Skip to main content
. 2022 Jan 27;76:103798. doi: 10.1016/j.ebiom.2021.103798

Figure 5.

Fig 5

Out-of-sample ROC curves of three classifiers to discriminate MS patients and Controls. The mean ROC curves generated from 200 iterations of model validation for the microbiome (a), immune profile (b), blood metabolome (c), and a combination of all the datasets (d). Means and 90% confidence intervals of AUC scores of ROC curves were listed at the bottom right for classification accuracy. (e) Top 20 important features from random forest (RF) model. Nineteen blood metabolites (indicated by mz) and memory Th17 cells were ranked as the top 20 features of important from RF model in classification of MS patients and controls.