Additional file 8 of Comparison study of differential abundance testing methods using two large Parkinson disease gut microbiome datasets derived from 16S amplicon sequencing

Additional file 8: Table S7. Pairwise concordances between methods with summary statistics calculations. Differential abundance testing was performed using 23 methods on both unfiltered (A-C, 445 genera dataset1, 561 genera dataset 2), and filtered (D-F, 133 genera dataset 1, 195 genera dataset 2) taxonomic data. Values in heatmap cells are pairwise concordances of differential abundance signature calls between two methods. Concordances within each dataset (A,B,D,E) were calculated by first creating a binary taxa by method matrix denoting which methods did (value of 1) or did not (value of 0) detect a certain differential abundance signature, then between each pair of methods, calculating the proportion of detected differential abundance signatures in common out of the total detected between the two methods. Calculation of concordances for replicated differential abundance signatures (C,F) were calculated by first creating a binary taxa by method matrix that denoted whether a signature was replicated (multiple testing corrected significant and same case:control mean relative abundance ratio in both datasets) across datasets (value of 1), or was not replicated (value of 0), then between each pair of methods, calculating the proportion of replicated signatures in common out of the total replicated between the two methods. Cells are colored by a blue (lower concordance) to white (around the mean pairwise concordance of each heatmap) to orange (higher concordance) color gradient. Methods are ordered from lowest (bottom, left) to highest (top, right) mean pairwise concordance. SAMseq did not successfully run for dataset 2 unfiltered data, therefore, table values for SAMseq have been put as "NA" and colored grey in B and C. KW: Kruskal-Wallis; GLM: generalized linear model; CLR: centered log-ratio transformation; TSS: total sum scaling; rCLR: robust CLR with matrix completion using OptSpace algorithm; TMM: trimmed mean of M-values; GLM NBZI: generalized linear model assuming negative binomial distribution with, or without, zero-inflation; RLE: relative log expression.