Skip to main content

Table 2 Comparison of model performance on the same benchmark datasets

From: Deep neural networks for human microRNA precursor detection

ModelhsaPseudo
This workCNN9688
This workRNN9092
AverageDT9793
ConsensusNB8686
ConsensusDT9990
DingNB8884
AverageNB8389
NgDT8989
Consensus9796
BatuwitaNB8683
BentwichNB9271
NgNB8681
  1. Note: Prediction results (percentage) of this work (rows 2–3) and the top ten models (rows 4–13) of the izMiR framework [19]. The values presented were true prediction rates (%) (TPR) achieved for each model and dataset. Pseudo: negative data, from the coding region of human RefSeq genes, 8492 hairpins; hsa: positive data, Homo sapiens miRNAs, 1881 sequences