Skip to main content

Table 4 Prediction accuracies for different numbers of species

From: Vertebrate gene finding from multiple-species alignments using a two-level strategy

 

Acceptors

Donors

Starts

Stops

Train set size

204,021

221,421

7,571

25,071

Eval set size

52,605

57,179

1,805

6,162

F scores (%)

    

   Human only

66.78

67.25

35.34

22.20

   Human+mouse

80.67

82.74

43.38

30.57

   All 4 mammals

82.53

83.99

44.02

31.88

   All 8 species

84.31

84.82

51.45

34.93

100-ROC (%)

    

   Human only

5.22

4.31

18.30

20.03

   Human+mouse

2.45

1.93

13.18

15.54

   All 4 mammals

2.21

1.81

11.77

14.75

   All 8 species

1.76

1.54

10.53

11.68

  1. The table shows the F score (geometric mean of sensitivity and specificity) and ROC error rate (area not under the ROC curve) for the horizontal component of Classifier Two trained on different numbers of informant species and running on the challenging evaluation (Eval) set. All scores are percentages.