Skip to main content

Table 7 Bootstrap resampling performance analysis of deep-learning based machine learning methods on unbalanced dataset (1:2)

From: Struct2Graph: a graph attention network for structure based predictions of protein–protein interactions

Method

Performance (%) on Unbalanced training set—1:2

Accuracy

Precision

Recall

Specificity

Decision trees

94.86 ± 0.58

95.67 ± 1.29

89.48 ± 0.79

97.79 ± 0.67

Random forest

98.74 ± 0.15

99.10 ± 0.17

97.33 ± 0.44

99.45 ± 0.09

DeepPPI

97.91 ± 0.38

98.37 ± 0.69

95.32 ± 1.40

99.21 ± 0.34

DeepFE-PPI

98.53 ± 0.20

98.41 ± 0.62

97.37 ± 0.31

99.16 ± 0.31

Struct2Graph

98.91 ± 0.24

99.17 ± 0.15

97.89 ± 0.17

99.52 ± 0.27

Method

MCC

F1-score

ROC-AUC

NPV

Decision trees

88.69 ± 1.28

92.47 ± 0.82

93.66 ± 0.58

94.47 ± 0.50

Random forest

97.25 ± 0.33

98.20 ± 0.22

99.71 ± 0.08

98.56 ± 0.24

DeepPPI

95.29 ± 0.85

96.81 ± 0.60

99.29 ± 0.25

97.70 ± 0.67

DeepFE-PPI

96.76 ± 0.45

97.88 ± 0.31

99.41 ± 0.17

98.59 ± 0.09

Struct2Graph

97.59 ± 0.51

98.43 ± 0.30

99.73 ± 0.18

98.87 ± 0.16

  1. Bold face numbers indicate the best performance