Impact of bioinformatic pipelines and reference databases on 16S rRNA gene sequencing in patients with cognitive impairment

Abstract

AbstractBackground Sequencing of 16S ribosomal RNA (rRNA) gene is the most used technique for the identification of Alzheimer’s specific differences in the relative abundance of specific organisms. Several bioinformatic pipelines have been developed so far to characterize complex gut microbial communities. This study aims at evaluating the effect of bioinformatic pipelines and reference databases on the taxonomic classification of bacteria in the stools of patients with cognitive impairment.MethodDNA from frozen stool samples of 40 subjects was extracted using QIAamp Fast DNA stool kit after mechanical disruption (bead‐beating) and amplified using primers targeting the V3‐V4 region of the bacterial 16S rRNA gene. Sequencing was performed according to the 16S‐protocol on MiSeq (Illumina). Sequence read files were analysed using 4 commonly used pipelines run on a Linux workstation: QIIME2 (Bokulich, Microbiome, 2018), Bioconductor (Callahan, F1000Research 2016, version 29 oct 2018), UPARSE (Edgar, Nat Methods, 2013) and mothur (Kozich, Applied and Environmental Microbiology, 2013). SILVA 132 and RDP v16 reference databases were used in all analyses. For UPARSE and mothur, sequences were clustered into OTUs at 99% identity threshold. Differences of relative abundances of phyla and genera were evaluated across pipelines using Friedman rank sum test.ResultThe identification of phyla but not their quantification was consistent among all the pipelines using both databases (p<0.001). At genus level, all pipelines identified Bacteroidetes, Alistipes, Faecalibacterium and Ruminococcus as the most abundant genera, regardless of the database applied for taxonomic assignment. Again, their relative abundance was different across pipelines and databases (p<0.001).ConclusionThe comparison between different studies should take analytical pipeline and database used for taxonomic assignation into account. To overcome these limitations, the scientific community should quickly converge towards a consensus pipeline for guarantee research reproducibility.