Related Experiment Video
Updated: Jul 12, 2026
Cancer-Associated Fibroblasts from Mouse Mammary Tumors as Tools for Molecular and Computational Studies
Published on: July 3, 2025
A single-cell and machine learning framework identifies CAFs-associated signatures linking stromal heterogeneity to
Faliang Xing1, Jia Sun1, Chun Li1
1Department of General Surgery, State Key Laboratory of Complex Severe and Rare Diseases, Peking Union Medical College Hospital, Chinese Academy of Medical Science and Peking Union Medical College, Beijing, China.
Background:
Cancer-associated fibroblasts (CAFs) are key components of the tumor microenvironment (TME) in pancreatic ductal adenocarcinoma (PDAC), contributing to tumor progression, metabolic reprogramming, and immune suppression. However, the functional heterogeneity of CAFs and their prognostic and immunological significance remain incompletely understood. This study aimed to characterize CAFs heterogeneity in PDAC and develop a robust CAFs-associated signature (CAFAS) for prognostic prediction and immune stratification.
Methods:
Single-cell RNA sequencing (scRNA-seq) datasets from PDAC were analyzed to identify and characterize CAFs subpopulations. Distinct CAFs clusters were annotated, and prognostic CAFs-associated genes were screened to construct a CAFAS through benchmarking seven machine learning algorithms under a nested cross-validation framework. The predictive performance of CAFAS was validated across five independent cohorts. Comprehensive analyses, including immune infiltration assessment, pathway enrichment, and drug sensitivity prediction, were performed to elucidate the biological and clinical implications of CAFAS.
Results:
Five CAFs subtypes with distinct molecular and functional features were identified, among which ADM+ALDOA+ CAFs were associated with poor prognosis and enriched in glycolytic and proliferative pathways. The resulting CAFAS demonstrated strong and consistent prognostic performance across multiple cohorts, accurately stratifying patients by overall survival and therapeutic responsiveness. High CAFAS scores correlated with cell cycle activation and glycolytic pathways, whereas low CAFAS scores were associated with immune activation and lipid metabolism. CAFAS also effectively predicted immune infiltration, immunotherapy response, and drug sensitivity.
Conclusions:
This study establishes a robust CAFs-associated prognostic model that integrates single-cell transcriptomic insights with machine learning to capture CAFs heterogeneity in PDAC. CAFAS provides a valuable framework for precision prognosis, immunotherapy stratification, and the identification of potential therapeutic targets aimed at remodeling the immunosuppressive stroma in PDAC.
