Improving plant DNA metabarcoding accuracy with ecological filters and Angiosperms353: Field and pollen microscopy validation.

Premise: Metabarcoding has become a successful tool for the identification of species in ecological assemblages. However, the usefulness of metabarcoding for identifying plant species has been hampered due to a lack of universal gene regions that work across all taxa, limiting the applications of metabarcoding in ecology. Methods: Here, we outline a spatiotemporal approach that combines Angiosperms353 baits with species distribution models and phenological analyses to generate a list of candidate species to increase metabarcoding accuracy. To evaluate the ecological realism of our framework, we compared the results of DNA metabarcoding pollen loads of wild bumble bees to long-term field observations of bee-plant interactions and visual pollen identification. Results: We show that metabarcoding bumble bee pollen loads was most accurate when combined with a candidate taxa list of plants flowering when the bumble bees were foraging, which improved the accuracy and taxonomic precision of 77.5% of samples relative to non-filtered matches. Discussion: With the proliferation of species occurrence and phenology data and advances in computing and software, spatiotemporal filtering provides an improved approach for interpreting metabarcoding results. Additionally, we demonstrate that Angiosperms353 offers significant promise for metabarcoding projects to reveal species interactions.