Beyond Biodiversity: Incorporating Uncertainty Into Metabarcoding Data for Improved Inference of Ecological...

Metabarcoding sequence data from environmental DNA (eDNA) is rapidly expanding as a powerful method for biodiversity surveys. In order to interpret these data, tools are needed that account for the uncertainty associated with eDNA sampling, sequencing and analysis. The data resulting from eDNA marker gene analysis differ from many traditional methods of biodiversity surveys because they are highly complex, sparse and compositional. Methodological biases produce uncertainty at every step of the sampling and sequencing process. Thus, it is critical that users have a way of interpreting eDNA results that accounts for their compositional nature and models the uncertainty resulting from factors like patchy sampling, PCR amplification biases and variable sequencing depth. Here, we introduce MAMBO: Metabarcoding Analysis using Modeled Bayesian Occurrences. MAMBO simulates in silico replication and models the uncertainty surrounding the sequencing and analysis process. Further, it uses these modelled sequence count data to correlate two sets of marker genes with a Bayesian regression, facilitating the linkage of different groups targeted by these assays. Compared with correlational network analyses, MAMBO overcomes many of the limitations to robust statistical analyses of eDNA marker gene data and provides an opportunity for new insight into ecological patterns over space and time.