Integrated analysis of 1804 samples of six centers to construct and validate a robust immune-related prognostic signature associated with stromal cell abundance in tumor microenvironment for gastric cancer

Background Stromal cells in tumor microenvironment could promote immune escape through a variety of mechanisms, but there are lacking research in the field of gastric cancer (GC). Methods We identified differential expressed immune-related genes (DEIRGs) between the high- and low-stromal cell abundance GC samples in The Cancer Genome Atlas and GSE84437 datasets. A risk score was constructed basing on univariate cox regression analysis, LASSO regression analysis, and multivariate cox regression analysis in the training cohort ( n =772). The median value of the risk score was used to classify patients into groups with high and low risk. We conducted external validation of the prognostic signature in four independent cohorts (GSE26253, n =432; GSE62254, n =300; GSE15459, n =191; GSE26901, n =109) from the Gene Expression Omnibus (GEO) database. The immune cell infiltration was quantified by the CIBERSORT method. Results The risk score contained 6 genes (AKT3, APOD, FAM19A5, LTBP3, NOV, and NOX4) showed good performance in predicting 5-year overall survival (OS) rate and 5-year recurrence-free survival (RFS) rate of GC patients. The risk death and recurrence of GC patients growing with the increasing risk score. The patients were clustered into three subtypes according to the infiltration of 22 kinds of immune cells quantified by the CIBERSORT method. The proportion of cluster A with the worst prognosis in the high-risk group was significantly higher than that in the low-risk group; the risk score of cluster C subtype with the best prognosis was significantly lower than that of the other two subtypes. Conclusion This study established and validated a robust prognostic model for gastric cancer by integrated analysis 1804 samples of six centers, and its mechanism was explored in combination with immune cell infiltration characterization.