PFC: An Efficient Soft Graph Clustering Method for PPI Networks Based on Purifying and Filtering the Coupling Matrix

Abstract

One of the most pressing problems of the post genomic era is identifying protein functions. Clustering Protein-Protein-Interaction networks is a systems biological approach to this problem. Traditional Graph Clustering Methods are crisp, and allow only membership of each node in at most one cluster. However, most real world networks contain overlapping clusters. Recently the need for scalable, accurate and efficient overlapping graph clustering methods has been recognized and various soft (overlapping) graph clustering methods have been proposed. In this paper, an efficient, novel, and fast overlapping clustering method is proposed based on purifying and filtering the coupling matrix (PFC). PFC is tested on PPI networks. The experimental results show that PFC method outperforms many existing methods by a few orders of magnitude in terms of average statistical (hypergeometrical) confidence regarding biological enrichment of the identified clusters.