Five-long non-coding RNA risk score system for the effective prediction of gastric cancer patient survival

The prognosis for patients with gastric cancer (GC) is usually poor, as the majority of patients have reached the advanced stages of disease at the point of diagnosis. Therefore, revealing the mechanisms of GC is necessary for the identification of key biomarkers and the development of effective targeted therapies. The present study aimed to identify long non-coding RNAs (lncRNAs) prominently expressed in patients with GC. The GC dataset (including 384 GC samples) was downloaded from The Cancer Genome Atlas database as the training set. A number of other GC datasets were obtained from the Gene Expression Omnibus database as validation sets. Following data processing, lncRNAs were annotated, followed by co-expression module analysis to identify stable modules, using the weighted gene co-expression network analysis (WGCNA) package. Prognosis-associated lncRNAs were screened using the 'survival' package. Following the selection of the optimal lncRNA combinations using the 'penalized' package, risk score systems were constructed and assessed. Consensus differentially-expressed RNAs (DE-RNAs) were screened using the MetaDE package, and an lncRNA-mRNA network was constructed. Additionally, pathway enrichment analysis was conducted for the network nodes using gene set enrichment analysis (GSEA). A total of seven modules (blue, brown, green, grey, red, turquoise and yellow) were obtained following WGCNA analysis, among which the green and turquoise modules were stable and associated with the histological grade of GC. A total of 12 prognosis-associated lncRNAs were identified in the two modules. Combined with the optimal lncRNA combinations, risk score systems were constructed. The risk score system based on the green module [including ITPK1 antisense RNA 1 (ITPK1-AS1), KCNQ1 downstream neighbor (KCNQ1DN), long intergenic non-protein coding RNA 167 (LINC00167), LINC00173 and LINC00307] was the more efficient at predicting risk compared with those based on the turquoise, or the green + turquoise modules. A total of 1,105 consensus DE-RNAs were identified; GSEA revealed that LINC00167, LINC00173 and LINC00307 had the same association directions with 4 pathways and the 32 genes involved in those pathways. In conclusion, a risk score system based on the green module may be applied to predict the survival of patients with GC. Furthermore, ITPK1-AS1, KCNQ1DN, LINC00167, LINC00173 and LINC00307 may serve as biomarkers for GC pathogenesis.