Gastric adenocarcinoma (GA) is a significant global health issue with poor prognosis, despite advancements in treatment. Although molecular classifications, such as The Cancer Genome Atlas (TCGA), provide valuable insights, their clinical utility remains limited. We performed a multi-layered functional analysis using TCGA RNA sequencing data to better define molecular subtypes and explore therapeutic implications. We reanalyzed TCGA RNA-seq data from 142 GA patients with localized disease who received adjuvant chemotherapy. Our approach included probabilistic graphical models and recurrent sparse k-means/consensus cluster algorithms for layer-based analysis. Our findings revealed survival differences among TCGA groups, with the GS subtype showing the poorest prognosis. We identified twelve functional nodes and seven biological layers, each with distinct functions. The combined molecular layer (CML) classification identified three prognostic groups that align with TCGA subtypes. CML2 (GS-like) displayed gene expression related to lipid metabolism, correlating with worse survival. Transcriptomic heterogeneity within the CIN subtype revealed clusters tied to proteolysis and lipid metabolism. We identified a subset of CIN tumors with profiles similar to MSI, termed CIN-MSI-like. Claudin-18, a key gene in proteolysis, was overexpressed across TCGA subtypes, suggesting it is a potential therapeutic target. Our study advances GA biology, enabling refined stratification and personalized treatment. Further studies are needed to translate these findings into clinical practice.
Keywords: CIN-MSI-like group; TCGA subtypes; consensus cluster analysis; gastric adenocarcinoma; molecular classification.