Construction of an RNA modification-related gene predictive model associated with prognosis and immunity in gastric cancer

Abstract

Abstract Background Gastric cancer (GC) is one of the most common causes of cancer-related fatalities worldwide, and its progression is associated with RNA modifications. Here, using RNA modification-related genes (RNAMRGs), we aimed to construct a prognostic model for patients with GC. Methods Based on RNAMRGs, RNA modification scores (RNAMSs) were obtained for GC samples from The Cancer Genome Atlas and were divided into high- and low-RNAMS groups. Differential analysis and weighted correlation network analysis were performed for the differential expressed genes (DEGs) to obtain the key genes. Next, univariate Cox regression, least absolute shrinkage and selection operator, and multivariate Cox regression analyses were performed to obtain the model. According to the model risk score, samples were divided into high- and low-risk groups. Enrichment analysis and immunoassays were performed for the DEGs in these groups. Four external datasets from Gene Expression Omnibus data base were used to test the accuracy of the predictive model. Results We identified SELP and CST2 as key DEGs, which were used to generate the predictive model. The high-risk group had a worse prognosis compared to the low-risk group (p < 0.05). Enrichment analysis and immunoassays revealed that 144 DEGs related to immune cell infiltration were associated with the Wnt signaling pathway and included hub genes such as ELN. Overall mutation levels, tumor mutation burden, and microsatellite instability were lower, but tumor immune dysfunction and exclusion scores were greater (p < 0.05) in the high-risk group than in the low-risk group. The validation results showed that the prediction model score can accurately predict the prognosis of GC patients. Finally, a nomogram was constructed using the risk score combined with the clinicopathological characteristics of patients with GC. Conclusion This risk score from the prediction model related to the tumor microenvironment and immunotherapy could accurately predict the overall survival of GC patients.