Figure 2

ROC curves of the three prediction models tested for DisTer. ROC curves (showing 1-Sp, where Sp is the proportion of true DTU pairs correctly classified as DTU by our method, as a function of Se, the proportion of true STU pairs correctly classified as STU by our method) are evaluating the resubstitution error rate (left) and the model performance when the model is trained on 80% of the E. coli data and tested on the other 20% (middle), or using the leave-one-out technique (right). The red lines correspond to the second predictor model (DisTer), which was chosen for Buchnera operon map prediction. The grey arrows indicate the nearest point of the curves to the left upper corner. For the description of the models 1, 2 and 3 see the Methods section.