Figure 5From: MiRPara: a SVM-based software tool for prediction of most probable microRNA coding regions in genome scale sequencesROC curves for training and test data. ROC curves for training and test data at different ratios of positive to negative data. Level 1 corresponds to a 1:1 ratio of Positive to Negative data, whereas Level 20 refers to a 1:20 ratio of positive to negative data. From left to right curves are shown for Level 1, Level 5, Level 10 & Level 20. Top row: ROC curves for training sets Overall (green), Animal (blue), Plant (red) &Virus (black). Middle Row: ROC curves for 100nt test datasets. Negative datasets comprises sequences that are predicted to form a hairpin loops but which are not in miRBase. Bottom Row: ROC curves for 10000nt test dataset. Positive dataset contains known pre-miRNAs from miRBase which have 5000nt flanking sequences identified by BLASTing against the NCBI nt database. Negative datasets comprises sequences that are predicted to form a hairpin loops but which are not in miRBase and flanking sequences were identified in the same manner.Back to article page