Prediction of Prognosis in Patients with Prostate Cancer by DNA Methylation Classification

Background: Prostate cancer is the second most frequently diagnosed cancer and the fifth leading cause of cancer-related death. It is estimated that the incidence of prostate cancer is on the rise worldwide. Epigenetic changes in tumors play an important role in the occurrence and development of prostate cancer. DNA methylation is one of the mechanisms of tumor epigenetic regulation and may be a new biomarker that has great potential in early tumor screening, treatment guidance and prognosis prediction. The purpose of this study was to explore a classification method from the perspective of DNA methylation.Methods: The least absolute shrinkage and selection operator (LASSO) method was used to analyze DNA methylation and RNA-seq data from the Cancer Genome Atlas (TCGA). The methylation sites with small differences were eliminated, and the 21 methylation sites with the most significant differences were retained for analysis. Using their corresponding gene expression levels, a recurrence prediction model for prostate cancer patients was constructed to distinguish high-risk, medium-risk, and low-risk cases. Immune cell abundance analysis, gene enrichment analysis, Tumor burden mutation analysis and gene copy number variation analysis were then used to analyze the differences among these three subtypes and their underlying mechanisms. Results: We observed the difference in disease-free survival (DFS) of the three methylated subtypes in the test set, which was verified in the validation set. We found three subtypes have different proportions of immune cells, especially in memory B cells, M2 macrophages, Treg cells. GSVA and GSEA analysis revealed that the relevant metastasis gene sets of prostate cancer were enriched in high-risk cases. In addition, the mutation frequencies of TP53, TTN and KMT2D were the highest, and gradually increased in the three genotypes according to Tumor burden mutation (TMB) analysis. Gene copy number variation (CNV) showed that AR, LAPTM4B, and MTDH were significantly amplified, while ATP1B2 and FAM92B were significantly deleted. Finally, univariate and multivariate analysis showed that there were statistical differences among the three methylation subtypes, which can be used as an index to predict prostate cancer recurrence.Conclusions: Our study suggests that classification based on DNA methylation is an independent factor for predicting tumor recurrence in patients with prostate cancer.