Skip to main content

Table 3 Prediction performance of RLRNPI-AEN in comparison to other methods in terms of area under ROC curve (AUC)

From: Regularized logistic regression with network-based pairwise interaction for biomarker identification in breast cancer

Dataset RLR-EN RLR-AEN RLRNPI-EN limma ENCAPP SVM-AEP RLRNPI-AEN
GSE2034 0.638 0.663 0.657 0.627 0.681 0.647 0.690
  (0.0219) (0.0516) (0.0516) (0.0036) (0.5966) (0.0516) (–)
GSE1456 0.724 0.734 0.711 0.619 0.722 0.717 0.736
  (0.9768) (0.9920) (0.5600) (0.0114) (0.9920) (0.6024) (–)
GSE11121 0.736 0.725 0.739 0.542 0.750 0.695 0.820
  (0.0171) (0.0076) (0.0250) (0.0012) (0.0250) (0.0050) (–)
GSE6532 0.721 0.725 0.725 0.643 0.730 0.715 0.747
  (0.0451) (0.0424) (0.0422) (0.0012) (0.4725) (0.0219) (–)
GSE4922 0.620 0.611 0.611 0.606 0.593 0.622 0.614
  (1.0000) (1.0000) (1.0000) (1.0000) (0.1032) (1.0000) (–)
GSE12093 0.571 0.518 0.613 0.685 0.616 0.607 0.845
  (0.0012) (0.0012) (0.0012) (0.0208) (0.0034) (0.0012) (–)
  1. The median AUC obtained for each method on the six datasets over ten times ten-fold cross validation. The adjusted p-values calculated using a Mann-Whitney U test are shown within parentheses, which evaluate the significance of difference in classification performance between RLRNPI-AEN and the other six methods. For each dataset, they are corrected using Holm-Bonferroni method for multiple testing. The best two median AUCs and the adjusted p-values that are less than 0.05 are shown in boldface