Deep Learning-Enhanced Generation and Screening of Antihyperuricemic Peptides from Chickpea Proteins: from Multienzyme Optimization to Molecular Mechanisms

Food-derived bioactive peptides have emerged as promising functional ingredients for hyperuricemia management. However, multienzyme hydrolysis strategies remain underexplored because of inefficient screening methods. Herein, a large language model (LLM)-guided strategy integrating deep learning-assisted enzyme selection with experimental validation was developed to generate antihyperuricemic peptides from chickpea proteins. The optimal enzyme combination (Flavourzyme-Pepsin-Pancreatin) produced a hydrolysate (MGI) with strong xanthine oxidase (XO) inhibitory activity (94.1% at 10 mg/mL), outperforming single-enzyme treatments. MGI retained 91.5% activity after simulated digestion and significantly reduced intracellular uric acid, oxidative stress, and inflammation in HK-2 cells. Molecular docking identified four tripeptides (LLF, GFM, FSF, and SWL) with favorable binding to XO through hydrogen bonding and hydrophobic interactions. This study provides a practical strategy for enhancing peptide bioactivity and supports the development of chickpea-derived peptides for hyperuricemia management.