Skip to main content

Table 4 Experiments on the GR426 dataset with different input schemes.

From: Predicting β-turns and their types using predicted backbone dihedral angles and secondary structures

Input

l 1

l 2

MCC

Accuracy (%)

PSSM-only

7

0

0.369

69.7

PSSM-only

9

0

0.387

70.3

PSSM-only

11

0

0.387

69.9

PSSM + SS

9

1

0.404

72.2

PSSM + Dih

9

1

0.398

71.4

PSSM + SS + Dih

9

1

0.413

73.2

PSSM + SS + Dih

9

3

0.419

74.2

PSSM + SS + Dih

9

5

0.424

76.0

PSSM + SS + Dih

9

7

0.421

76.7

  1. Window sizes, l1 and l2, are the windows for PSSM values and predicted secondary structures and/or dihedral angles, respectively. PSSM-only: only scaled PSSM values are used in the input vector; PSSM + SS: the input vector is augmented with predicted secondary structures; PSSM + Dih: the input vector is augmented with predicted dihedral angles; PSSM + SS + Dih: the input vector is augmented with both predicted secondary structures and predicted dihedral angles.