Fig. 3

ROC curves for Human (PAR-CLIP) and Mouse (HITS-CLIP). The figures in the first row are for target sites involving canonical seed matches while the second row shows results for non-canonical seed match target sites. The legend "Human train" in the ROC curves for mouse data indicates the model which was trained on human data while the mouse data was used as the test dataset. Similarly the legend "Mouse train" in the ROC curves for human data indicates the model which was trained on mouse data while the human data was used as the test dataset. MirSVR [26], PITA [36], and TargetScan [37] only generate predictions for seed match sites in the 3’ UTR region. Note that for seedless sites in human, although mirSVR appears to perform slightly better than MIRZA, it generates very few seedless target sites, thereby resulting in a very jagged ROC curve. The markers indicate points on the curve where the difference between the TPR and FPR is maximum