Table 18

Impact of augmenting inferred interactions on the performance of MCL, MCL-CAw, CMC and HACO

Method

PPI Network

#Predicted complexes

#Matched predictions

Precision

#Derivable benchmarks

#Derived benchmarks

Recall


MCL

G+K

242

55

0.226

182

62

0.338

I

50

2

0.040

31

3

0.097

G+K+I

249

55

0.221

189

58

0.307

ICD(G+K+I)

115

53

0.461

156

58

0.372

FSW(G+K+I)

89

54

0.607

141

61

0.433


MCL-Caw

G+K

310

77

0.248

182

77

0.423

I

42

2

0.048

31

3

0.097

G+K+I

315

78

0.247

189

78

0.412

ICD(G+K+I)

118

82

0.694

156

82

0.525

FSW(G+K+I)

95

84

0.884

141

84

0.596


CMC

G+K

113

60

0.531

182

60

0.330

I

10

3

0.300

31

5

0.161

G+K+I

119

60

0.504

189

63

0.333

ICD(G+K+I)

184

77

0.418

156

83

0.532

FSW(G+K+I)

186

74

0.398

141

80

0.567


HACO

G+K

278

78

0.281

182

85

0.467

I

12

2

0.167

31

2

0.064

G+K+I

309

78

0.252

189

84

0.444

ICD(G+K+I)

119

66

0.589

156

75

0.481

FSW(G+K+I)

98

61

0.622

141

70

0.496


Most algorithms showed marginal dip in performance on Gavin+Krogan+Inferred compared to Gavin+Krogan. However, upon scoring the augmented network, their performance was better compared to Gavin+Krogan. This indicated that inferred interactions were useful for complex detection provided affinity scoring is employed to reduce the impact of the noise present in them.

Srihari et al. BMC Bioinformatics 2010 11:504   doi:10.1186/1471-2105-11-504

Open Data