Machine learning-based prognostic signature integrating mitochondrial function and programmed cell death patterns in pancreatic adenocarcinoma.

The interplay between mitochondria and programmed cell deaths (PCD) is associated with tumor pathogenesis. However, the specific roles of genes related to mitochondrial function and PCD patterns in pancreatic adenocarcinoma (PAAD) remain unclear. Though integrated multi-omics and bioinformatic analyses, we identified PCD activity-associated mitochondrial function genes (PCDMGs), and constructed a robust prognostic signature containing eight PCDMGs via CoxBoost+StepCox algorithms[both] algorithms. The constructed nomogram was validated to have superior prognostic predictive efficacy. Multiple-dimensional analyses demonstrated that the prognostic signature was significantly correlated with oncogenic pathways activation, tumor microenvironment (TME) reconstruction, and genomic variants. Drug sensitivity analysis indicated the high-risk group derived more benefits from immunotherapy but less from chemotherapy and targeted therapy. A ceRNA regulatory network was also constructed to elucidate the potential molecular mechanism of PAAD, and molecular docking simulations revealed favorable binding of the target proteins to prevalent anticancer agents. Finally, qRT-PCR in 6 PAAD patients confirmed the elevated expression levels of the eight model genes in tumor tissues, corroborating the bioinformatic findings. In conclusion, the prognostic signature and nomogram can stratify PAAD patients by prognosis and drug sensitivity, thus assisting clinicians in treatment optimization and clinical decision-making .