关键词: 罗伊氏乳杆菌；人工神经网络；遗传算法；生物信息学；冷胁迫；抗冻性能

Abstract: In order to improve the freezing resistance of Lactobacillus reuteri LTR1318 during freeze-drying, four lyoprotectant ingredients, namely skim milk, glutamic acid, fructo-oligosaccharide and sorbitol were identified as main factors affecting the freeze-drying survival rate of this strain by single factor experiments. The survival rates of freeze-dried cells with different protectant combinations designed by Box-Behnken design were used for training and testing to construct a radial basis function neural network model whose fitting degree was 0.984 4. The genetic algorithm was used to optimize the fitting results of neural network with 50 iterations, and the optimal lyoprotectant composition was obtained as follows: skim milk 10.90%, glutamic acid 1.20%, fructo-oligosaccharide 1.30% and sorbitol 0.80%. The optimized formulation gave a freeze-drying survival rate of (95.74 ± 5.07)%. The activities of lactate dehydrogenase and β-galactosidase in freeze-dried L. reuteri with the lyoprotectant increased by 53.19% and 353% compared to the control without the lyoprotectant, respectively, which accounted for 78.25% and 86.43% of the enzyme activities under normal culture conditions, respectively. In addition, the relative expression of cold shock protein gene was slightly increased by 8.29%. The lyoprotectant developed in this study could be a new promising probiotic lyoprotectant. Our results provide a reference and guidance for the development of freeze-dried preparations of lactic acid bacteria and more extensive applications of vacuum freeze-drying technology.

Key words: Lactobacillus reuteri; neural network; genetic algorithm; bioinformatics; cold stress; freezing resistance