A comparison of bioinformatic approaches for 16S rRNA gene profiling of food bacterial microbiota

The different pipelines that may be used in 16S rRNA gene profiling of bacterial communities are known to have a significant impact on alpha and beta diversity measures and this may prevent direct comparison of results obtained in studies using different bioinformatic approaches to analyse raw sequences. To evaluate the feasibility of meta-studies on food bacterial communities, we compared four analysis procedures, varying in OTU picking and taxonomy assignment strategies. A closed reference OTU picking resulted in the most divergent results in terms of both alpha and beta diversity, compared to open reference methods. Nevertheless, when OTUs were collapsed at the genus level, a high correlation was obtained among the estimated abundances of taxa for most studies. Aggregation of samples by their nature and occurrence of food spoilage or fermentation resulted in a very similar classification using two beta diversity analysis methods. We conclude that comparisons of data obtained from different studies are feasible at the genus level, when the same OTU picking strategy is used. Finally, we provide a new version of FoodMicrobionet (Parente et al., 2016), including data from 26 recent studies on food bacterial communities, together with R scripts allowing both the extraction of data in formats which can be used in several analysis tools (including the R package phyloseq and the Cytoscape app CoNet) and the statistical and graphical analysis using common alpha- and beta-diversity analysis methods.