Skip to main content

Table 6 The prediction performance of different methods on the EVA benchmark datasets.

From: Improving protein secondary structure prediction based on short subsequences with local structure similarity

EVA_Set1 (80 proteins)

Q 3

ERRsig Q 3

sov

ERRsig sov

sovH

sovE

sovC

SymPred

78.8

±1.4

76.4

±1.9

85.0

76.5

70.4

SAM-T99sec

77.2

±1.2

74.6

±1.5

80.9

72.5

71.2

PSIPRED

76.8

±1.4

75.4

±2.0

82.1

72.3

65.2

PROFsec

75.5

±1.4

74.9

±1.9

78.3

75.9

71.3

PHDpsi

73.4

±1.4

69.5

±1.9

73.7

73.9

65.2

EVA_Set2 (212 proteins)

Q 3

ERRsig Q 3

sov

ERRsig sov

sovH

sovE

sovC

SymPred

79.2

±0.9

76.0

±1.2

85.1

77.7

71.3

PSIPRED

77.8

±0.8

75.4

±1.1

80.6

72.6

70.4

PROFsec

76.7

±0.8

74.8

±1.1

79.2

76.2

71.8

PHDpsi

75.0

±0.8

70.9

±1.2

77.0

72.4

67.0

  1. sovH/E/C values are the specific SOV accuracies of the predicted helix, strand and coil, respectively. The prediction results of other methods on EVA_Set 1 and EVA_Set2 are reported at http://cubic.bioc.columbia.edu/eva/sec/common3.html.