Integrative multimodal transcriptomics identifies a cancer-associated fibroblast membrane signature for predicting prognosis and therapeutic response in pancreatic ductal adenocarcinoma

Cancer-associated fibroblasts (CAFs) are central to the pancreatic ductal adenocarcinoma (PDAC) microenvironment, promoting tumor progression and therapeutic resistance. However, the expression landscape of CAF membrane proteins in PDAC remains poorly defined. We integrated scRNA-seq (n = 33; 87,949 cells), spatial transcriptomics (n = 2; 7,011 spots), and bulk RNA-seq (n = 7; 642 samples) to systematically identify PDAC-specific CAF membrane genes. A LASSO-based Cox model was developed to construct a prognostic signature, PaFMS, and evaluated through multi-cohort validation. Functional enrichment, immune infiltration, drug sensitivity, and immunotherapy response analyses were further conducted. Validation was performed using multiple database-driven analyses. We identified a PDAC-enriched myoCAF-c1 cluster closely associated with epithelial-mesenchymal transition (EMT) and angiogenesis. From this cluster, 33 candidate CAF membrane genes were defined, whose protein-protein interactions were predominantly linked to extracellular matrix organization and collagen remodeling, and spatially colocalized with myoCAF-c1 and EMT regions. An 11-gene prognostic signature, PaFMS that robustly stratified patients across six independent cohorts, achieving high predictive accuracy for overall survival. High-risk patients exhibited proliferative signaling activation, immune suppression, and reduced T/B-cell infiltration. PaFMS was associated with responses to 33 anticancer agents and predicted enhanced benefit from anti-PD-L1 immunotherapy in the low-risk group. Multi-cohort validation confirmed the expression specificity of PaFMS genes, including PLAU, TMEM158, and TRIM59. Together, these findings reveal that myoCAF-c1 promotes angiogenesis and tumor progression, and establish PaFMS as a robust CAF membrane-based prognostic model in PDAC with potential utility for precision prognosis and therapeutic decision-making. KEY MESSAGES: Integrated single-cell, spatial, and bulk RNA-seq analyses identified PDAC-specific CAF membrane genes. Discovered a PDAC-enriched myoCAF-c1 subtype linked to EMT and angiogenesis. Developed an 11-gene CAF membrane-based prognostic model (PaFMS) validated across six cohorts. PaFMS predicts patient survival, drug sensitivity, and immunotherapy response in PDAC.