Leveraging Diverse Regulated Cell Death for Prognostic Prediction of Pancreatic Cancer Using Machine Learning and Multi-omics Analysis

Abstract

Abstract Pancreatic cancer is an aggressive form of cancer with poor prognosis. Recent findings indicate that regulated cell death (RCD) significantly contributes to tumor growth, which could be a potential prognostic indicator for pancreatic cancer. We utilized weighted gene co-expression network analysis to extract 103 genes shared by different subtypes of RCD (apoptosis, ferroptosis, necroptosis, pyroptosis, and entotic cell death). Following our analysis, we conducted cross-validation on 15 machine learning algorithms, testing a total of 167 combinations to establish a predictive model. The combination of deep learning and random survival forest achieved the highest C-index among the 167 evaluated machine learning algorithms. Therefore, it has been selected for further research. Patients were divided into high- and low-risk group. The high-risk group of pancreatic cancer patients had poorer prognosis when validated in three independent datasets. Further analysis revealed that individuals in the high-risk group were more likely to respond to immunotherapy and chemotherapy. Based on the proteome data, the protein expression of shared genes was significantly differentially expressed between pancreatic cancer and normal control. For MYOF, a shared gene, expression levels were significantly higher in pancreatic cancer tissues compared to adjacent tissues by PCR. Additional experimental results demonstrate MYOF's involvement in various processes, including proliferation, viability, invasion, and migration in pancreatic carcinoma cells. These results highlight its potential as a significant target for further investigation.