Constrained Competitive Learning Algorithm for DNA Microarray Gene Expression Data Analysis

Cluster analysis is an important tool for discovering the structures and patterns hidden in gene expression data. In this paper, a new algorithm for clustering gene expression profiles is proposed. In this method, we find natural clusters in the data based on a competitive learning strategy. Using partially known modes as constraints in our method, we reduce the sensitivity of the clustering procedure to the algorithm initialization and produce more reliable results. Also the proposed algorithm can give the correct estimation of the number of clusters in the data. Experiments on simulated and real gene expression data demonstrate the robustness of our method. Comparative studies with several other clustering algorithms illustrated the effectiveness of our method.