Sustainable probiotic production via AI: medium optimization and metabolic mechanisms in Bifidobacterium animalis ssp. lactis BB-12 using agricultural waste

Bifidobacterium animalis ssp. lactis BB-12 (BB-12) is a well-established probiotic with notable health benefits and broad applications. However, its conventional MRS medium is expensive and poses safety and religious concerns. Agricultural wastes represent sustainable alternatives for microbial cultivation. This study aimed to optimize the BB-12 culture medium using agricultural waste with artificial intelligence (AI) and to investigate their metabolic impact through metabolomic analysis. AI approaches, including RSM (Response Surface Methodology), machine learning, deep learning, and evolutionary optimization methods, were employed to model and optimize the effects of medium composition on OD₆₀₀, growth rate (μ _max), and cost. The optimized media were further evaluated through organic acids analysis and metabolomic profiling to elucidate how variations in nitrogen and carbon sources affect the metabolic responses of BB-12. The O1 medium optimized by Ridge-NSGAII (Non-dominated Sorting Genetic Algorithms II) significantly (p < 0.05) enhanced BB-12 production of lactic, acetic and propionic acids. Metabolomic analysis indicated the involvement of nucleotide salvage, starch and sucrose metabolism, pentose phosphate pathway, and glutathione metabolism. This study highlighted the utility of AI in optimizing BB-12 medium formulations. The optimized medium improved strain performance and enabled the valorization of agricultural wastes, offering a scalable strategy for sustainable probiotic production and advancing the circular bioeconomy.