Virtual screening of umami peptides during sufu ripening based on machine learning and molecular docking to umami receptor T1R1/T1R3.

Umami peptides might significantly contribute to the taste of sufu. However, the inefficiencies of traditional identification methods had great limitations. This study explored a new approach for umami peptides characterization in sufu. Combining peptidomics with machine learning, 637 umami peptides were identified, with their abundance gradually increased during ripening. These peptides were derived from the hydrolysis of 319 precursor proteins from soybeans at various positions, and over 30 % of them derived from 11 major precursor proteins. Thus, five novel umami peptides (DFEGDV, GRGPTVTDP, NDDRDSYNL, RVPAGTTY, and SDNFEY) in ripened sufu were selected via molecular docking. Results indicated the identified peptides could interact with key residues of the umami receptor T1R1/T1R3 through hydrogen bonding and hydrophobic interactions. Sensory evaluation confirmed their umami taste, with thresholds ranging from 0.22 to 0.38 mmol/L. These results broaden our understanding of umami peptide formation during sufu ripening and provide novel insights into their identification.