BDB-Lab June 2026 Updates
The elusive resistome and upcoming travel/conferences
Looking forward
Several lab members will be presenting in major conferences over the coming months. Please feel free to get in touch if you will be there.
August 2–7: 8th Environmental Dimensions of Antimicrobial Resistance (EDAR8) Conference: Faith, Juan, and Luis will attend. Juan and Luis are hosting a workshop on Tuesday: Pitfalls and Directions: Interpreting the Output of Current ARG Annotation Tools.
August 16–21: ISME20 in Auckland, New Zealand: Alexandre will present “Microbial functional role prediction: an activity-based, context-aware mechanistic approach to microbial characterisation” in the Computational approaches in microbial ecology session; Catarina will present a poster in the Computational approaches in microbial ecology session; Juan will present a poster on the elusive resistome preprint (see below); while Luis is one of the conveners for the AI-based tools to predict key microbiome traits session on Monday.
Focus of the Quarter: The elusive resistome by Inda-Díaz et al. (2026)
The elusive resistome: a global comparison reveals large discrepancies among detection pipelines by Juan S. Inda-Díaz, Faith Adegoke, Ulrike Löber, Víctor Hugo Jarquín-Díaz, Yiqian Duan, Johan Bengtsson-Palme, Svetlana Ugarcina Perovic, Luis Pedro Coelho bioRxiv 2026.05.11.724158; doi: https://doi.org/10.64898/2026.05.11.724158
Antimicrobial resistance surveillance increasingly relies on metagenomics, allowing researchers to observe resistance genes directly from environmental and host-associated microbial communities. However, we asked how much do the conclusions depend on the resistance gene detection pipeline being used?
To answer this question, we performed a large comparative study of antimicrobial resistance annotation pipelines. We analysed more than 270 million microbial genes from the Global Microbial Gene Catalogue, spanning over 11,500 metagenomes from 13 distinct habitats, including human-associated microbiomes, animal microbiomes, wastewater, soil, freshwater, and marine environments.
Key findings
Up to a 45-fold difference in the number of reported ARGs. Some pipelines reported only a few thousand resistance genes, while others reported more than one hundred thousand.
Limited agreement between methods. The average Jaccard Index between pipelines was only 16%, meaning that any two pipelines mostly identify unique sets of ARGs.
Different pipelines tell different biological stories. The estimated size and composition of pan- and core-resistomes varied substantially depending on the chosen method. Pipelines supported contrasting interpretations regarding the distribution and prevalence of resistance genes across habitats, even those using the same reference gene database!
Class-level biases are substantial. Certain pipelines reported mainly efflux pumps or van genes, while others emphasised β-lactamases, macrolide, and aminoglycoside resistance genes.
A few gene classes dominate the results
DeepARG and RGI reported dramatically larger resistomes than other tools, largely because they accept low-identity matches. In the case of RGI, this occurred despite our use of its “perfect” and “strict” categories. Much of the resistome reported by RGI consisted of efflux pumps and van genes (vancomycin-resistance), while DeepARG reported large numbers of poorly characterised genes, including sequences assigned to broad multidrug-resistance categories that the tool developers themselves recommend manually curating (Gustavo Arango-Argoty, et al., 2018), genes for which the tools failed to assign a class, and even housekeeping homologs with uncertain resistance function.
The widespread detection of efflux pumps, van genes, and rpoB homologs across most studied habitats raises an important question: is antimicrobial resistance truly ubiquitous in microbial ecosystems, or are many of these sequences homologs perform essential cellular functions without conferring a resistance phenotype?
Why this matters
The implications extend far beyond technical benchmarking. Despite the potential of metagenomics for antimicrobial resistance (AMR) surveillance, persistent limitations and uneven performance across computational approaches for resistome inference constrain the reliable implementation of genomic AMR surveillance at a global scale. This lack of consistency also propagates into downstream evolutionary interpretations, including inferences about bacterial selection, adaptation, and the origins of resistance genes.
We advocate for greater transparency in methodological choices and for explicit recognition that different tools make distinct—and often defensible—trade-offs between sensitivity and specificity.
Papers & preprints
Juan’s preprint “The elusive resistome: a global comparison reveals large discrepancies among detection pipelines” is now online (see Focus). The study compares commonly used ARG detection pipelines using over 270 million prokaryotic genes from the GMGC.
“Unbinned contigs expand known diversity in the global microbiome” co-authored by Luis has been published in Nature Microbiology. The study analysed around 92k metagenomes and showed that much of the discoverable bacterial and archaeal diversity remains hidden among unbinned contigs.
“Storage and soil depth, in addition to wastewater treatment, govern microbiota, and mobile genetic element and antibiotic resistance markers during reclaimed water irrigation” has been published in Water Research. It discusses how reclaimed water treatment, storage, and soil depth affect microbial communities, and ARGs. Yiqian and Luis are co-authors on this work.
Conferences & presentations
Anil presented at the Australian Society for Microbiology Annual National Meeting (June 15-18) at the Melbourne Convention & Exhibition Centre.
Kruthi presented a poster on small proteins at the CIPPS Annual Retreat 2026 in the Blue Mountains, Australia.
Anil presented at the AIFST 2026 Food Microbiology Conference (May 6-7) in Sydney in the NextGen Microbiologists session.
Luis presented “Big data and small genes. The small proteins of the global microbiome (May 16-19)” at the 55th Annual Meeting of SBBq in Águas de Lindóia, São Paulo, Brazil, and at CNPEM, the Brazilian Centre for Research in Energy and Materials, in Campinas.
Luis spoke at the Oxford Nanopore User Group Meeting in Brisbane (May 28), where he presented work on the urban soil and dog microbiome projects.
Software
SemiBin has been updated to version 2.3.0. The update fixes problems in semi-supervised training, long-read clustering, and multi-sample binning.
Other
Luis has been selected as a 2026 TRI Translation Fellow. The fellowship provides tailored coaching and support to strengthen research translation and collaborations.
 | A guest post by
|