Truace
Subscribe

Study of 30,554 E. coli genomes used AI to predict antibiotic resistance in 126 countries

Study title: Global genomic surveillance of β-lactam resistance in Escherichia coli across human, animal, and environmental reservoirs

A One-Health study analyzed 30,554 E. coli whole-genome sequences from human, animal and environmental sources across 126 countries from 2000 to 2025, using AMRFinderPlus and MLST to map resistance genes and clones, and applied gradient-boosted machine learning to predict MICs from gene profiles and chromosomal features.

Read the study abstract

Background Escherichia coli poses a global health threat from increasing β-lactam resistance. This study uses genomic and One-Health data to map resistance patterns and enhance antimicrobial resistance (AMR) prediction and management. Methods This study performed a One-Health whole-genome analysis of 30,554 E. coli isolates from human, animal, and environmental sources, spanning from 2000 to 2025. Publicly available genomic data were retrieved from NCBI, encompassing β-lactam resistance genes, including extended…

By the Truvace desk1 min read1 sourceEvidence: Moderate · 65/100 — Observational study · n = 30,554 · 1 independent sourcePeer-reviewedTRV-2026-0584· cite & verify
Global genomic surveillance of β-lactam resistance in Escherichia coli across human, animal, and environmental reservoirs
Journal.ppat.1008211 by Sarah Ben Maamar, Adam J. Glawe, Taylor K. Brown, Nancy Hellgeth, Jinglin Hu, Ji-Ping Wang, Curtis Huttenhower, Erica M. Hartmann. CC BY 4.0 · https://creativecommons.org/licenses/by/4.0
Trace impact readingPositive state
P 57The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 62The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

The work matters because it shows AI can accurately predict carbapenem resistance for surveillance, with imipenem R2 0.88, yet remains unreliable for ampicillin and inhibitor combinations, leaving uncertainty about clinical deployment across all beta-lactams and about generalizability beyond the curated NCBI dataset.

Main points

  1. Analysis covered 30,554 E. coli whole-genome sequences from 126 countries between 2000 and 2025, including 14,320 human, 8184 animal, and 3941 environmental isolates.
  2. MLST identified dominant sequence types ST131, ST73 and ST1193 in human isolates, ST10 in animal isolates, and ST155 in environmental isolates.
  3. Temporal analysis showed rising prevalence of bla CTX-M-15 , bla NDM-5 , and bla OXA-1 driving beta-lactam resistance in high-risk clones.

The gain

Gradient-boosted machine learning predicted carbapenem MICs from resistance gene profiles with strong performance for imipenem and ertapenem, supporting AMR surveillance.

The problem

The same gradient-boosted models showed poor agreement with observed MICs for ampicillin and piperacillin-tazobactam and struggled with beta-lactamase inhibitor combinations.

The rundown

Researchers retrieved 30,554 E. coli genomes from NCBI spanning 2000 to 2025, with an average length of 5,120,983 bp, and curated beta-lactam resistance genes using AMRFinderPlus alongside MLST and pathotype assignment.

By publication date July 24 2026, the study reported observed model performance strongest for carbapenems and weakest for ampicillin and piperacillin-tazobactam, concluding that integrating genomics and One-Health data can guide AMR surveillance and control.

What this doesn’t fix

Model accuracy was uneven across drug classes, with strong results for carbapenems but poor agreement for ampicillin and beta-lactamase inhibitor combinations, limiting broad clinical use.

Sources

  1. Peer-reviewedInfection, Genetics and Evolution2026-07-24

The debate