Various machine learning approaches coupled with molecule simulation in the screening of natural compounds with xanthine oxidase inhibitory activity

Abstract

Gout is a common inflammatory arthritis associated with various comorbidities, such as cardiovascular disease and metabolic syndrome. Xanthine oxidase inhibitors (XOIs) have emerged as effective substances to control gout. Much attention has been given to the search for natural XOIs. In this study, a molecular database of natural XOIs was created for modeling purposes. Quantitative structure–activity relationship models were developed by combining various machine learning approaches and three descriptor pools. The models revealed several features of XOIs, including hydrophobicity and steric molecular structures. Experimental results showed the xanthine oxidase (XO) inhibitory activity of predicted compounds. Vanillic acid was identified as a promising new XOI candidate, with an IC₅₀ of 0.593 μg mL⁻¹. The functions of hydrogen bonds and hydrophobic interactions in XO activity inhibition were confirmed by molecular docking. This study fills knowledge gaps pertaining to the discovery of natural XOIs and to the interaction mechanisms between XOIs and XO.