Identifying biomarkers of diabetes with gene coexpression networks

Diabetes is one of the most prevalent diseases worldwide, and hundreds of millions of patients are suffering from diabetes and its serious complications. Early detection and early treatment are urgent needed for clinical diagnosis of diabetics. In this work, we establish a gene coexpression network framework to identify biomarkers of transcripts with highly different gene coexpression patterns in disease and control samples. We category the differentially expressed genes into four groups by their pair-wise gene coexpression dynamics, i.e., the gene pairs with (and/or without) high correlation in disease samples, while without (and/or with) high correlation in control samples. And we then identify the key genes significantly associated with the pathogenesis of diabetes mellitus. We evaluate the classification abilities of these gene groups by a support vector machine based method. The genes with good classification accuracy are further investigated in functional enrichment analysis, which are served as identified biomarkers.