Prediction of prognosis of gastric cancer by a five-microRNA risk score model / 中华消化杂志

ABSTRACT

Objective:To analyze and screen microRNA (miRNA) related to the prognosis of gastric cancer(GC) by bioinformatics analysis, and to construct and validate a risk score model.Methods:The human genome miRNA sequencing data and corresponding clinicopathological data of the 491 samples (446 GC tissue samples and 45 normal gastric tissue samples) were downloaded from the cancer genome atlas (TCGA) database. The differentially expressed microRNA (DEM) was analyzed with edgeR package of R 4.0.2 software and the obtained DEM’s profile was randomly divided into training set and test set according to the ratio of 1∶1. The miRNA related to prognosis were analyzed and screened with univariate Cox regression and least absolute shrinkage and selection operator (LASSO) regression, and multivariate Cox regression analysis was further performed to analyze the screened prognostic-related miRNA and then the prognostic risk score model was constructed. Kaplan-Meier curve, receiver operating characteristic curve (ROC), and dynamic area under the ROC were drawn to evaluate the predictive power of the model.Results:A total of 175 DEM in GC tissues were screened out based on the cut-off criteria of |log2 Fold Change|>1.5 and P<0.01. Six DEMs related to the overall survival rate of patients with GC were screened out by univariate Cox regression and LASSO regression analysis, and then a five-miRNA risk score model was successfully constructed by multivariate Cox regression. The risk score=0.183×hsa-miRNA-184+ 0.086×hsa-miRNA-675-0.231×hsa-miRNA-2115+ 0.548×hsa-miRNA-3943-1.455×hsa-miRNA-1246. In the training set, test set and overall data set, the cumulative survival rates of the patients with higher risk score were lower than those of the patients with lower risk score, respectively, and the differences were statistically significant ( χ2=18.90, 9.50 and 26.70, all P<0.05). The prediction power of the model was better than that of TNM stage. And the results of stratified analysis showed the predictive ability of the model in patients with early GC. The results of univariate Cox regression and multivariate Cox regression demonstrated that the risk score of the model, gae and M stage were independent risk factors for poor prognosis in patients with GC (hazard ratio(95% confidence interval)1.19(1.07 to 1.32), 1.20(1.06 to 1.40), 1.50(1.01 to 2.23), 1.90(1.28 to 2.90), 1.34(1.15 to 1.57), 2.10(1.05 to 4.40); all P<0.05). Conclusion:The 5-miRNA risk score model based on 5 miRNAs which was an independent prognostic factor had high accuracy in predicting the prognosis of patients with GC.