Development and Verification of Ferroptosis-Related Gene Signature for Predicting the Prognosis and Immune Microenvironment in Gastric Cancer

Objective: The study is to explore the role of ferroptosis-related genes (FRGs) in the occurrence and development of gastric cancer (GC), and to construct a new prognosis signature to predict the prognosis and immune microenvironment in GC. Method: We downloaded RNA sequencing data and related clinical information of GC from the Cancer Genome Atlas (TCGA-STAD) database as a training cohort. Microarray GSE84426 and GSE84437 were downloaded from Gene expression synthesis (GEO) database as validation cohorts. FRGs were come from FerDb. Prognostic genes associated with ferroptosis were identified in the training cohort using univariate Cox analysis and venn diagram, and two different molecular subtypes were identified by consistent clustering. Kaplan-Meier survival curve verified the prognostic value. ESTIMATE, CIBERSORT, McpCounter and TIMER algorithm were used to analyze the infiltration of immune cells in each sample. LASSO algorithm and multiple Cox regression analysis were used to construct a prognostic risk signature and verified it. Finally, gene set enrichment analysis (GSEA) revealed several important ways to participate in GC. Result: We obtained 16 prognostic genes for GC associated with ferroptosis, and divided GC patients into two subgroups by consistent clustering. Cluster C2 showed obvious median survival advantage, while Cluster C1 showed poor prognosis. Compared with Cluster C2, GC patients in Cluster C1 have significantly higher ESTIMATE score, higher immune cell infiltration and higher matrix score. MCPCounter, TIMER and CIBERSORT algorithms also showed that Cluster C1 had more immune cell infiltration, suggesting that the tumor immune microenvironment (TIME) of GC patients in high-risk group Cluster C1 accorded with immune exclusion subtype. Based on LASSO analysis, a risk signature was established, and it was found that TUBE1, NFE2L2 and ACSL4 genes were protective factors, while ZFP36, NOX5 and MIR9-3 genes were risk factors. Through the verification of GEO cohort, the risk signature of FRGs had a great potential to predict the prognosis of GC patients, and it was a strong correlation between the prognosis signature and GC immunity. The nomogram combined with risk signature and clinical features can accurately predict the prognosis of GC patients. GSEA analysis showed that compared with Cluster C2, Cluster C1 had lower expression in lipid metabolism and glutathione metabolism, which may be related to poor prognosis of GC patients. Conclusion: We constructed a new prognostic signature based on ferroptosis-related prognostic genes to predict the prognosis and TIME in GC, and the development of this signature had provided a new clue for determining the relationship between ferroptosis and immunity, which can be used to accurately predict the prognosis of GC patients and provide new strategies for immunotherapy of GC patients.