A Genomic Algorithm for the Molecular Classification of Common Renal Cortical Neoplasms: Development and Validation

Abstract

Accurate discrimination of benign oncocytoma and malignant renal cell carcinoma is useful for planning appropriate treatment strategies for patients with renal masses. Classification of renal neoplasms solely based on histopathology can be challenging, especially the distinction between chromophobe renal cell carcinoma and oncocytoma. In this study we develop and validate an algorithm based on genomic alterations for the classification of common renal neoplasms. Using TCGA renal cell carcinoma copy number profiles and the published literature, a classification algorithm was developed and scoring criteria were established for the presence of each genomic marker. As validation, 191 surgically resected formalin fixed paraffin embedded renal neoplasms were blindly submitted to targeted array comparative genomic hybridization and classified according to the algorithm. CCND1 rearrangement was assessed by fluorescence in situ hybridization. The optimal classification algorithm comprised 15 genomic markers, and involved loss of VHL, 3p21 and 8p, and chromosomes 1, 2, 6, 10 and 17, and gain of 5qter, 16p, 17q and 20q, and chromosomes 3, 7 and 12. On histological rereview (leading to the exclusion of 3 specimens) and using histology as the gold standard, 58 of 62 (93%) clear cell, 51 of 56 (91%) papillary and 33 of 34 (97%) chromophobe renal cell carcinomas were classified correctly. Of the 36 oncocytoma specimens 33 were classified as oncocytoma (17 by array comparative genomic hybridization and 10 by array comparative genomic hybridization plus fluorescence in situ hybridization) or benign (6). Overall 93% diagnostic sensitivity and 97% specificity were achieved. In a clinical diagnostic setting the implementation of genome based molecular classification could serve as an ancillary assay to assist in the histological classification of common renal neoplasms.