Ruiz et al. (2024) - Fish eDNA metabarcodes' taxonomic resolution

Even though environmental DNA (eDNA) metabarcoding is revolutionizing biomonitoring, many critical steps remain unstandardized leading to arbitrary choices. Among them, choosing PCR primers and similarity thresholds for clustering can drastically influence inferences on biodiversity. For fishes, current<i> in silico </i>comparisons have generally focused on primer sets properties, but few have compared the taxonomic resolution for a comprehensive set of metabarcodes while more than 20 exists. In addition, these studies were hampered by genetic database biases, uncertain taxonomic reference, as well as unstandardized and simplistic taxonomic resolution definitions. To overcome these issues, we developed a robust framework based on the comparison of metabarcodes extracted from the same mitogenomes (all available for fishes in NCBI) to a standardized taxonomic reference baseline based on COI Barcode Index Numbers (BINs), allowing to quantify both false-positive (same taxon splitting) and false-negative (different taxa merging) as well as their determinants. Although each metabarcode exhibited various sensitivities to false-negative or false-positive errors, clustering threshold appears as the most important factor influencing biodiversity estimates, leading us to propose optimal similarity thresholds per metabarcode for taxonomic levels delineation (metabarcode gaps). In addition, we also found that taxonomic resolution significantly varied among gene, orders and community properties (diversity/redundancy), but not with metabarcode length as previously thought. Overall, our study strongly prevents against the blind choice of metabarcode and clustering threshold, which must be instead adapted the goal of each study. A set of commented R functions makes it easy to apply this taxonomic resolution evaluation to any other animal taxa.