Lumping Protein Complexes to Reduce the Complexity of Human Protein-Protein Interaction Network

Abstract

Aim: For network biology applications in health sciences, most of the time it is impossible tohandle the whole protein-protein interaction map at once, and accordingly there have beenvarious approaches to network reduction for computational efficiency. In this study we aimed touse protein complexes as a base for network reduction, proposing a node-lumping procedure.Materials and Methods: The comprehensive resource of mammalian protein complexes wasused to extract protein complex interactions. Human protein-protein interaction map was retrieved from the Agile Protein Interactomes Data (APID) server. A novel lumping procedure wasintroduced to create a reduced network. Topological analysis of the resulting context-specificnetwork and examination of the highly connected nodes were compared with the original network.Results: After lumping we obtained a heterogeneous map of 9,888 proteins and 304 lumpednodes with 41,940 interactions. Total number of nodes and interactions were reduced by 9.7%and 16%, respectively. The resulting network preserves the scale-free topology.Discussion and Conclusion: The results indicated that the procedure was helpful in networkreduction without disturbing the biologically relevant structure of the network.