Fig. 6

Multifactor feature selection and classification modelling using clinical or methylation data. A Variable importance plot for model 1 (clinical data available at diagnosis). B Variable importance plot for model 2 (methylation data at 1708 CpGs). C Distribution of weighted methylation risk scores (wMRS) by Sample Group (p < 2.72 × 10^–16). D Receiver Operator Curves (ROCs) of disease severity classification using clinical data (AUC = 0.74), methylation data at 1708 CpGs (AUC = 0.91) or wGRS (AUC = 0.77). Models were trained on 164 samples (70%) and tested on 71 samples (30%)