequential comparisons as well as lack residue information. Like any sequence independent model, these methods ignore the correspondence of residues in similarity calculation. This work proposes protein structure representation method, Triangular Spatial Relationships in 3D (TSR 3-D). TSR 3-D is local-scale sensitive, reference frame insensitive, protein structure descriptor that incorporates the residue information. It can be used to establish a strict pairwise equivalence that acknowledges the pivotal role played by corresponding residues in determining protein 3-D structure. Protein structures represented by TSR-3D can be used for flexible non-sequential, pairwise structure comparison using local-global equivalences.