Construction of an immune-related gene signature for overall survival prediction and immune infiltration in gastric cancer

Abstract

BACKGROUND Treatment options for patients with gastric cancer (GC) continue to improve, but the overall prognosis is poor. The use of PD-1 inhibitors has also brought benefits to patients with advanced GC and has gradually become the new standard treatment option at present, and there is an urgent need to identify valuable biomarkers to classify patients with different characteristics into subgroups. AIM To determined the effects of differentially expressed immune-related genes (DEIRGs) on the development, prognosis, tumor microenvironment (TME), and treatment response among GC patients with the expectation of providing new biomarkers for personalized treatment of GC populations. METHODS Gene expression data and clinical pathologic information were downloaded from The Cancer Genome Atlas (TCGA), and immune-related genes (IRGs) were searched from ImmPort. DEIRGs were extracted from the intersection of the differentially-expressed genes (DEGs) and IRGs lists. The enrichment pathways of key genes were obtained by analyzing the Kyoto Encyclopedia of Genes and Genomes (KEGGs) and Gene Ontology (GO) databases. To identify genes associated with prognosis, a tumor risk score model based on DEIRGs was constructed using Least Absolute Shrinkage and Selection Operator and multivariate Cox regression. The tumor risk score was divided into high- and low-risk groups. The entire cohort was randomly divided into a 2:1 training cohort and a test cohort for internal validation to assess the feasibility of the risk model. The infiltration of immune cells was obtained using ‘CIBERSORT,’ and the infiltration of immune subgroups in high- and low-risk groups was analyzed. The GC immune score data were obtained and the difference in immune scores between the two groups was analyzed. RESULTS We collected 412 GC and 36 adjacent tissue samples, and identified 3627 DEGs and 1311 IRGs. A total of 482 DEIRGs were obtained. GO analysis showed that DEIRGs were mainly distributed in immunoglobulin complexes, receptor ligand activity, and signaling receptor activators. KEGG pathway analysis showed that the top three DEIRGs enrichment types were cytokine-cytokine receptors, neuroactive ligand receptor interactions, and viral protein interactions. We ultimately obtained an immune-related signature based on 10 genes, including 9 risk genes (LCN1 , LEAP2 , TMSB15A mRNA, DEFB126 , PI15 , IGHD3-16 , IGLV3-22 , CGB5 , and GLP2R ) and 1 protective gene (LGR6 ). Kaplan-Meier survival analysis, receiver operating characteristic curve analysis, and risk curves confirmed that the risk model had good predictive ability. Multivariate COX analysis showed that age, stage, and risk score were independent prognostic factors for patients with GC. Meanwhile, patients in the low-risk group had higher tumor mutation burden and immunophenotype, which can be used to predict the immune checkpoint inhibitor response. Both cytotoxic T lymphocyte antigen4+ and programmed death 1+ patients with lower risk scores were more sensitive to immunotherapy. CONCLUSION In this study a new prognostic model consisting of 10 DEIRGs was constructed based on the TME. By providing risk factor analysis and prognostic information, our risk model can provide new directions for immunotherapy in GC patients.