A new method for internal urinary metabolite exposure and dietary exposure association assessment of 3-MCPD and glycidol and their esters based on machine learning

Abstract

3-Monochloropropane-1,2-diol (3-MCPD) and glycidol along with their esters are commonly found in chemical production, wastewater treatment, food processing, and exhibit toxicity. Accurate exposure assessment is essential for evaluating the environmental hazards and health risks posed by these contaminants. We collected demographic data from 1587 participants and developed seven models using machine-learning algorithms to investigate urinary metabolite exposure and dietary exposure associations of 3-MCPD and glycidol and their esters. Urinary dihydroxypropyl mercapturic acid concentrations, edible oils, and total energy were identified as key predictors of dietary exposure to these contaminants (p < 0.001). The seven machine learning models demonstrated strong predictive capabilities for internal urinary metabolite exposure and dietary exposure associations (average R > 0.6). Among these, generalized additive model and extreme gradient boosting exhibited the strongest correlation and highest accuracy in predicting the associations. We utilized machine learning techniques to link dietary exposure to 3-MCPD, glycidol, and their esters with internal urinary metabolite exposure, providing an innovative and accurate method for risk exposure assessment.