Skip to main content

Advertisement

Table 1 Prediction performance of three meta-predictors using three different sets of techniques

From: Significant improvement of miRNA target prediction accuracy in large datasets using meta-strategy based on comprehensive voting and artificial neural networks

Dataset C-I C-II mirTarDANN
Sens Spec Acc Sens Spec Acc Sens Spec Acc
D4 0.240 ± 0.120 0.780 ± 0.128 0.735 ± 0.108 0.357 ± 0.093 0.733 ± 0.087 0.702 ± 0.072 0.568 ± 0.049 0.555 ± 0.031 0.556 ± 0.025
D3–1 0.475 ± 0.063 0.602 ± 0.036 0.584 ± 0.022 0.297 ± 0.032 0.789 ± 0.044 0.716 ± 0.031 0.602 ± 0.048 0.530 ± 0.041 0.541 ± 0.032
D3–2 0.671 ± 0.097 0.235 ± 0.108 0.268 ± 0.094 0.433 ± 0.128 0.690 ± 0.041 0.669 ± 0.036 0.511 ± 0.176 0.605 ± 0.111 0.593 ± 0.092
D3–3 0.087 ± 0.018 0.693 ± 0.042 0.557 ± 0.028 0.693 ± 0.052 0.704 ± 0.069 0.701 ± 0.047 0.660 ± 0.037 0.730 ± 0.056 0.715 ± 0.037
D3–4 0.383 ± 0.058 0.434 ± 0.088 0.420 ± 0.050 0.622 ± 0.061 0.752 ± 0.030 0.713 ± 0.012 0.650 ± 0.051 0.722 ± 0.051 0.700 ± 0.026
D2–1 0.700 ± 0.073 0.221 ± 0.040 0.334 ± 0.029 0.402 ± 0.065 0.653 ± 0.040 0.593 ± 0.013 0.312 ± 0.066 0.722 ± 0.078 0.624 ± 0.037
D2–2 0.519 ± 0.049 0.444 ± 0.040 0.480 ± 0.023 0.668 ± 0.202 0.728 ± 0.017 0.701 ± 0.094 0.685 ± 0.102 0.796 ± 0.020 0.745 ± 0.038
D2–3 0.306 ± 0.137 0.307 ± 0.062 0.312 ± 0.035 0.527 ± 0.055 0.849 ± 0.019 0.738 ± 0.042 0.558 ± 0.047 0.849 ± 0.026 0.749 ± 0.017
D2–4 0.398 ± 0.024 0.659 ± 0.038 0.509 ± 0.004 0.777 ± 0.081 0.393 ± 0.182 0.608 ± 0.050 0.745 ± 0.023 0.540 ± 0.029 0.656 ± 0.023
D2–5 0.388 ± 0.175 0.188 ± 0.174 0.276 ± 0.044 0.700 ± 0.079 0.837 ± 0.060 0.779 ± 0.011 0.721 ± 0.086 0.849 ± 0.022 0.797 ± 0.050
D2–6 0.700 ± 0.155 0.255 ± 0.032 0.466 ± 0.063 0.678 ± 0.065 0.724 ± 0.048 0.700 ± 0.038 0.622 ± 0.074 0.769 ± 0.060 0.693 ± 0.017
  1. C-I integrates individual predictors with a decision tree, C-II uses ANNs to integrate individual predictors, and mirTarDANN is a combination of individual predictors, decision tree, and ANNs. The performance is measured by sensitivity (Sens), specificity (Spec), and accuracy (Acc) under multi-fold cross-validation. The high-lighted values are the highest among three meta-predictors in the same subset