Figure 2

ROC graph. The Disprot database was used to establish the positive and negative datasets to find the most efficient algorithm combination. The TPR and FPR values of all possible combinations among the selected disordered predictors are presented. The chosen combination is indicated with a red arrow. The chosen algorithm combination for structural disorder prediction included REM465, GlobPipe, IUPred and VSL2B (True Positive Rate = 0.670 and False Positive Rate = 0.402). This particular combination presented the lowest count of false positives and a high count of false negatives among the five best predictor combinations.