Table 4 Table of previous operon prediction methods.

Predictor method

Data Types

Applied species

Operon Data

Sen (%)

Spe (%)

Accuracy

Probability

Current study, from Bowers et al., 2004

Intergenic distance

E. coli K12

2684 TU E. coli K12 Jun07

78

_

_

B. subtilis

(831 multiprotein operons)

(precision = PPV = 82% E. coli K12)

1823 pairs in same operon

770 TU E. coli K12 Jan06

75

_

_

(356 mulitprotein operons)

(precision = 47%)

905 pairs in same operon

94

_

_

1115 TU B. subtilis

(precision = 45% B. subtilis)

(419 multiprotein operons)

972 pairs in same operon

HMM

Yada et al., 1999

Sequence information

E. coli K12

390 TU

59

_

_

Naïve Bayes

Craven et al., 2000

Sequence information

E. coli K12

365 TU

75

91

83%*

Log likelihood

Salgado et al., 2000

Intergenic distance, functional classes

E. coli K12

361 TU (237 multi)

88

88

82%* Distance

572 pairs in same operon

88%* Both

346 pairs at TU border

Probability

Ermolaeva et al., 2001

Conserved gene clusters across 34 genomes

E. coli K12

389 TU

48

92

70%*

541 pairs in same operon

263 pairs at TU border

(pair if ≤ 200 bp apart)

HMM

Tjaden et al., 2002

Expression data

E. coli K12

463 pairs in same operon

63

99

81%*

Graph analysis

Zheng et al., 2002

Metabolic pathway information

E. coli K12 (also applied to 42 other genomes)

128 TU metabolism related

89

87

88%*

Log likelihood

Moreno-Hagelsieb et al., 2002

Intergenic distance

B. subtilis (trained on E. coli K12, applied to 68 genomes)

100 TU B. subtilis

88

88

88%* B. subtilis

310 pairs in same operon

82%* E. coli K12

123 pairs at TU border

Bayesian posterior probability

Sabatti et al., 2002

Intergenic distance, co-expression

E. coli K12

257 TU

82

70

76%* Co-expr

604 pairs in same operon

84

82

83%* Distance

151 pairs at TU border

88

88

88%* Both

Bayesian network

Bockhorst et al., 2003

Intergenic distance, sequence information, expression data

E. coli K12

365 TU

78

90

84%*

Machine learning Romero et al., 2004

Intergenic distance, functional information

B. subtilis

(trained on E. coli K12)

100 TU B. subtilis

446 TU E. coli K12

81

48

65%* B. subtilis

87

86

87%* E. coli K12

(91)

(87)

(89%* if use all info on E. coli)

Bayesian classifier

De Hoon et al., 2004

Intergenic distance, operon length, gene expression

B. subtilis

635 TU

82^¶

89^¶

83^¶ distance

582 pairs in same operon

80^¶

79^¶

80^¶ expression

91 pairs at TU border

88^¶

88^¶

89^¶ all

Machine learning without extensive training data

Westover et al.,2005

Intergenic distance, functional classes, conserved gene clusters

E. coli K12

(validated by known operons)

E. coli K12:

88

80

84%* E. coli K12

B. theta

(validated by co-expression)

797 pairs in same operon

294 pairs at TU border

B. theta:

73

80

76.5%* B. theta

936 concordant pairs

106 discordant pairs