Patterns of flavonoid digestion influenced by molecular descriptors: machine learning reveals flavonoid structure-conversion rate relationships in blood orange juice.

Blood orange juice (BOJ) is a processed product rich in flavonoids, offering multiple health benefits. This study investigated the transformation patterns of structurally diverse flavonoids during gastrointestinal digestion of BOJ. Analyses were conducted through in vitro digestion models, random forest algorithms, and correlation studies. Findings: total flavonoid content decreased with digestion duration, showing a significant decline after 0.5 h of gastric digestion. Four key flavonoids were identified, with their transformation patterns exhibiting three distinct digestive modes: anthocyanins predominated in the stomach, flavonones in the intestine, and flavonols and chalcones in continuous gastrointestinal digestion. Descriptive factors including ring number, double bond number, oil-water partition coefficient, and hydroxyl electro-topological state showed significant correlations (p < 0.05) with conversion rates at different digestive sites. This study provides theoretical foundations and design references for stabilising flavonoid protection, elucidating physiological functions, and structure-directed optimization.