Effects of Modified Atmosphere Package on the Chilling Injury of Postharvest Atemoya Fruits Based on Physiochemical Analysis and Machine Learning

Abstract

The study aimed to investigate the effects of modified atmosphere package combined with low-temperature storage on the chilling injury of atemoya and analyze the relationship between the physiochemical indexes and chilling stress by machine learning. The fruits were stored at room temperature (RT, 25 ℃), low temperature (LT, 10 ℃) and low temperature combined with modified atmosphere package (CA, 10 ℃), and quality indicators such as flesh firmness, chilling injury index (CI), total phenols, relative electrolyte conductivity (EC), and phenylalanineammonialyas (PAL) activity were measured within 7 days. The results indicated that low-temperature storage effectively delayed the ripening of atemoya, retarding the decrease of flesh firmness. But atemoya stored at low temperature suffered from chilling stress, emerging soaking spots, EC increased and the content of MDA in the flesh accumulated. On this basis, several machine learning algorithms were established and the optimal model chosen for predicting CI was Ridge regression. Explanation analysis (SHapley Additive exPlanations，SHAP) showed that storage time, soluble solids, water loss rate, soluble proteins, and total phenols contributed significantly to the model and were closely related to chilling stress. Low temperature combined with modified atmosphere could increase the content of soluble substances and phenols, enhance the osmoregulation capacity and clearing-reactive- oxygen ability, and thus maintain the integrity of membrane and mitigate the chilling injury.

Keywords

• atemoya
• modified atemosphere package
• chilling injury
• machine learning
• ridge regression