The Prognosis Prediction Model for Endometrial Cancer Based on DNA Methylation Signature

Background

DNA methylation alteration is a common event during the carcinogenesis and progression of endometrial cancer (EC). Our study aimed to investigate the value of DNA methylation-related genes in predicting the prognosis and immunotherapy response for EC patients.

Methods

The clinical information and the expression of DNA methylation-related genes of 544 endometrial cancers were obtained from the Cancer Genome Atlas (TCGA) database. The univariate Cox regression analysis and the LASSO regression analysis were subsequently used to identify prognosis-related methylation regulators and construct a risk model. Gene functional enrichment analysis, immune infiltration analysis, drug sensitivity analysis, and molecular feature analysis were performed in different subgroups.

Results

25 methylation-Related gene signatures were found in EC patients and are correlated to tumor differentiation and tumor metastasis. By LASSO-Cox regression analyses, a recurrence prediction model and a prognostic-related model were constructed based on methylation-related genes in the TCGA training cohort. The Area Under the Curve (AUC) values of the recurrence prediction model were 0.671, 0.708, and 0.689 for 1-, 3-, and 5-year time points, respectively, while those of the prognostic model were 0.731, 0.717, and 0.725. The relationship of risk score (RS) with ER/PR-related genes, immune checkpoint expressions, and IC50s of paclitaxel, cisplatin, tamoxifen, and cetuximab was investigated. The results showed That patients in the low-risk group are more effective in cetuximab and immune checkpoint blockade (ICB) treatment.

Conclusions

The model based on the methylation-related genes showed promising outcomes in predicting the recurrence and treatment response of EC. The patients with high-risk scores showed a poorer prognosis and may benefit more from the treatment of cetuximab or immune checkpoint inhibitors.