Development and Validation of an RNA-Seq-Based Prognostic Signature in Neuroblastoma

Objective: Stratification of neuroblastoma (NBL) prognosis remains difficult. RNA-based signatures might predict the prognosis, but independent cross-platform validation is still rare. Methods: RNA-seq-based profiles were acquired from NBL and then analyzed. RNA-Seq Prognostic Index (RPI) and clinical adjusted RPI (RCPI) were successively established in the training cohort (TARGET-NBL), and then verified in the validation cohort (GSE62564). Survival prediction was assessed using a time-dependent ROC curve and AUC. Functional enrichment analysis of genes was conducted using bioinformatics methods. Results: In the training cohort, 10 gene pairs eventually integrated into RPI. In both cohorts, a high-risk group had poor OS (all P < 0.001) and favorable EFS (P = 0.00032, P = 0.06, respectively). ROC curve analysis also showed that RPI predicting OS (60-month AUC values of 0.718 and 0.593, respectively) and EFS (60-month AUC values of 0.627 and 0.852, respectively) well in the training and validation cohorts. By univariate and multivariate regression analysis, clinicopathological indicators associated with prognosis were identified and added to RPI to form RCPI. RCPI also can divide population into different risk groups and a high-risk group had poor OS (all P < 0.001) and EFS (all P < 0.05). Finally, RCPI had higher accuracy for OS (60-month AUC values of 0.730 and 0.852, respectively) and EFS (60-month AUC values of 0.663 and 0.763, respectively) prediction in the training and validation cohorts. Moreover, these differentially expressed genes may be involved in certain NBL-related events. Conclusions: RCPI was performed and proposed to reliably categorize NBL into different death risk.