Screening and identification of potential long non-coding RNA for the prognosis evaluation of glioma patients based on bioinformatics analysis / 肿瘤研究与临床

ABSTRACT

Objective:To investigate the possibility of screened long non-coding RNA (lncRNA) as a prognostic marker in evaluating glioma.Methods:A total of 694 glioma samples and 5 tumor-adjacent tissues were selected in the Cancer Genome Atlas (TCGA) database from the establishment of the database to December 2018. The differential lncRNA, microRNA (miRNA) and mRNA between glioma tissues and 5 tumor-adjacent tissues were screened out, the correlation between the three and the prognosis of glioma patients was analyzed, and a competitive endogenous RNA (ceRNA) network was constructed. The biological functions of mRNA were enriched and analyzed based on Gene Ontology (GO) database and Kyoto encyclopedia of genes and genomes (KEGG) database. The survival analysis of patients with different expression levels of lncRNA, miRNA or mRNA were performed by using Kaplan-Meier method to obtain lncRNA, miRNA and mRNA related to prognosis. Univariate and multivariate Cox proportional hazards regression models were used to analyze the different lncRNAs in the ceRNA network, and an lncRNA prognosis model for predicting the 5-year overall survival rate of patients was constructed. According to the constructed model, the risk value of each sample in 694 samples in TCGA database was calculated. Taking the median risk value as the critical value, patients were divided into high-risk group (≥ median risk value) and low-risk group (< median risk value), and the survival curves of the two groups were drawn. The receiver operating characteristic (ROC) curve was drawn for predicting the 5-year overall survival rate of glioma patients in TCGA database according to the risk value of lncRNA prognosis model. The heat map of lncRNA gene expression levels in the prognostic model of patients in high-risk and low-risk groups was drawn with pheatmap R software package.Results:A total of differential 44 lncRNAs, 19 miRNAs and 54 mRNAs between glioma and tumor-adjacent tissues were obtained from TCGA database, and the ceRNA network map was drawn. Kaplan-Meier method showed that among them, 22 differential lncRNAs, 7 miRNAs and 38 mRNAs were related to the overall survival of patients. The univariate Cox regression analysis obtained 28 lncRNAs related to prognosis. After multivariate Cox regression analysis, Akaike information criterion (AIC) was used to seek the optimal prognostic risk model involving 16 lncRNAs, that was, risk value = ARHGAP31-AS1×(-0.357 7)+LY86-AS1×(0.155 1)+WARS2-IT1×(0.206 4)+PART1×(-0.110 0)+AC110491.1×(-0.142 6)+CACNA1C-IT2×(-0.381 3)+HAS2-AS1×(0.128 8)+AC092171.1×(-0.161 3)+CCDC26×(-0.144 2)+HCG15×(0.384 0)+AL359541.1×(-0.289 2)+GRM5-AS1×(0.120 5)+LINC00237×(-0.085 1)+LINC00310×(-0.213 0)+VCAN-AS1×(-0.090 3)+ LINC00303×(0.091 5). The median risk value was 0.758 calculated by the constructed model. The 5-year overall survival rate in the high-risk group was 16.8% (95% CI 11.4%-24.7%) and 75.7% (95% CI 68.5%-83.7%) in the low-risk group. The area under of ROC curve of 5-year overall survival predicted by lncRNA model was 0.893. Through the heat map, it can be found that the expression level of all lncRNAs in the model was different between high-risk and low-risk patients. Conclusions:The prognostic risk model constructed based on the screened lncRNAs can better evaluate the prognosis of glioma patients. These lncRNAs are expected to become new candidate biomarkers related to the prognosis of glioma.