Skip to main content

Table 1 Comparison of performance of pre-trained embedding methods for amino acids for sub-classifiers \(f_{5}, f_{10}, f_{15}\)

From: Protein–protein interaction prediction based on ordinal regression and recurrent convolutional neural networks

Embedding

Accuracy (%)

Precision (%)

Sensitivity (%)

Specificity (%)

\(F_{1}\)Score (%)

Sub-classifier \(f_{5}\)

\({\mathbf{a}}_{co}\)

90.35

93.25

94.42

79.61

93.83

\({\mathbf{a}}_{eh}\)

90.78

93.94

94.21

79.60

94.08

One-hot

89.94

93.28

93.81

77.95

93.55

Sub-classifier \(f_{10}\)

\({\mathbf{a}}_{co}\)

89.15

89.24

89.90

89.05

89.57

\({\mathbf{a}}_{eh}\)

88.86

88.85

89.77

88.88

89.31

One-hot

88.39

88.55

89.12

88.22

88.83

Sub-classifier \(f_{15}\)

\({\mathbf{a}}_{co}\)

96.17

93.03

92.10

97.25

92.57

\({\mathbf{a}}_{eh}\)

95.51

90.63

92.18

97.25

91.40

One-hot

95.92

92.17

92.07

97.23

92.11

  1. The values in each column represents the experimental results for each criterion of performance. Maximal values in each column is shown in bold