Feed

Gastric cancer (GC), a highly aggressive and heterogeneous malignancy, remains challenging in immunotherapy despite recent advancements. This study aims to identify novel biomarkers and construct a prognostic model to improve outcome prediction and therapeutic strategies. Mendelian randomization (MR) analysis identified immune cell subtypes linked to GC using FinnGen and GWAS cohorts. CIBERSORT and WGCNA algorithms were applied to define M2 tumor-associated macrophage (TAM)-related gene modules. Key prognostic genes were selected via Lasso-Cox regression to establish a risk model, validated using GEO datasets. Biological function disparities, tumor microenvironment heterogeneity, and therapeutic sensitivities were assessed via GSEA and immune infiltration analysis. Protein-level validation was performed using TCGA, HPA, and Western blot. MR analysis revealed 26 immune cell subtypes associated with GC. WGCNA identified 20 gene modules, with the most M2 TAM-correlated module prioritized. A prognostic signature incorporating SEC61G, BGN, and STC1 was developed, stratifying patients into distinct risk groups with divergent survival outcomes (1-/3-/5-year, all P < 0.05). High-risk patients exhibited enriched calcium signaling pathways, reduced immunotherapy responsiveness, and increased sensitivity to veriparib and palbociclib. Protein overexpression of key genes was validated in GC tissues. This integrated bioinformatics-MR framework establishes a TAM-driven prognostic model for GC, demonstrating clinical utility in survival prediction, immunotherapy efficacy evaluation, and personalized therapeutic targeting. The findings provide actionable insights for advancing precision immunotherapy in GC.