Area Under the Curve (AUC) based on the training dataset.

Area Under the Curve (AUC) was obtained from the ROC curves of 9 predictors: AUC cannot be computed for AMYLPRED2 as it provides only a binary prediction. For each method, the global AUC, the AUC for the False Positive Rate range of 0–20% and the AUC for the False Positive Rate range of 0–5% are reported. Numbers in brackets correspond to 95% confidence intervals (95% C.I.) that were obtained using bootstrap replicates [43]. The comparison of MetAmyl AUC and the other methods is summarized by the p.value obtained with Delong's method [49]. For the MetAmyl classifier, results were obtained using a Leave-One-Out Cross Validation.