Figure 3

Multivariate ROC curve based exploratory analysis. (A) comparison of ROC curves to discriminate patients with CF from controls (CTR) using from 2 to 4 variables; (B) predictive accuracies of models from 2 to 4 variables; (C) predicted class probabilities for each sample using the 4 features-model; (D) rank features for average importance. AUC: area under the ROC curve; CF: cystic fibrosis; ROC: receiver operating characteristic.