Table 1 Mean test and train accuracies for prediction of protein–protein interaction interfaces.

From: Classification and prediction of protein–protein interaction interface using machine learning algorithm

PPI category

FNAT

NC > 0.80; NNC ≤ 0.25

NC > 0.80; NNC < 0.25 and ≤ 0.5

NC > 0.80; NNC < 0.50 and ≤ 0.80

NC > 0.80; NNC ≤ 0.80

Mean test accuracy (%)

Mean test AUC

Mean test accuracy (%)

Mean test AUC

Mean test accuracy (%)

Mean test AUC

Mean test accuracy (%)

Mean test AUC

Homo

99.28 ± 0.41

0.99 ± 0.00

98.78 ± 0.51

0.99 ± 0.00

97.50 ± 0.60

0.99 ± 0.00

98.52 ± 0.50

0.99 ± 0.00

Hetero

99.04 ± 0.49

0.99 ± 0.00

98.31 ± 0.71

0.99 ± 0.00

96.93 ± 0.961

0.99 ± 0.00

98.05 ± 0.55

0.99 ± 0.00

PPI category

iRMSD

NC < 5 Å; NNC ≥ 15 Å

NC < 5 Å; NNC ≥ 10 Å and < 15 Å

NC < 5 Å; NNC ≥ 5 Å and < 10 Å

NC < 5 Å; NNC ≥ 5 Å

Mean test accuracy (%)

Mean test AUC

Mean test accuracy (%)

Mean test AUC

Mean test accuracy (%)

Mean test AUC

Mean test accuracy (%)

Mean test AUC

Homo

97.55 ± 1.34

0.99 ± 0.00

97.12 ± 1.15

0.99 ± 0.00

94.90647 ± 1.14

0.98 ± 0.00

97.00 ± 0.92

0.99 ± 0.00

Hetero

96.66 ± 0.87

0.99 ± 0.00

96.82 ± 0.59

0.99 ± 0.00

96.83 ± 0.65

0.99 ± 0.00

97.37 ± 0.53

0.99 ± 0.00

  1. NC, native cutoff; NNC, non-native cutoff. Mean and SD were calculated from 100 randomized cross-validations using the 20% testing datasets.