Fig. 4.

Comparisons between our method and others on our test set. The performance of the multi-label model with ROC curves on a ubiquitylation site prediction and b SUMOylation site prediction, and PR curves on c ubiquitylation site prediction, and d SUMOylation site prediction. The red line represents the average result of the tenfold cross-validation for our tool. The dashed lines and marks in other colors represent the results of popular Ubiquitylation and SUMOylation site prediction tools. The reason we used marks here is that those tools returned classification results instead of predictive probabilities. We are unable to plot ROC and PR curves with the predictive probabilities