Screening of umami peptides from fermented grains by machine learning, molecular docking and molecular dynamics simulation

Abstract

Listen

Translate article

The intense flavor compounds in baijiu give it a unique taste and umami. This study used metagenomics, machine learning, experimental validation, and molecular docking to identify three candidate umami peptides: EFFSNYGTRV, EFFSNYDTRL, and GCWGRGRLCQW, from fermented grains (jiupei). Sensory evaluation showed that these peptides enhance umami with a threshold range of 0.09–0.116 mmol/L, and EFFSNYDTRL exhibited the best umami properties. Molecular docking and molecular dynamics simulations revealed that T1R1/T1R3 and the three candidate peptides can form stable complexes, with hydrogen bonding and hydrophobic interactions playing essential roles. The study also identified Glu, Tyr, and Ser as the critical residues to which the three candidate peptides bind in the umami receptor T1R1/T1R3. Our study demonstrates the potential of machine learning approaches to mine umami peptides from metagenomics data that can accelerate the development and in-depth study of umami peptides.