Identifying functional modules in tissue specific protein interaction network

Abstract

Identifying functional modules in protein interaction network is an important problem for computational biologists. As well known, protein interaction is affected by environment, cell cycle and many other factors, but most existing methods only focus on how to identify modules from static protein interaction network. In this paper, we proposed a method to identify protein functional modules from tissue specific protein interaction networks. Tissue specific protein networks of human is built by combining the tissue specific gene expression data and static protein networks. Meanwhile, a virtual tissue specific protein interaction network is also constructed. Then, CFinder clustering algorithm is applied to original PPI network, the virtual protein interaction network and each tissue specific network to identify functional modules. Finally, topological analysis and GO enrichment analysis are performed to compare the functional modules identified from above three kinds of PPI networks. Analysis results show that the functional modules identified from the tissue specific PPI networks we built have more specific biological meaning than those from the original protein interaction network and virtual tissue specific network. Some huge functional modules identified from the original PPI network are split into some comparable smaller, but more specific biological meaningful functional modules in the results from tissue specific networks. These smaller modules have more specific biological function. In sum, our study shows that tissue specificity of protein interactions and function modules is an important factor that must be seriously considered in the study of network-based identification of protein functional modules.