ModularBoost: an efficient network inference algorithm based on module decomposition

BMC Bioinformatics

Table 6 Module identification evaluation of the DREAM5 networks

Methods	Module gold	DREAM5 E. coli	DREAM5 Yeast
Methods	Module gold	DREAM5 E. coli	Sisima	Macisaac
ICA-FDR	Minimal	0.206	0.102	0.107
	Strict	0.193	0.090	0.073
	Interconnected	0.199	0.102	0.094
ICA-FDR2	Minimal	0.203	0.093	0.099
	Strict	0.190	0.082	0.069
	Interconnected	0.189	0.094	0.087
ICA-zscore	Minimal	0.192	0.086	0.096
	Strict	0.202	0.070	0.064
	Interconnected	0.183	0.081	0.086
PCA decomposition	Minimal	0.101	0.047	0.047
	Strict	0.100	0.046	0.042
	Interconnected	0.097	0.047	0.048
K-means	Minimal	0.173	0.070	0.076
	Strict	0.178	0.059	0.055
	Interconnected	0.146	0.063	0.066

\(F_{rr}\) indexes were calculated by comparing the predicted and known modules. High \(F_{rr}\) values demonstrate high consistency between the predictions and module gold standards. Three types of gene modules, i.e. Minimal, Strict and Interconnected were taken into consideration. Highest values in each type of gene modules for networks were displayed in bold

ISSN: 1471-2105