Machine learning-based exploration of Umami peptides in Pixian douban: Insights from virtual screening, molecular docking, and post-translational modifications

Abstract

Pixian Doubanjiang (PXDB)'s distinctive umami profile is primarily attributed to its unique peptides; however, their structural characteristics, sensory mechanisms, and biosynthetic pathways during aging remain poorly understood. This study employed a machine learning-based approach to investigate umami peptides in 1–2 year aged PXDB. We identified 117 peptides, predicting 69 with umami potential. Sensory analysis confirmed VEGGLR's remarkably low umami threshold (0.22 mmol/L). Molecular docking further elucidated VEGGLR's interaction with T1R1/T1R3 receptors via salt bridges and hydrogen bonds, enhancing umami perception. Observed post-translational modifications, including phosphorylation and acetylation on protein N6U2M1/N6UWT4, suggest a potential regulatory role in umami peptide biosynthesis. These findings offer key molecular insights into PXDB umami development, enhancing our understanding of its flavor chemistry.