To develop prognostic markers related to drug resistance in pancreatic cancer patients based on multiple machine learning methods

This study aimed to identify novel prognostic biomarkers and therapeutic strategies by focusing on drug resistance-related phenotypes. We integrated multi-omics data from various databases encompassing sequencing, clinical information, resistance-related gene sets, and genomic alteration data. ssGSEA was employed to calculate resistance scores for individual samples, which were subsequently applied in survival analysis. Furthermore, we utilized machine learning algorithms to develop a robust prognostic model validated across multiple independent datasets. Our findings revealed 12 genes consistently linked to pancreatic adenocarcinoma (PAAD) prognosis in diverse datasets. Pathway enrichment analysis indicated that the high-risk group was enriched in pathways associated with systemic lupus erythematosus and the cell cycle, whereas the low-risk group showed significant enrichment in neuroactive ligand-receptor interaction pathways. Additionally, immune cell infiltration analysis exhibited substantial differences between risk groups, with the high-risk cohort presenting lower levels of activated CD8+ T cells but higher levels of regulatory T cells. The random survival forest model demonstrated superior predictive performance, achieving a concordance index of 0.634 and time-dependent receiver operating characteristic area under the curve values of 0.973, 0.978, and 0.996 at 1, 2, and 3 years, respectively. In conclusion, this study identifies 12 critical drug resistance genes in PAAD and highlights the associated immune differences in patient risk, paving the way for targeted immunotherapy research to improve therapeutic strategies against this formidable disease.