UNITE (https://unite.ut.ee/) is a web-based database and sequence management environment for the molecular identification of fungi. It targets the formal fungal barcode—the nuclear ribosomal internal transcribed spacer (ITS) region—and offers all ∼1 000 000 public fungal ITS sequences for reference. These are clustered into ∼459 000 species hypotheses and assigned digital object identifiers (DOIs) to promote unambiguous reference across studies. In-house and web-based third-party sequence curation and annotation have resulted in more than 275 000 improvements to the data over the past 15 years. UNITE serves as a data provider for a range of metabarcoding software pipelines and regularly exchanges data with all major fungal sequence databases and other community resources. Recent improvements include redesigned handling of unclassifiable species hypotheses, integration with the taxonomic backbone of the Global Biodiversity Information Facility, and support for an unlimited number of parallel taxonomic classification systems.

Here, we describe the taxon hypothesis (TH) paradigm, which covers the construction, identification, and communication of taxa as datasets. Defining taxa as datasets of individuals and their traits will make taxon identification and most importantly communication of taxa precise and reproducible. This will allow datasets with standardized and atomized traits to be used digitally in identification pipelines and communicated through persistent identifiers. Such datasets are particularly useful in the context of formally undescribed or even physically undiscovered species if data such as sequences from samples of environmental DNA (eDNA) are available. Implementing the TH paradigm will to some extent remove the impediment to hastily discover and formally describe all extant species in that the TH paradigm allows discovery and communication of new species and other taxa also in the absence of formal descriptions. The TH datasets can be connected to a taxonomic backbone providing access to the vast information associated with the tree of life. In parallel to the description of the TH paradigm, we demonstrate how it is implemented in the UNITE digital taxon communication system. UNITE TH datasets include rich data on individuals and their rDNA ITS sequences. These datasets are equipped with digital object identifiers (DOI) that serve to fix their identity in our communication. All datasets are also connected to a GBIF taxonomic backbone. Researchers processing their eDNA samples using UNITE datasets will, thus, be able to publish their findings as taxon occurrences in the GBIF data portal. UNITE species hypothesis (species level THs) datasets are increasingly utilized in taxon identification pipelines and even formally undescribed species can be identified and communicated by using UNITE. The TH paradigm seeks to achieve unambiguous, unique, and traceable communication of taxa and their properties at any level of the tree of life. It offers a rapid way to discover and communicate undescribed species in identification pipelines and data portals before they are lost to the sixth mass extinction.

Different distance-based threshold selection approaches were used to assess and compare use of the internal transcribed spacer (ITS) region to distinguish among 901 Cortinarius species represented by >3000 collections. Sources of error associated with genetic markers and selection approaches were explored and evaluated using MOTUs from genus and lineage based-alignments. Our study indicates that 1–2% more species can be distinguished by using the full-length ITS barcode as compared to either the ITS1 or ITS2 regions alone. Optimal threshold values for different picking approaches and genetic marker lengths inferred from a subset of species containing major lineages ranged from 97.0–99.5% sequence similarity using clustering optimization and UNITE SH, and from 1–2% sequence dissimilarity with CROP. Errors for the optimal cut-off ranged from 0–70%, and these can be reduced to a maximum of 22% when excluding species lacking a barcode gap. A threshold value of 99% is suitable for distinguishing species in the majority of lineages in the genus using the entire ITS region but only 90% of the species could be identified using just the ITS1 or ITS2 region. Prior identification of species lacking barcode gaps and their subsequent separate analyses, maximized the accuracy of threshold approaches.

The article presents an AI-based fungi species recognition system for a citizen-science community. The system’s real-time identification too — FungiVision — with a mobile application front-end, led to increased public interest in fungi, quadrupling the number of citizens collecting data. FungiVision, deployed with a human-in-the-loop, reaches nearly 93% accuracy. Using the collected data, we developed a novel fine-grained classification dataset — Danish Fungi 2020 (DF20) — with several unique characteristics: species-level labels, a small number of errors, and rich observation metadata. The dataset enables the testing of the ability to improve classification using metadata, e.g., time, location, habitat and substrate, facilitates classifier calibration testing and finally allows the study of the impact of the device settings on the classification performance. The continual flow of labelled data supports improvements of the online recognition system. Finally, we present a novel method for the fungi recognition service, based on a Vision Transformer architecture. Trained on DF20 and exploiting available metadata, it achieves a recognition error that is 46.75% lower than the current system. By providing a stream of labeled data in one direction, and an accuracy increase in the other, the collaboration creates a virtuous cycle helping both communities.

Nomenclatural type definitions are one of the most important concepts in biological nomenclature. Being physical objects that can be re-studied by other researchers, types permanently link taxonomy (an artificial agreement to classify biological diversity) with nomenclature (an artificial agreement to name biological diversity). Two proposals to amend the International Code of Nomenclature for algae, fungi, and plants (ICN), allowing DNA sequences alone (of any region and extent) to serve as types of taxon names for voucherless fungi (mainly putative taxa from environmental DNA sequences), have been submitted to be voted on at the 11^th International Mycological Congress (Puerto Rico, July 2018). We consider various genetic processes affecting the distribution of alleles among taxa and find that alleles may not consistently and uniquely represent the species within which they are contained. Should the proposals be accepted, the meaning of nomenclatural types would change in a fundamental way from physical objects as sources of data to the data themselves. Such changes are conducive to irreproducible science, the potential typification on artefactual data, and massive creation of names with low information content, ultimately causing nomenclatural instability and unnecessary work for future researchers that would stall future explorations of fungal diversity. We conclude that the acceptance of DNA sequences alone as types of names of taxa, under the terms used in the current proposals, is unnecessary and would not solve the problem of naming putative taxa known only from DNA sequences in a scientifically defensible way. As an alternative, we highlight the use of formulas for naming putative taxa (candidate taxa) that do not require any modification of the ICN.