Discussion
Whether used alone or in combination with existing reference databases, FishCARD reference barcodes dramatically improve the accuracy of eDNA metabarcoding assignments from California Current coastal waters, including species for recreational and commercial fishing and marine ecosystem assessments (Allen & Horn, 2006; Pondella et al., 2015; Sprague et al., 2013). In a test eDNA dataset from 3 sites on Santa Cruz Island, FishCARD performed better, identifying ASVs to species for an additional 15 California Current fishes that were not identified by theCRUX-12S database. This increase in accuracy greatly improves the utility of eDNA for monitoring California Current coastal ecosystems, echoing previous research on the importance of complete reference databases in metabarcoding (Leray, Boehm, Mills, & Meyer, 2012; Machida et al., 2017).
Unexpectedly, almost half of the ASVs and a quarter of all sequences generated in our eDNA test datasets could not be assigned to species. While other metabarcoding studies report similar results (Leray & Knowlton, 2017) (Supplemental Table 1), further investigation showed that the vast majority of unassigned ASVs were not fish. Instead, they mapped to uncultured bacteria 16S loci derived from marine shotgun sequencing metagenomic studies (Bork et al., 2015). This unexpected result highlights that the MiFish Teleost 12S primer set, while extremely useful for targeting vertebrate 12S loci, can also amplify non-target 16S genes, potentially inflating the number of ASVs unassigned to species using this primer set.