An Improved Ant Colony Optimization Algorithm for Clustering Proteins in Protein Interaction Network

-Recently, we have introduced the first version of ACOPIN, an Ant Colony Optimization (ACO) algorithm for clustering proteins in the protein interaction network (PIN). Based on the experimental results, we have discovered serious pitfalls; very difficult to define the boundaries among clusters and unable to identify the overlapping clusters. These pitfalls contribute to low accuracy of detected cluster especially for the larger size. In this paper, we improve ACOPIN to overcome these pitfalls. The improvement includes the construction tour of ant based on topological weight and functional similarity weight. Artificial ants themselves perform the ‘exploration’ and ‘exploitation’ process in the PIN with the assistance from pheromone trails. We simulate the improved version onto one Yeast datasets and the result shows that the detected clusters has very high accuracy in terms of functional homogeneity of proteins.