2026/04/03 by Huizi Cui, Cheng Cheng, Yuxuan Wang +4 · 1 voice
Biochemistry, Genetics and Molecular Biology · Engineering · Nursing · #Advanced Chemical Sensor Technologies #Biochemical Analysis and Sensing Techniques #Machine Learning in Bioinformatics
paper · pdf · doi:10.3168/jds.2026-28183
openalex publication_date 2026/04/03 · openalex created_date 2026/04/04 · openalex updated_date 2026/07/11
Bioactive peptides derived from dairy proteins are increasingly recognized for their dual capacities to enhance sensory attributes and provide health-promoting functions. In fermented dairy products such as yogurt, proteolysis releases a diverse array of peptides, including sour-taste peptides that can refine flavor profiles and reduce the need for added acids or synthetic flavorings. However, systematic discovery and characterization of such peptides remain limited by the time and cost of traditional screening methods. Here, we developed an integrated computational-experimental workflow to identify yogurt-derived sour peptides with potential multifunctional applications. A curated dataset of labeled sour and non-sour peptides was analyzed for amino acid composition, sequence length, and physicochemical properties, revealing distinct molecular signatures associated with sourness. Multiple machine learning models, including multilayer perceptron, random forest, and XGBoost, were trained on diverse feature sets, with XGBoost consistently achieving the highest predictive performance (area under the receiver operating characteristic curve >0.95). Four top-predicted candidates (DSEPV, EPVLL, QEPVL, and RYLGYLE) were synthesized and evaluated using an electronic tongue, confirming DSEPV as the most potent sour-taste peptide, with others exhibiting balanced profiles incorporating umami and astringency. To explore potential bioactivities beyond taste, we assessed peptide interactions with the proton channel OTOP1, a protein involved in sour taste transduction and broader physiological processes, using molecular docking and 100-ns molecular dynamics simulations. Stable peptide-protein complexes were observed, with EPVLL showing the most favorable binding free energy, suggesting possible modulation of channel conformation. These findings establish a rapid, cost-effective strategy for discovering flavor-modifying peptides from dairy streams, offering a sustainable route to upcycle fermentation by-products and highlighting the potential for dual-purpose applications in both flavor optimization and functional food development.