Skip to main content

Table 4 Resulting performance measures when best-feature sets were used in the classification tasks, using two different feature selection algorithms

From: A machine learning strategy for predicting localization of post-translational modification sites in protein-protein interacting regions

Dataset

Relief-F

ACC

AUC

MCC

PPV

Sn

Sp

#Features used

Acetylation

0.88

0.92

0.78

0.95

0.82a

0.95

144

Phosphorylation

0.91

0.93

0.83

0.99a

0.82

1.00a

73

Ubiquitylation

0.88

0.91

0.77

0.96

0.80a

0.96

512

 

Information Gain

Acetylation

0.88

0.90a

0.78

0.97

0.80a

0.97

82

Phosphorylation

0.91

0.93

0.84

0.99a

0.83

1.00a

35

Ubiquitylation

0.88

0.91

0.77

0.96

0.80a

0.96

343

  1. aExcept for these values, others were significantly increased (t-test, α < 0.05) when compared with the results using optimized sets of indices shown in Table 3