Most complex diseases can be summarized by a small number of “disease processes.” Identifying those processes can lead to improved understanding and therapeutic options. We have devised a new procedure for disease process discovery using diabetic nephropathy (DN) as the example. Our method consists of three steps: 1) an effect size is calculated for the disease of interest and all other publicly‐available Affymetrix gene array experimental conditions (using a condor grid system), 2) a novel distance statistic is calculated for each gene to identify genes that are similarly affected in the experimental condition of interest and DN 3)the counts of the gene ontology (GO) terms associated with these similarly affected genes serve as the basis for class discovery (using positioning around medioids) followed by variable selection using a Wilcox test ranking procedure. We divide DN into 3 different disease processes (cytoskeltal/morphological changes, inflammation, and response to hyperglycemia). Important GO terms in distinguishing between these 3 factors included actin filament bundle assembly, defense response, and response to nutrient. Since these disease processes are well known components of the pathology of diabetic nephropathy, this suggests we have developed a powerful method for identifying the major components of complex disease.