Skip to main content

Table 3 Change in MCC and F1 of predictions on the full set of 392 proteins and the better labelled NI1 subset of 71 proteins for various machine learning algorithms (shown in Table 2 ) with or without feature selection

From: Transient protein-protein interface prediction: datasets, features, algorithms, and the RAD-T predictor

 

Without feature selection

With feature selection

 

â–µMCC

%â–µMCC

 

â–µF1

%â–µF1

 

â–µMCC

%â–µMCC

 

â–µF1

%â–µF1

 

LR

0.044

35.106

***

0.141

47.960

***

0.058

44. 183

***

0.150

50.43

***

BN

0.052

38.682

***

0.157

52.233

***

0.063

44.726

***

0.165

54.78

***

FT

0.029

21.798

*

0.140

47.035

***

0.057

40.724

***

0.160

52.48

***

RT

0.022

20.943

*

0.133

46.655

***

0.057

47.932

***

0.154

52.67

***

AT

0.048

35.068

***

0.154

51.107

***

0.059

41.926

***

0.158

52.05

***

avg

0.039

30.319

 

0.145

48.998

 

0.059

43.898

 

0.157

52.48

 
  1. Algorithm abbreviations in Table 1. Statistical significance was calculated using 1-sided Wilcoxon’s signed rank test. (* P < 0.05; ** P< 0.01; *** P < 0.001).