TFCpGRegulon defines a fibroblast-driven epigenetic–transcriptional subtype associated with therapeutic resistance in gastric cancer

Gastric cancer (GC) has a complex etiology and diverse pathological subtypes, which pose challenges for treatment and prognostic evaluation. Transcriptional dysregulation mediated by epigenetic regulation varies across patients and tumor regions, potentially driving the formation of molecular heterogeneity in GC. Therefore, systematically elucidating the interplay between DNA methylation (CpGs) and transcription factors (TFs) can help uncover the molecular basis underlying GC heterogeneity. We developed TFCpGRegulon, an integrative computational framework that employs an optimized WGCNA-based metric (CorComP) to identify epigenetically coordinated gene modules. This approach stratified GC into mesenchymal-like and epithelial-like subtypes. Single-cell RNA sequencing (scRNA-seq), spatial transcriptomics (ST), a cross-dataset cancer-associated fibroblast (CAF) atlas, motif enrichment analyses, immunotherapy cohort analysis, and in silico drug sensitivity prediction were integrated to characterize subtype-associated biological and therapeutic features. The mesenchymal-like subtype was characterized by widespread CpG hypomethylation, fibroblast infiltration, and aggressive clinical behavior. Multi-layer analyses revealed a fibroblast-enriched, immunosuppressive microenvironment characterized by TGF-β activation and EMT signaling. Single-cell and spatial transcriptomic analyses showed preferential enrichment of the TFCpGRegulon transcriptional output in fibroblasts and fibroblast-rich regions rather than epithelial compartments. A cross-dataset CAF atlas identified a TFhigh/CpG-genehigh fibroblast state linked to WNT/TGF-β signaling and extracellular matrix remodeling. Motif analysis provided sequence-level validation, revealing high-affinity binding of core TFs (e.g., ZEB1, GLI3) to hypomethylated target promoters. Temporal analysis indicated that these early molecular alterations were accompanied by a stage-dependent escalation of fibroblast infiltration and EMT. This fibroblast-orchestrated microenvironment was associated with poor late-stage survival and reduced therapeutic responsiveness, particularly to immune checkpoint inhibitors. Our study provides a robust framework for dissecting coordinated epigenetic–transcriptional regulation in GC. We highlight a critical stromal–epigenetic axis that is associated with tumor progression, immunosuppression, and clinically relevant heterogeneity. These findings provide a valuable framework for improved prognostic stratification and the development of personalized therapeutic strategies in GC.