Measuring Protein Structural Similarity by Maximum Common Edge Subgraphs

Abstract

It is known that the function of a protein is determined by its structure. Thus, structural similarity between proteins plays an important role as a good predictor of functional similarity. Many methods focus on solving the protein structure alignment problem. In this paper, we propose a graph-based approach to measure the similarity of two proteins. We first transfer a protein into a labeled graph according to its secondary structures, chemical properties, and 3D topology. For two graphs, we then find their maximum common edge subgraph for measuring the structural similarity of the corresponding proteins. By using a practical technique, the maximum common edge subgraph of two graphs can be found efficiently. Finally, by a backtracking, we can find the common substructure of the given proteins. Experimental results show that our method outperforms the RMSD method. This graph-based approach provides a practical direction for measuring protein structural similarity.