Abstract
Gout is a common metabolic disorder caused by hyperuricaemia, with xanthine oxidase (XO) playing a key role in uric acid production. Food-derived XO inhibitors are therefore of interest as candidate chemotypes, although any nutraceutical application will require in-vivo validation. In this study, we developed an integrated in-silico workflow that combines machine learning, molecular docking, and molecular dynamics (MD) to prioritize potential inhibitors. From a library of 3,142 medicine-food homology compounds, we trained multiple fingerprint-algorithm classifiers; the topological-torsion Random Forest (TT-RF) model performed best, achieving an AUC of 0.992 and a precision of 0.98 on a held-out test set. Applying this model yielded 128 high-confidence hits, ten of which showed docking scores ≤ -9.0 kcal/mol. Subsequent 200-ns MD simulations indicated that luteolin-7-glucuronide, 5,4'-dihydroxyflavone, and uralenol form stable, compact complexes with XO. In-vitro assays further confirmed XO inhibition, with IC50 values of 26.15, 39.06, and 34.64 µM, respectively, and negligible cytotoxicity in HepG2 cells up to 100 µM. Together, these results identify food-derived compounds with reproducible in-vitro XO inhibition. Their significance for gout management remains to be established in in-vivo studies. More broadly, this study illustrates a scalable framework for natural-product-based inhibitor discovery that can guide future preclinical validation.