Abstract
BACKGROUND: Tumor microenvironment (TME) contributes to tumor development, progression, and treatment response. In this study, we detailed the cell composition of the TME in neuroblastoma (NB) and constructed a cell risk score model to predict the prognosis of NB. METHODS: xCell score was calculated through transcriptomic data from the datasets GSE49711 and GSE45480 based on the xCell algorithm. The random forest method was employed to select important features and the coefficient was obtained via multivariate cox regression analysis to construct a prognostic model, and the performance was validated in another two independent datasets, GSE16476 and TARGET-NBL. RESULTS: We found that both immune and non-immune cells varies significantly in different prognostic groups, and were correlated with survival time. The proposed prognostic cell risk score (pCRS) model we constructed can be an independent prognostic indicator for overall survival (OS) and event-free survival (EFS) (training: OS, HR 1.579, EFS, HR 1.563; validation: OS, HR 1.665, 3.848, EFS, HR 2.203, all p-values < 0.01) and only independent prognostic factor in International Neuroblastoma Risk Group high risk patients (HR 1.339, 3.631; p-value 1.76e-2, 3.71e-5), rather than MYCN amplification. Besides, pCRS model showed good performance in grouping, in discriminating MYCN status, the area under the curve (AUC) was 0.889, 0.933, and 0.861 in GSE49711, GSE45480, and GSE16476, respectively. In separating high risk groups, the AUC was 0.904 in GSE49711. CONCLUSION: This study details the cellular components in the TME of NB through gene expression data, the proposed pCRS model might provide a basis for treatment selection of high risk patients or targeting cellular components of TME in NB.