Skip to main content

Table 1 Comparison of the four methods using the three different evaluation schemes for classification

From: Towards sequence-based prediction of mutation-induced stability changes in unseen non-homologous proteins

Method

Evaluation

MCC

Q2

Se

Sp

PPV

NPV

AUC

SEQ-NEIGHB

cross-validation ∗

0.41

77.65

55.09

85.57

57.35

84.45

0.78

 

unseen-mutation

0.36

75.80

51.16

84.48

53.74

83.08

0.78

 

unseen-residue

0.14

71.65

21.28

89.42

41.52

76.30

0.67

 

unseen-protein

0.14

69.09

29.45

83.13

38.21

76.89

0.67

SEQ-FREQ

cross-validation ∗

0.33

72.63

57.37

77.98

48.00

83.92

0.75

 

unseen-mutation

0.29

69.27

58.64

73.02

43.36

83.36

0.74

 

unseen-residue

0.22

69.30

43.40

78.44

41.54

79.70

0.68

 

unseen-protein

0.18

65.71

47.70

72.09

37.71

79.55

0.66

EASE

cross-validation ∗

0.45

78.54

60.95

84.72

58.33

86.08

0.81

 

unseen-mutation

0.41

77.05

57.91

83.80

55.74

84.96

0.81

 

unseen-residue

0.26

73.26

37.63

85.84

48.40

79.59

0.72

 

unseen-protein

0.23

71.27

39.76

82.44

44.50

79.44

0.70

EASE-AA

cross-validation ∗

0.43

76.69

63.35

81.37

54.43

86.35

0.80

 

unseen-mutation

0.38

74.71

60.64

79.66

51.23

85.17

0.79

 

unseen-residue

0.34

73.23

56.11

79.27

48.86

83.65

0.76

 

unseen-protein

0.35

73.24

58.79

78.36

49.04

84.30

0.77

  1. ∗ cross-validation folds were created by dividing mutations randomly (unseen-mutation cross-validation)