Optimizing food protein stability through flavonoid complexation: An efficient computational-experimental approach

The combination of proteins and flavonoids has been proven to enhance protein stability, increase small molecule solubility, and exhibit additional advantageous properties. However, traditional experimental methods for studying protein-flavonoid binding are inefficient due to complex influencing factors. To address multivariate nonlinear interactions and improve mechanistic reliability, this study established an integrated framework combining machine learning, molecular docking, and systematic experimental validation using phycocyanin—a natural blue pigment with significant bioactivity but prone to degradation under heat, light, and pH variations, limiting its food applications—and two structurally similar flavonoids (rutin and hesperidin) for experimental validation. LightGBM was identified as the optimal predictive model (R 2  > 0.84), revealing key determinants such as molecular weight. The model quantitatively identified structure–property relationships, which were subsequently verified through spectroscopic, physicochemical, and stability analyses. Experimental results demonstrated that both flavonoids induced significant fluorescence quenching in phycocyanin. FT-IR analysis confirmed enhanced hydrogen bonding and hydrophobic interactions, which tightened local protein structures. Surface hydrophobicity measurements further demonstrated reduced exposure of hydrophobic patches after complexation. The complexes exhibited uniformly dispersed particle sizes consistent with model predictions. Critically, rutin and hesperidin improved phycocyanin's UV stability, heat resistance, and color retention. Molecular docking further indicated their binding to critical phycocyanin domains via hydrogen bonding, van der Waals forces, and other interactions. This bidirectional strategy, in which machine learning guides experimental interpretation and experimental results validate predictive outcomes, provides a robust framework for rational design of optimized protein-flavonoid systems in food science.