Comparison of Mothur and QIIME for the Analysis of Rumen Microbiota Composition Based on 16S rRNA Amplicon Sequences

Background: Microbiome studies need to analyze massive sequencing data, which requires the use of sophisticated bioinformatics pipelines. Up to date, several tools are available, although the literature is scarce on studies that compare the performance of different bioinformatics pipelines on rumen microbiota when 16S rRNA amplicons are analyzed. The impact of the pipeline on the outcome of the results is also unknown, mainly in terms of the output from studies using these tools as an intermediate phenotype (pseudophenotypes). This study compares two commonly used software (Quantitative Insights Into Microbial Ecology) (QIIME) and mothur, and two microbial gene data bases (GreenGenes and SILVA) for 16S rRNA gene analysis, using metagenome read data collected from rumen content of a cohort of dairy cows. Results: We compared the relative abundance (RA) of the identified OTUs at the genus level. Both tools presented a high degree of agreement at identifying the most abundant genera: Bifidobacterium, Butyrivibrio, Methanobrevibacter, Prevotella, and Succiniclasticum (RA > 1%), regardless the database. There were no statistical differences between mothur and QIIME (P > 0.05) at estimating the overall RA of the most abundant (RA > 10%) genera, either using SILVA or GreenGenes. However, differences were found at RA < 10% (P < 0.05) when using GreenGenes as database, with mothur assigning OTUs to a larger number of genera and in larger RA for these less frequent microorganisms. With this database mothur resulted in larger richness (P < 0.05), more favorable rarefaction curves and a larger analytic sensitivity. These differences caused significant and relevant differences between tools at identifying the dissimilarity of microbiotas between pairs of animals. However, these differences were attenuated, but not erased, when SILVA was used as the reference database. Conclusion: The findings showed that the SILVA database seemed a preferred reference dataset for classifying OTUs from rumen microbiota. If this database was used, both QIIME and mothur produced comparable richness and diversity, and also in the RA of most common rumen microbes. However, important differences were found for less common microorganisms which impacted on the beta diversity calculated between pipelines. This may have relevant implications at studying global rumen microbiota.