Integrative transcriptomics and single-cell analysis uncover prostaglandin-linked communication in high-risk pancreatic cancer

Background Pancreatic cancer has an extremely poor prognosis due to late diagnosis, early metastasis, and limited benefit from systemic therapies. Metabolic reprogramming is central to tumor progression and immune microenvironment remodeling. Methods Metabolism-related genes (MRGs) were systematically analyzed using TCGA as the training cohort and four GEO datasets (GSE28735, GSE57495, GSE79668, GSE85916) for external validation. Data were normalized to TPM, log-transformed, and batch-corrected with ComBat; PCA assessed cross-cohort consistency. Prognosis-associated MRGs were screened by differential expression analysis and univariate Cox regression. Model selection was performed via 10-fold cross-validation across 101 machine-learning algorithm combinations. Immune characteristics were evaluated by pathway and cell-recruitment/activation analyses. Single-cell RNA-seq (GSE155698) was used to examine prostaglandin-biosynthesis programs and cell–cell communication. In vitro assays assessed gene function. Results Sixty-one prognosis-associated MRGs were identified. CoxBoost + SuperPC achieved the highest concordance and was chosen for the final metabolism-related prognostic model (MRM). The risk score robustly stratified patients into high- and low-risk groups with consistent survival discrimination across cohorts, supported by time-dependent ROC and PCA analyses. High-risk tumors showed reduced T-cell priming/activation and CD8⁺ T-cell recruitment but increased myeloid-cell recruitment, consistent with an immunosuppressive (“immune-cold”) phenotype. Single-cell analysis revealed broad activation of a prostaglandin biosynthesis–related program across multiple cell types and more frequent cell–cell communication in the PB-high state. Among the five model genes (PLOD2, PLCB4, HNMT, DERA, B4GALT5), B4GALT5 correlated positively with risk and poorer prognosis; its knockdown inhibited pancreatic cancer cell proliferation. Conclusions The MRM provides a metabolism-informed tool for prognostic stratification and highlights metabolic–immune coupling as a potential therapeutic avenue in pancreatic cancer.