Cindy Boer

The Gut Microbiome of Childen and Adults | 221 5.1 and decrease diversity in the samples. We, therefore, analyzed the effects of stool sam- ple collection at ambient temperature in both cohorts. Within our own initial datasets, we addressed this issue by using several approaches and observed increased abundanc- es of Escherichia/Shigella when stool samples were mailed to the research laboratory and received after 3 days (RS) or after 5 days (GenR) in the post. Therefore, to avoid the influence of any delayed processing time, we excluded samples that had been in the mail for longer than 3 days in RS and 5 days in GenR cohorts. Next to the increase in abundance of Escherichia/Shigella upon prolonged times in the mail, we observed several taxa that decreased over time ( Coprococcus and Rose- buria ). These decreases were, however, relatively low and likely a consequence of the compositionality of the data: if one OTU increases, other OTUs decrease. We decided to adjust for these technical artifacts in further analyses by including time in mail as a technical cofactor. The actual explanation for the difference in microbiota composition stability over time between RS and GenR cohorts is likely due to the fact that GenR participants were asked to keep their samples in their home fridge at 4 °C (for post- ing on Monday) if they were produced at the weekend, allowing better preservation of samples compared to RS participants who mailed their samples over the weekend. As well as the exclusion of 3 day and 5 day samples, we also included TIM as a technical covariate in all analyses. Next to excluding samples that had been in the mail for too long, we also excluded recent antibiotic users from our datasets (196 samples from GenR and 7 samples from RS). Antibiotic use had a significant effect on alpha diversity and overall composition in GenR ( Supplementary Table 1 ), whereas in RS the number of users was too small to have a detectable effect. Probiotic use and travel abroad were also recorded, but we could not detect significant effects on alpha diversity and composition in both cohorts. Given the small effect size of probiotic use, this study might be underpowered to con- clude a lack of effect. Recent studies indicate that the method of DNA isolation is the main source of technical variance in microbiome studies[69-71]. We, therefore, analyzed the effect of DNA isolation throughout the 391 runs (134 runs for RS and 257 runs for GenR) that were performed in this study and observed a batch effect causing a reduction in aver- age DNA yield per sample in a proportion of the hundreds of runs performed. As we could not trace any clear cause for this technical artifact we introduced a “Batch” vari- able that allowed us to discriminate between low yield and the high yield runs. This “Batch” variable was significantly associated with overall profiles in the GenR and RS cohorts and was included as technical covariate in all analyses. Another, more general