Skip to main content

Table 2 LOOCV performance of the individual models

From: Integrative approach for detecting membrane proteins

Encoding

ML algorithm

Sensitivity

Specificity

Accuracy

MCC

AAC

OET-KNN

71.34

81.08

76.28

0.5271

KNN

75.72

74.87

75.29

0.5058

SVM

70.96

83.47

77.30

0.5492

GBM

71.86

83.75

77.89

0.5606

RF

68.11

85.13

76.73

0.5409

PseAAC

OET-KNN

73.05

81.38

77.27

0.5465

KNN

74.24

79.38

76.84

0.5370

SVM

70.59

83.98

77.37

0.5511

GBM

74.99

86.07

80.60

0.6149

RF

68.84

84.86

76.95

0.5446

PAAC

OET-KNN

68.94

72.09

70.53

0.4105

KNN

72.96

66.26

69.57

0.3930

SVM

76.15

84.22

80.24

0.6060

GBM

71.33

85.01

77.84

0.5661

RF

71.00

81.67

76.41

0.5301

SAAC

OET-KNN

66.63

72.88

69.80

0.3960

KNN

69.75

68.81

69.28

0.3856

SVM

72.51

85.85

79.27

0.5895

GBM

73.90

85.95

80.00

0.6034

RF

67.82

87.02

77.54

0.5595

Pse-PSSM, \(\lambda =0\)

OET-KNN

86.57

92.75

89.70

0.7953

KNN

85.22

90.44

87.86

0.7580

SVM

83.23

90.05

86.68

0.7350

GBM

83.41

90.45

86.98

0.7409

RF

79.45

92.53

86.08

0.7269

Pse-PSSM, \(\lambda =1\)

OET-KNN

85.92

91.79

88.89

0.7788

KNN

85.89

89.06

87.50

0.7501

SVM

86.75

92.22

89.52

0.7912

GBM

85.00

92.19

88.64

0.7744

RF

79.86

93.66

86.85

0.7433

Pse-PSSM, \(\lambda =2\)

OET-KNN

85.51

91.90

88.75

0.7762

KNN

85.65

88.28

86.98

0.7397

SVM

86.83

92.06

89.48

0.7904

GBM

84.86

91.72

88.34

0.7682

RF

79.80

93.70

86.84

0.7432

  1. This table shows microaverage LOOCV performance of the different protein encodings on different machine learning algorithms. The SAAC with SVM, highlighted in italics, reflects the LOOCV performance of the iMem-2LSAAC method [3] on DS-M. Only the Pse-PSSMs where \(\lambda \in (0, 1, 2)\) are shown here; the complete performance of all the Pse-PSSMs (\(\lambda \in (0, \ldots , 49) )\) can be found in Additional file 1