Fig. 4

ROC curve and waterfall plots for performance of the top 3 CpG diagnostic model in a training and b validation datasets. The value of the classifier is given by 6.52–17.04*cg00054525 + 24.18*cg16794576–13.82*cg24581650, where the intercept and coefficients have been regressed by a binomial generalized linear model. A threshold value of this classifier was chosen to yield maximal non-unity specificity in the training set. The red dot on the ROC curve corresponds to the sensitivity and specificity of the classifier at the chosen threshold. The dashed line on the waterfall plots is drawn at the chosen threshold value of the classifier