Abstract
Umami peptides are naturally occurring bioactive peptides with distinctive umami taste and promising applications in the food industry. However, conventional identification methods based on sensory evaluation, chromatography, and mass spectrometry are time-consuming, costly, and low-throughput. Recent advances in artificial intelligence (AI) have enabled efficient computational frameworks for large-scale umami peptide screening. This review summarizes recent progress in AI-driven umami peptide prediction and mechanistic insights from molecular docking analyses. We overview available umami peptide databases, feature representation strategies, and state-of-the-art AI models, including machine learning, deep learning, and multi-model fusion approaches. In addition, molecular docking studies elucidating the interactions between umami peptides and taste receptors, particularly T1R1/T1R3, are discussed to support rational peptide design. Finally, current challenges and future perspectives in AI-assisted umami peptide research are highlighted, providing guidance for the intelligent development and application of umami peptides in the food industry.