Skip to main content

Table 3 The performance of ensemble PPI-BioBERT-x10 on the test and validation set

From: Large-scale protein-protein post-translational modification extraction with distant supervision and confidence calibrated BioBERT

Dataset

Interaction

P

R

F1

ECE

SD

Support

Test

Acetylation

100.00

100.00

100.00

0.49

0.25

1

Test

Dephosphorylation

50.00

16.67

25.00

0.67

0.40

6

Test

Methylation

25.00

25.00

25.00

0.60

0.28

4

Test

Phosphorylation

62.50

34.09

44.12

0.79

0.26

44

Test

Ubiquitination

0.00

0.00

0.00

–

–

1

Test

ECE

–

–

-

0.75

–

31

Test

Average SD

–

–

–

–

0.28

31

Test

Macro avg

47.50

35.15

38.82

–

–

56

Test

Micro avg

58.06

32.14

41.38

–

-

56

Val

Acetylation

100.00

100.00

100.00

0.53

0.16

1

Val

Dephosphorylation

66.67

20.00

30.77

0.61

0.37

10

Val

Methylation

0.00

0.00

0.00

0.53

0.29

1

Val

Phosphorylation

77.78

66.67

71.79

0.78

0.26

21

Val

Ubiquitination

0.00

0.00

0.00

–

–

1

Val

ECE

–

–

–

0.73

–

24

Val

Average SD

–

–

–

–

0.28

24

Val

Macro avg

48.89

37.33

40.51

–

–

34

Val

Micro avg

70.83

50.00

58.62

–

–

34

  1. ECE is the expected calibration error. SD denotes the average standard deviation within the ensemble