Skip to main content

Table 2 Retrospective evaluation performance

From: Industry-scale application and evaluation of deep learning for drug target prediction

 MetricNull Hyp.: DNN AUC < row AUC
AlgorithmROC-AUCKappaF1Wilcoxon testSign test
DNN0.83 ± 0.110.39 ± 0.230.58 ± 0.30  
XGB0.81 ± 0.110.36 ± 0.210.56 ± 0.308.01e−487.90e−50
MF0.78 ± 0.110.15 ± 0.200.45 ± 0.341.80e−711.14e−84
  1. Retrospective evaluation performance values (mean and standard deviation across targets) for the considered machine learning algorithms together with p-values of tests comparing the ROC-AUC of the respective algorithm in the table row with that of DNNs