Identification and evaluation of potential probiotics with favorable fermentation and storage characteristics for...

Brown milk beverages have garnered widespread attention due to their unique flavor and potential as a functional matrix for probiotic delivery. This study aimed to identify Lacticaseibacillus paracasei strains suitable for fermenting brown milk beverages by combining artificial intelligence-driven predictive screening with experimental validation and metabolomic analysis. Using the iProbiotics platform, 27 L. paracasei strains underwent initial screening, yielding seven candidates for further evaluation. These trains were systematically evaluated for gastrointestinal tolerance and bile salt tolerance, fermentation performance, storage stability, and metabolic profiles. Strains PC724 and PC646 demonstrated superior probiotic and technological properties, outperforming the commercial control strain LC-01 in acidification capacity, viable cell counts, and textural attributes. Over 35 days of refrigerated storage, both strains maintained high viability (>10⁸ CFU/mL) and stable metabolic activity, confirming remarkable storage stability. Non-targeted metabolomics revealed sustained enrichment of carbohydrate metabolism, amino acid metabolism, and global metabolic regulation in PC724 and PC646. These metabolic signatures likely underpin their robust stress tolerance long-term survival. Collectively, integrating artificial intelligence prediction with metabolomic profiling provides a scalable strategy for identifying industrially viable probiotic candidates, offering a valuable framework for developing functional fermented milk beverages.