Method for predicting composition of bacterial community in white wine pit mud based on physical and chemical factors

Abstract

The invention relates to the technical field of flora prediction, in particular to a method for predicting bacterial community composition in white wine pit mud based on physical and chemical factors, which comprises the following steps: (1) sample collection: collecting pit mud with different masses in a pit according to vertical and horizontal evolution dimensions; (2) sample analysis: determining physicochemical factors of the pit mud sample and sequencing amplicons of bacterial communities to respectively form data sets consisting of the physicochemical factors and the bacterial communities; (3) model construction: respectively screening physical and chemical factors for forming an input set and bacterial OTU for forming an output set through collinearity analysis and difference analysis, and determining a better machine learning method according to a training set result; and (4) result verification: randomly collecting pit mud samples, analyzing physical and chemical factors and bacterial communities of the pit mud samples, and verifying the accuracy of the prediction result.

Description

Technical Field

[0001] The invention relates to the technical field of bacterial community prediction, and specifically relates to a method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors. Background Art

[0002] The cellar mud needs to go through the processes of cultivation, evaluation, maintenance, and replacement to ensure that its quality can meet the high standards of brewing in the long term. However, to date, the cellar mud still has problems such as backward cultivation technology, limited evaluation methods, and insufficient control of bacterial flora, which seriously hinders the controllability of liquor brewing and restricts the development of theoretical research and production practice.

[0003] At present, the research on the evaluation of cellar mud has been widely carried out, and many scientific and effective evaluation methods have been proposed. These methods usually start from indicators of multiple aspects such as microorganisms, physicochemical factors and sensory characteristics. The constructed cellar mud evaluation system can judge the quality of cellar mud, and the results are highly consistent with the actual situation. Among the evaluation indicators of cellar mud, the bacterial community can continuously produce short-chain fatty acids for the synthesis of flavor substances such as ethyl caproate, ethyl butyrate and ethyl valerate, which plays a direct role in liquor brewing and can best reflect the changes in the quality of cellar mud. Due to the inconvenience of using microbial sequencing technology in actual production, the microbial indicators in cellar mud evaluation tend to use the direct counting method under a microscope to evaluate the number of caproic acid-producing bacteria or anaerobic bacteria in the cellar mud. Although this method is simple and convenient, it cannot obtain the structural information of the bacterial community and has obvious limitations in the evaluation of cellar mud.

[0004] Machine learning is a technology that can identify complex relationships in high-dimensional data sets and perform decision-making tasks such as regression and clustering. Supervised machine learning is particularly good at using existing data sets to build prediction models and then quickly identify and predict target indicators for new data. It has shown excellent accuracy in predicting changes in environmentally responsive microbial communities. The close interaction between bacterial communities and physical and chemical factors gives cellar mud typical environmental response characteristics. Machine learning has great application potential in quickly obtaining information on the bacterial community structure of cellar mud and developing accurate cellar mud evaluation methods. Summary of the invention

[0005] The purpose of the present invention is to provide a method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors comprises the following steps:

[0008] (1) Sample collection: Sampling points were set up in the cellar along the vertical and horizontal evolution dimensions to collect cellar mud of different qualities;

[0009] (2) Sample analysis: The physicochemical factors of the pit mud samples were measured and the amplicon sequencing of the bacterial community was performed to form data sets of physicochemical factors and bacterial community composition, respectively;

[0010] (3) Model construction: Through collinearity analysis and difference analysis, the physicochemical factors used to form the input set and the bacterial OTUs used to form the output set were screened respectively, and the optimal machine learning method was determined based on the training set results;

[0011] (4) Result verification: randomly collect pit mud samples, analyze their physical and chemical factors and bacterial communities, and verify the accuracy of the prediction results.

[0012] Specifically, step (1) comprises selecting three horizontal positions as sampling points, vertically digging a 10 cm deep sample at each sampling point and dividing the sample into five layers for stratified sampling; the horizontal positions are (center, middle, corner); and the five layers are 0-2 cm, 2-4 cm, 4-6 cm, 6-8 cm, and 8-10 cm, respectively.

[0013] Among them, the center is the center of the bottom of the cellar, the midpoint is the midpoint of the four sides of the bottom of the cellar, and the corners are the four corners of the bottom of the cellar; among them, the midpoint and the corner are samples from the four sampling points mixed as one sample.

[0014] 100 g of each sample was collected, mixed thoroughly and stored in a sterile anaerobic bag, of which 20 g was stored at -80°C for analysis of bacterial communities, and the remaining 80 g was stored at 4°C for detection of physicochemical factors.

[0015] Wherein, the step (3) includes, through Spearman correlation analysis, screening water, pH, total acid, available phosphorus, ammonium nitrogen, humus, caproic acid, butyric acid as characteristic prediction variables to form an input data set.

[0016] Wherein, the step (3) includes, through the Wilcoxon rank-sum test and the Kruskal-Wallis H test, selecting the 10 bacterial OTUs with the highest differences under different evolutionary dimensions as prediction targets to form an output data set.

[0017] In step (3), the determination coefficients (R-Square, R 2) and Normalized Root Mean Square Error (NRMSE), it is determined that Random Forest is the better machine learning method.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] The prediction method of the present invention can accurately predict the composition of the cellar mud bacterial community when the sample size is sufficient, the sample representativeness is strong and the data is accurate. It provides a technical basis for quickly obtaining the cellar mud bacterial community structure information and accurately evaluating the cellar mud quality. It can further promote the controllability of liquor brewing and contribute to the high-quality development of the liquor industry.

[0020] Based on the high-dimensional characteristics of the physicochemical factors of cellar mud and the bacterial community data set and the relationship between the two, this paper explores the application potential of machine learning in the field of liquor brewing technology, and can provide certain theoretical guidance for the optimization of cellar mud evaluation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The present invention is a method flow chart of a method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors;

[0022] Figure 2 Schematic diagram of sample collection for the training set;

[0023] Figure 3 This is the collinearity diagram of physical and chemical factors;

[0024] Figure 4 For the prediction model R 2 and NRMSE line chart;

[0025] Figure 5 This is a regression plot of actual values ​​and predicted values. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] like Figure 1 As shown, a method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors, the specific steps are:

[0028] (1) Sample collection: Sampling points are set up in the cellar along the vertical and horizontal evolution dimensions to collect cellar mud of different qualities.

[0029] Specifically include:

[0030] 1) Six pits were randomly selected for the collection of training set samples. Three horizontal positions (center, middle, corner) were selected as sampling points at the bottom of each pit. A 10 cm deep sample was dug vertically at each sampling point and divided into five layers (0-2, 2-4, 4-6, 6-8, 8-10 cm) for stratified sampling.

[0031] 2) Each cellar has only one central sampling point, and the midpoint and corner samples are taken from four sampling points and mixed as one sample. The specific sampling method is as follows: Figure 2 As shown;

[0032] 3) About 100 g of each sample was collected, mixed thoroughly and stored in a sterile anaerobic bag, of which 20 g was stored at -80°C for analysis of bacterial communities, and the remaining 80 g was stored at 4°C for detection of physical and chemical factors.

[0033] (2) Sample analysis: The physicochemical factors of the pit mud samples were measured and the amplicon sequencing of the bacterial community was performed to form data sets of physicochemical factors and bacterial community composition, respectively.

[0034] 1) Determination of physical and chemical factors of pit mud: water content was determined according to "HJ 613-2011 Determination of soil dry matter and water content by weight", pH was determined according to "NY / T 1377-2007 Determination of soil pH", and total acid was determined according to

[0035] The determination is carried out according to GB 12456-2021 National Food Safety Standard Determination of Total Acid in Foods. The effective phosphorus is based on

[0036] The determination was carried out according to "DB34 / T 2266-2014 Determination of available phosphorus in cellar mud - ammonium fluoride-hydrochloric acid colorimetric method", the ammonium nitrogen was determined according to "NY / T 1849-2010 Determination of ammonium nitrogen, available phosphorus and available potassium in acidic soil - combined extraction-colorimetric method", the humus was determined according to "DB34 / T 2265-2014 Determination of humus in cellar mud - potassium dichromate oxidation method", lactic acid and acetic acid were determined by high performance liquid chromatography, and caproic acid and butyric acid were determined by gas chromatography.

[0037] 2) Shanghai Meiji Biopharmaceutical Technology Co., Ltd. was commissioned to analyze the bacterial communities in the cellar mud samples using 16S amplicon sequencing. The total genomic DNA of the bacterial communities in the cellar mud samples was extracted using the FastDNA Spin kit, and after quality assessment, the primers 338F (5'-ACTCCTACGGGAGGCAGCAG-3') and 806R (5'-GGACTACHVGGGTWTCTAAT-3'), the sequences of which are shown in SEQ ID NO: 1-2, were used for polymerase chain reaction (PCR) to amplify the V3-V4 region of the 16S rRNA gene.

[0038] Amplification process: ① Pre-denaturation at 95°C for 3 minutes; ② 35 cycles of amplification, the cycle steps include denaturation at 95°C for 30 seconds, annealing at 55°C for 30 seconds, and extension at 72°C for 45 seconds; ③ Extension at 72°C for 10 minutes; ④ Stop the reaction at 10°C.

[0039] The PCR products were eluted with Tris-HCl buffer and identified by 2% agarose gel electrophoresis. According to the electrophoresis results, QuantiFluorTM-ST, Rapid DNA-Seq Kit, and Illumina MiSeqPE300 were used for product quantification, library construction, and sequencing. Finally, Fastp 0.19.6 and FLASH 1.2.11 were used for quality control and splicing of sequence data, and the sequences of representative operational taxonomic units (OTUs) were annotated based on the SILVA 138-16S rRNA database.

[0040] (3) Model construction: The physicochemical factors used to form the input set and the bacterial OTUs used to form the output set were screened through collinearity analysis and difference analysis, and the optimal machine learning method was determined based on the training set results.

[0041] 1) Spearman correlation analysis of physicochemical factors was performed using the Hmisc package in R 4.2.2 software, as follows: Figure 3 As shown, water, pH, total acid, available phosphorus, ammonium nitrogen, humus, caproic acid, and butyric acid were selected as characteristic prediction variables based on the threshold value of Spearman ρ < 0.8 to form the input data set;

[0042] 2) The Wilcoxon rank-sum test and Kruskal-Wallis Htest of the bacterial community were performed using the wilcox.test() function and kruskal.test() function in R 4.2.2 software, respectively, and the 10 bacterial OTUs with the highest differences under different evolutionary dimensions were selected as prediction targets to form an output data set;

[0043] 3) Four different prediction models, namely Random Forest, k-Nearest Neighbor, Multivariable Linear Regression and Support Vector Regression, were constructed using the randomForest package, caret package, lm() function and e1071 package in R 4.2.2 software;

[0044] 4) Randomly select 10% of the training set samples as the validation set for cross-validation, and calculate the determination coefficient (R-Square, R 2 ) and Normalized Root Mean Square Error (NRMSE), R of Random Forest 2 The highest and the lowest NRMSE are better machine learning methods.

[0045] (4) Result verification: randomly collect pit mud samples, analyze their physical and chemical factors and bacterial communities, and verify the accuracy of the prediction results.

[0046] 1) Randomly select cellars to collect test set samples to test the accuracy of the prediction model. A total of 10 test set samples were collected;

[0047] 2) About 100 g of each sample was collected, mixed thoroughly and stored in a sterile anaerobic bag, of which 20 g was stored at -80 °C for analysis of bacterial communities, and the remaining 80 g was stored at 4 °C for detection of physical and chemical factors;

[0048] 3) Determination of physicochemical factors and analysis of bacterial communities, the relevant specific methods are consistent with the methods described in step (2);

[0049] 4) Regression analysis of the actual and predicted values ​​of bacterial OTU in the test set Figure 5 As shown, the regression is extremely significant (p<0.001) and R 2The prediction model can accurately predict the bacterial community structure and change trend in cellar mud, and the average error between the predicted value and the actual value of each OTU is 0.91%.

[0050] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors, characterized in that: The following steps are involved: (1) Sample collection: Sampling points were set up in the cellar along the vertical and horizontal evolution dimensions to collect cellar mud of different qualities; (2) Sample analysis: The physicochemical factors of the pit mud samples were measured and the amplicon sequencing of the bacterial community was performed to form data sets of physicochemical factors and bacterial community composition, respectively; (3) Model construction: Through collinearity analysis and difference analysis, the physicochemical factors used to form the input set and the bacterial OTUs used to form the output set were screened respectively, and the optimal machine learning method was determined based on the training set results; (4) Result verification: randomly collect pit mud samples, analyze their physical and chemical factors and bacterial communities, and verify the accuracy of the prediction results.

2. The method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors according to claim 1, characterized in that: Specifically, step (1) comprises selecting three horizontal positions as sampling points, vertically digging 10 cm deep samples at each sampling point and dividing the samples into five layers for stratified sampling; the horizontal positions are the center, the midpoint, and the corner; and the five layers are 0-2 cm, 2-4 cm, 4-6 cm, 6-8 cm, and 8-10 cm, respectively.

3. The method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors according to claim 2, characterized in that: The center is the center of the bottom of the cellar, the midpoint is the midpoint of the four sides of the bottom of the cellar, and the corners are the four corners of the bottom of the cellar; wherein the midpoint and the corner are samples from the four sampling points mixed as one sample.

4. The method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors according to claim 3, characterized in that: 100 g of each sample was collected, mixed thoroughly and stored in a sterile anaerobic bag, of which 20 g was stored at -80°C for analysis of bacterial communities, and the remaining 80 g was stored at 4°C for detection of physicochemical factors.

5. The method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors according to claim 1, characterized in that: The step (3) includes screening water, pH, total acid, available phosphorus, ammonium nitrogen, humus, caproic acid, and butyric acid as characteristic prediction variables through Spearman correlation analysis to form an input data set.

6. The method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors according to claim 5, characterized in that: The step (3) includes, through the Wilcox rank sum test and the Kruskal-Wallis rank sum test, selecting the 10 bacterial OTUs with the highest differences in different evolutionary dimensions as prediction targets to form an output data set.

7. The method for predicting bacterial community composition in liquor cellar mud based on physicochemical factors according to claim 6, characterized in that: In step (3), by calculating the coefficient of determination and normalized root mean square error of different prediction models, it is determined that random forest is a better machine learning method.

Images