Fig. 7

Evaluation of differential abundance analysis (DAA) methods based on 106 experimental datasets. a Heat map showing the numbers of significant taxa discovered by each DAA method in each dataset. Each row represents one dataset. The sidebars on the left show the sample size and taxa number for each dataset. The color scale for detection power is based on the standardized (scaled and mean centered) number of findings for each dataset. Datasets are hierarchically clustered based on Euclidean distance with the complete linkage. Box plots at the bottom show the distribution of the standardized number of findings across all datasets for each method. b Overlap of significant taxa (5% FDR) between DAA methods. Color and dot size indicate the percentage of overlap. Methods are hierarchically clustered based on Euclidean distance with the complete linkage. c The distribution of the percentage of taxa detected by at least one method at FDR 5%, 10%, and 20%. d The distribution of the observed false discovery rate (FDR) across the 106 real datasets when the group labels are randomly shuffled. The observed FDR level is calculated as the percentage of the 1000 repetitions making any false discoveries