Construction and verification of tumor microenvironment-related gene prognostic model for adrenocortical carcinoma / 肿瘤研究与临床

ABSTRACT

Objective:Bioinformatics method was used to screen out prognostic model constructed by the tumor microenvironment (TME)- related genes of adrenocortical carcinoma (ACC), and the prognostic model was verified to provide clinical guidances and related biomarkers for the diagnosis and treatment of ACC.Methods:Transcriptome and clinicopathological data of 79 ACC patients were collected from the Cancer Genome Atlas (TCGA) database. ESTIMATE algorithm was used to calculate immune score, stromal score (both reflect TME) and ESTIMATE score; VennDiagram was used to select differentially expressed genes among immune score, high and low stromal score groups (grouped by median value); Gene Ontology (GO) database and Kyoto Encyclopedia of Genes and Genome (KEGG) database were used to perform functional enrichment analysis on selected genes and to explore the potential function and pathway of genes. Univariate Cox analysis, lasso regression analysis and multivariate Cox analysis were used to screen out genes related to ACC TME and to establish risk score (RS) model for ACC patients. The receiver operating characteristic (ROC) curve was used to evaluate the prognostic value of RS. The data sets GSE33371 and GSE19750 of Gene Expression Omnibus (GEO) were used as external validation sets to validate the prognostic model. The data of 79 ACC patients were extracted from the TCGA database, and the clinicopathological factors and the RS of the established prognostic model were included in the Cox regression analysis to obtain the prognostic factors of ACC patients.Results:According to the immune score and stromal score, 1 205 differentially expressed genes from intersection of both scores were screened out by using VennDiagram, including 833 up-regulated genes and 372 down-regulated genes. After continuing the regression analysis and screening of differentially expressed genes, the ACC prognostic model containing 9 TME-related genes (GREB1, POU4F1, HIC1, HOXC9, CACNB2, RAB27B, ZIC2, C3, CYP2D6) was finally constructed, that was, RS = GREB1×0.223 6+POU4F1×0.671 7+HIC1×0.167 5+HOXC9×0.211 3+CACNB2×0.156 0+RAB27B×0.956 5+ZIC2×0.582 7+C3×(-0.003 1)+CYP2D6×0.819 3. The area under the curve (AUC) of ROC for the 1, 3, and 5-year overall survival of 79 ACC patients predicted by the model in the TCGA database was 0.876, 0.919, 0.917, respectively. In the GEO validation set, the AUC of the 1, 3, and 5-year overall survival for 45 ACC patients predicted by the model was 0.689, 0.704, and 0.708, respectively, indicating that the model had a high prediction accuracy for survival results of ACC patients. Cox regression analysis on the data of 79 ACC patients in the TCGA database showed that the TME-related gene prognostic model RS was an independent factor influencing the prognosis of ACC patients ( HR = 1.011, 95% CI 1.005-1.016, P < 0.01). Conclusions:The established ACC TME-related gene prognostic model can be used to predict the prognosis of ACC patients. The model including 9 genes may become a new target for studying the pathogenesis and immunotherapy of ACC, and it is worthy of further research.