Which DNA Marker Best Identifies Marine Fish? A Landmark Benchmark Puts eDNA Barcoding to the Test

Environmental DNA metabarcoding has rapidly become one of the most transformative tools in aquatic ecology. By sampling water and sequencing the genetic traces that fish shed into it, scientists can inventory biodiversity without ever hauling in a net or disturbing a fragile habitat. Yet behind the promise lies an unresolved technical question with real consequences for fisheries management: which genetic marker, and which fragment of that marker, actually delivers reliable species-level identifications? A new study published in Metabarcoding and Metagenomics provides one of the most comprehensive answers to date, benchmarking the discrimination power of the three markers most commonly used in marine fish (e)DNA (meta)barcoding — 12S rRNA, 16S rRNA and cytochrome oxidase subunit I (COI) — along with two widely used amplicons for each.

The research team, led by investigators mining public sequence data from NCBI GenBank, analysed sequences spanning ten orders of ray-finned fishes (Actinopterygii), including many commercially and ecologically important species. Their approach rested on two complementary datasets. The first drew on mitochondrial genomes, allowing the extraction of all nine target regions — the full 12S, 16S and COI genes plus the MiFish-U, Teleo, Berry-Fish, Ac16S, Folmer and Leray-Lobo amplicons — from the same individual organisms. This design eliminated a major confound: differences in reference sequence quality between markers could not distort the comparison, because every fragment came from identical source DNA. The second dataset used independently deposited sequences for each region, totalling thousands of species, to reflect what practitioners actually encounter when querying environmental reads against real-world reference libraries.

To quantify discriminatory power, the researchers constructed Neighbour-Joining trees for each fragment and measured the percentage of species that emerged as monophyletic — meaning all sequences of a species clustered together in a single clade. Monophyly is a stringent and intuitive criterion: if a marker cannot keep a species’ sequences together, downstream taxonomic assignment of environmental reads risks false positives and false negatives. The mitochondrial genome dataset comprised 2,094 sequences across 771 species, while the independent-sequence analysis expanded to 3,879 species, with some regions, such as Leray-Lobo, represented by nearly 50,000 locally aligned sequences.

The results from the mitochondrial genome analysis were striking. Among the metabarcoding amplicons, the COI Folmer and Leray-Lobo regions achieved the highest discriminatory power, with 89.2% and 87.0% of species monophyletic, respectively. At the opposite end, the 12S Teleo region — a fragment averaging under 70 base pairs — managed only 71.6%, the lowest of any amplicon tested. Full-length genes generally outperformed their shorter derivatives, as expected, since longer fragments harbour more informative sites. Yet the pattern was not simply a matter of length: the full 12S gene, roughly 950 base pairs, was outperformed by the shorter Folmer region (~658 bp) in every fish group, and even by the 313-base-pair Leray-Lobo fragment in three of five groups. The evolutionary rate of the region, not just its span, clearly matters.

That conclusion was reinforced by the congeneric divergence analysis. COI regions showed the highest average pairwise distances between species of the same genus, consistent with its status as a fast-evolving protein-coding gene. The authors attribute part of COI’s advantage to the accumulation of synonymous substitutions, which purifying selection tolerates more readily than amino-acid-changing mutations. Intriguingly, the Teleo region also displayed high congeneric divergence — it is a highly variable stretch of DNA — but its brevity prevents it from accumulating enough informative sites to build a strong phylogenetic signal. High variability in a very short fragment, it turns out, is no substitute for adequate length.

When the researchers turned to the independent-sequence datasets, the rankings inverted dramatically. Now the Ac16S amplicon led with 83.0% monophyletic species, while the Folmer and Leray-Lobo regions fell to the bottom, at 64.8% and 63.5%. The explanation lies not in the markers themselves but in the composition of public databases. COI’s enormous sequence volume — tens of thousands of records and the highest number of sequences per species — means it inevitably accumulates more misidentified or mislabelled entries. A single outlier sequence, likely stemming from a taxonomic error, can break a species’ monophyly. The authors documented cases such as the grouper Epinephelus areolatus, whose 79 sequences would have been perfectly monophyletic but for one likely misidentification. Because GenBank is frequently the direct source of reference data in eDNA workflows, these database-quality effects translate directly into real-world identification errors.

The study also probed why species fail to discriminate, categorising non-monophyly into species sharing identical haplotypes, sequences intruding into another species’ clade, outliers clustering away from their own clade, or combinations thereof. Short fragments such as Berry, Teleo and MiFish-U showed the highest proportions of non-monophyletic species sharing haplotypes with other species — a signature of genuinely insufficient resolution. In contrast, the COI regions showed elevated shares of species with both intruders and outliers, pointing more towards data-quality problems than intrinsic biological limits. This distinction matters: reference library gaps and errors can be fixed through curation, but a fragment that simply lacks informative sites cannot be improved by more data.

Biology itself imposes hard limits in some cases. The researchers found species that no mitochondrial marker could resolve, including the shads Alosa alosa and Alosa fallax, which hybridise extensively, and several tuna species in the genus Thunnus, where introgression and taxonomic ambiguity blur species boundaries. Scomber japonicus and Scomber colias were likewise indistinguishable across all nine regions, possibly compounded by outdated labels in sequence databases that predate the clarification of their geographic distributions. These cases underscore that mitochondrial DNA alone cannot untangle recently diverged or interbreeding species, regardless of the fragment chosen.

The practical implications are significant for anyone designing an eDNA survey. The authors argue that amplicon choice should weigh species discriminatory power alongside technical considerations such as primer specificity, amplification efficiency and fragment preservation in environmental samples. COI, despite its superior resolution, suffers from co-amplification of prokaryotes and non-target eukaryotes, reducing the fraction of usable reads; 12S and 16S amplicons like MiFish-U and Berry offer better fish specificity and shorter fragments that survive better in degraded environmental DNA. For species-poor systems where the fauna is known in advance, a single well-chosen amplicon may suffice. But in species-rich waters teeming with close congeners — precisely the conditions where fisheries stock assessment demands precision — the study advocates a multi-marker approach as standard practice, using one fragment’s results to rectify ambiguities in another’s.

Ultimately, the research delivers a sober but constructive message: eDNA metabarcoding’s effectiveness should not be overestimated when it relies on a single marker, especially under high diversity. Gaps in reference coverage will narrow as sequencing efforts continue, but differences in intrinsic discriminatory power are rooted in biology and will persist. By benchmarking nine genomic regions across hundreds of species and ten fish orders, this study gives the field the evidence base it needs to move from ad hoc marker choices towards consensus guidelines — a crucial step if environmental DNA is to fulfil its potential as a backbone of marine biodiversity monitoring and fisheries management.

Subject of Research: Species discrimination power of mitochondrial DNA markers and amplicons in marine fish eDNA metabarcoding

Article Title: Benchmarking the discrimination power of commonly used markers and amplicons in marine fish (e)DNA (meta)barcoding

Article References: Fontes, J. T., Katoh, K., Pires, R., Soares, P., & Costa, F. O. (2024). Benchmarking the discrimination power of commonly used markers and amplicons in marine fish (e)DNA (meta)barcoding. Metabarcoding and Metagenomics, 8, Article e128646. https://doi.org/10.3897/mbmg.8.128646

Image Credits: AI Generated

DOI: 10.3897/mbmg.8.128646

Keywords: eDNA, metabarcoding, marine fish, COI, 12S rRNA, 16S rRNA, species discrimination, monophyly, reference databases, mitochondrial genomes, fisheries management, amplicon selection