BIOTRAM: Biochemistry-informed Transformer for Relative Abundance via Metabarcoding
Abstract
Arthropod monitoring at scale begins with metabarcoding of bulk samples: trap collections of hundreds to thousands of specimens processed as one unit, extraction of DNA from the bulked specimens, then sequencing of the DNA from all specimens together. Metabarcoding reveals which taxa such a sample contains, but due to factors such as differing DNA amplification rates and body sizes, its read counts do not neatly correspond to how abundant the different taxa are relative to one another. This lack of relative abundance limits the data's utility for the multitude of ecological analyses that depend on it. We propose BIOTRAM: a neural network that estimates per-taxon specimen counts by dividing the observed read counts by a learned read yield per specimen, which it factors into the DNA each specimen contributes and how efficiently that DNA is amplified. On a paired dataset of 1,188 bottles (800 k specimens) processed both by metabarcoding and by comprehensive individual barcoding for ground-truthing, BIOTRAM reduces the Jensen-Shannon divergence between the estimated and the true composition from 0.3586 for raw read proportions (the usual proxy for abundance) to 0.1200. Trained on the same objective and inputs as four generic deep architectures, all with tuned loss weights, it reaches a count error 3.2% below that of the strongest of them and the lowest composition divergence. It also generalizes to out-of-domain data better, beating all baselines on bottles from unseen regions, later collection dates and unseen sequencing batches, and transferring to data from an unseen laboratory. Structuring the mapping from reads to counts around the steps of the experimental pipeline thus yields more accurate abundance estimates than leaving that mapping unconstrained.
Related
- Trichoderma reduces the relative abundance of pathogen-associated fungi while maintaining fungal community stability in Rhododendron soils
- Trait-informed metabarcoding captures tardigrade community trait shifts but not functional dispersion along a land-use gradient
- SeCAT (Sequence Consensus Amplicon Trimming): a simulation-informed Nextflow pipeline for cross-study harmonisation of 16S rRNA metabarcoding datasets
- Strain resolution and wide relative abundance quantification of bacterial microbiomes with nanopore sequencing of the ribosomal RNA operon
- Differential sensitivity of k-mer based identification tools using closely related database sequences affects relative abundance estimation
BibTeX
@misc{boehly2026biotram,
title = {{BIOTRAM: Biochemistry-informed Transformer for Relative Abundance via Metabarcoding}},
author = {Emma Boehly and Scott C. Lowe and Ken A. Thompson and Anthony Fuller and Will Armstrong and Iuliia Zarubiieva and Neelu Madan and Tiancheng Gao and Sameer Padhye and Jayme Sones and Kate H. J. Perez and Bess Hardwick and Brendan Furneaux and Evgeny V. Zakharov and Dirk Steinke and Jeremy R. deWaard and Stephanie L. deWaard and Jaclyn McKeown and Norman Monkhouse and Cole Pace and Sujeevan Ratnasingham and Raina Schmuck and Amy Thompson and Jorge L. León-Cortés and Luciano Cagnolo and Julieta Soledad Arena and Jeremías Domínguez Masciale and Errol Douwes and Elli Groner and Masae Iwamoto Ishihara and Shunsuke Matsuoka and Mailyn A. Gonzalez and Alexandre Antonelli and Renata S. Sousa-Lima and Armin Niessner and Patrícia Guedes and Elizabeth G. Biro and Zuania Colón-Piñeiro and Dezene P. W. Huber and Santosh Mazumdar and Magda Bou Dagher Kharrat and Eddie P. Mondejar and Pablo Luis Tubaro and Eugene Marais and Marie-Eugénie Maggia and Darío Alejandro Lijtmaer and Sean W. J. Prosser and Andrea Murillo-Vázquez and Paola Pulido-Santacruz and Paula Leitman and Thiago Arnt Dorigo and Nao Ito and Paul Knoops and Maria L. Kuzmina and Hikaru Morishita and Keishi Kimoto and Dimby Raharinjanahary and Rhian J. Smith and Solny A. Adalsteinsson and Kevin G. Borja-Acosta},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.05418},
eprint = {2611.05418},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.05418},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - BIOTRAM: Biochemistry-informed Transformer for Relative Abundance via Metabarcoding
AU - Emma Boehly
AU - Scott C. Lowe
AU - Ken A. Thompson
AU - Anthony Fuller
AU - Will Armstrong
AU - Iuliia Zarubiieva
AU - Neelu Madan
AU - Tiancheng Gao
AU - Sameer Padhye
AU - Jayme Sones
AU - Kate H. J. Perez
AU - Bess Hardwick
AU - Brendan Furneaux
AU - Evgeny V. Zakharov
AU - Dirk Steinke
AU - Jeremy R. deWaard
AU - Stephanie L. deWaard
AU - Jaclyn McKeown
AU - Norman Monkhouse
AU - Cole Pace
AU - Sujeevan Ratnasingham
AU - Raina Schmuck
AU - Amy Thompson
AU - Jorge L. León-Cortés
AU - Luciano Cagnolo
AU - Julieta Soledad Arena
AU - Jeremías Domínguez Masciale
AU - Errol Douwes
AU - Elli Groner
AU - Masae Iwamoto Ishihara
AU - Shunsuke Matsuoka
AU - Mailyn A. Gonzalez
AU - Alexandre Antonelli
AU - Renata S. Sousa-Lima
AU - Armin Niessner
AU - Patrícia Guedes
AU - Elizabeth G. Biro
AU - Zuania Colón-Piñeiro
AU - Dezene P. W. Huber
AU - Santosh Mazumdar
AU - Magda Bou Dagher Kharrat
AU - Eddie P. Mondejar
AU - Pablo Luis Tubaro
AU - Eugene Marais
AU - Marie-Eugénie Maggia
AU - Darío Alejandro Lijtmaer
AU - Sean W. J. Prosser
AU - Andrea Murillo-Vázquez
AU - Paola Pulido-Santacruz
AU - Paula Leitman
AU - Thiago Arnt Dorigo
AU - Nao Ito
AU - Paul Knoops
AU - Maria L. Kuzmina
AU - Hikaru Morishita
AU - Keishi Kimoto
AU - Dimby Raharinjanahary
AU - Rhian J. Smith
AU - Solny A. Adalsteinsson
AU - Kevin G. Borja-Acosta
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.05418
UR - https://arxiv.org/abs/2611.05418
AB - Arthropod monitoring at scale begins with metabarcoding of bulk samples: trap collections of hundreds to thousands of specimens processed as one unit, extraction of DNA from the bulked specimens, then sequencing of the DNA from all specimens together. Metabarcoding reveals which taxa such a sample contains, but due to factors such as differing DNA amplification rates and body sizes, its read counts do not neatly correspond to how abundant the different taxa are relative to one another. This lack of relative abundance limits the data's utility for the multitude of ecological analyses that depend on it. We propose BIOTRAM: a neural network that estimates per-taxon specimen counts by dividing the observed read counts by a learned read yield per specimen, which it factors into the DNA each specimen contributes and how efficiently that DNA is amplified. On a paired dataset of 1,188 bottles (800 k specimens) processed both by metabarcoding and by comprehensive individual barcoding for ground-truthing, BIOTRAM reduces the Jensen-Shannon divergence between the estimated and the true composition from 0.3586 for raw read proportions (the usual proxy for abundance) to 0.1200. Trained on the same objective and inputs as four generic deep architectures, all with tuned loss weights, it reaches a count error 3.2% below that of the strongest of them and the lowest composition divergence. It also generalizes to out-of-domain data better, beating all baselines on bottles from unseen regions, later collection dates and unseen sequencing batches, and transferring to data from an unseen laboratory. Structuring the mapping from reads to counts around the steps of the experimental pipeline thus yields more accurate abundance estimates than leaving that mapping unconstrained.
ER -