Method for optimizing formula of lactobacillus reuteri composite freeze-drying protective agent based on neural network model

Abstract

The invention discloses a method for optimizing a formula of a lactobacillus reuteri composite freeze-drying protective agent based on a neural network model, and belongs to the technical field of optimization of the formula of the composite freeze-drying protective agent. The method comprises the following steps: by taking lactobacillus reuteri as a test strain and the freeze-drying survival rate of the strain as an evaluation index, firstly, carrying out a single-factor test on a candidate lactobacillus reuteri composite freeze-drying protective additive; on the basis, the lactobacillus reuteri composite freeze-drying protective agent which has obvious influence on the freeze-drying survival rate of the strain is screened out through a Plackett-Burman test; finally, a Box-Behnken response surface test and a BP-GA neural network model optimization test are combined, and formula optimization of the lactobacillus reuteri composite freeze-drying protective agent is completed. According to the method, on the basis of a single-factor test, a Plackett-Burman test and a BP-GA neural network model are introduced, the defects of traditional experimental design can be systematically made up, and the globality, nonlinear adaptability and robustness of formula optimization are remarkably improved, so that the direct vat set leavening agent with high survival efficiency is obtained.

Description

Technical Field

[0001] This invention belongs to the field of compound freeze-drying protectant formulation optimization technology, specifically involving a method for optimizing the formulation of Lactobacillus reuteri compound freeze-drying protectant based on a neural network model. Background Technology

[0002] Lactobacillus reuteri is abundant in the gastrointestinal tracts of mammals and is an important component of the beneficial gut microbiota in humans, contributing to the stability of the human gastrointestinal microecology. As a probiotic with numerous health benefits, Lactobacillus reuteri shows great application potential in the food, pharmaceutical, and health product industries. However, in current production and application, the use of Lactobacillus reuteri as a direct-inoculation starter for yogurt is relatively rare, mainly due to problems such as low survival rate after freeze-drying, long fermentation time, and poor yogurt quality. The low freeze-drying survival rate not only increases production costs but also limits the number of live bacteria in the product, thus affecting the quality of fermented yogurt. The long fermentation time results in low production efficiency, failing to meet the growing market demand for this strain. Therefore, developing a method for preparing a direct-inoculation starter that improves the fermentation efficiency, acid production capacity, and water-holding capacity of Lactobacillus reuteri in yogurt preparation is an urgent practical need.

[0003] Direct-inoculation starter cultures refer to bacterial or microbial cultures that can be directly applied to the fermentation of yogurt, cheese, and other fermented products. Direct-inoculation starter cultures directly impact the quality of fermented products. Using them effectively maintains strain activity, reduces the probability of strain degeneration or mutation, significantly shortens the fermentation cycle in factory production, improves product quality, protects consumer health, and enriches consumers' dietary needs. Currently, the most commonly used direct-inoculation starter culture is a freeze-dried bacterial powder preparation. The key process in preparing this powder is vacuum freeze-drying of the bacterial strain to obtain the freeze-dried powder. Therefore, the quality of the starter culture is directly related to the survival rate of the bacterial strain during the vacuum freeze-drying process. Vacuum freeze-drying primarily involves freezing the wet material below its eutectic point, at which point the water inside the material solidifies into ice. Through appropriate vacuum and temperature, the ice sublimates directly into water vapor, which is then condensed by a condenser inside the apparatus to obtain the dried product. Vacuum freeze-drying can cause some damage to bacterial cells. Therefore, maintaining the viability of the strain to the greatest extent possible during freeze-drying is extremely important. Adding a suitable freeze-drying protectant before freeze-drying can greatly improve the survival rate of the strain and maintain its viability. A freeze-drying protectant is a protective substance added to the bacterial cells before vacuum freeze-drying. It can maintain the osmotic pressure balance inside and outside the cell, stabilize the protein structure, and reduce the contact between the cell and the external environment. Through these effects, it can improve the survival rate of the strain and maintain its activity.

[0004] Currently, most freeze-drying protectants are optimized through single-factor orthogonal experiments or single-factor response surface methodology experiments. Orthogonal experiments are essentially screening designs to obtain "discrete optimal combinations," but they cannot approximate extreme points through continuous models like response surface methodology experiments. Response surface methodology has disadvantages such as multiple experimental groups, long cycles, low efficiency, and difficulty in reproducibility. Therefore, the survival rate of bacteria in freeze-drying protectants screened through single-factor orthogonal experiments or single-factor response surface methodology experiments is low, resulting in low fermentation efficiency, acid production capacity, and water holding capacity of yogurt. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, the present invention aims to provide a method for optimizing the formulation of Lactobacillus reuteri compound freeze-drying protectant based on a neural network model, so as to solve the technical problem that the existing screening methods cannot guarantee the survival rate of bacteria in the freeze-drying protectant.

[0006] To achieve the above objectives, the present invention employs the following technical solution: The first aspect of this invention discloses a method for optimizing the formulation of a compound lyophilization protectant based on a neural network model, using *Lactobacillus reuteri* (…). Limosilactobacillus reuteri Using the strain as the test strain and its freeze-drying survival rate as the evaluation index, a single-factor experiment was first conducted on the candidate *Lactobacillus reuteri* compound freeze-drying protectant. Based on the results of the single-factor experiment, the Plackett-Burman experiment was used to screen out the *Lactobacillus reuteri* compound freeze-drying protectant that had a significant impact on the freeze-drying survival rate of the strain. Finally, the formulation optimization of the *Lactobacillus reuteri* compound freeze-drying protectant was completed by combining the BP-GA neural network model optimization experiment.

[0007] Preferably, the survival rate detection method is as follows: *Lactobacillus reuteri* bacterial sludge is mixed with *Lactobacillus reuteri* compound freeze-drying protectant and then freeze-dried to obtain bacterial powder; the bacterial powder is rehydrated with an equal volume of sterile PBS solution before freeze-drying, mixed well, diluted with a bacterial suspension, and then spread onto MRS solid medium. The survival rate is counted according to the freeze-drying survival rate formula; the freeze-drying survival rate formula is: .

[0008] Preferably, the specific steps of the single-factor experiment are as follows: The *Lactobacillus reuteri* bacterial sludge, *Lactobacillus reuteri* compound freeze-drying protectant, and distilled water are mixed at a mass-to-volume ratio of 1:5 to prepare solutions of different concentrations. These solutions are then freeze-dried to obtain bacterial powder. After rehydrating the bacterial powder by an equal volume, the freeze-drying survival rate of the strains under different components and concentrations is calculated.

[0009] More preferably, the *Lactobacillus reuteri* complex lyophilization protectant is one or more of the following: a basic protectant, a sugar, an alcohol, and an antioxidant.

[0010] More preferably, the basic protective agent is skim milk powder; the sugar is trehalose, fructooligosaccharide, or sucrose; the alcohol is glycerol, mannitol, or sorbitol; and the antioxidant is ascorbic acid or L-cysteine ​​hydrochloride.

[0011] Preferably, the Plackett-Burman experiment steps are as follows: based on the results of the single-factor experiment, the significant factors affecting the freeze-drying survival rate are screened out, and each factor is taken at two levels, the lowest and the highest, with -1 as the lowest level and 1 as the highest level, to screen the main effect factors.

[0012] Preferably, the steps of the Box-Behnken response surface methodology are as follows: using the results obtained from the Plackett-Burman experiment as the independent variable of the Box-Behnken experiment, and designing the experiment with the freeze-dried survival rate as the response value.

[0013] Preferably, the BP-GA neural network model optimization experiment is as follows: a single hidden layer BP neural network model is constructed using MATLAB 2024a. The significant factors affecting the freeze-drying survival rate, determined through single-factor experiments and Plackett-Burman experiments, are used as the input layer neurons of the BP neural network model, and the freeze-drying survival rate of *Lactobacillus reuteri* is used as the output layer neurons. After repeated testing to determine the topology, the BP-GA neural network model is constructed using a genetic algorithm toolbox. Based on the BP-GA neural network model, the formulation of the compound freeze-drying protectant for *Lactobacillus reuteri* strain is optimized.

[0014] A second aspect of the present invention discloses the application of the above method in the preparation of direct-inoculation fermentation agents.

[0015] A third aspect of the present invention discloses the application of the above method in the preparation of yogurt.

[0016] A fourth aspect of the present invention discloses a method for preparing yogurt, comprising the following steps: 1) Using the above method, the optimal formulation of the compound lyophilization protectant for *Lactobacillus reuteri* was obtained; 2) Mix *Lactobacillus reuteri* bacterial sludge and *Lactobacillus reuteri* compound freeze-drying protectant at a mass-to-volume ratio of 1:5, freeze-dry, and obtain *Lactobacillus reuteri* direct-inoculation fermentation agent; the formula of the *Lactobacillus reuteri* compound freeze-drying protectant is obtained by the method in step 1). 3) Add white sugar to milk, mix well, pasteurize, cool, and inoculate with the Lactobacillus reuteri direct inoculation starter obtained in step 2) to obtain yogurt.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for optimizing the formulation of a compound lyophilization protectant based on a neural network model, using *Lactobacillus reuteri* (…). Limosilactobacillus reuteri This study optimized the formulation of freeze-drying protectants. Based on single-factor experiments, Plackett-Burman experiments were conducted to screen for significant factors affecting freeze-drying survival rate. Box-Behnken response surface methodology was used to optimize the dosage of each protectant. Furthermore, a BP-GA neural network model was introduced for optimization experiments, systematically compensating for the shortcomings of traditional Box-Behnken response surface methodology and significantly improving the globality, nonlinear adaptability, and robustness of formulation optimization, thereby obtaining a direct-inoculation starter with high survival efficiency. The dual optimization of Box-Behnken response surface methodology and BP-GA neural network model maximized the synergistic effect of each component of the protectant. Experiments demonstrated that the freeze-drying protectant formulated using this method achieved a higher freeze-drying survival rate than the freeze-drying protectant formulation obtained through a combination of traditional single-factor and response surface methodology experiments (skim milk only), thus reducing production losses. The Lactobacillus reuteri direct-inoculation starter culture prepared based on this can improve the fermentation efficiency, acid production capacity and water holding capacity of yogurt, reduce whey separation in yogurt, give yogurt a thicker texture, improve yogurt quality, ensure a sufficient number of live bacteria in the product, and greatly enhance product quality and market competitiveness. Attached Figure Description

[0018] Figure 1 This is a graph showing the results of screening basic protective agents according to the present invention; Figure 2 This is a graph showing the results of screening carbohydrates according to the present invention; Figure 3 This is a graph showing the results of screening alcohols according to the present invention; Figure 4 This is a graph showing the results of screening L-cysteine ​​hydrochloride according to the present invention; Figure 5 This is a graph showing the results of the screening of ascorbic acid according to the present invention; Figure 6 The graph shows the optimization results of the BP-GA neural network model of the present invention; the left side is a bar chart of the optimal formula parameters, and the right side is a scatter plot of the model prediction accuracy. Figure 7 This is a photograph of the fermented yogurt of the present invention; In the figure, a, b, and c represent significant differences. p<0.05). Detailed Implementation

[0019] To enable those skilled in the art to understand the features and effects of the present invention, the following descriptions and definitions are only general descriptions of the terms and expressions mentioned in the specification and claims. Unless otherwise specified, all technical and scientific terms used herein have the ordinary meaning understood by those skilled in the art regarding the present invention, and in the event of any conflict, the definitions in this specification shall prevail.

[0020] The theories or mechanisms described and disclosed herein, whether right or wrong, should not in any way limit the scope of the invention, that is, the contents of the invention can be implemented without being limited by any particular theory or mechanism.

[0021] In this document, all features defined by numerical ranges or percentage ranges, such as numerical values, quantities, contents, and concentrations, are for the sake of brevity and convenience only. Accordingly, descriptions of numerical ranges or percentage ranges should be considered as covering and specifically disclosing all possible sub-ranges and individual numerical values ​​(including integers and fractions) within those ranges.

[0022] In this article, unless otherwise specified, “contains,” “includes,” “containing,” “has,” or similar terms cover the meanings of “composed of” and “mainly composed of,” for example, “A contains a” covers the meanings of “A contains a and others” and “A contains only a.”

[0023] For the sake of brevity, not all possible combinations of the technical features in each implementation scheme or embodiment are described herein. Therefore, as long as there is no contradiction in the combination of these technical features, the technical features in each implementation scheme or embodiment can be combined arbitrarily, and all possible combinations should be considered within the scope of this specification.

[0024] This invention provides a method for optimizing the formulation of a compound lyophilization protectant based on a neural network model, using *Lactobacillus reuteri* (… Limosilactobacillus reuteri Using the *Lactobacillus reuteri* strain as the test strain and its survival rate after freeze-drying as the evaluation index, single-factor experiments were first conducted on candidate *Lactobacillus reuteri* compound freeze-drying protectants (including basic protectants, sugar protectants, alcohol protectants, and antioxidants). Based on the results of these single-factor experiments, Plackett-Burman experiments were used to screen for *Lactobacillus reuteri* compound freeze-drying protectants that significantly affected the freeze-drying survival rate of the strains. Finally, the formulation optimization of the *Lactobacillus reuteri* compound freeze-drying protectant was completed by combining Box-Behnken response surface methodology and BP-GA neural network model optimization experiments.

[0025] The present invention will be further illustrated below using *Lactobacillus reuteri* (accession number CGMCC No. 24755) as an example. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading this description, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0026] The following examples use conventional instruments and equipment in the art. The *Lactobacillus reuteri* strain used in the following examples is deposited at the China General Microbiological Culture Collection Center (CGMCC) with accession number CGMCC No. 24755 and publication number CN114990023A. This strain can be obtained through public channels such as cultural heritage institutions. Other experimental methods, unless otherwise specified, are generally performed under conventional conditions or according to the manufacturer's recommendations. All raw materials used in the following examples are conventional commercially available products with specifications in the art, unless otherwise stated. The optimized culture medium formulation used in the following examples is: 37.8 g sucrose, 38.8 g soybean peptone, 2.5 g sodium citrate, 1 g citric acid, 60 mg L-cysteine ​​hydrochloride, 5 g beef meal, 4 g yeast extract, 0.2 g magnesium sulfate, 0.05 g manganese sulfate, and 1 mL Tween 80, with distilled water to a final volume of 1 L.

[0027] I. A Method for Optimizing the Formulation of a Compound Lyophilization Protectant for Lactobacillus reuteri Based on a Neural Network Model 1. Preparation of Lactobacillus reuteri mycelium mud Lactobacillus reuteri (CGMCC No. 24755) was inoculated into MRS solid medium for activation and cultured at 37°C for 12 h. Single colonies were picked and transferred to optimized medium and cultured overnight at 37°C. The culture was then inoculated into fresh optimized medium at a 10% (v / v) inoculation rate and cultured for another 8 h. The culture was centrifuged at 4500×g for 20 min, the supernatant was discarded, and the culture was washed with sterile PBS to obtain Lactobacillus reuteri sludge.

[0028] 2. Single-factor experiment The *Lactobacillus reuteri* bacterial sludge obtained in step 1 was uniformly mixed with the protectant at a ratio of 1:5 (w / v), then dispensed and pre-frozen at -20℃ for 12 h. After pre-freezing, it was placed in a vacuum freeze dryer and freeze-dried at -50℃ and a vacuum degree <15 Pa for 24 h to obtain a flocculent dried bacterial powder. The obtained bacterial powder was rehydrated with an equal volume of sterile PBS and thoroughly shaken to obtain a bacterial suspension. 100 μL of the bacterial suspension was accurately pipetted and diluted to three appropriate dilution factors, and then spread onto MRS solid medium. The medium was placed in a constant temperature incubator at 37℃ for 24 h. After incubation, the medium was removed and the survival rate was counted according to the following freeze-drying survival rate formula:

[0029] The specific steps are as follows: 1) Effect of basic protectant concentration on the freeze-dried survival rate of strains Using skim milk powder as the base protectant, the *Lactobacillus reuteri* bacterial sludge prepared in step 1 was mixed evenly with 0%, 5%, 10%, 15%, 20%, and 25% skim milk solutions prepared from skim milk powder and distilled water at a ratio of 1:5 (w / v). The mixtures were then pre-frozen at -20℃ for 12 h. After pre-freezing, the mixtures were placed in a vacuum freeze dryer and freeze-dried at -50℃ under a vacuum of <15 Pa for 24 h to obtain a flocculent dried bacterial powder. The obtained bacterial powder was rehydrated by an equal volume, and its freeze-dried survival rate was determined. Three parallel experiments were performed to select the protectant and concentration with the best protective effect against *Lactobacillus reuteri*.

[0030] The results are as follows Figure 1 As shown, comparing the survival rates of bacterial strains with different concentrations of preservatives, it can be seen that the highest survival rate of 26.42% is achieved when the skim milk concentration is 20%. This is because skim milk is a macromolecular preservative, which forms a protective film on the cell surface, thereby mitigating the damage to bacterial cells caused by changes in the external environment. At the same time, it can also enhance the freeze-drying protection effect of low-molecular-weight substances on *Lactobacillus reuteri*.

[0031] 2) Effects of sugar content and concentration on the freeze-drying survival rate of bacterial strains Using skim milk solution as the base protectant, sugars (trehalose / sucrose / fructooligosaccharides) at concentrations of 2%, 4%, 6%, 8%, and 10% were mixed with the skim milk solution to obtain sugar-containing protectants of different concentrations. The *Lactobacillus reuteri* bacterial sludge prepared in step 1 was mixed evenly with different concentrations of sugar-containing protectants at a ratio of 1:5 (w / v) of bacterial sludge to sugar-containing protectant, and pre-frozen at -20℃ for 12 h. After pre-freezing, it was placed in a vacuum freeze dryer and freeze-dried at -50℃ and a vacuum degree <15 Pa for 24 h to obtain a flocculent dried bacterial powder. The obtained bacterial powder was rehydrated by an equal volume, and its freeze-dried survival rate was measured. Three parallel experiments were performed to select the protectant and concentration with the best protective effect on the bacterial strain.

[0032] The results are as follows Figure 2 As shown, comparing the survival rates of bacterial strains corresponding to different concentrations of sugar-containing protective agents, it can be seen that the fructooligosaccharide group had the highest survival rate, reaching 73.28% at a fructooligosaccharide concentration of 8%. This is because the molecular structure of sugars contains multiple hydroxyl groups, which can form hydrogen bonds with proteins to increase protein stability. Furthermore, by binding with water molecules in the solution, the free water is reduced, the viscosity of the solution is increased, and the growth process of crystal nuclei is slowed down, resulting in smaller ice crystals and ultimately reducing cell damage.

[0033] 3) Effects of alcohols and their concentration on the freeze-drying survival rate of bacterial strains Using skim emulsion as the base protectant, polyols (glycerol / sorbitol / mannitol) at concentrations of 1%, 2%, 3%, 4%, and 5% were mixed with the skim emulsion to obtain alcohol-containing protectants of different concentrations. The *Lactobacillus reuteri* bacterial sludge prepared in step 1 was mixed evenly with different concentrations of alcohol-containing protectants at a ratio of 1:5 (w / v) of bacterial sludge to alcohol-containing protectant, and pre-frozen at -20℃ for 12 h. After pre-freezing, it was placed in a vacuum freeze dryer and freeze-dried at -50℃ and a vacuum degree <15 Pa for 24 h to obtain a flocculent dried bacterial powder. The obtained bacterial powder was rehydrated by an equal volume, and its freeze-dried survival rate was measured. Three parallel experiments were performed to select the protectant and concentration with the best protective effect on the bacterial strain.

[0034] The results are as follows Figure 3 As shown, comparing the survival rates of bacterial strains corresponding to different concentrations of alcohol-containing preservatives, it can be seen that the glycerol group had the highest survival rate, reaching 64.39% at a glycerol concentration of 4%. This is because alcohols, as permeable preservatives, can enter cells and inhibit the formation of ice crystals, thereby mitigating cell damage during the freeze-drying process.

[0035] 4) Effect of antioxidants on the lyophilized survival rate of bacterial strains a. Using skim milk solution as the base protectant, L-cysteine ​​hydrochloride at concentrations of 0.5%, 1.0%, 1.5%, 2.0%, and 2.5% was mixed with the skim milk solution to obtain protectants containing L-cysteine ​​hydrochloride at different concentrations. The *Lactobacillus reuteri* bacterial sludge prepared in step 1 was mixed evenly with different concentrations of L-cysteine ​​hydrochloride protectants at a ratio of 1:5 (w / v) of bacterial sludge to L-cysteine ​​hydrochloride protectant, and pre-frozen at -20℃ for 12 h. After pre-freezing, it was placed in a vacuum freeze dryer and freeze-dried at -50℃ under a vacuum of <15 Pa for 24 h to obtain a flocculent dried bacterial powder. The obtained bacterial powder was rehydrated by an equal volume, and its freeze-dried survival rate was determined. Three parallel experiments were performed to select the protectant and concentration with the best protective effect on the bacterial strain.

[0036] b. Using skim milk solution as the base protectant, ascorbic acid at concentrations of 1.0%, 1.5%, 2.0%, 2.5%, and 3% was mixed with the skim milk solution to obtain protectants containing ascorbic acid at different concentrations. The *Lactobacillus reuteri* bacterial sludge prepared in step 1 was mixed evenly with different concentrations of protectants containing ascorbic acid at a ratio of 1:5 (w / v) of bacterial sludge to protectant containing ascorbic acid. The mixture was then pre-frozen at -20℃ for 12 h. After pre-freezing, it was placed in a vacuum freeze dryer and freeze-dried at -50℃ under a vacuum of <15 Pa for 24 h to obtain a flocculent dried bacterial powder. The obtained bacterial powder was rehydrated by an equal volume, and its freeze-dried survival rate was determined. Three parallel experiments were performed to select the protectant and concentration with the best protective effect on the bacterial strain.

[0037] The results for L-cysteine ​​hydrochloride are as follows: Figure 4 As shown, comparing the bacterial survival rates corresponding to different concentrations of L-cysteine ​​hydrochloride, it can be seen that the highest bacterial survival rate (44.30%) was achieved when the L-cysteine ​​hydrochloride concentration was 2.0%. The ascorbic acid results are as follows... Figure 5 As shown, comparing the bacterial survival rates corresponding to different concentrations of ascorbic acid, it can be seen that the highest bacterial survival rate (44.48%) is achieved when the ascorbic acid concentration is 2.5%. This is because during freeze-drying and storage, intracellular proteins may undergo oxidation with oxygen in the air, leading to deterioration. Adding an appropriate amount of antioxidant can improve the stability of freeze-dried samples and extend their shelf life.

[0038] 3. Plackett-Burman test Based on the single-factor results, significant factors affecting freeze-drying survival rate were identified. For each factor, two levels, the lowest (-1) and the highest (1), were selected for main effect factor screening. The specific steps are as follows: First, based on the single-factor results obtained in step 2, the lyophilization survival rate of *Lactobacillus reuteri* (CGMCC No. 24755) was used as the response value to evaluate four influencing factors: fructooligosaccharides, glycerol, L-cysteine ​​hydrochloride, and ascorbic acid. Significant factors affecting the lyophilization survival rate were screened, with each factor having two levels: the lowest (-1) and the highest (1), for a total of 12 experiments. Finally, the Plackett-Burman factor variance analysis results were obtained (Table 1).

[0039] Table 1 shows the results of this model. p A value less than 0.01 indicates that the fitted model is highly significant. Fructooligosaccharides, glycerol, and L-cysteine ​​hydrochloride were selected based on importance ranking. Therefore, subsequent response surface methodology optimization experiments focused on investigating the interaction between these three factors and the basic protective agent (skimmed milk). L. reuteri The effect of freeze-drying survival rate of CGMCC strain 24755.

[0040] Table 1. Results of ANOVA for the Plackett-Burman Experiment.

[0041] 4. Box-Behnken response surface methodology Based on the importance ranking of Plackett-Burman assays, three substances—fructooligosaccharides, glycerol, and L-cysteine ​​hydrochloride—were selected for subsequent response surface methodology (RSM) optimization experiments. The experiments focused on investigating the interaction between these three factors and the basic protectant (skimmed milk) on the freeze-dried survival rate of *Lactobacillus reuteri* (accession number CGMCC No. 24755).

[0042] The effects of a composite freeze-drying protectant on the survival rate of *Lactobacillus reuteri* (CGMCC No. 24755) were investigated. Skim milk concentration (A), fructooligosaccharide (B), glycerol (C), and L-cysteine ​​hydrochloride (D), selected by single-factor experiments and Plackett-Burman experiments, were used as independent variables in the Box-Behnken experiment, with freeze-drying survival rate as the response value. The Box-Behnken response surface methodology was used for optimization (Table 2).

[0043] Table 2 Regression Model and Analysis of Variance for Box-Behnken Response Surface Experiment

[0044] Multiple regression analysis was performed on the response surface methodology to obtain the regression equation for the freeze-dried survival rate of Lactobacillus reuteri (CGMCC No. 24755) (represented by Y) in relation to the amount of skim milk A, fructooligosaccharides B, glycerol C, and L-cysteine ​​hydrochloride D in the culture medium: Y = 72.99 - 3.71A - 1.01B - 3.14C - 3.38D - 0.3725AB - 0.7625AC - 3.48AD + 0.2875BC - 2.25BD - 2.39CD - 5.16A² - 9.62B² - 7.49C² - 8.98D².

[0045] Using Design Expert 13 software for optimization, the optimal conditions were obtained as follows: 19.68 g skim milk, 6.23 g fructooligosaccharides, 3.54 g glycerol, 1.65 g L-cysteine ​​hydrochloride, and 100 mL distilled water. Under these conditions, the predicted response value (lyophilized survival rate) can reach 74.0407%.

[0046] 5. BP-GA Neural Network Model Experiment A backpropagation (BP) neural network model was constructed using the Neural Network Tool in MATLAB R2024a. The model consisted of one input layer, one hidden layer, and one output layer. The concentrations of skim milk, fructooligosaccharides, glycerol, and L-cysteine ​​hydrochloride, determined through single-factor experiments and Plackett-Burman experiments, were used as the four input neurons of the BP neural network model, with one hidden layer in the middle. The lyophilized survival rate of *Lactobacillus reuteri* (CGMCC No. 24755) served as one output neuron. After multiple training iterations using MATLAB, the optimal network topology for this experiment was determined to be 4-10-1 with 10 neurons in the hidden layer. 80% of the Box-Behnken data was used for training, 10% for testing, and 10% for validation to complete the self-learning training and prediction of the BP artificial neural network.

[0047] The genetic algorithm toolbox in MATLAB was used to simulate natural phenomena such as reproduction, hybridization, variation, competition, and selection in species for genetic algorithm optimization. Artificial neural network simulations were performed on four factors during the freeze-drying process: skim milk, fructooligosaccharides, glycerol, and L-cysteine ​​hydrochloride, within their respective addition ranges. The model fit values ​​were used to establish an individual fitness function for global optimization. Higher strain survival rates resulted in higher individual fitness values. The maximum number of generations was set to 100, the number of binary bits for variables was 8, the mutation probability was 0.01, the crossover probability was 0.8, the generation gap was 0.25, and other parameters were set to default values. The constructed BP neural network structure was combined with the GA genetic algorithm to build a BP-GA neural network model; the BP-GA neural network model was then used to optimize the freeze-drying protectant formulation for the strains.

[0048] The optimal freeze-drying protectant formulation for *Lactobacillus reuteri* strain (CGMCC No. 24755) was predicted using a constructed neural network model. The optimal formulation was found to be: 15 g skim milk powder, 9.33 g fructooligosaccharides, 5 g glycerol, 1.98 g L-cysteine ​​hydrochloride, and 100 mL distilled water, with a predicted freeze-drying survival rate of 82.27%. Figure 6 ).

[0049] II. Preparation of Lactobacillus reuteri direct-inoculation fermentation agent using an optimized Lactobacillus reuteri composite lyophilization protectant. According to the ratio of bacterial sludge to protectant of 1:5 (W / V), the *Lactobacillus reuteri* bacterial sludge was mixed evenly with the protectant formulated by the neural network model in step one, and pre-frozen at -20℃ for 12 h. After pre-freezing, it was placed in a vacuum freeze dryer and freeze-dried at -50℃ and vacuum degree <15 Pa for 24 h to obtain the direct-inoculation fermentation agent of *Lactobacillus reuteri*.

[0050] III. Evaluation of Optimization Methods According to the ratio of bacterial sludge to protectant of 1:5 (w / v), *Lactobacillus reuteri* bacterial sludge was mixed evenly with the protectant of the response surface methodology (RSM) optimized formulation and the neural network model optimized formulation, and then pre-frozen at -20℃ for 12 h. After pre-freezing, it was placed in a vacuum freeze dryer and freeze-dried at -50℃ and a vacuum degree <15 Pa for 24 h to obtain the corresponding flocculent dried bacterial powder. The obtained bacterial powder was rehydrated by equal volume and its freeze-drying survival rate was measured. Three parallel experiments were performed to select the protectant and concentration with the best protective effect on the strain, so as to compare the freeze-drying survival rate of the RSM optimized formulation and the neural network model optimized formulation.

[0051] The comparison results are shown in Table 3. By comparing the two optimization models, it can be seen that the BP-GA neural network optimization model has a higher degree of fit to the actual experimental values, smaller relative error, and a higher freeze-drying survival rate of the strains. In conclusion, the BP-GA neural network model is more effective in optimizing the freeze-drying protectant formulation.

[0052] Table 3 Comparison of predicted and actual values ​​under BP-GA neural network model optimization and response surface optimization models

[0053] IV. Application of Lactobacillus reuteri direct-inoculation starter Example 1 A method for fermenting yogurt using Lactobacillus reuteri direct inoculation starter includes the following steps: 1. Preparation of direct-inoculation fermentation agent of Lactobacillus reuteri A compound lyophilization protectant for *Lactobacillus reuteri* was prepared by mixing 15 g of skim milk powder, 9.33 g of fructooligosaccharides, 5 g of glycerol, 1.98 g of L-cysteine ​​hydrochloride, and 100 mL of distilled water. *Lactobacillus reuteri* bacterial sludge and the compound lyophilization protectant were mixed evenly at a ratio of 1:5 (w / v) and pre-frozen at -20℃ for 12 h. After pre-freezing, the mixture was placed in a vacuum freeze dryer and freeze-dried at -50℃ under a vacuum of <15 Pa for 24 h to obtain a direct-inoculation fermentation broth for *Lactobacillus reuteri*.

[0054] 2. Preparing yogurt Add 7 wt% white sugar to 250 mL of fresh milk and stir continuously until well mixed. Pasteurize and cool to approximately 40°C. Inoculate with 6 wt% of the *Lactobacillus reuteri* direct-inoculation starter culture prepared in step 1, ensuring complete dissolution. Dispense into sterilized yogurt cups and incubate at 38°C until curd forms. After curdling, transfer to a 4°C refrigerator for 12 hours of post-fermentation to obtain yogurt. The fermented yogurt is shown below. Figure 7 As shown.

[0055] Comparative Example 1 The difference from Example 1 is that only 20 wt% skim milk was added as a preservative. The fermented yogurt is as follows: Figure 7 As shown.

[0056] V. Quality Assessment of Yogurt Produced Using Lactobacillus reuteri Direct Inoculation Starter The quality of the yogurts prepared in Example 1 and Comparative Example 1 was examined, and the steps were as follows: 1. pH and acidity detection The pH value of each group of yogurt samples was determined using a pH meter; the titratable acidity was determined using the phenolphthalein indicator method in GB5009.239-2016.

[0057] 2. Water-holding capacity test First, take a 15 mL centrifuge tube, and take 5 g of yogurt from each group of yogurt samples. Centrifuge at 5000 r / min for 20 min at room temperature, and discard the supernatant. Then calculate the water-holding capacity using the following formula:

[0058] 3. Viable bacteria count detection After diluting the fermented yogurt of each group to a certain ratio with sterile water, viable bacteria were counted using MRS agar medium and incubated at 38°C for 24 h to calculate the viable bacteria count.

[0059] Table 4 Quality Assessment of Fermented Yogurt

[0060] The test results for each group are shown in Table 4. It can be seen that, under the same inoculation amount of direct-inoculation starter, the yogurt prepared in Example 1 group is superior to that in Comparative Example 1 in terms of fermentation completion time, pH, acidity, water holding capacity, and viable cell count. This indicates that the freeze-drying protectant optimized based on the neural network model can improve the quality of the compound freeze-drying protectant formulation. Consequently, the prepared Reuteria mucilage CGMCC 24755 direct-inoculation starter improves the yogurt fermentation efficiency, acid production capacity, and water holding capacity.

[0061] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for optimizing the formulation of a compound lyophilization protectant for *Lactobacillus reuteri* based on a neural network model, characterized in that... Lactobacillus reuteri ( Limosilactobacillus reuteri Using the strain as the test strain and its freeze-drying survival rate as the evaluation index, a single-factor experiment was first conducted on the candidate *Lactobacillus reuteri* composite freeze-drying protectant. Based on the results of the single-factor experiment, the Plackett-Burman assay was used to screen for the *Lactobacillus reuteri* composite freeze-drying protectant that had a significant impact on the freeze-drying survival rate of the strain. Finally, the formulation optimization of the *Lactobacillus reuteri* composite freeze-drying protectant was completed by combining Box-Behnken response surface methodology and BP-GA neural network model optimization experiments.

2. The method for optimizing the formulation of Lactobacillus reuteri composite lyophilization protectant based on a neural network model according to claim 1, characterized in that, The survival rate detection method is as follows: *Lactobacillus reuteri* bacterial sludge is mixed with *Lactobacillus reuteri* compound freeze-drying protectant and then freeze-dried to obtain bacterial powder; the bacterial powder is rehydrated with an equal volume of sterile PBS solution before freeze-drying, mixed well, diluted with a bacterial suspension, and then spread on MRS solid medium. The survival rate is counted according to the freeze-drying survival rate formula; the freeze-drying survival rate formula is: 。 3. The method for optimizing the formulation of Lactobacillus reuteri composite lyophilization protectant based on a neural network model according to claim 1, characterized in that, The specific steps of the single-factor experiment are as follows: Following a mass-to-volume ratio of *Lactobacillus reuteri* bacterial sludge to *Lactobacillus reuteri* compound freeze-drying protectant of 1:5, *Lactobacillus reuteri* bacterial sludge, *Lactobacillus reuteri* compound freeze-drying protectant, and distilled water are mixed to prepare solutions of different concentrations. These solutions are then freeze-dried to obtain bacterial powder. After rehydrating the bacterial powder by an equal volume, the freeze-drying survival rate of the strains under different components and concentrations is calculated.

4. The method for optimizing the formulation of the Lactobacillus reuteri composite lyophilization protectant based on a neural network model according to claim 3, characterized in that, The Lactobacillus reuteri complex freeze-drying protectant is one or more of the following: a base protectant, a sugar, an alcohol, and an antioxidant.

5. The method for optimizing the formulation of Lactobacillus reuteri composite lyophilization protectant based on a neural network model according to claim 1, characterized in that, The steps of the Plackett-Burman experiment are as follows: Based on the results of the single-factor experiment, the significant factors affecting the freeze-drying survival rate are screened out. For each factor, two levels are taken: the lowest level and the highest level. The lowest level is -1 and the highest level is 1. The main effect factors are screened.

6. The method for optimizing the formulation of Lactobacillus reuteri composite lyophilization protectant based on a neural network model according to claim 1, characterized in that, The steps of the Box-Behnken response surface methodology experiment are as follows: using the results obtained from the Plackett-Burman experiment as the independent variable of the Box-Behnken experiment, and designing the experiment with the freeze-dried survival rate as the response value.

7. The method for optimizing the formulation of a compound lyophilization protectant for *Lactobacillus reuteri* based on a neural network model according to claim 1, characterized in that, The optimization experiment of the BP-GA neural network model was carried out as follows: A single hidden layer BP neural network model was constructed using MATLAB 2024a. The significant factors affecting the freeze-drying survival rate, determined through single-factor experiments and Plackett-Burman experiments, were used as the input layer neurons of the BP neural network model, and the freeze-drying survival rate of *Lactobacillus reuteri* was used as the output layer neurons. After repeated testing to determine the topology, the BP-GA neural network model was constructed using the genetic algorithm toolbox. Based on the BP-GA neural network model, the formulation of the compound freeze-drying protectant for *Lactobacillus reuteri* strain was optimized.

8. The application of the method according to any one of claims 1 to 7 in the preparation of direct-inoculation fermentation agents.

9. The application of the method according to any one of claims 1 to 7 in the preparation of yogurt.

10. A method for preparing yogurt, characterized in that, Includes the following steps: 1) Using the method described in any one of claims 1 to 7, the optimal formulation of the Lactobacillus reuteri composite freeze-drying protectant is obtained; 2) Mix *Lactobacillus reuteri* bacterial sludge and *Lactobacillus reuteri* compound freeze-drying protectant at a mass-to-volume ratio of 1:5, freeze-dry, and obtain *Lactobacillus reuteri* direct-inoculation fermentation agent; the formula of the *Lactobacillus reuteri* compound freeze-drying protectant is obtained by the method in step 1). 3) Add white sugar to milk, mix well, pasteurize, cool, and inoculate with the Lactobacillus reuteri direct inoculation starter obtained in step 2) to obtain yogurt.

Images