The ITS region of nuclear ribosomal DNA serves as the standard DNA barcode for identifying fungi. Because many fungal species resist cultivation in the laboratory, researchers typically amplify and sequence this marker directly from environmental samples, a technique known as amplicon sequencing or metabarcoding. The raw reads, however, are not species themselves; they must be processed computationally into units that approximate species. OTU approaches cluster reads at a similarity cutoff, traditionally 97 percent, while ASV approaches, exemplified by the widely used DADA2 algorithm, denoise the data to recover the true underlying sequences. The two methods differ conceptually: ASVs are supposed to represent actual biological sequences and can be compared across studies, whereas OTUs are clusters of similar sequences that are specific to each study. Despite years of debate, no consensus has emerged on which approach best represents the true fungal community.

Previous studies have reached conflicting conclusions. Some work on bacterial 16S data found no difference between ASVs and OTUs at the family level, and broad-scale ecological patterns have proven robust to the choice of method for both bacterial and fungal amplicons. Yet concerns persist that fungal ASVs might inflate estimates of alpha diversity because ribosomal DNA exists in multiple copies and varies within species. Other researchers have reported the opposite, that OTUs exaggerate diversity through false positives; one analysis of a 189-species mock community identified between 577 and 1413 OTUs. Tosadori and Bosch designed their study to cut through this confusion by exploiting a key advantage unavailable in real-world experiments: perfect knowledge of the ground truth.

The team downloaded more than 1500 fungal genomes from 859 species in the Ensembl Fungi database and used them to simulate amplicon sequencing of the ITS1 and ITS2 regions. They built mock communities spanning 50 to 800 fungal genomes, with abundances following a power law distribution, and simulated sequencing errors, chimeras, and quality profiles modeled on real Illumina data. Because the true composition of each community was known exactly, any discrepancy between the analyzed results and reality could be attributed directly to the analysis pipeline. The researchers then processed these data through six pipelines combining two read-processing strategies with three ASV or OTU methods, evaluating diversity metrics, taxonomic composition, sensitivity, specificity, and distance from the true community using Aitchison, Hellinger, and Bray-Curtis distance measures.

The results were striking. In the simulated communities, the choice between ASV and OTU methods had a statistically significant but comparatively small effect, whereas read processing, specifically whether the ITS region was extracted from merged paired-end reads or whether only forward reads were used, explained a larger share of the variation in community composition. On average, sensitivity was 12.7 percent higher for data that had undergone ITS extraction than for forward-read-only processing, and OTU methods were 15.9 percent more sensitive than ASV methods. Specificity, by contrast, varied little between treatments. The taxonomic level of analysis was consistently the single most important factor determining how closely a pipeline’s output matched the true community, with distances increasing at finer taxonomic resolutions.

Richness estimation revealed another clear pattern. Small sequencing libraries of 10,000 reads were insufficient to capture the full diversity of the simulated communities, underestimating species richness at nearly all diversity levels. Medium libraries performed better, and large libraries of 50,000 reads allowed both ASV and OTU pipelines with ITS extraction to produce accurate richness estimates, except in the highest-diversity communities containing multiple strains of the same species. The forward-read-only pipelines never managed to estimate richness accurately, and one of them, forward-read OTU clustering, actually overestimated richness, with the overestimate growing worse as library size increased. This finding directly contradicts fears that ASVs systematically inflate fungal diversity; in these simulations, ASVs tended to underestimate richness while OTUs tended to overestimate it, but the discrepancy largely vanished when ITS extraction and large library sizes were combined.

To confirm that their simulation results transferred to real sequencing data, the researchers reanalyzed a previously published 189-strain fungal mock community originally used to compare 260 bioinformatics pipelines. That earlier study had recommended using only forward reads with DADA2 ASVs and deemed ITS extraction unnecessary. Tosadori and Bosch reached very different conclusions, which they attribute to methodological differences: the original analysis required strict read-merging parameters with no mismatches allowed, used the older ITSx tool rather than ITSxpress, and relied on a custom database of Sanger sequences rather than the large reference databases standard in environmental microbiology. With more relaxed merging requirements, the joined reads retained far more information, and the benefit of ITS extraction became apparent. In the reanalysis, certain taxa were detected only in ITS-extracted samples while others appeared only in forward-read processing, and no orders or genera were specific to either ASVs or OTUs.

The practical implications are substantial. The authors recommend that researchers prioritize large library sizes, ideally around 50,000 reads per sample, which modern sequencing technology makes achievable with little extra effort. They further advise merging forward and reverse reads and performing ITS extraction unless there is a compelling reason not to, such as poor-quality reverse reads or species whose ITS regions are too long to merge, like chanterelles with ITS1 regions approaching 1,100 bases or the oomycete Plasmopara halstedii with an ITS2 region of roughly 2,200 bases. For pathogen surveillance or other applications requiring detection of specific taxa, analyzing data with both processing strategies may maximize the chances of correctly identifying the target organism. Notably, the team found no statistical benefit to fungal-specific modifications of DADA2 parameters that had been proposed to improve read retention, suggesting default settings remain appropriate.

The study does carry limitations the authors acknowledge. Their simulated communities were drawn only from fungi with fully sequenced genomes, a subset biased toward species that are easy to culture or of medical and commercial interest, and therefore not fully representative of natural fungal diversity. They also tested a limited, though realistic, set of processing options rather than exhaustively exploring every possible pipeline configuration. Standard short reads, moreover, lacked the resolution to distinguish strains of the same species, a task that requires full-length ITS sequences or alternative markers. Even so, the combination of simulated data with known ground truth and validation against a real mock community gives the findings unusual force in a field where method comparisons typically cannot determine which pipeline is closer to reality.

For a discipline that has invested considerable energy in the ASV-versus-OTU debate, the message is quietly liberating: for most fungal community studies, either method will yield broadly similar ecological conclusions, and researchers can choose based on their specific goals, favoring OTUs when sensitivity to particular taxa matters most and ASVs when minimizing false positives is the priority. The real leverage lies elsewhere, in sequencing deeply enough, joining paired-end reads, and carefully extracting the ITS barcode before analysis. As fungal metabarcoding continues to underpin research on soil health, plant pathology, climate change responses, and biodiversity monitoring, these evidence-based recommendations offer a clearer roadmap for study design than the field has previously possessed.

Subject of Research: Comparison of ASV and OTU bioinformatics pipelines for fungal ITS amplicon sequencing using simulated communities

Article Title: Decoding fungal communities: ITS extraction matters more than ASVs vs. OTUs

Article References: Tosadori, G., & Bosch, J. (2026). Decoding fungal communities: ITS extraction matters more than ASVs vs. OTUs. Web Ecology, 26(1), 47-59. https://doi.org/10.5194/we-26-47-2026

Image Credits: AI Generated

DOI: 10.5194/we-26-47-2026

Keywords: fungal ecology, metabarcoding, ITS region, ASVs, OTUs, amplicon sequencing, DADA2, bioinformatics pipelines, alpha diversity, ribosomal DNA, microbiome, study design