Cindy Boer

The Gut Microbiome of Childen and Adults | 211 5.1 the technical covariates. Linear regression models intending to determine significantly different genera between both cohorts were performed in MaAsLin. During analysis we adjusted for BMI, sex, technical covariates (TIM and Batch), and multiple testing by FDR (q-value<0.05). To compare the functional metagenome of gut bacteria between children and adults, we used the PICRUSt (v.1.1.0) tool[45] to obtain predicted bacterial functions. HUMAnN2 (v0.99) was used to identify the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways[46]. For the identification of the specific pathway biomarkers distin- guishing the gut microbiome of the children from those of adults, we performed a linear discriminant effect size (LEfSe) analysis[47] with the default settings: α (ANOVA) =0.05 and logarithmic LDA (linear discriminant analysis) score=2.0. Results Stool microbiota 16S rRNA data generation in the GenR and RS cohorts Selection of subjects For GenR 4,959, and for RS 2,440 participants were invited to provide a stool sample. In total, 2,921 (response rate=59%) and 1,691 (response rate=69%) were received at Erasmus MC for GenR and RS stool samples, respectively ( Supplementary Figure 1 ). Excluding antibiotic users resulted in exclusion of 196 samples from GenR and 7 sam- ples from RS. In GenR, individuals who used antibiotics in the last year had a significant- ly lower microbiome diversity and altered microbiome composition ( Supplementary Table 1 ). For probiotic use and recent travelling activity outside the Netherlands we could not detect significant effects on diversity or composition in the two cohorts ( Sup- plementary Table 1 ). After quality control ( Supplementary Figure 1 ), 16S rRNA data of 2,214 subjects in the initial dataset of GenR and 1,544 subjects in the initial dataset of RS were included. Assessing the influence of technical co-variates and sample exclusion Since stool samples were collected at ambient temperature via regular mail, we as- sessed the effects of time in the mail (TIM) on the microbiome profiles. Furthermore, since DNA-yields varied across different DNA isolation batches ( Supplementary Fig- ure 3 ), we assessed its effect on the microbiome profiles as well. Upon longer periods of TIM (except for day 7 in GenR), an increase in the relative abundance of phylum Proteo- bacteria was observed in the average profiles of both cohorts ( Supplementary Figure 4 ). For RS only, an increase in phylum Bacteroidetes was observed between days 6 and