Fifteen-gene expression based model predicts the survival of clear cell renal cell carcinoma

Clear-cell renal cell carcinoma (ccRCC) is the major renal cell carcinoma subtype, but its postsurgical prognosis varies among individual patients. We used gene expression, machine learning (random forest variable hunting), and Cox regression analysis to develop a risk score model based on 15 genes to predict survival of patients with ccRCC in the The Cancer Genome Atlas dataset (N = 533). We validated this model in another cohort, and analyzed correlations between risk score and other clinical indicators. Patients in the high-risk group had significantly worse overall survival (OS) than did those in the low-risk group (P = 5.6e-16); recurrence-free survival showed a similar pattern. This result was reproducible in another dataset, E-MTAB-1980 (N = 101, P = .00029). We evaluated correlations between risk score and other clinical indicators. Risk was independent of age and sex, but was significantly associated with hemoglobin level, primary tumor size, and grade. Radiation therapy also had no effect on the prognostic value of the risk score. Cox multivariate regression showed risk score to be an important indicator for ccRCC prognosis. We plotted a nomogram for 3-year OS to facilitate use of risk score and other indicators. The risk score model based on expression of the 15 selected genes can predict survival of patients with ccRCC.