Abstract
Pancreatic adenocarcinoma (PAAD) has a poor prognosis and responds poorly to immunotherapy. Growing evidence indicates that autophagy promotes tumor progression, and autophagy-related genes (ARGs) have been established as prognostic markers in various cancers. Based on consensus clustering of ARGs, 177 PAAD cases were classified into two subtypes, with cluster B exhibiting weaker anti-tumor immunity. Using univariate COX and Least absolute shrinkage and selection operator (LASSO) regression, 13 ARGs were identified to construct a prognostic model. A risk score formula was developed and validated using the GSE183795 dataset. Patients were divided into high- and low-risk groups, with the high-risk group associated with poorer survival. The model showed strong predictive accuracy, with time-dependent area under the curve (AUC) values ranging from 0.762 to 0.848 in the training cohort and 0.658 to 0.714 in the validation cohort. Further analysis revealed that high-risk patients had suppressed immunity but increased sensitivity to certain chemotherapy drugs. Additionally, high tumor mutational burden (TMB) and RNA stemness score (RNAss) were correlated with worse prognosis in high-risk patients. This ARG-based model effectively predicts PAAD outcomes and may guide clinical treatment strategies.