Development of a processing factor prediction model for pesticides in processed tomato foods using elastic net regularization

• Published in: Food chemistry 2024-07, Vol.447, p.138943-138943, Article 138943
• Main Authors: Yamasaki, Yuki, Nakamura, Kosuke, Kashiwabara, Nao, Chiba, Shinji, Akiyama, Hiroshi, Tsutsumi, Tomoaki
• Format: Article
• Language: English
• Subjects: Elastic net; Pesticide; Physicochemical property; Predictive model; Processing factor
• ISSN: 0308-8146; 1873-7072
• DOI: 10.1016/j.foodchem.2024.138943

A novel regularized elastic net regression model was developed to predict processing factor (PF) for pesticide residues, which represents a change in the residue levels during food processing. The PF values for tomato juice, wet pomace and dry pomace in the evaluations and reports published by the Joint FAO/WHO Meeting on Pesticide Residues significantly correlated with the physicochemical properties of pesticides, and subsequently the correlation was observed in the present tomato processing study. The elastic net regression model predicted the PF values using the physicochemical properties as predictor variables for both training and test data within a 2-fold range for 80–100% of the pesticides tested in the tomato processing study while overcoming multicollinearity. These results suggest that the PF values are predictable at a certain degree of accuracy from the unique sets of physicochemical properties of pesticides using the developed model based on a processing study with representative pesticides. [Display omitted] •A prediction model was developed for processing factor (PF) of pesticide residues.•The development of the model was examined for tomato juice, wet and dry pomace.•Elastic net regression was employed for the model to overcome multicollinearity.•The physicochemical properties of pesticides were used as predicter variables.•The established model could predict PF within a 2-fold range for many pesticides.