Construction and multi-omics validation of a five-gene ferroptosis-based model for predicting prognosis and therapy response in pancreatic ductal adenocarcinoma with immune landscape analysis.
TL;DR
A concise and externally validated five-gene ferroptosis-related prognostic signature was developed, which effectively stratifies survival outcomes in PDAC patients and reflects ferroptosis-associated alterations in tumor proliferation, metabolic reprogramming, immune evasion, and drug response.
Abstract
Background
Ferroptosis is an iron-dependent programmed cell death, which plays a complex role in pancreatic ductal adenocarcinoma (PDAC), regulating both tumor development and immune interaction. However, the clinical significance and potential molecular mechanism of ferroptosis in PDAC have not been fully clarified, which limits its application in treatment.
Methods
In order to identify ferroptosis-related genes (FRGs), we integrated transcriptome data from TCGA-PDAC and GTEx databases, as well as supplementary data from three GEO datasets (GSE62452, GSE78229 and GSE183795). We constructed a prognostic risk model by sequential analysis, which included univariate Cox regression, LASSO regression and multivariate Cox regression. Through Kaplan-Meier survival curve, time-specific ROC curve analysis and correlation study with clinicopathological features, we strictly evaluated the prediction accuracy and clinical relevance of this model. Multi-omics analyses included GO/KEGG/GSEA/GSVA pathway enrichment, CIBERSORT immune deconvolution, ESTIMATE scores, TIDE/ICB response prediction, tumor mutational burden (TMB) profiling, oncoPredict drug sensitivity estimation, single-cell RNA-seq (GSE212966), spatial transcriptome (GSM8452850), pseudotime trajectory, and HPA immunohistochemistry validation.
Results
A robust five-gene ferroptosis-related prognostic signature (BCAR3, GSK3B, STAT1, MAGED2, MYEOV) was established. Patients categorized into the high-risk group demonstrated a markedly inferior overall survival compared to those in the low-risk group across both the TCGA cohort (5-year AUC = 0.87) and three external validation cohorts (5-year AUC 0.804-0.833). High-risk tumors showed enrichment in mitotic spindle, PI3K-AKT-mTOR, G2M checkpoint, glycolysis, and p53 pathways, markedly elevated KRAS/TP53 mutation rates, higher TMB, an "inflammatory yet immunosuppressive" microenvironment (increased resting NK cells/plasma cells, decreased activated NK cells, upregulated checkpoints including CD44/HHLA2/LGALS9), and differential chemotherapy sensitivity (greater sensitivity to gemcitabine, erlotinib, gefitinib, trametinib in high-risk group). Single-cell and spatial analyses confirmed predominant expression in malignant ductal cells, with dynamic pseudotime-dependent patterns (especially GSK3B upregulation in late-stage cells) and enhanced MIF signaling-mediated crosstalk in high-score subpopulations. Protein-level heterogeneity was verified by HPA-IHC.
Conclusion
In this study, a concise and externally validated five-gene ferroptosis-related prognostic signature was developed, which effectively stratifies survival outcomes in PDAC patients and reflects ferroptosis-associated alterations in tumor proliferation, metabolic reprogramming, immune evasion, and drug response. This model provides a framework for mechanism-informed precision treatment strategies, such as ferroptosis induction combined with immunotherapy or risk-adapted chemotherapy, and may offer new insights into improving outcomes in this highly lethal malignancy.