Identification of gene interactions associated with disease from gene expression data using synergy networks

Analysis of microarray data has been used for the inference of gene-gene interactions. If, however, the aim is the discovery of disease-related biological mechanisms, then the criterion for defining such interactions must be specifically linked to disease.

Here we present a computational methodology that jointly analyzes two sets of microarray data, one in the presence and one in the absence of a disease, identifying gene pairs whose correlation with disease is due to cooperative, rather than independent, contributions of genes, using the recently developed information theoretic measure of synergy. High levels of synergy in gene pairs indicates possible membership of the two genes in a shared pathway and leads to a graphical representation of inferred gene-gene interactions associated with disease, in the form of a "synergy network." We apply this technique on a set of publicly available prostate cancer expression data and successfully validate our results, confirming that they cannot be due to pure chance and providing a biological explanation for gene pairs with exceptionally high synergy.

Thus, synergy networks provide a computational methodology helpful for deriving "disease interactomes" from biological data. When coupled with additional biological knowledge, they can also be helpful for deciphering biological mechanisms responsible for disease.