Table 2 Comparison of observed and expected consistency in differentially abundant genera across five diarrhea datasets.

From: Microbiome differential abundance methods produce different results across 38 datasets

Tool

No. sig. genera

Max overlap

Mean exp.

Mean obs.

Fold diff.

p

ALDEx2

57

3

1.141

1.544

1.353

<0.001

limma voom (TMM)

76

4

1.22

1.618

1.326

<0.001

MaAsLin2 (rare)

74

3

1.204

1.595

1.325

<0.001

ANCOM-II

15

3

1.033

1.333

1.29

<0.001

MaAsLin2

79

3

1.215

1.557

1.281

<0.001

Wilcoxon (rare)

88

4

1.269

1.625

1.281

<0.001

metagenomeSeq

66

3

1.164

1.485

1.276

<0.001

Wilcoxon (CLR)

82

3

1.22

1.549

1.27

<0.001

limma voom (TMMwsp)

85

4

1.239

1.565

1.263

<0.001

t-test (rare)

62

3

1.145

1.403

1.225

<0.001

corncob

87

5

1.275

1.552

1.217

<0.001

DESeq2

82

4

1.246

1.512

1.213

<0.001

LEfSe

119

5

1.408

1.613

1.146

<0.001

edgeR

138

5

1.509

1.667

1.105

0.002

  1. No. sig. genera: Number of genera significant in at least one dataset; Max overlap: Maximum number of datasets where a genus was called significant by this tool; Mean exp.: Mean number of datasets that each genus is expected to be significant in (of the genera that are significant at least once); Mean obs.: Mean number of datasets that each genus was observed to be significant in (of the genera that are significant at least once); Fold diff.: Fold difference of mean observed over mean expected number of times significant genera are found across multiple datasets; p: p-value based on one-tailed permutation test that used the ‘Mean obs.’ as the test statistic. Note that <0.001 is indicated instead of exact values, because 0.001 was the minimum non-zero p-value we could estimate based on our permutation approach. Source data are provided as a Source Data file.
Back to article page