Clustering in a fixed manifold to detect groups of genes with similar expression patterns

Clustering genes into groups that exhibit similar expression patterns is one of the most fundamental issues in microarray data analysis. In this paper, we present a normalized Expectation-Maximization (EM) approach for the problem of gene-based clustering. The normalized EMclustering also follows the framework of generative clustering models but for the data in a fixed manifold. We illustrate the effectiveness of the normalized EM on two real microarray data sets by comparing its clustering results with the ones produced by other related clustering algorithms. It is shown that the normalized EM performs better than the related algorithms in term of clustering outcomes. © Springer-Verlag Berlin Heidelberg 2008