Coexpression network analysis of gastric adenocarcinoma identifies hub genes as biomarker candidates and their tumor microenvironment associations.

Gastric adenocarcinoma (GAC) is characterized by molecular heterogeneity that limits early detection and targeted treatment. We applied weighted gene coexpression network analysis (WGCNA) to paired RNA-seq data from 119 GAC and peritumoral tissue (PTT) samples and identified six coexpression modules with distinct biological identities. Four modules were positively correlated with GAC and two were negatively correlated. Among the 30 hub genes evaluated, several outperformed established clinical biomarkers in accuracy. Specifically, SPARC (AUC = 0.89), COL3A1 (0.87), and COL1A2 (0.85) exceeded MUC5AC (0.76), VEGFA (0.68), and ERBB2 (0.64). Validation in the TCGA-STAD cohort confirmed concordant expression trends for MEblack (5/5 genes) and MEmagenta (4/5), with an overall fold-change correlation of ρ = 0.58 (p = 6.96 × 10⁻⁴). Immune deconvolution delineated two opposing microenvironmental axes, with an adaptive-immune-epithelial program (MEblack) associated with B-cell abundance, and a fibroblast-collagen program (MEmagenta) associated with cancer-associated fibroblast enrichment. DepMap CRISPR screening identified ribosomal hub genes as cell-intrinsic dependencies in gastric cancer cell lines. Among all hub genes, SPARC, COL3A1, COL1A2, GKN1, and GKN2 emerged as the potential biomarker candidates, with SPARC additionally showing a validated unfavorable prognostic association in STAD.