Message passing clustering with stochastic merging based on kernel functions

In this paper, we propose a new Stochastic Message Passing Clustering (SMPC) algorithm for clustering biological data based on the Message Passing Clustering (MPC) algorithm, which we introduced in earlier work. MPC has shown its advantage when applied to describing parallel and spontaneous biological processes. SMPC, as a generalized version of MPC, extends the clustering algorithm from a deterministic process to a stochastic process, adding three major advantages. First, in deciding the merging cluster pair, the influences of all clusters are quantified by probabilities, estimated by kernel functions based on their relative distances. Second, the proposed algorithm property resolve the "tie" problem, which often occurs for integer distances as in the case of protein interaction data. Third, clustering can be undone to improve the clustering performance when the algorithm detects objects which don't have good probabilities inside the cluster and moves them outside. The test results on colon cancer gene-expression data show that SMPC performs better than the deterministic MPC.