Machine learning models to predict the bioaccessibility of parent and substituted polycyclic aromatic hydrocarbons (PAHs) in food: Impact on accurate health risk assessment

Food intake is the primary pathway for polycyclic aromatic hydrocarbons (PAHs) to enter the human body. Once ingested, PAHs tend to accumulate, posing health risks. To accurately assess the risk of PAHs from food, concentrations of 10 parent PAHs (PPAHs) and 15 substituted PAHs (SPAHs) were detected across 34 commonly consumed foods. Results indicated that SPAHs concentrations (3.89–11.6 ng/g dw) were higher than PPAH concentrations (1.66–3.43 ng/g dw) in shrimp and shellfish and freshwater fish. Four machine learning algorithms were used to predict the bioaccessibility of PAHs in foods, with the random forest model performing the best (R2 =0.987, RMSE=5.99). Feature variable importance analysis revealed that lipid and protein contents in food are critical variables influencing PAH bioaccessibility. Subsequently, the bioaccessibility of 25 PAHs in various foods was predicted to explore its impact on health risk assessment. Consequently, the carcinogenic risks considering bioaccessibility (5.62 ×10−5-7.12 ×10−5) was about an order of magnitude lower than that ignoring bioaccessibility (1.52 ×10−4-1.69 ×10−4), yet it still exceeded 10⁻6, indicating potential carcinogenic risks. Although PPAHs and alkylated PAHs were predominant in foods, 6-nitrochrysene was the main compound inducing both non-carcinogenic and carcinogenic risks owing to its high toxicity. This study developed a novel method for assessing pollutant bioaccessibility and evaluating its impact on health risk assessment, which provides a valuable model for managing massive hazardous pollutants and is essential for improving the accuracy of health risk assessment. [Display omitted] •SPAH concentrations in shrimp and fish were higher than those of PPAHs.•The RF model performed the best in both internal and external validations.•Lipid and protein content in food have a significant impact on PAH bioaccessibility.•Considering bioaccessibility, the HQs of PAHs increased and the ILCRs exceeded 10−6.•The highly toxic 6-nitrochrysene was the primary source of both HQs and ILCRs.