Figure 7

Random Forest machine learning analyses of ³¹P MRS data. An imbalanced Random Forest machine learning algorithm was trained on the ³¹P MRS data to predict the group (MMD patients before Mg supplementation, control family members) with metabolites and brain region (hippocampus, premotor cortex, and motor cortex) as predictors. Panel A depicts importance (predictive power) of used predictors in the whole group, in the subgroup of MMD patients, and in the subgroup of healthy control family members. ROC curve with AUC and 95% CI (panel B) is based on Out-Of-Bag data. Abbreviations: AUC, area under curve; CI, confidence interval; MMD, Miyoshi myopathy/dysferlinopathy; MRS, magnetic resonance spectroscopy; ROC, receiver operating characteristic.