Skip to main content
Fig. 3 | BMC Genomics

Fig. 3

From: Statistical evaluation of methods for identification of differentially abundant genes in comparative metagenomics

Fig. 3

Most methods have a biased p-value distribution under the null hypothesis. The p-value distributions on the Qin dataset with no added effect and a group size of 6 + 6 averaged over 100 resampled data sets. For the quantile-quantile-plots, each grey line represent a resampled metagenome, the solid black line represents the average value and the dotted line the line with slope one corresponding to a uniform p-value distribution. The panels correspond to edgeR (a), DESeq2 (b), overdispersed Poisson GLM (c), metagenomeSeq (d), metastats (e) and voom (f). (See Additional file 10: Figure S6 for the additional eight methods)

Back to article page