2.3. Construction and evaluation of epigenetic risk score based on MDGs

We conducted the univariate Cox regression analysis to select prognostic MDGs with P < .05 as the threshold. Stepwise regression analysis was performed to select 17 hub MDGs with the minimum Akaike information criterion (AIC) value. Meanwhile, we utilized the multivariate Cox method to obtain the β value of hub genes, which was the coefficient of each variable and represented the respective weight. Hazard ratio and 95% confidence interval (95% CI) were calculated and shown in forest plot by survminer package. Then, risk score based on the epigenetic MDGs could be calculated as the follows: Risk score = Ʃ (β _i × Exp_i) (i = 17), where Exp represented the expression data of 17 hub genes. Based on this, 405 BCA patients from TCGA cohort can be divided into high‐ and low‐ risk groups using the median risk score as the cutoff. Importantly, we further used the receiver operating characteristic curve (ROC) to evaluate the predictive efficiency of MDGs with survivalROC package.²² Significant difference between high‐ and low‐ risk groups was assessed by Kaplan‐Merier analysis using survival package.