Table 8

Comparisons between the different methods on the unscored Gavin+Krogan network

Method

MCL

MCL-CA

MCL-CAw

COACH

CORE

CMC

HACO


#Predicted

242

219

310

447

386

113

278


Wodak

(#182)

#Matched

55

49

77

62

83

60

78

Precision

0.226

0.224

0.248

0.139

0.215

0.531

0.281

#Derived

62

49

77

49

83

60

85

Recall

0.338

0.269

0.423

0.269

0.456

0.330

0.467


MIPS

(#177)

#Matched

35

42

53

45

59

41

45

Precision

0.143

0.192

0.171

0.101

0.153

0.363

0.162

#Derived

40

42

53

38

59

41

57

Recall

0.226

0.237

0.300

0.215

0.333

0.232

0.322


Aloy

(#76)

#Matched

43

41

52

54

59

43

59

Precision

0.179

0.187

0.168

0.121

0.153

0.381

0.212

#Derived

42

41

52

37

59

43

59

Recall

0.556

0.539

0.684

0.487

0.776

0.566

0.776


Methods considered: MCL, MCL-CA, MCL-CAw, COACH, CORE, CMC and HACO. CMC performed the best in terms of precision, while HACO and CORE performed the best in terms of recall. MCL-CAw stood third among of the seven algorithms in both precision and recall. #Matched: #Predictions matching some benchmark complex(es). #Derived: #Benchmark complexes derived by some predicted complex(es).

The unscored Gavin+Krogan network

#Proteins 2964; #Interactions 13507

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

Open Data