Elucidating the Material Basis and Receptor Mechanism of Bitterness in Castanopsis fissa Honey Using Machine Learning, Metabolomics, and Molecular Docking

= 6), this study investigated bitterness-associated compounds and their potential receptor interactions by integrating sensory evaluation, machine learning, untargeted metabolomics, electronic tongue analysis, targeted UPLC-QQQ-MS/MS quantification, and molecular docking. A Random Forest model combined with untargeted metabolomics screened 71 candidate bitter compounds, among which alkaloid-related metabolites were prominently represented. Electronic tongue analysis showed that several compounds exhibited higher bitterness-related sensor responses than quinine under the tested conditions. Targeted UPLC-QQQ-MS/MS analysis identified and quantified five key compounds, among which kynurenic acid was the most abundant, reaching approximately 4500 ppm (mg/kg). Molecular docking suggested that these compounds could favorably interact with the human bitter taste receptor TAS2R46, with binding affinities ranging from -5.4 to -6.5 kcal/mol, mainly through hydrogen bonding, hydrophobic interactions, and π-related interactions. Overall, this study provides chemical evidence and mechanistic clues for understanding the bitterness of LSZH and offers an integrated analytical framework for screening bitterness-associated compounds in complex food systems.