Identification of Functional Modules and Gene Networks

In addition to classification and feature selection tasks, we are interested in the identification of interactive networks between genes, which determine the concerted action of different mRNAs or proteins, and eventually the status of the cell. Genes do not generally act independently, but their action has to be considered inside a framework of interactive modules, that form a complex network. The identification of those modules and the comprehension of the affinities between modules is a priority for the study of every disease, or generally for a better understanding of cellular functions.

A module is defined as a set of genes that perform a certain function. Therefore, the expression profile of each gene-member in the module must be regulated in a way that serves the functionality of the module. A reasonable assumption is that expression profiles of gene-members in a module must be interrelated. In reality many functional modules in a cell have genes that are co-regulated. This is achieved by using common transcriptional factors for all or most of the genes in a module. This is, for example, the case for many metabolic pathways, as well as for genes whose products form a multiprotein complex.

Our goal is to detect the kind of interrelations that govern each module, as well the rules for between modules interactions. Towards this target we currently use gene expression data in conjuction with several statistical approaches, such as linear and higher-order correlation functions, since the distributions of gene expression data cannot always be modelled by the normal distribution. Future work in this area includes the combination of gene expression data, sequence information and protein-to-protein interaction data to guide the search for gene modules and their charecterization.