Fig. 7

Selected WD phenotype blood DMRs classify hepatic from neurologic WD in an independent test cohort. A classifier using the AdaBoost algorithm was trained with methylation values at WD phenotype DMRs in WD patient blood (WDH n = 20, WDN n = 20). a Feature importance was determined for each of the 2860 DMRs after training. b Heatmap of 44 DMRs used in classifier using methylation in training samples. c Genes near DMRs with feature importance ≥ 0.04. d Classification results in test cohort of WD patients. In the plot, the column represents true phenotype, while the color represents predicted phenotype (WDH n = 5, WDN n = 5, precision = 0.9, ROC AUC = 0.8)