Skip to main content
Fig. 2 | Genome Biology

Fig. 2

From: treeclimbR pinpoints the data-dependent resolution of hierarchical hypotheses

Fig. 2

The performance of methods on parametric simulated microbial data. a A toy tree to display a schematic example of three simulated scenarios: BS, US, and SS. Signal branches (i.e., with DA between groups) are in turquoise (decreased) and gold (increased), and larger points represent bigger change. b The average TPR and FDR (over 100 repetitions) of methods on three scenarios under different sample sizes (10, 25, and 50 per group) (see Additional file 1: Table S2 for a tabular format of the figure). Methods are in colors. Each method has three points that represent imposed FDR cutoffs at 0.01, 0.05, and 0.1. c DA branches identified by methods in one of the 100 repetitions. The tree has 956 leaves. The non-DA branches of the phylogenetic tree are represented with dashed lines to save space. Nodes identified by treeclimbR are labeled on the tree, and other methods are in Additional file 1: Fig. S2. OTU counts are shown in the heatmap with samples split by groups. Rows of the heatmap corresponding to dashed lines (consisting of many OTUs) are more gold (abundant). All OTUs (rows of heatmap) identified as DA by methods are summarized in the results panel. The results of simulations with varying signal branches are in Additional file 1: Fig. S9–14

Back to article page