Skip to main content

Table 5 Target identification results for 18 test sequences with SHAPE profiles from [8]

From: On an enhancement of RNA probing data using information theory

RNA

Length

Hajdin et al. [8]

Our method

Sens

PPV

Acc

Sens

PPV

Acc

Pre-Q1 riboswitch, B. subtilis

34

100

100

100

100

100

100

Fluoride riboswitch, P. syringae

66

93.8

93.8

93.8

100

100

100

Adenine riboswitch, V. vulnificus

71

100

100

100

100

100

100

tRNA(asp), yeast

75

95.2

95.2

95.2

100

100

100

tRNA(phe), E. coli

76

100

84.0

91.7

100

100

100

TPP riboswitch, E. coli

79

95.5

87.5

91.4

95.6

100

97.7

SARS corona virus pseudoknot

82

84.6

88.0

86.3

97.4

84.4

90.5

cyclic-di-GMP riboswitch, V. cholerae

97

89.3

86.2

87.7

100

96.6

98.3

5S rRNA, E. coli

120

85.7

76.9

81.2

100

97.3

98.6

M-Box riboswitch, B. subtilis

154

87.5

91.3

89.4

89.7

97.8

93.5

P546 domain, bI3 group I intron

155

94.6

96.4

95.5

98.2

100

99.1

Lysine riboswitch, T. maritima

174

87.3

88.7

88.0

94.8

100

97.3

Group I intron, Azoarcus sp.

214

92.1

95.1

93.6

96.5

96.5

96.5

Hepatitis C virus IRES domain

336

92.3

96.0

94.1

98.0

97.0

97.5

Group II intron, O. iheyensis

412

93.2

97.6

95.4

94.4

100

97.1

Group I Intron, T. thermophila

425

93.9

91.2

92.5

96.8

92.3

94.5

5′ domain of 23S rRNA, E. coli

511

97.2

76.8

86.4

98.7

83.4

90.4

5′ domain of 16S rRNA, E. coli

530

93.0

83.6

88.2

97.6

89.7

93.5

Average

 

93.0

90.4

91.6

97.6

96.4

96.9

  1. We consider the accepted secondary structure excluding pseudoknots as the target. Our method identifies the distinguished structure from a Boltzmann sample of \(2^{10}\) structures via 10 base-pair queries on the target. For each base-pair query, we consider the error rates of accepting or rejecting a base pair, \(e_0=0.05\) and \(e_1=0.01\). To compare the structure prediction approach in [8] and our method, we present three measures of performance: sensitivities (Sens), the fraction of pairs in the accepted structure that are predicted (\(\text {Sens}=\frac{\text {FP}}{\text {TP}+\text {FN}}\)); positive predictive value (PPV), the proportion of predicted pairs that are in the accepted structure (\(\text {PPV}=\frac{\text {FP}}{\text {TP}+\text {FP}}\)); and accuracy (Acc), the harmonic mean of Sens and PPV (\(\text {Acc}=\frac{2\cdot \text {Sens} \cdot \text {PPV}}{\text {Sens}+\text {PPV}}\)). The average accuracy of both methods is displayed in italics