Skip to main content

Table 14 Classification performance for different choices of length

From: Secondary structural entropy in RNA switch (Riboswitch) identification

Variable Length

     

Features

T P %

F P 1 %

F P 2 %

Sensitivity

Sig.

L,MFE,GC,RND

69.6

39.1

7.7

61

MFE,GC

L,MFE,GC,BJK

87.0

34.8

0.0

71.2

GC

L,MFE,GC

87.0

30.4

0.0

76.3

L,MFE

100

     

Features

TP%

FP 1 %

FP 2 %

Sensitivity

Sig.

MFE,GC,RND

69.6

26.1

7.7

66.1

-

MFE,GC,BJK

65.2

21.7

7.7

64.4

-

MFE,GC

56.5

21.7

15.4

64.4

-

150

     

Features

TP%

FP 1 %

FP 2 %

Sensitivity

Sig.

MFE,GC,RND

69.6

26.1

23.1

57.6

MFE,RND

MFE,GC,BJK

69.6

39.1

7.7

59.3

MFE

MFE,GC

69.6

39.1

0.0

64.4

-

200

     

Features

TP%

FP 1 %

FP 2 %

Sensitivity

Sig.

MFE,GC,RND

65.2

34.8

7.7

62.7

MFE

MFE,GC,BJK

78.3

34.8

7.7

66.1

MFE

MFE,GC

82.6

39.1

0.0

76.3

MFE

  1. Sub-section Variable Length refers to results of actual sequence lengths for both training and test sets (equal number of varying sequence lengths of 100, 150, and 200 from E. coli UTR chosen as negative set). Sub-sections 100, 150, and 200 refers to results where all sequences in the training and test sets have a constant length. Column Features denotes features used from the training set. T P % denotes percentage of true positives. F P 1 % and F P 2 % represent the percentages of antisense sequences and E. coli UTRs that are misclassified as riboswitches, respectively. Sensitivity denotes overall percentage of correctly classified sequences. Sig. denotes significant (less than 0.05 in the training set) features of the multinomial classifier.