Skip to main content

Table 1 F1-score on CDS, transcript, and gene level for BRAKER1 (RNA-seq hints), BRAKER2 (protein hints of type (iii)), TSEBRA_EVM, EVM using comparable evidence, and TSEBRA (default hyperparameter) with hints generated by the BRAKER runs

From: TSEBRA: transcript selector for BRAKER

CDS level F1-score

 

BRAKER1

BRAKER2

EVM

TSEBRA _EVM

TSEBRA

A. tha.

81.87

84.01

84.41

86.21

86.90

B. ter.

76.12

72.84

  

77.80

C. ele.

85.87

81.13

86.14

85.13

84.48

D. mel.

79.82

76.79

79.67

79.89

81.66

D. rer.

74.00

72.23

  

78.40

M. tru.

71.46

75.11

  

80.98

P. tep.

68.61

63.90

  

67.96

P. tri.

78.32

83.40

  

87.60

R. pro.

53.54

54.49

  

56.30

T. nig.

53.95

57.97

  

58.70

X. tro.

74.96

75.89

  

79.44

Transcript level F1-score

 

BRAKER1

BRAKER2

EVM

TSEBRA_EVM

TSEBRA

A. tha.

53.78

56.63

57.32

61.35

62.00

B. ter.

33.15

26.49

  

35.02

C. ele.

53.30

42.71

52.76

54.46

55.94

D. mel.

51.33

46.94

49.90

53.76

55.18

D. rer.

24.99

22.17

  

33.43

M. tru.

39.04

44.09

  

51.72

P. tep.

26.14

18.04

  

28.89

P. tri.

47.04

55.96

  

62.31

R. pro.

12.84

12.65

  

15.22

T. nig.

5.74

7.93

  

9.78

X. tro.

22.88

23.84

  

31.83

Gene level F1-score

 

BRAKER1

BRAKER2

EVM

TSEBRA_EVM

TSEBRA

A. tha.

65.51

70.58

70.88

78.35

79.69

B. ter.

38.91

32.18

  

44.71

C. ele.

63.13

52.29

63.98

68.90

70.78

D. mel.

64.44

61.25

64.94

71.34

73.93

D. rer.

31.49

27.37

  

44.13

M. tru.

40.03

44.96

  

54.05

P. tep.

28.59

19.99

  

33.83

P. tri.

53.11

63.88

  

73.45

R. pro.

13.64

12.91

  

16.21

T. nig.

6.59

8.87

  

11.46

X. tro.

26.40

30.58

  

41.26

  1. For A. thal, C. ele., D. mel., a set of genome partitions, each totaling 90% of the genome size, was sampled for the evaluation of all methods. For all other species, the tests were run on the full genomes for BRAKER1, BRAKER2, and TSEBRA. (See Additional file 1: Table S1 for full species names and Additional file 1: Table S2 for the results with different protein sets.)
  2. The highest F1 score in each row is printed in bold face