A novel approach for predicting aflatoxin B₁ production using regression models and whole-cell biosensors in moldy maize and peanut kernels

Lu Sun ¹, Junning Ma ², Yongping Jiang ²

⁰Institute of Food Science and Technology, Chinese Academy of Agricultural Sciences /Key Laboratory of Agro-Products Quality and Safety Control in Storage and Transport Process, Ministry of Agriculture and Rural Affairs, Beijing 100193, PR China; Gembloux Agro-Bio Tech, University of Liège, Passage des Déportés 2, Gembloux 5030, Belgium.

Journal of Hazardous Materials +

|

September 20, 2025

View abstract on PubMed

Summary

A new biosensor array accurately detects aflatoxin contamination in maize and peanuts using transcriptome-guided promoters and machine learning. This cost-efficient platform enables scalable, real-time food safety monitoring.

Area Of Science

Microbiology
Biosensing Technology
Computational Biology

Background

Aspergillus flavus contamination causes significant post-harvest losses in crops like maize and peanuts due to aflatoxin B1 (AFB1).
Sensitive, scalable, and early detection methods for AFB1 are crucial for food safety and agricultural economics.
Existing detection methods often lack the sensitivity, scalability, or mechanistic interpretability required for comprehensive monitoring.

Purpose Of The Study

To develop a novel whole-cell biosensor array for sensitive and scalable detection of aflatoxin B1 (AFB1).
To integrate transcriptome-guided promoters with machine learning for quantitative prediction of fungal infection stages and AFB1 levels.
To establish a cost-efficient, non-invasive platform for real-time aflatoxin risk assessment in agroecosystems.

Main Methods

Developed a biosensor array using eight infection-induced promoters identified from E. coli transcriptomic responses to volatile organic compounds.
Immobilized bioreporters in calcium alginate and utilized time-resolved bioluminescence signals.
Employed ensemble machine learning regressors (XGBoost, CatBoost, RandomForest) for quantitative prediction and feature importance analysis.

Main Results

The transcriptome-guided biosensor array, particularly using XGBoost, demonstrated high predictive accuracy (R² > 0.91) for infection staging and AFB1 quantification in both maize and peanuts.
The novel biosensors outperformed previous designs using general stress-responsive promoters, especially in external validation.
Feature importance analysis identified host transcriptional regulation and biofilm formation as key predictive indicators, offering mechanistic insights.

Conclusions

The developed biosensing platform provides a biologically informed, non-invasive, and cost-efficient solution for real-time aflatoxin risk assessment.
This approach integrates transcriptomic data with ensemble learning for robust and scalable food safety monitoring.
The findings offer a versatile tool for diverse agroecosystems, enhancing agricultural product safety and reducing economic losses.

Keywords:

Aspergillus flavus CatBoost Post-harvest storage Volatile organic compound Whole-cell biosensor array XGBoost