Enzyme-specific release dynamics of umami and bitter peptides from shrimp processing by-products: a machine learning-aided study.

Shrimp processing annually generates substantial underutilized by-products (esp. head and shell). This study aimed to integrate in silico, machine learning and experimental strategies to investigate the dynamic release of umami and bitter peptides during enzymatic hydrolysis of shrimp by-products. Bromelain and Flavourzyme were selected via in silico and empirical analysis. Degree of hydrolysis (DH), molecular weight distribution, amino acid composition, peptide profile, and taste characteristics were determined. After 7 h, Flavourzyme achieved higher DH (33.3% vs. 8.4%) and yielded more <1 kDa peptides (89.17% vs. 76.02%, P < 0.05) than bromelain. Flavourzyme-treated hydrolysates showed increased umami amino acids and reduced bitter amino acids. E-tongue and sensory evaluation confirmed stronger umami in Flavourzyme-treated samples throughout hydrolysis. Peptidomics (Peptigram) and machine learning analyses indicated Flavourzyme preferentially released umami peptides and progressively degraded bitter peptides. These findings provide a theoretical basis for enzyme-specific production of umami peptides and value-added utilization of shrimp by-products.