Method for predicting and screening two-to-two interaction of lactobacillus bulgaricus and streptococcus thermophilus

Abstract

The invention provides a two-to-two interaction prediction and screening method for lactobacillus bulgaricus and streptococcus thermophilus, and belongs to the technical field of food production. The method comprises the following steps: calculating whole genome k-mer data of two strains of lactobacillus bulgaricus and two strains of streptococcus thermophilus to form respective feature vectors, and calculating a gene copy number to construct a KEGG matrix; and generating a comprehensive feature vector by combining the KEGG feature frequency and the k-mer feature frequency. The first 200 important features are screened out from a real label sample by utilizing chi-square test, gradient lifting and variance analysis. And then, generating a pseudo-label sample through a generative adversarial network (GAN), and constructing a machine learning model in combination with a real label sample for predicting the interaction effect of the strain combination. And finally, verifying the accuracy of model prediction through a fermentation test, and selecting an optimal model. According to the method, the interaction symbiotic potential of the strain combination can be efficiently predicted, and the yoghurt production efficiency and quality are improved.

Description

Technical Field

[0001] The invention relates to the technical field of food production, specifically to the technical field of yogurt production, and in particular to a method for predicting and screening a two-to-two interaction between Lactobacillus bulgaricus and Streptococcus thermophilus. Background Art

[0002] Yogurt is a curd-like product made by fermenting Lactobacillus bulgaricus and Streptococcus thermophilus in milk powder (sterilized milk or concentrated milk) or without milk powder (or skim milk powder). The finished product contains a large number of corresponding active microorganisms. Yogurt is rich in nutrients such as calcium, protein, riboflavin and vitamins. It has been confirmed that yogurt can balance intestinal flora, improve immunity, lower cholesterol and delay aging. More and more people consume yogurt on a daily basis, which puts higher requirements on the quality of yogurt production.

[0003] The existing technology puts two strains of thermophilic Streptococcus and two strains of Bulgarian Lactobacillus together, randomly combines them through biological experiments, and finally determines whether these four strains can interact with each other to accelerate the fermentation speed in the yogurt production process and improve the fermentation characteristics such as viscosity and water holding capacity by testing the acid production rate and protein hydrolysis ability phenotype data. This method is time-consuming and labor-intensive, and the output is small. It takes 3-4 months to determine whether a group of bacteria interacts with each other. Summary of the invention

[0004] The purpose of the present invention is to provide a two-to-two interaction prediction and screening method for Lactobacillus bulgaricus and Streptococcus thermophilus, which can achieve high-throughput and efficient prediction under the premise of ensuring prediction accuracy.

[0005] To achieve the above object, the present invention provides a method for predicting and screening a two-to-two interaction between Lactobacillus bulgaricus and Streptococcus thermophilus, comprising the following steps: Step S1, respectively calculate the k-mer data of the whole genomes of two strains of Lactobacillus bulgaricus and two strains of Streptococcus thermophilus, and calculate their respective dimensional feature vector, calculate the gene copy number of each strain to form a KEGG matrix; Step S2: The KEGG features of the four strains were merged according to the principle of adding the overlapping gene copy numbers and duplicating the non-overlapping gene copy numbers to obtain Features; the k-mer feature frequencies of the four strains are accumulated to obtain feature; Features and Features concatenated Features Step S3: Set the number of true label samples to , according to this True label sample pairs The first 200 features in the feature importance ranking list were selected through three feature selection methods: chi-square test, gradient boosting and variance analysis. Step S4: The real label samples are generated by alternating the generator and discriminator of GAN to complete the iterative process of generating fake data and distinguishing true and false data. Pseudo-labeled samples; Step S5, constructing a machine learning model based on the real label samples and the pseudo label samples, and then using the constructed machine learning model to predict Lactobacillus bulgaricus and Streptococcus thermophilus in a two-to-two combination; Step S6: Select several combinations from the predicted results for fermentation test, comprehensively evaluate the fermentation effect of the strain combination according to the fermentation characteristics, compare the experimental results with the predicted results of the machine learning model, and select the optimal model with the highest prediction accuracy.

[0006] Preferably, in step S1, k=5-9.

[0007] Preferably, in step S5, the machine learning model includes logistic regression (LR), support vector machine (SVM), random forest (RF), K nearest neighbor (KNN) and Gaussian naive Bayes (GNB).

[0008] Therefore, the present invention adopts the above-mentioned two-to-two interaction prediction and screening method of Lactobacillus bulgaricus and Streptococcus thermophilus, and the beneficial technical effects are as follows: Through in-depth analysis of the genomes of Lactobacillus bulgaricus and Streptococcus thermophilus, combined with a series of operations such as KEGG operations, k-mer feature extraction, fine feature selection, and GAN (generative adversarial network) data enhancement, a high-precision prediction model for the interaction between two strains of Lactobacillus bulgaricus and two strains of Streptococcus thermophilus was successfully constructed. This model can efficiently predict in batches whether any combination of these four strains can achieve mutual symbiosis.

[0009] In the feature selection process, the feature combinations with the most significant impact on the interaction are accurately screened, thereby ensuring that the prediction model can focus on the most critical information. At the same time, with the help of data enhancement technology, the efficiency of machine learning modeling is further improved, which not only improves the prediction efficiency and throughput, but also ensures the accuracy of the prediction results. This series of optimization measures jointly promotes the application potential of the present invention in dairy fermentation and other related fields. DETAILED DESCRIPTION

[0010] The technical solution of the present invention is further illustrated by the following examples.

[0011] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0012] Embodiment 1 The method for predicting and screening the two-to-two interaction between Lactobacillus bulgaricus and Streptococcus thermophilus comprises the following steps: Step S1, feature extraction of Lactobacillus bulgaricus and Streptococcus thermophilus.

[0013] Calculate the k-mer (k=5-9) data of the whole genomes of two strains of Lactobacillus bulgaricus and two strains of Streptococcus thermophilus respectively, and calculate their respective The gene copy number of each strain was calculated using CENSOR, CNVnator and other software to form a KEGG matrix.

[0014] Step S2: characteristic combination of 2 strains of Lactobacillus bulgaricus and 2 strains of Streptococcus thermophilus.

[0015] The KEGG features of the four strains were merged according to the principle of adding the overlapping gene copy numbers and duplicating the non-overlapping gene copy numbers. Features.

[0016] The k-mer feature frequencies of the four bacterial strains are accumulated to obtain Features.

[0017] Will Features and Features concatenated Features.

[0018] Step S3: Perform feature selection based on the existing small amount of label data.

[0019] If the number of labeled samples is , according to this A number of labeled sample pairs The first 200 features were selected from the feature importance ranking lists of the three methods through three feature selection methods: chi-square test, gradient boosting and variance analysis.

[0020] Step S4: data enhancement.

[0021] right The GAN generator and discriminator complete the iterative process of generating fake data and distinguishing true and false data through two steps of alternating implementation, and finally generate Pseudo-labeled data.

[0022] Step S5: Model construction.

[0023] use Samples ( is generated, The real labeled positive samples) were modeled using LR, SVM, RF, KNN, and GNB to build five machine learning models. The 265,364,100 2:2 combinations of 181 strains of Lactobacillus bulgaricus and 181 strains of Streptococcus thermophilus available in the laboratory were predicted, and the model prediction results of all combinations were obtained (0 or 1, 0 means no interaction, 1 means interaction), which were then submitted to the laboratory for verification.

[0024] Step S6: laboratory verification and optimal model determination.

[0025] 30 groups are randomly selected from step S5 to carry out fermentation experiments. According to the fermentation characteristics such as fermentation time, viscosity, water holding capacity, etc. obtained from the fermentation experiments, the fermentation labels of the 30 strain combinations are comprehensively evaluated to be 0 or 1.

[0026] The results of laboratory verification were compared with the prediction results of five machine learning models, and the optimal model was selected as the logistic regression model.

[0027] Step S7, using the optimal model obtained in step S6, a set of experiments were carried out, that is, the interaction between two strains of Lactobacillus bulgaricus (IMAU20360 and IMAU20428) and two pairs of Streptococcus thermophilus (IMAU10630 and IMAU40145) was predicted, and the prediction result was output in 5 seconds: interaction. It shows that the time consumed by the present invention for screening the 2:2 starter strain is much less than the time consumed by the laboratory.

[0028] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.

[0029] Therefore, the present invention adopts the above-mentioned two-to-two interaction prediction and screening method between Lactobacillus bulgaricus and Streptococcus thermophilus, which can achieve high-throughput and efficient prediction while ensuring the prediction accuracy.

[0030] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for predicting and screening a two-to-two interaction between Lactobacillus bulgaricus and Streptococcus thermophilus, characterized in that: The following steps are involved: Step S1, respectively calculate the k-mer data of the whole genomes of two strains of Lactobacillus bulgaricus and two strains of Streptococcus thermophilus, and calculate their respective dimensional feature vector, calculate the gene copy number of each strain to form a KEGG matrix; Step S2: The KEGG features of the four strains were merged according to the principle of adding the overlapping gene copy numbers and duplicating the non-overlapping gene copy numbers to obtain Features The k-mer feature frequencies of the four strains are accumulated to obtain Features Will Features and Features concatenated Features Step S3: Set the number of true label samples to , according to this True label sample pairs The first 200 features in the feature importance ranking list were selected through three feature selection methods: chi-square test, gradient boosting and variance analysis. Step S4: The real label samples are generated by alternating the generator and discriminator of GAN to complete the iterative process of generating false data and distinguishing true and false data. Pseudo-labeled samples; Step S5, constructing a machine learning model based on the real label samples and the pseudo label samples, and then using the constructed machine learning model to predict Lactobacillus bulgaricus and Streptococcus thermophilus in a two-to-two combination; Step S6: Select several combinations from the predicted results for fermentation test, comprehensively evaluate the fermentation effect of the strain combination according to the fermentation characteristics, compare the experimental results with the predicted results of the machine learning model, and select the optimal model with the highest prediction accuracy.

2. The method for predicting and screening two-to-two interactions between Lactobacillus bulgaricus and Streptococcus thermophilus according to claim 1, characterized in that: In step S1, k=5-9.

3. The method for predicting and screening two-to-two interactions between Lactobacillus bulgaricus and Streptococcus thermophilus according to claim 1, characterized in that: In step S5, the machine learning models include logistic regression, support vector machine, random forest, K nearest neighbor and Gaussian naive Bayes.