Extended Data Fig. 7: DEBIAS-M inference yields biological insights into sequencing bias.

Analyses of a fitted DEBIAS-M model on the collection of HIV studies used in Figs. 2a, 4, with bias-correction factors for species not found in a certain study imputed to the largest observed factor across all datasets. a, Heatmap illustrating the presence (blue) and absence (orange) of each OTU across each of the HIV studies analyzed, displayed using agglomerative clustering (Methods). The OTU detection patterns of the different studies cluster according to the 16S region amplified. b, Adonis PERMANOVA explained variance and p values for the effect of different experimental factors (Supplementary Table 1) on these bias correction factors. c, PCA plot of the bias-correction factors inferred by DEBIAS-M, same as Fig. 4c, but with bias-correction factors for OTUs not found in a certain study imputed to the largest observed factor across all datasets. Color represents extraction kit type and shape the 16S rRNA region used. d, Box and swarm plots (Box, IQR; line, median; whiskers, nearest point to 1.5*IQR) showing the standard deviation of bias-correction factors for each of 17 studies (Supplementary Table 1), comparing those with manual and robotic processing. p, one-sided Mann-Whitney U test. e, Scatterplot showing the bias-correction factors inferred by DEBIAS-M plotted versus the 16S copy number of the same species.