Skip to main content

Table 4 Comparison of receiver operator characteristics (< = 10 false positives) and sensitivity for different fold recognition methods based on all LiveBench-2008.1 targets.a

From: DescFold: A web server for protein fold recognition

 

Receiver operator characteristics (< = 10 false positives)b

Sensitivityc

 

1

2

3

4

5

6

7

8

9

10

All

Trivial

Easy

Hard

FFASd

85

94

119

133

135

139

140

140

140

140

150

8

103

39

Inubd

73

89

106

116

120

121

121

121

121

121

134

6

91

37

Fugued

61

79

81

85

87

96

101

102

104

104

135

8

95

32

mGenThreaderd

77

89

89

90

90

93

97

97

98

98

143

8

97

38

3D-PSSMd

48

55

72

75

78

80

86

86

87

89

102

5

75

22

DescFolde

87

89

99

103

104

108

111

114

115

116

134

8

92

34

  1. a LiveBench-2008.1 contains 283 targets, which can be divided into 9 trivial, 109 easy and 165 hard targets. As defined by the developer of LiveBench, trivial targets means those proteins sharing strong sequence similarity to the other previously known structures, as measured by Blast using an e-value < 0.001. The division of easy and hard targets is based on whether a structural template can be identified by Psi-blast with an e-value < 0.001.
  2. b1-10: number of correct predictions with higher reliability than the 1-10 false prediction.
  3. c Number of correct predictions for all, trivial, easy and hard targets, respectively.
  4. dThe results for FFAS, Inub, Fugue, mGenThreader, and 3D-PSSM were cited from http://meta.bioinfo.pl/results.pl?comp_name=livebench-2008.1
  5. eThe performance was evaluated based on the number of correctly assigned folds. We considered two hits as similar, provided that the Z-Score obtained by applying the CE structural alignment algorithm [51] was > = 4.0.