EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Exploring Machine Learning for the Curation of DNA Barcode Reference Libraries: A Case Study With Marine Fish Species of the Eastern Atlantic and Adjacent Waters

João T Fontes iD, André O Ferreira iD, Ana S Lavrador iD, Pedro Soares iD, Filipe O Costa iD, Carlos A Canchaya iD

Molecular Ecology Resources · 2026-11 · 26 (8)

DOI: 10.1111/1755-0998.70223 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Molecular Ecology Resources, online erschienen am 08.11.2026. Verfasst von João T Fontes und 5 weiteren Autor:innen. Beteiligte Einrichtung: Centre of Molecular and Environmental Biology (CBMA) / ARNET-Aquatic Research Network, Department of Biology, University of Minho, Braga, Portugal. Hauptthemen laut MEDLINE: DNA Barcoding, Taxonomic, Fishes, Machine Learning. Die Arbeit nennt 72 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Portuguese Foundation of Science and Technology, Aga Khan Development Network, FCT, I.P..

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate species assignments using (e)DNA (meta)barcoding depend on the reliability of DNA barcode reference libraries. Manual curation can help resolve pseudo‐discordances and identify incongruences, but it is time‐consuming and increasingly difficult to scale as reference sequence data continue to expand. We explored the potential of automated approaches to support the curation of DNA barcode records within discordant Barcode Index Numbers (BINs). Records were mined from the Barcode of Life Data System (BOLD) for marine fish species occurring in the Eastern Atlantic and adjacent waters, and features describing taxonomic consistency, morphospecies‐BIN congruence, and associated metadata were extracted. Within a subset comprising 52,207 records, 3055 species names, and 1138 discordant BINs, each record was manually classified as supported or inconclusive using the taxonomic and metadata information associated with the records as a heuristic guide. We compared a simple rule‐based approach with logistic regression, extreme gradient boosting (XGBoost) and artificial neural networks (ANNs). The rule‐based approach achieved accuracies ranging from 80.4% to 90.9% across the evaluated thresholds. Logistic regression achieved the highest overall accuracy (92.1%), followed closely by ANN (91.7%) and XGBoost (91.4%). These results support the use of machine‐learning approaches to assist the triage of discordant records by helping identify cases that warrant closer human inspection. Such approaches could be integrated into a human‐in‐the‐loop framework that retains manual assessment where taxonomic judgement is required, while helping to scale the screening and curation of rapidly growing DNA barcode reference libraries.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

DNA Barcoding, TaxonomicFishesMachine LearningAnimalsAtlantic Ocean

Stichwörter der Autor:innen

DNA barcodingDNA metabarcodingautomationcurationeDNAmachine learningmarine fishspecies‐level identification

Autor:innen und Institutionen

  • Centre of Molecular and Environmental Biology (CBMA) / ARNET-Aquatic Research Network, Department of Biology, University of Minho, Braga, Portugal.
  • Institute of Science and Innovation for Bio-Sustainability (IB-S), University of Minho, Braga, Portugal.
  • Department of Biochemistry, Genetics, and Immunology, University of Vigo, Vigo, Spain.
  • Marine Research Centre (CIM-UVIGO), Universidade de Vigo, Vigo, Spain.

ORCID-Profile: João T Fontes, André O Ferreira, Ana S Lavrador, Pedro Soares, Filipe O Costa, Carlos A Canchaya

Förderung

  • Portuguese Foundation of Science and Technology
  • Aga Khan Development Network
  • FCT, I.P.
  • Collaboration Protocol for Financing the Multiannual Research Grants Plan for Doctoral Students
  • European Social Fund under the Northern Regional Operational Program-Norte2020

Zitieren

APA

Fontes, J. T., Ferreira, A. O., Lavrador, A. S., Soares, P., Costa, F. O., & Canchaya, C. A. (2026). Exploring Machine Learning for the Curation of DNA Barcode Reference Libraries: A Case Study With Marine Fish Species of the Eastern Atlantic and Adjacent Waters. Molecular Ecology Resources, 26(8). https://doi.org/10.1111/1755-0998.70223

Vancouver

Fontes JT, Ferreira AO, Lavrador AS, Soares P, Costa FO, Canchaya CA. Exploring Machine Learning for the Curation of DNA Barcode Reference Libraries: A Case Study With Marine Fish Species of the Eastern Atlantic and Adjacent Waters. Molecular Ecology Resources. 2026;26(8). doi:10.1111/1755-0998.70223

BibTeX
@article{Fontes2026,
  author = {Fontes, João T and Ferreira, André O and Lavrador, Ana S and Soares, Pedro and Costa, Filipe O and Canchaya, Carlos A},
  title = {Exploring Machine Learning for the Curation of DNA Barcode Reference Libraries: A Case Study With Marine Fish Species of the Eastern Atlantic and Adjacent Waters},
  journal = {Molecular Ecology Resources},
  year = {2026},
  volume = {26},
  number = {8},
  doi = {10.1111/1755-0998.70223}
}
Kostenlos starten →