Estimating absolute DNA abundance from replicated metabarcoding

Datasets derived from mixed‐template DNA amplification and sequencing (‘metabarcoding’) are common in environmental DNA (eDNA), microbiome or microbial ecology analyses. Such metabarcoding data have been broadly used to inform estimates of biodiversity, distribution and abundance across a wide range of species and habitats. Generally, data from metabarcoding consist of a table of read‐counts assigned to each of a set of taxa or sequence variants. As such, they appear to contain quantitative information about the environment under study, immediately eligible for traditional community ecology analyses. In fact, metabarcoding data are compositional, can be biased by variation in PCR amplification rates among taxa, and have other statistical attributes that are distinct from many other common ecological observations. The difficulty in understanding and accounting for the complexity of metabarcoding observations has hindered their ecological interpretation and limited their integration with other ecological data types. Here, we use both simulations and empirical observations to show that the variability among replicate metabarcoding observations is closely tied to the absolute quantities of DNA present in a sample. We then develop a general statistical model for estimating DNA abundance from metabarcoding observations using the relationship between mean and variance of metabarcoding reads. We demonstrate that the model can effectively estimate DNA abundances using simulations and from experimentally created DNA samples. The model can be applied to a single marker or to multiple markers simultaneously, allowing the direct integration of information from multiple genetic markers derived from a common biological sample. Estimates of absolute DNA abundance are the logical unit to connect DNA observations to organismal abundance or biomass. Our methods can be used to link together multiple types of DNA observations or to connect DNA observations with non‐genetic ecological methods for surveying ecosystems. However, important aspects of the ‘ecology of DNA’ remain unresolved and may limit our ability to link observed concentrations to organismal abundances in a general way.