M ROSIKIEWICZ¹, KJ BURTON², G PIMENTEL, U BUTIKOFER, U VON AH U³, MJ VOIROL², S AEBY¹, G VERGÈRES³, N VIONNET², FP PRALONG², G GREUB<sup>1

1</sup>Institute of microbiology of the University of Lausanne and University Hospital, Lausanne, Switzerland, ²Service of Endocrinology, Diabetes and Metabolism, University Hospital, Lausanne, Switzerland, ³Institute of Food Science, Agroscope, Federal Office of Agriculture, Berne, Switzerland

The intestinal microbiota has been described as a dynamic ecosystem that can influence both health and disease. Recent developments in technologies such as 16S sequencing provided fast methods to assess the changes in the intestinal microbiota. Different methods exist for the processing of such data including automated and user-defined pipelines. In this work we evaluate three data processing approaches and their impact on downstream analyses conducted in a dietary intervention study. A randomised, cross-over study design was used to assess the impact of a probiotic yoghurt and milk acidified with a prebiotic (gluconic acid) on the gut microbiota of fourteen healthy male volunteers. Each product was consumed on a daily basis (400g/day) for a two-week period, with a three-week wash-out period separating the two test phases. Faecal samples were taken at eight time points during the study. 16S rRNA sequencing was performed using the Illumina MiSeq. Three different pipelines were used to obtain taxonomic assignment of reads: automated Illumina Metagenomic Application with Illumina-customized Greengenes database, QIIME closed reference approach and de-novo clustering using UPARSE algorithm (USEARCH package) with SILVA database. Changes in the microbiota were assessed by differential analysis using DESeq2. In the pooled data, differences were observed in the total number of taxa identified depending on the data processing method used. Notably, total number of genera that were identified using the three processing techniques was highly divergent with 616 genera in the Illumina processed data, 303 genera in the QIIME processed data and 208 genera obtained in USEARCH analysed dataset. The differential analysis showed differences depending on the processing approach used but was robust with respect to the expected increases in Lactobacillus and Streptococcus after probiotic yoghurt intake and increased in Bifidobacteria after prebiotic acidified milk. The differences in taxa abundance associated with data processing methodology appears to be linked to the more stringent criteria and better pre-processing of the reads applied during analysis with customisable pipelines.