ROC curve showing the performance of tetranucleotide frequency-based discrimination. The ROC curves show the performance in 10-fold cross validations for the Oli method on the AURA_dataset and on PUM2+. The negative data are in both cases provided by 3K-. The further the ROC curve advances towards the upper-left corner, the better the classification ability of the model. A curve near the 45-degree diagonal reflects a random classification.