Skip to main content

Table 3 AUPR scores (mean and standard deviation obtained by nested 5-fold cross-validation) on the DBEColi dataset in the four \(S_1\), \(S_2\), \(S_3\), or \(S_4\) settings, and for a test sample positive:negative ratio 1:5

From: Evaluation of deep and shallow learning methods in chemogenomics for the prediction of drugs specificity

 \(S_1\)\(S_2\)\(S_3\)\(S_4\)
kronSVM\(\mathit{61.55} \pm \mathit{3.11}\)\(\mathit{28.95} \pm \mathit{7.66}\)\(\mathit{60.96} \pm \mathit{3.73}\)\(24.4 \pm 8.8\)
NRLMF\(\mathit{59.89} \pm \mathit{4.35}\)\(\mathit{35.62} \pm \mathit{8.07}\)\(\mathit{60.06} \pm \mathit{2.73}\)\(\mathit{34.5} \pm \mathit{9.75}\)
FNN\(51.55 \pm 2.55\)\(24.26 \pm 3.91\)\(49.34 \pm 3.11\)\(22.97 \pm 6.59\)
Chemogenomic neural network (CN)\(49.08 \pm 4.31\)\(\mathit{28.38} \pm \mathit{5.81}\)\(46.14 \pm 4.92\)\(\mathit{27.0} \pm \mathit{6.96}\)
RF with proteochemometric features\(53.49 \pm 2.19\)\(23.6 \pm 4.13\)\(51.16 \pm 4.32\)\(22.51 \pm 5.36\)