Machine learning and molecular docking-driven identification of novel umami peptides from pea protein and electroencephalography (EEG) neural mechanisms analysis of their umami-enhancing effects.

This study developed an integrated strategy combining machine learning prediction, molecular docking, and EEG-based neural characterization to screen and validate novel umami peptides derived from pea protein, including QEGEK, QEEEEQSH, and ATTETVDALR. The results showed that the umami recognition thresholds of the three peptides ranged from 0.203 to 0.328 mM. Sensory evaluation and electronic tongue analysis further indicated that QH8 and AR10 exhibited significant umami-enhancing effects when combined with MSG. EEG analysis provided objective neurophysiological evidence for the umami-enhancing effects of the peptides. Compared with MSG, QH8 and AR10 elicited stronger neural oscillatory responses, particularly in the theta, alpha, and beta bands, and enhanced functional connectivity in the alpha band. Source localization results showed significant activation in the parietal and parieto-occipital regions during umami recognition. In addition, molecular docking indicated that Asp108, Ser148, and Arg277 exhibited high contact frequencies and may represent key receptor-binding site. This innovative approach provides new scientific insights for the future discovery of food-derived umami peptides, taste modulation, and the neural sensory mechanisms underlying umami perception.