Fig. 1

Microbiota composition in a mock community and human milk samples using a clustering-based method (Qiime1) and a denoising algorithm (Qiime2) with and without contaminant removal. a Schematic of the study design. b Composition of the mock community by Qiime1 and Qiime2 prior to contaminant removal (each dataset = combined data from 8 replicates). c Comparison of milk microbiota richness (observed OTUs/ASVs) and diversity (inverse Simpson index) between Qiime1 and Qiime2 with and without contaminant removal. d Correlation of the relative abundances of milk genera between Qiime1 and Qiime2 prior to contaminant removal (See also Figures. S2 and S3 and Tables S2 and S3) (each dataset = combined data from 393 milk samples). Each dot represents a classified genus. Contaminant removal doesn’t impact the associations (not shown). e Comparison of the composition of abundant families (> 1% mean relative abundance) between Qiime1 and Qiime2 with or without contaminant removal. Contaminant removal reduced the relative abundance of certain low-abundance taxa (e.g. Caulobacteraceae and Rhodospirillaceae) and proportion of Other taxa (OTUs with less than 1% mean relative abundance) estimated by Qiime1, but generally did not affect the microbiota profile estimated by Qiime2. f Agreement and consistency between methods by intraclass correlation for alpha diversity and 13 most abundant families. * p < 0.05, *** p < 0.001