A comprehensive multi-evidence framework for network pharmacology-based prediction of dietary flavonoid effects

Dietary flavonoids associate with various aspects of disease prevention, yet systematic frameworks integrating computational prediction with experimental and epidemiological evidence remain limited. We develop a multi-tiered network pharmacology framework that quantitatively predicts flavonoid-related therapeutic properties and supports these predictions with integrated computational, experimental, and epidemiological evidence. We constructed a master network of 17,869 human proteins, 14 dietary flavonoids, and 1,496 FDA-approved drugs (278,768 interactions). Flavonoids averaged 45.3 target proteins compared to 16.8 for FDA-approved drugs (2.7-fold higher; p = 7.5 × 10^-4), reflecting multi-target architecture. Statistical analysis using target protein overlap (Fisher's exact test) revealed that 71.4% of flavonoids showed significant associations with cardiovascular drugs and 78.6% with anticancer drugs. Experimental validation in cancer cell models demonstrated high predictive accuracy: flavonoids with strong computational associations to anticancer drugs (luteolin: -log₁₀ p = 10.5; myricetin: -log₁₀ p = 9.9) exhibited potent cytotoxicity (LC₅₀ ~30 μM), whereas weakly associated flavonoids remained inactive (LC₅₀ > 200 μM). Computational association strength explained 84% of the variance in experimental potency (Pearson r = 0.918; R² = 0.843), providing quantitative experimental support of network pharmacology predictions for dietary bioactives. Translating predictions to 506 foods yielded 685 food-ATC therapeutic combinations. Systematic PubMed analysis identified literature-supported evidence for 96 associations (132 unique references), achieving 47.1% cardiovascular predictions showing observational consistency with published studies. Food category analysis identified tomato, cranberry, tea, orange, and blueberry products with strongest evidence (18-40 items). This multi-evidence framework enables evidence-based prediction of dietary polypharmacological effects and provides a computational foundation to generate hypotheses for precision nutrition.