Development of Machine-Learning Models for Predicting Escherichia coli O157:H7 Inactivation on Fresh-Cut Lettuce...

Area of Science:

• Food Microbiology
• Computational Biology
• Food Safety Engineering

Background:

• Minimally processed leafy greens frequently transmit Escherichia coli O157:H7, leading to outbreaks.
• Chlorine washing is a key industrial method to reduce contamination, but its efficacy is complex.
• Conventional kinetic models struggle to accurately predict chlorine inactivation due to interacting factors.

Purpose of the Study:

• To compare the predictive accuracy of machine learning (ML) algorithms against traditional models for E. coli O157:H7 inactivation on lettuce.
• To develop and validate data-driven models for optimizing chlorine wash parameters in food processing.
• To provide an accessible decision-support tool for the food industry and regulators.

Main Methods:

• Generated a laboratory dataset of E. coli O157:H7 reductions on iceberg lettuce under varying chlorine concentrations (0-150 mg L⁻¹) and contact times (0-150 s) at 4.5 °C.
• Fitted a traditional inactivation model and three ML algorithms: support-vector regression (SVR), random-forest regression (RFR), and Gaussian-process regression (GPR).
• Evaluated model performance using root-mean-square error (RMSE) and R-squared (R²) values, assessing generalisability across validation folds.

Main Results:

• Machine learning models significantly improved predictive accuracy over the traditional model (RMSE 0.392, R² 0.721).
• Gaussian-process regression (GPR) achieved the highest accuracy (RMSE 0.283, R² 0.852), followed closely by SVR (RMSE 0.298, R² 0.836).
• All ML models demonstrated strong generalisability and stability across validation datasets.

Conclusions:

• Data-driven ML approaches offer superior performance for modeling chlorine wash effectiveness compared to traditional methods.
• The developed ML models provide a reliable and accessible decision-support tool for optimizing food processing parameters.
• Implementing these tools can enhance food safety practices and reduce the public health impact of leafy green-associated outbreaks.