Clustering PPI Data Based on Ant Colony Optimization Algorithm

Predicting function of unknown proteins in PPI (Protein-protein interaction) network is an important task of bioinformatics. The traditional clustering and functional flow algorithms performed not well in clustering PPI networks. Therefore this paper introduced the concepts of pheromone and state transition probability in the Ant colony optimization (ACO) algorithm to optimize the process of forming functional modules. The pheromone on the paths which the ants have passed by was updated via the accumulative strategy instead of constants in order to generate clusters as completely as possible. The experiments on MIPS dataset turned out that our approach was superior to the flow methods in terms of precision, recall and f-measure value, meantime reduced the time consumed.