Skip to main content

Table 2 Prediction performance on different sliding windows for encoding input vectors on the CASP9 data set.

From: LigandRFs: random forest ensemble to identify ligand-binding residues from sequence information alone

Window length

All ligand sites

Partial ligand sites

 

Sen (%)

Spe (%)

Acc (%)

MCC

Prec (%)

F1 (%)

Sen (%)

Spe (%)

Acc (%)

MCC

Prec (%)

F1 (%)

5

52.97

93.82

91.76

0.36

36.24

34.85

58.32

93.05

91.73

0.32

24.86

29.40

7

56.96

93.21

91.31

0.37

34.97

35.99

54.80

93.38

91.99

0.32

27.80

30.45

9

56.43

92.66

90.76

0.35

32.31

34.66

49.93

95.42

93.69

0.32

28.75

29.95

11

58.40

91.67

89.94

0.35

31.05

33.85

50.95

95.10

93.46

0.32

28.20

30.41

17

62.44

91.35

89.72

0.36

29.61

34.42

47.37

96.95

94.95

0.34

32.39

33.09

27

46.66

95.85

93.34

0.34

34.13

34.93

49.75

96.37

94.51

0.33

29.11

32.04

37

45.48

96.53

93.90

0.36

37.37

36.00

56.00

94.58

93.00

0.32

25.74

30.95

47

48.55

96.17

93.67

0.37

37.63

36.95

55.15

95.12

93.44

0.33

26.82

31.42

57

41.75

96.93

94.18

0.35

39.69

35.81

44.24

96.47

94.42

0.31

30.30

29.78

Combine

42.07

97.91

95.06

0.40

47.85

38.93

48.34

97.02

95.07

0.34

32.80

33.18

  1. It contains prediction performance on the all ligand site group and the partial ligand site group. The ensemble of all sliding windows is shown at the last row of the table.
  2. The italic number denotes the best performance on the measure of MCC.