Reaserch On Function Module Detection From Large-scale And Dynamic Ppi Networks Based On Ant Colony Algorithm

Protein-Protein interaction networks(PPI) is a network of protein interactions in living organisms. Using computational methods for detecting functional modules is an important research issue in bioinformatics. With the improvement of technology to identify protein interactions, available data has become increasingly diverse. The scale of PPI networks is more and more large. Therefore, how to get functional modules more efficiently and robustly on large-scale PPI networks has become an important scientific problem in the era of big data. With further research, it is found that real PPI networks change over time and conditions. This is closely related to the creation and development of life activities. As a result, it is important to study the dynamic functional modules in dynamic PPI networks. In this paper, aiming at the detecting functional modules in large-scale PPI and dynamic PPI networks, we carry out the following two aspects of research:(1) Aiming at the weakness of the time performance when using ant colony optimization(ACO) algorithm to detect functional modules in large scale PPI networks, we propose a fast approach based on multiple grain representation and ant colony optimization. Firstly, a novel multiple grain representation model of PPI networks is proposed from the perspective of granular computing. Based on this model, a new algorithm which mainly contains three phases, i. e. a granularity partition process integrating functional and topological message, an ACO process on the coarse grain network, and a refinement and optimization process for solutions, is presented. On the basis of the model above, this paper also proposes a framework for detection of functional modules in PPI networks. The goal of the framework is to reduces the size of PPI network before algorithm with high time complexity applying on large-scale PPI network. Finally, we make some experiments both on the ant colony algorithm based on multiple grain model and the framework. The results of experiment demonstrate that the multiple grain model can not only significantly reduces the running time of functional module detection, but also effectively identifies modules while keeping some competitive performances.(2) Compared with static PPI network, protein interactions occur at the same time in dynamic PPI networks. Through the clustering analysis on dynamic PPI networks, it can get functional modules in real time. This has important implications for studying composition of modules and how modules change. To study this emerging research area, this paper presents a algorithm for detecting the dynamic functional modules based on temporal function continue feature and ant colony clustering. Firstly, based on the subnet structure in adjacent timestamps, the algorithm selects the set of seed which are active in adjacent timestamps. Secondly, based on the seed set by using timing evolution, we create an initial clustering set which has functional similarity with modules detected in last timestamp. Then the ant colony employs the picking and dropping operations to cluster the rest of proteins in dynamic PPI network. Finally, experiment results demonstrate that using temporal function continue feature can get more reliable and accurate functional modules. By comparison with other classical algorithms, ACC-DFM has a good performance at precision value.