Parallel Markov Clustering for Large-scale Protein-Protein Interaction Networks using MPI

Empirical study of networks has enlightened our understanding of many application domains including the topology of biological systems. The Markov clustering algorithm (MCL), originally developed for clustering graphs, has been adopted to clustering protein-protein interaction (PPI) networks. MCL is also becoming an effective algorithm for the extraction of complexes from interaction networks. However, increasingly large amounts of data and massive networks will lead to a sparse and very complex network structure. It is impossible to perform computations on networks with millions of nodes without specialized computer facilities. Here we introduce parallel implementation approaches to improve the performance of MCL, and to allow the analysis of very large PPI-network datasets.