A 4-gene panel predicting the survival of patients with glioblastoma.

ABSTRACT

BACKGROUND: To identify independently prognostic gene panel in patients with glioblastoma (GBM). MATERIALS AND METHODS: The Cancer Genome Atlas (TCGA)-GBM was used as a training set and a test set. GSE13041 was used as a validation set. Survival associated differentially expression genes (DEGs), derived between GBM and normal brain tissue, was obtained using univariate Cox proportional hazards regression model and then was included in a least absolute shrinkage and selection operator penalized Cox proportional hazards regression model. Thus, a 4-gene prognostic panel was developed based on the risk score for each patient in that model. The prognostic role of the 4-gene panel was validated using univariate and multivariable Cox proportional hazards regression model. RESULTS: A total of 686 patients with GBM were included in our study; 724 DEGs was identified, 133 of which was significantly correlated with the overall survival (OS) of patients with GBM. A 4-gene panel including NMB, RTN1, GPC5, and epithelial membrane protein 3 (EMP3) was developed. Kaplan-Meier survival analysis suggested that patients in the 4-gene panel low risk group had significantly better OS than those in the 4-gene panel high risk group in the training set (hazard ratio [HR] = 0.3826; 95% confidence interval [CI] 0.2751-0.532; P < 0.0001), test set (HR = 0.718; 95% CI 0.5282-0.9759; P = 0.033) and the independent validation set (HR = 0.6898; 95% CI 0.4872-0.9766; P = 0.035). Both univariate and multivariable Cox proportional hazards regression analysis suggested that the 4-gene panel was independent prognostic factor for GBM in the training set. CONCLUSION: We developed and validated 4-gene panel that was independently correlated with the survival of patients with GBM.