Identifying Epithelial-Cell Progression Signatures (ECPSs) from Multi-Resolution Multi-Omics Data for Translational...

Dynamic transcriptomic remodeling of epithelial cells represents a critical hallmark of lung adenocarcinoma (LUAD) progression, yet epithelial-cell progression signatures (ECPSs) linked to this process remain incompletely characterized, impeding our understanding of malignant LUAD transformation. Here, we performed a multi-resolution multi-omics study and identified two functionally opposing ECPSs: a Cancer-Promoting Signature (CPS) and a Cancer-Suppressing Signature (CSS). The CPS was progressively upregulated from normal to early- and advanced-stage lesions, while the CSS was gradually downregulated, and both were validated by multi-resolution transcriptomic data. Both the CPS and CSS demonstrated robust diagnostic value for LUAD, particularly in early-stage detection (median AUC > 0.95). In seven independent validation cohorts, the CPS and CSS served as robust prognostic risk and protective factors, respectively. Their combination could better stratify LUAD patients into distinct prognostic subgroups, with the CPS_high-CSS_low subgroup showing the worst prognosis, accompanied by high genomic instability, an immunosuppressive tumor microenvironment, and resistance to chemotherapy. Notably, the CPS and CSS exhibited broad translational value in other epithelium-derived cancers. HMGA1, a key CPS gene, showed epithelium- and advanced-stage-specific high expression, which was significantly associated with poor prognosis, genomic instability, and immune escape. Collectively, our study identifies the CPS and CSS as clinically reliable ECPSs, providing valuable biomarkers for clinical application to LUAD.