Protein complex detection based on integrated strategy

Protein complexes are of great importance in understanding the principles of cellular organization and function. The increase in available protein-protein interaction (PPI) data, subcellular information and other biological resources makes it possible to design computational methods for protein complex prediction. Most existing methods mainly focus on the topological structure of protein- protein interaction networks and fail to consider the influence of subcellular information. In this article, firstly, we combine non-redundant protein-protein interaction data from three different experiment sources with subcellular information and two matrices are constructed, which can effectively take into account the topological properties and space information. Then, an integrated strategy is designed to locate candidate clusters and merge these clusters based on the localization matrix. To validate our method quantitatively, two different real complex datasets is used. The extensive experimental results showed that (i) our integrated strategy can improve the performances of the four original methods and is robust to different thresholds. (ii) detailed comparison with functional annotations illustrates and certifies the efficiency of the spatial information and this strategy indicates to be helpful to find functional modules. Our method introduces new insight of topologicalspace information for protein complex detection and can also provide some useful reference for identification community structure in social network.