FAST Community Detection for Proteins Graph-Based Functional Classification

Abstract

In this paper we present and evaluate a fast and parallel method that addresses the problem of similarity assessment between node-labeled and edge-weighted graphs which represent the binding pockets of protein. In order to predict the functional family of proteins, graphs can be used to model binding pockets to depict their geometry and physiochemical composition without information loss. To facilitate the measure of similarity on graphs, community detection can be used. Our approach is based on a parallel implementation of community detection algorithm which is an adaptation and extension of Louvain method. Compared to the existing solutions, our method can achieve nearly well-balanced workload among processors and higher accuracy of graph clustering on real-world large graphs.