Ppi Network Information Flow Model Of Swarm Intelligence Algorithm

With the continuous research of bioinformatics, the researchers find that the biological functions can be displayed via the networks generated by means of the interactions of a large number of genes. Therefore, adopting the novel idea of interaction, network, function to integrate the gene and analyze the genetic functions from the various aspects of proteins is a new orientation in the field of current genomic research. The interactions among all the proteins in an organism are named after the protein-protein interaction network. The main purpose of studying PPI networks is to identify and analyze the interactions of biological molecules in the cellular environment, further have a profound understanding of the mechanisms of interacting and executing functions among the biological molecules. In addition, it is also beneficial to predict the functions of unknown proteins. The swarm intelligence optimizing algorithms have been intensively studied and widely applied. Therefore this paper makes an attempt to make use of the swarm intelligence mechanisms to solve the problem of clustering PPI networks.The main context of study and innovational points of this paper are as follows:Firstly, with regard to the drawbacks of traditional clustering methods in predefining the cluster number and performing not well in the cluster results of PPI networks, this paper takes the small-world and scale-free characters of PPI networks into account to propose a novel artificial bee colony (ABC) and breadth first traversing based clustering approach which uses the distance and density to automatically determine the cluster number and eliminate the noisy points. This method initially determines the cluster number and abandons the noisy points according to the distance and density clustering algorithm, and then clusters the PPI networks in accordance with the proposed breadth first traversing clustering method. Afterwards acquires an optimal merging threshold through ABC algorithm to merge the cluster modules possessing the higher similarities and obtain the final cluster results. Compared with the traditional clustering methods, this algorithm gets evident improvements in the cluster effect.Secondly, owing to the facts that the functional flow algorithm ignores the interactive effect of distance and the merging threshold is manually set, this paper takes advantages of the different functions of three types of bees in the ABC algorithm to propose a novel functional flow algorithm under the basis of the optimizing searching of bee colony, which makes improvements on the whole clustering procedure of original functional flow algorithm. This method adopts the network comprehensive feature value of nodes to preprocess and sorts the nodes according to the descending order of network comprehensive feature value of nodes, and then determines the cluster centers in according to network comprehensive feature value of nodes. Afterwards the position of food source is regarded as cluster center, the benefits of food source is considered to be the similarity between cluster modules, all the adjacent nodes of exploited bee node are sorted in a descending order according to network comprehensive feature value of nodes, which are visualized as the searching neighborhood of scouts. And then considering the network comprehensive feature value of nodes as the initial functional flow of annotated proteins and completing the clustering procedure via traversing the neighboring nodes according to the descending order of network comprehensive feature values of nodes, finally automatically obtains the merging threshold by means of ABC algorithm. The improved functional flow algorithm is superior to the original functional flow algorithm in terms of precision, recall and f-measure value.Finally, one protein node may belong to two or more functional modules and several protein nodes are unreachable in PPI networks, therefore this paper utilizes the property of both this and that existing in the fuzzy set theory to advance a novel algorithm combining the optimizing mechanism of ABC algorithm and the matrix of fuzzy membership degree. This approach improves the object function and updates the cluster center via ABC optimizing algorithm. Inspired by the chemotactic operation of bacteria foraging behavior, this paper takes advantage of the following bee to choose the node which arises from the cluster of original cluster center and contains a large amount of information as the new cluster center. If the following bee updates in failure, then uses the scouts to make a global searching and updates the cluster center. This method makes use of ABC algorithm to automatically optimize cluster center in order to overcome the drawbacks of sensitivities of fuzzy C means and intuitionistic fuzzy clustering algorithms towards cluster center. The simulation results turns out that the algorithm performs steadily and works better in clustering PPI networks.