Gene-Rave analysis of microarray data identifies molecular classifiers for 5 classes of parathyroid tumors. Free

3257

Gene-Rave is a powerful statistical tool for analyzing microarray data. Based on Bayesian methods, pair-wise discrimination of classes using Gene-Rave results in typically very small sets of discriminating genes. We have used Gene-Rave to identify genes able to differentiate between 5 clinico-pathological subtypes of hyperparathyroidism. Histopathological classification of these subtypes is inherently difficult and molecular events causing these lesions are poorly understood. cDNA microarray experiments were performed on 53 parathyroid tumors and a pool of normal parathyroid tissue*. All were hybridized against the Universal Human Reference RNA (Stratagene), and the expression results normalized using loess fitting and spatial smoothing. Gene-Rave analysis of 5 subtypes or classes of tumors (sporadic: carcinoma, adenoma and hyperplasia; and familial: Multiple Endocrine Neoplasia Type 1 (MEN 1) and Hyperparathyroidism-Jaw Tumor syndrome (HPT-JT)) was then performed. Cross validation and permutation testing, used to assess the prediction error rate, found the HPT-JT and sporadic carcinoma classes to be inseparable and these were thus combined into 1 class. The subsequent 4 classes could then be differentiated, with the HPT-JT / carcinoma class well separated from the other 3 classes, and the adenoma and MEN 1 classes close together. Nine genes, UCHL1, CD24, RALDH2, PVALB, clone DKFZp434E03, VCAM1, SPOCK3, MOX2 and GAD1, were able to discriminate between the classes. Further assessment of the discriminating power of a selection of these genes, in addition to the CaSR and HRPT2, was performed. The relative expression levels in 30 parathyroid samples including adenoma, hyperplasia, carcinoma, HPT-JT tumors and a pool of normal tissue, were assessed by real-time RT-PCR. The genes identified in this study by Gene-Rave will be useful as molecular classifiers for hyperparathyroidism, and will also assist in understanding the molecular events involved in the formation of these lesions. *All microarray experiments were performed at the VARI by CJH and VMH (Haven et al, PNAS, submitted). The VARI microarray core facility is acknowledged for the fabrication of the microarray slides.