Clustering PPI network based on functional flow model through artificial bee colony algorithm

The protein-protein interaction (PPI) network clustering is one of the accesses to identify the functional modules and predict the functions of unknown proteins. At present the functional flow based model is a common method to solve this problem. However, the algorithm doesn´t work well on the PPI networks. The precision and recall is relatively low. This paper proposes a new functional flow clustering algorithm based on the optimizing searching ability of artificial bee colony. The principle of the improved algorithm initially adopts the distance and density based method to determine the cluster number and eliminate the noise spots so that the clustering result is able to escape from the disturbance of noise spots. Afterwards the method takes advantage of the feature of clustering coefficient to select the cluster centers. Simultaneously, we introduce the artificial bee colony algorithm into the searching procedures of functional flow based clustering. Finally, we testify the algorithm on PPI data and compare the simulated result with the other algorithms. It turns out that the algorithm can effectively improve the precision and recall of clustering results.