Identification of natural inhibitors targeting polyphenol oxidase through machine learning-assisted virtual screening and investigation of molecular mechanisms using multispectral experiments

Abstract

The development of potent inhibitors of polyphenol oxidase (PPO) is crucial for preventing enzymatic browning, especially when derived from natural food ingredients. However, there are currently few natural PPO inhibitors available, and the underlying structure-activity relationship remains unclear. In this study, a quantitative structure-activity relationship (QSAR) model utilizing machine learning (ML) was developed and integrated with molecular docking to screen for potential PPO inhibitors from a database of over 70,000 food compounds. Subsequently, 17 structurally diverse compounds were validated through experiments, and it was found that 16 of them could inhibit PPO activity at micromolar concentrations, with ε-viniferin demonstrating the most potent inhibitory effect on PPO at 3.69 μM. Furthermore, some molecular fingerprints that may significantly contribute to PPO inhibitors have been discovered through structure-based analysis. Ultimately, the molecular mechanism was elucidated through a series of enzyme kinetics and multispectral experiments. This study aims to promote the diverse development and widespread application of natural PPO inhibitors in industrial production in the future.