Skip to main content

Table 4 The cross-validated accuracy of dihedral prediction on CB513 dataset.

From: Prediction of backbone dihedral angles and protein secondary structure using support vector machines

CB513

 

PSSM-only

SSE-1st run

SSE-2nd run

SSE-3rd run

NC

EM (%)

k-Means (%)

EM (%)

k-Means (%)

EM (%)

k-Means (%)

EM (%)

k-Means (%)

2

81.4

81.7

81.8

82.1

83.2

83.4

81.8

83.5

3

79.3

79.6

79.7

79.8

81.2

81.1

79.6

81.2

4

78.7

74.5

79.0

74.4

80.5

76.1

79.0

75.8

5

65.0

63.8

65.2

64.1

66.9

65.3

65.2

65.0

6

63.7

59.2

63.8

59.3

65.5

60.4

63.7

60.1

7

56.5

54.6

56.8

54.7

58.3

56.0

56.8

55.4

8

53.8

53.7

54.0

53.8

55.4

55.1

53.9

54.6

9

53.8

51.1

54.0

51.0

55.3

52.3

54.0

51.7

10

52.9

50.2

53.1

50.3

54.5

51.6

53.0

51.0

11

50.3

48.5

50.6

48.5

51.8

49.7

50.6

49.1

12

47.0

41.2

47.2

41.5

48.4

42.3

47.2

42.1

  1. NC: the number of clusters,
  2. PSSM-only: input vector with only PSSM values,
  3. SSE: input vector augmented with predicted secondary structure elements.