GIPSCo: A Method for Comparison of Protein Structures Based on Geometric Invariants

Protein structure comparison is important for elucidation of evolutionary relationships, function and functionally important amino acid residues. We propose Geometric Invariant based Protein Structure Comparison (GIPSCo), an approach that compares a protein pair by using geometric invariants of local geometry of the backbone structures. The geometric invariants capture enough structural information so that the method has a relatively low false positive rate. The method first generates a list of aligned fragment pairs (AFPs) using the geometric invariants of the local geometry. These structurally similar AFPs are assembled using a graph theoretic approach to maximum weighted clique to obtain global structural alignment between two proteins. The proposed method is purely geometric in nature and is independent of sequence alignment and secondary structure assignment. Results for sample protein pairs were compared with well established methods of protein structure alignment. Further, our method could detect some structural similarities that were missed by other algorithms.