Table 1

From: Correction: Corrigendum: Network deconvolution as a general method to distinguish direct dependencies in networks

Regulatory network application

Reported ND performance improvement

Actual ND performance improvement

Average ND improvement over mutual information and corr. in DREAM5 overall (p. 729, paragraph 1 and Fig. 2 legend)

59%

66%

Average ND improvement over mutual information and corr. in DREAM in silico (p. 729, paragraph 1)

53%

61%

ND improvement in DREAM5 over ARACNE (p. 729, paragraph 1)

75%

87%

ND improvements in DREAM5 over community network (p. 729 paragraph 2 and Fig. 2 legend)

22%

25%

Protein folding application

Reported ND discrimination

Actual ND discrimination

ND discrimination in protein application (p. 730, Fig. 3b correction explained below)

15% (fivefold improvement)

22% (sevenfold improvement)

Social network application

Reported correlation with Newman weight

Actual correlation with Newman weight

Correlation coefficient between ND weights and Newman weights (p. 731, paragraph 1 and Fig. 4 legend)

R2 = 0.76

R2 = 0.74

Back to article page