Truace
Subscribe

Study in Nigeria finds AI model predicts food-borne drug resistance better than single-data methods

Study title: Multi-omics-driven predictive modelling of antimicrobial resistance emergence in African food value chains under climate change scenarios.

Researchers built a machine learning framework that combines genomics, metagenomics, transcriptomics, proteomics and metabolomics data to predict antimicrobial resistance in Nigerian food value chains. Their best model, XGBoost, reached an AUROC of 0.942 with 0.91 sensitivity and 0.89 specificity, a 16.3% improvement over single-omics approaches.

Read the study abstract

Antimicrobial resistance (AMR) poses a critical threat to food safety, particularly in African food systems where surveillance infrastructure is limited and climate change is accelerating microbial adaptation. Traditional microbiological risk assessment approaches fail to capture the biological complexity of AMR emergence across food value chains. This study developed an integrated multi-omics and machine learning framework to predict AMR emergence across Nigerian food value chains under projected climate scenar…

Truvace record2 min read1 sourceEvidence: Moderate · 68/100 — Controlled or validated comparison · n = 679 · 1 independent sourcePeer-reviewedTRV-2026-2392 · cite or check it
Multi-omics-driven predictive modelling of antimicrobial resistance emergence in African food value chains under climate change scenarios."Medical Laboratory" by ben.dracup, CC BY 2.0.

In brief

The team collected 487 food and 192 environmental samples in Delta, Benin and Port Harcourt. Beta-lactamase and tetracycline resistance genes were the most common, and carbapenemase genes turned up in 8.6% of samples. Temperature was the strongest predictor of resistance gene abundance (β = 0.42).

Combined with CMIP6 climate projections, the work suggests Vibrio prevalence could rise 47-82% under the high-emission SSP5-8.5 scenario. The authors say the approach could support climate-aware food safety surveillance in low- and middle-income countries where monitoring capacity is limited. The climate figures are projections, not measured outcomes.

Main points

  1. The study was cross-sectional and covered Delta, Benin and Port Harcourt in Nigeria.
  2. Samples totalled 487 food and 192 environmental specimens.
  3. XGBoost outperformed Random Forest and Support Vector Machines, with sensitivity 0.91 and specificity 0.89.
  4. The multi-omics model improved on single-omics approaches by 16.3%.
  5. Beta-lactamase and tetracycline resistance genes predominated; carbapenemase genes appeared in 8.6% of samples.
  6. Temperature was the strongest predictor of resistance gene abundance (β = 0.42).
  7. Projected Vibrio prevalence rises 47-82% under SSP5-8.5.

The gain

In a study of 487 food and 192 environmental samples from three Nigerian locations, an XGBoost model combining multiple omics datasets predicted antimicrobial resistance emergence with an AUROC of 0.942, a 16.3% improvement over single-omics approaches.

The full story

The framework integrated five omics layers with CMIP6 climate projections. XGBoost, Random Forest and Support Vector Machines were trained and externally validated, and explainable AI methods were used to identify which factors drove predictions.

Quantitative microbial risk assessment incorporated omics-derived parameters. The authors argue that traditional microbiological risk assessment fails to capture the biological complexity of resistance emergence across food value chains.

The authors say the framework could enable predictive, climate-resilient surveillance and support One Health governance in low- and middle-income countries, where surveillance infrastructure is limited.

Sources

  1. Peer-reviewedFood Research International2026-12-01

The debate