A novel cancer-associated membrane signature predicts prognosis and therapeutic response for lung adenocarcinoma.

Lung adenocarcinoma (LUAD) is a leading cause of cancer-related death, and reliable biomarkers for prognosis and treatment guidance remain limited. Membrane proteins play key roles in tumor progression and therapeutic response, yet their clinical utility in LUAD remains underexplored. We integrated scRNA-seq, spatial transcriptomics, and bulk RNA-seq datasets from multiple LUAD cohorts to identify cancer-specific membrane proteins derived from epithelial subpopulations. Based on these results, we constructed a prognostic signature, LCaMPS, and evaluated its predictive performance using multiple datasets. The expression of model genes was confirmed at both the bulk RNA and protein levels. Associations with the tumor microenvironment (TME) and drug sensitivity were further analyzed. A distinct LUAD-enriched epithelial cluster (Epi_c0) exhibiting hypoxic and EMT signatures was identified. 35 cancer-specific membrane proteins were defined, several of which, including TSPAN8, BACE2, and COX16, showed strong spatial localization within the tumor regions. LCaMPS, a 9-membrane gene-based prognostic model, stratified patient prognosis and predicted 5- and 10-year survival rates with high accuracy. High LCaMPS scores were associated with increased infiltration of neutrophils, endothelial cells, and fibroblasts in the TME and predicted higher sensitivity to 66 chemotherapeutic agents, including Gemcitabine and Sorafenib. Low-risk patients were predicted to respond better to drugs, such as Cisplatin and Parthenolide. This study highlights the importance of membrane expression patterns in LUAD at single-cell and spatial resolution. The LCaMPS model provides a robust prognostic and therapeutic stratification tool with potential applications in personalized cancer management.