Characterizing cell types through differentially expressed gene clusters using a model-based approach

Understanding the regulatory basis of a cell is a central question in molecular biology. Although all cells in an organism have the same genetic material different genes are expressed. Measuring the gene expression with DNA microarrays can reveal the cell specific pattern of expression. Various bioinformatics methods attempt to cluster these expression patterns in order to find groups of genes that have similar expression over numerous samples. These groups are thought to comprise genes that have similar function or that build a module in the regulatory network of the cell which is switched on or off in response to the environment. Moreover, one tries to find marker genes that are sufficient to discriminate the cell type from others and characterize the phenotype of the cell. However, analyzing the sets of marker genes to find over-represented functional annotations is difficult as the marker genes are only those genes that are most differently expressed among the cell types and most functionally related genes are not present in the marker gene set. In this book, we propose a method based on model-based clustering to detect modules, i. e. clusters of genes, that distinguish one cell type from another. Subsequently, marker modules that are most important in classifying the cells are determined. We show at the example of Acute Lymphoblastic Leukemia that these modules are coherent in their expression across samples and that they give more biological insight into the different cell types than using just marker genes.