Machine learning and molecular dynamics reveal anti-osteoporosis potential of dietary flavonoids

Osteoporosis (OP) has imposed a heavy burden on global health. Current treatments often come with side effects, which has spurred interest in dietary bioactive compounds with preventive potential. Flavonoids, typical representatives of both medicine and food, possess nutritional and pharmacological properties related to bone health. However, the key active components and mechanisms of action of their anti-OP effects still require systematic research. Data from the National Health and Nutrition Examination Survey and the Food and Nutrient Database for Dietary Studies were integrated. The flavonoid intake and OP status of 5789 American adults (aged ≥50) were analyzed. Five machine-learning algorithms (Boruta, LassoCV, RFECV, mRMR, and ReliefF) were employed to screen flavonoid sub-classes, followed by ADMET (absorption, distribution, metabolism, excretion, toxicity) analysis. The XGBoost model was used to predict the risk of OP and was validated through the area under the receiver operating characteristic curve and SHapley Additive exPlanations analysis. Network pharmacology was used to identify the common targets of OP and flavonoids, and the binding stability was verified by molecular docking and molecular dynamics simulation. Eleven key flavonoids were identified by machine learning. Five compounds with high bioavailability and low toxicity were prioritized through ADMET screening: daidzein, quercetin, catechin, apigenin, and kaempferol. Five core targets related to the pathogenesis of OP were identified through network pharmacology and protein-protein interaction network: signal transducer and activator of transcription 3, estrogen receptor 1, carbonic anhydrase 2, androgen receptor (AR), and estrogen receptor 2. Molecular docking confirmed strong binding between the 5 screened flavonoids and the core targets, especially the binding between apigenin and AR. Molecular dynamics simulation further indicated excellent stability of the AR-apigenin complex. Daidzein, quercetin, catechin, apigenin, and kaempferol may inhibit bone resorption and promote bone formation by regulating signal transducer and activator of transcription 3, estrogen receptor 1, carbonic anhydrase 2, AR, and estrogen receptor 2. The high stability of the AR-apigenin complex supports the potential of flavonoids as therapeutic agents for OP through a multi-target mechanism and their value in nutritional intervention.