Skip to main content

Table 2 Performance evaluation metrics for miTAR1 and miTAR2

From: miTAR: a hybrid deep learning-based approach for predicting miRNA targets

Model: dataset Accuracy Sensitivity Specificity F-score PPVc NPVc Brier score
DeepMirTar: DeepMirTarRawa 0.9348 NA NA 0.9348 NA NA NA
miTAR1: DeepMirTar Test set 0.9781 0.9648 0.9921 0.9783 0.9922 0.9641 0.0214
miTAR1: DeepMirTar (30 times)b [95% CI] 0.9787 [0.9714–0.9836] 0.9717 [0.9615–0.9801] 0.9857 [0.9759–0.9921] 0.9786 [0.9717–0.9837] 0.9858 [0.9756–0.9922] 0.9719 [0.9610–0.9807] 0.0193 [0.0144–0.0265]
miRAW: miRAWRawa 0.935 0.935 0.938 0.935 NA NA NA
miTAR2: miRAW Test set 0.9654 0.9609 0.9697 0.9652 0.9695 0.9613 0.0283
miTAR2: miRAW (30 times)b [95% CI] 0.9649 [0.9601–0.9686] 0.9616 [0.9562–0.9678] 0.9683 [0.9618–0.9740] 0.9651 [0.9604–0.9693] 0.9687 [0.9623–0.9742] 0.9610 [0.9558–0.9676] 0.0271 [0.0246–0.0296]
  1. NA represents the value is not reported in the corresponding study
  2. aDeepMirTarRaw and miRAWRaw present the dataset used in the DeepMirTar and the miRAW study. The best performance values are selected for the performance metrics if multiple values are reported in the respective study for different conditions
  3. bEvaluation was done by randomly running on the DeepMirTar and the miRAW datasets 30 times. The average value and the 95% confidence interval (given in []) were reported here. Details are in “Results” section
  4. cPPV represents positive predictive value; NPV presents negative predictive value