A Five-miRNA Risk Score Model Predicts Prognosis In Patients With Gastric Cancer

Objective : To identify micro RNAs(miRNAs)associated with the prognosis of gastric cancer(GC)patients,construct and validate the miRNA risk score model through bioinformatics analysis.Methods:1.The miRNA 、 messenger RNA(mRNAs)expression profile and corresponding clinical data of the GC patients were obtained from the Cancer Genome Atlas(TCGA)database.Differentially expressed miRNAs and mRNAs were identified with “edge R” package based on the cut-off criteria of| log2 FC | > 1.5,P < 0.01 and | log2 FC | > 1,P < 0.05;and further all the patients with differentially expressed miRNAs’ profile and complete survival time,status was then randomly divided into train set and test set.2.The miRNAs related to prognosis were screened with Univariate Cox regression and LASSO regression,which were analyzed with multivariate Cox regression to construct a prognostic assessment model.3.In the train set、test set and overall dataset,Receiver operating characteristic(ROC)curve,Kaplan-Meier curve and dynamic area under the curve(AUC)were drawn to evaluate the effectiveness of the model.4.The mRNAs that miRNAs might bind to were analyzed by Gene Ontology(GO)enrichment analysis and Kyoto Encyclopedia of Genes and Genomes(KEGG)signaling pathway.5.Univariate Cox regression and multivariate Cox regression considering other clinical factors were displayed to identify the whther the miRNA signature serve as an independent prognostic factor.Results:1.175 differentially expressed miRNAs in GC tissues were screened based on the cut-off criteria of |log2FC| > 1.5 and P < 0.01,2219 differentially expressed mRNAs were screened based on the cut-off criteria of|log2FC| > 1.0 and P < 0.05;then all the patients(388)with differentially expressed miRNAs and complete survival time,status and stage were randomly divided into train group(194)and the test group(194);2.Six differentially expressed miRNAs with P < 0.05 were revealed in univariate Cox regression and LASSO regression analysis of the train set,then stepwise multivariate Cox regression was applied for constituting a five-miRNA signature.The risk score = hsa-mi R-184 * 0.183 + hsa-mi R-675 * 0.086 +hsa-mi R-2115 *(-0.231)+ hsa-mi R-3943 * 0.548+ hsa-mi R-1246 *(-1.455);3.The risk score of each patient was calculated using the signature and the median risk score of the training set was set as cut-off value to name the high-risk group and low-risk group.In the train set,test set and overall dataset,patients with higher risk score had poorer prognosis(P < 0.0001、P =0.002、P < 0.001),and the prognostic model demonstrated better predictive power than tumor-node-metastasis(TNM)stage in all the three sets(The3-year AUC: 0.748 > 0.592、 0.678 > 0.569、0.704 > 0.581),stratified analysis also demonstrated the predict ability in patients with each stage of GC and was independent of TNM-stage,the P-values of four stage of GC were 0.019、0.0051、0.03 and 0.05;4.GO and KEGG analysis showed miRNAs were involved in a few signaling pathways of gastric cancer;5.Univariate Cox regression and multivariate Cox regression considering other clinical factors displayed that high risk score was an independent risk factor for poor prognosis of GC patients.The area under the ROC curve of the signature was 0.681,better than age(0.616)、gender(0.491)、TNM-stage(0.619)、T(0.570)、M(0.534)、N(0.585).Conclusion: The 5-miRNA signature with high predictive performance could serve as an independent prognostic factor for GC patients.