Abstract
Emerging evidence suggests that epigenetic alterations are responsible for the oncogenesis and progression of cancer. However, the role of epigenetic reprogramming in pancreatic cancer is still not clear. In this study, we used the limma R package to identify differentially expressed protein-coding genes (PCGs) between pancreatic cancer tissues and normal control tissues. The cell-type identification by the estimating relative subsets of RNA transcripts (CIBERSORT) package was used to quantify relative cell fractions in tumors. Prognostic molecular clusters were constructed using ConsensusClusterPlus analysis. Furthermore, the least absolute shrinkage and selection operator and stepAIC methods were used to construct a risk model. We identified 2351 differentially expressed PCGs between pancreatic cancer and normal control tissues in The cancer genome atlas dataset. Combined with histone modification data, we identified 363 epigenetic PCGs (epi-PCGs) and 19,010 non-epi-PCGs. Based on the epi-PCGs, we constructed three molecular clusters characterized by different expression levels of chemokines and immune checkpoint genes and distinct abundances of various immune cells. Furthermore, we generated a 9-gene model based on dysfunctional epi-PCGs. Additionally, we found that patients with high risk scores showed poorer prognoses than patients with low risk scores (p < 0.0001). Further analysis showed that the risk score was significantly related to survival and was an independent risk factor for pancreatic cancer patients. In conclusion, we constructed a 9-gene prognostic risk model based on epi-PCGs that might serve as an effective classifier to predict overall survival and the response to immunotherapy in pancreatic cancer patients.