Fig. 5

Selection of the optimal ASCVD-related metabolic/clinical variable panel by using Random Forest algorithm. a Top important ASCVD-associated variable selection by using the values of mean decrease accuracy in Random Forest analysis. b ROC curves generated from MCCV-based multivariate Random Forest models. The AUC-ROC value and its 95% CI are shown. c Predictive accuracies of multivariable Random Forest models with different numbers of top important variables. d Kaplan–Meier survival analysis of eight top important metabolites in HoFH patients with and without ASCVD events during the follow-up. Log-rank test P values are shown. LDLR status: at least one null mutation. Other abbreviations are shown in Table 2 and Figs. 2 and 3