Protein Complexes Identification Based On The Best Neighbor Node Measure

In the post genome era, one of the most important challenge is to systematically analyze and understand how life activities be done through interactions among proteins in vivo. Through analysis of structure characteristic of protein interaction networks, identifying protein complexes or function modules and predicting unknown protein function are becoming the focus of current proteomics research.Based on an in-depth analysis of structure characteristics of protein interaction networks, according to their common structural properties, there are some excellent methods put forward for identifying protein complexes, the main original works include:In the most exist methods, the best neighbor node of a module is usually defined as the most close node with all nodes within the module. In this paper, the best neighbor node of a node or module is redefined, calculating the number of common neighbors between nodes as an important factor of the best neighbor node measure, the best neighbor node of a module is defined as the best neighbor node of the node within the module which satisfies certain conditions.A hybrid modular metric LGQ which takes global and local feature of the entire network into consideration is proposed to overcome the limitation of the global modular measure Q which cannot effectively identify smaller modules, and consider the global features of the entire network which is ignored by the local modular measure LQ. Next, a best neighbor node mining method BN-LGQ based on the measure LGQ is put forward for detecting protein complexes from the protein interaction networks. The new method does not need other auxiliary information, which can be implemented with less time and operations. The BN-LGQ method is applied to the protein interaction networks. The experimental results show that our method can identify more match known protein complexes and more protein complexes of biological significance.Through the research of the formation law of network community in complex social networks, a best neighbor node mining method BN-MNE(Best Neighbour-Multistage Nuclear Extension) based on multistage nuclear extension is proposed to mining the significant module from complex networks. Firstly, we apply our BN-MNE method to community structure identification of several typical complex social networks and find that it can be better to identify the community structure of complex networks than other algorithms. Secondly, we use the yeast protein interaction networks to verify the method BN-MNE. Our method identify more protein complexes of biological significance. The result has good consistency with the previous result found in social complex networks. Meanwhile, our research can provide valuable reference information for complex identification of other protein interaction networks based on complex network structure.