Long Nanopore reads sharpen protist diversity surveys and cut sequencing noise

Protists — the microbial eukaryotes that drift through oceans, lakes and sediments — are among the most diverse organisms on Earth, yet their small size and stubborn resistance to laboratory culture have long frustrated efforts to catalogue them. Over the past two decades, environmental metabarcoding has transformed the field, allowing researchers to sequence DNA extracted straight from water samples and infer which protist species are present, where they live and how their communities change over time. Almost all of that work, however, has relied on short-read sequencing platforms that read only a small slice of the protist ribosomal RNA gene. A new study published in Metabarcoding and Metagenomics argues that the time has come to go longer, and it delivers both the evidence and the software to make that transition practical.

The research team, working with samples from Polish lakes and from the Mediterranean water column, compared the two sequencing strategies head to head. On the short-read side, they used Illumina MiSeq to sequence the V4 region of the 18S ribosomal RNA gene, the roughly 380-base-pair fragment that has become the de facto standard for protist metabarcoding. On the long-read side, they turned to Oxford Nanopore’s MinION platform, sequencing an amplicon of about 3,200 base pairs that spans the entire 18S gene, the internal transcribed spacers, the 5.8S gene and the beginning of the 28S gene. The comparison included a defined mock community of seven cultured protist species and complex environmental communities from ten freshwater and twelve marine samples.

Nanopore sequencing has long been viewed with caution in biodiversity research because of its higher error rate, which produces a substantial number of insertions and deletions that complicate the construction of clean taxonomic units. Denoising pipelines such as DADA2, which reliably separate biological signal from sequencing error in Illumina data, cannot model Nanopore’s less predictable error profile. The team’s answer is BaNaNA — the Barcoding Nanopore Neat Annotator — a Snakemake-based pipeline that clusters Nanopore reads into Operational Taxonomic Units rather than trying to denoise them. The pipeline basecalls reads with Dorado’s duplex super-accurate model, filters them for length and quality, extracts rDNA fragments with Barrnap, verifies the correct 18S-5.8S-28S structure with a custom Python script, and adjusts the clustering threshold to the average error rate of each sample. Consensus sequences are built with MAFFT alignments within clusters, polished with Minimap2 and Racon, screened for chimeras against the PR2 reference database and de novo, and finally clustered at 99 percent identity to produce the finished OTUs.

The mock community provided the clearest test of what each technology gets right and wrong. Illumina sequencing of the V4 region produced 268 ASVs with an average length of 379 base pairs, while Nanopore sequencing, processed through BaNaNA, yielded 147 OTUs averaging 1,843 base pairs. The Illumina ASVs gave a taxonomic composition that diverged substantially from the known community structure: dinoflagellates appeared most abundant even though they were added at low concentration, while Euglena gracilis, which made up nearly a third of the cells by count, was underestimated by three orders of magnitude. The full-length Nanopore OTUs tracked the expected composition far more closely, with haptophytes and euglenozoans appearing at roughly the right proportions. Crucially, the Illumina data also contained the highest number of false positives — sequences assigned to genera that were never part of the mock community — suggesting that the short-read denoising approach inflates apparent diversity with background noise.

Species-level identification told a similar story. Neither technology detected all seven mock species perfectly, but the full-length Nanopore OTUs achieved the highest number of species-level classifications. The advantage of sequence length became especially evident when the researchers extracted the V4 region from their Nanopore OTUs — creating so-called V4-tags — and compared species assignments. For Paramecium bursaria, the full-length OTUs placed the sequence correctly at species level, while many V4-tags could only be assigned to the genus level, causing the species’ apparent abundance to drop dramatically. The lesson is that longer fragments carry enough information for accurate species-level annotation, provided the reference database is adequate, while shorter fragments lose that resolution even when amplified with the same primers.

Environmental samples from the Masurian lakes and the Gulf of Lion confirmed the pattern in complex natural communities. Illumina sequencing recovered 1,668 ASVs from freshwater and 2,716 from marine samples, while BaNaNA processing of Nanopore data produced 1,068 freshwater and 1,230 marine OTUs averaging nearly 1,800 base pairs. At higher taxonomic levels, both technologies painted very similar pictures of community composition. In the lakes, dinoflagellates and perkinseans dominated, followed by ciliates, cryptophytes and chlorophytes, with little difference between sunlit and dark zones. In the marine samples, dinoflagellates and radiolarians prevailed, and the parasitic diplonemids were detected exclusively by Nanopore sequencing. Bray-Curtis beta diversity analysis grouped samples from both platforms into three distinct clusters corresponding to freshwater, marine photic and marine deep environments, indicating that the ecological patterns were robust regardless of which technology generated the data.

One of the study’s most striking findings concerns the sheer amount of undocumented protist diversity revealed by the long reads. The researchers identified 747 full-length 18S rDNA OTUs whose sequence identity to their closest PR2 reference fell between 80 and 97 percent — a range suggesting they represent taxa not yet in the database. Freshwater environments yielded 513 of these novel OTUs, representing 48 percent of all freshwater OTUs obtained, while marine samples contributed 234, or 19 percent of the marine total. Together, the novel sequences amount to 7.6 percent of all eukaryotic sequences currently in PR2 for freshwater and 3.5 percent for marine environments. Because the OTUs are full-length, they can be aligned and used to reconstruct phylogenetic trees. The resulting tree, built from 721 novel OTUs with IQ-TREE, grouped the sequences into clades spanning six supergroups and fifteen divisions, with most novel taxa belonging to perkinseans and dinoflagellates. The authors note that some of these sequences may represent entirely new clades, something impossible to detect with short amplicons that lack sufficient phylogenetic signal.

The study also highlights a persistent imbalance in protist reference databases. Freshwater samples showed roughly twice the proportion of low-identity sequences compared with marine samples — 41 percent versus 14 percent in the Nanopore data — a discrepancy the authors attribute directly to the underrepresentation of freshwater protists in PR2. Marine ecosystems have historically received more molecular attention, and the gap is visible at every taxonomic level. The researchers argue that filling this gap is essential for the future of long-read metabarcoding, since accurate annotation of full-length sequences requires databases with longer reference sequences. They also point out that Nanopore’s ability to sequence the entire rDNA operon will demand operon-spanning databases that are only now beginning to emerge.

Perhaps the most practically important result is the demonstration that short- and long-read data can be integrated. Because Nanopore OTUs cover the whole 18S gene, researchers can extract any primer-defined fragment — such as the V4 region — in silico and compare it directly with Illumina datasets generated years earlier with different primers. The V4-tags extracted from Nanopore OTUs proved highly comparable to Illumina V4 ASVs, opening the door to long-term biodiversity studies that combine vast archives of legacy Illumina data with new Nanopore surveys. The authors recommend retaining the taxonomic assignment from the full-length 18S rDNA before extracting any specific region, to fully exploit the resolution that long reads provide.

Taken together, the findings position Nanopore metabarcoding not merely as a cheaper complement to Illumina but as a reliable replacement for protist diversity studies. The full-length OTUs offer greater taxonomic resolution, richer phylogenetic information, fewer false positives and direct compatibility with existing short-read datasets. With BaNaNA freely available on GitHub and benchmarked on both mock and environmental communities, the study removes a key bioinformatic barrier that has kept long-read approaches on the margins of protist ecology. As Nanopore chemistry continues to improve — accuracy has risen from 97.6 percent with R9.4.1 flow cells to 99.1 percent with R10.4 — and as reference databases grow to match, the era of reading protist genes in full may finally be at hand.

Subject of Research: A new pipeline called BaNaNA shows that Nanopore long-read sequencing can match Illumina metabarcoding for protist diversity surveys while offering better species resolution, fewer false positives and phylogenetic power that short reads cannot deliver.

Article Title: From short to long reads: enhanced protist diversity profiling via Nanopore metabarcoding

Article References: Chwalińska, M., Karlicki, M., Romac, S., Not, F., & Karnkowska, A. (2025). From short to long reads: enhanced protist diversity profiling via Nanopore metabarcoding. Metabarcoding and Metagenomics, 9, Article e163750. https://doi.org/10.3897/mbmg.9.163750

Image Credits: AI Generated

DOI: 10.3897/mbmg.9.163750

Keywords: protists, Nanopore sequencing, metabarcoding, 18S rDNA, biodiversity, BaNaNA, long-read sequencing, microbial ecology, eDNA, reference databases, short, long