A method and system for predicting dynamic changes of cold spring long-term enriched bacterial communities

Abstract

The present invention provides a method and system for predicting the dynamic changes of a cold spring long-term enriched bacterial community. The method comprises: collecting Raman spectra of bacterial culture samples at different enrichment stages and analyzing them by using an MCR-ALS algorithm to obtain metabolite data, which are saved together with environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups as an original data set; using a CART decision tree algorithm to perform importance calculation and data screening on the original data set; using the screened data set to train a random forest prediction model for prediction; and finally using the trained random forest model to predict the dynamic changes of the community. The present invention forms a data set by conventional sampling and collection of a complete bacterial community succession process in the early stage of enrichment culture, and constructs an accurate prediction model by using a machine learning algorithm, thereby improving the sampling accuracy and efficiency of the long-term enrichment culture operation of the cold spring bacterial community.

Description

Technical Field

[0001] The present invention relates to the technical field of microbial monitoring, and more specifically, to a method and system for predicting dynamic changes in a cold spring long-term enriched bacterial community. Background Art

[0002] Cold spring ecosystems contain a large number of functional bacteria such as methanotrophs, sulfate-reducing bacteria, and methanogens, which rely on cold spring seepage to survive and reproduce. Their unique biochemical processes and metabolites greatly alleviate the greenhouse effect caused by the migration of methane into the atmosphere. At the same time, they also provide important scientific basis for exploring the biogeochemical cycles in extreme environments and even the origin of life.

[0003] However, the cold seep environment has extreme deep-sea characteristics such as high pressure, low oxygen, and high-concentration methane seepage, which makes it difficult for cold seep bacteria to grow and reproduce normally in the terrestrial environment. Existing studies have achieved successful enrichment and culture of key cold seep bacterial groups by collecting cold seep samples with fidelity and simulating high pressure and low temperature conditions in the laboratory. During the enrichment culture, regular monitoring of the dynamic changes of metabolites and bacterial communities in the enrichment system is an important indicator of the successful enrichment of key bacterial groups. However, the reproduction iteration cycle of cold seep bacteria is generally higher than that of normal environmental bacteria. Some studies have reported that the enrichment cycle of cold seep bacteria takes months or even years.

[0004] Existing technologies usually use regular sampling, DNA extraction, library construction and sequencing methods to monitor enriched bacterial communities. However, the amount of enriched samples required for DNA extraction is high, and the extremely long enrichment culture cycle (such as several years) limits the sampling volume and frequency of the enrichment system. At the same time, the collected enriched samples may be inefficiently sampled because the bacteria have not reached the reproduction cycle and fail to capture the "inflection point" of key bacterial growth metabolism and community dynamic changes in the enriched samples, which greatly affects the sampling and monitoring efficiency of enriched samples. Summary of the invention

[0005] In order to overcome the problems of large sampling volume, low frequency and low efficiency in the dynamic monitoring of traditional cold spring enriched bacterial communities in the above-mentioned prior art, the present invention provides a method and system for predicting the dynamic changes of cold spring long-term enriched bacterial communities. A data set is formed by conventional sampling and collection of the complete bacterial community succession process (adaptation, growth, reproduction and stability) in the initial stage of enrichment culture, and an accurate prediction model is constructed using a machine learning algorithm, thereby improving the sampling efficiency of the long-term enrichment culture operation of cold spring bacterial communities.

[0006] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0007] A method for predicting the dynamic changes of a cold spring long-term enriched bacterial community comprises the following steps:

[0008] S1: Raman spectra, environmental parameters, α-diversity of bacterial communities and relative abundance of key bacterial groups of bacterial culture samples collected at different enrichment stages;

[0009] The Raman spectrum is analyzed using the MCR-ALS algorithm (multivariate curve resolution-alternating least squares method) to obtain metabolite data;

[0010] The metabolite data, environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups are jointly saved as an original data set;

[0011] S2: Use the CART decision tree algorithm to calculate the importance and filter the data of the original data set to obtain the filtered data set;

[0012] S3: Optimize and train the random forest prediction model used for prediction using the filtered data set to obtain a trained random forest prediction model;

[0013] S4: Obtain metabolite data and environmental parameters of the bacterial culture sample to be predicted and input them into a trained random forest prediction model. The trained random forest prediction model outputs prediction results of bacterial community alpha diversity and relative abundance of key bacterial groups.

[0014] Preferably, in step S1, the bacterial enrichment stage includes: an adaptation stage, a growth stage, a reproduction stage and a stabilization stage.

[0015] Preferably, in step S1, the environmental parameters include: water quality physical and chemical parameters, nutrient salt content and OD600 bacterial concentration parameters;

[0016] The bacterial community alpha diversity includes: Shannon index, Chao1 index, Ace index and Simpson index;

[0017] The relative abundance of the key bacterial groups includes: the relative abundance of anaerobic methane-oxidizing archaea ANME, sulfate-reducing bacteria SRB and methanogens;

[0018] The metabolite data include: the contents of formic acid, acetic acid, methanol, formaldehyde and sulfate.

[0019] Preferably, the water quality physical and chemical parameters include: dissolved methane and carbon dioxide content, temperature, salinity, pH, dissolved oxygen content;

[0020] The nutrient salt content includes: nitrate, dissolved organic matter DOM and total organic carbon TOC content.

[0021] Preferably, the Raman spectrum is collected using a micro-confocal Raman spectroscopy probe;

[0022] Use multi-parameter water quality probes to collect water quality physical and chemical parameters;

[0023] The nutrient salt content and OD600 bacterial concentration parameters were collected using an ultraviolet spectrophotometer;

[0024] DNA extraction, PCR amplification, library construction and sequencing are performed on the bacterial culture samples at different enrichment stages in sequence to obtain DNA sequencing data; and the alpha diversity of the bacterial community and the relative abundance of key bacterial groups are obtained based on the DNA sequencing data.

[0025] Preferably, the step S1 further comprises: evaluating the parsing accuracy of the MCR-ALS algorithm by using the MCR-BANDS algorithm.

[0026] Preferably, step S2 comprises:

[0027] A decision tree is constructed using the original data set, and the Gini coefficient of each decision node in the decision tree is calculated using the CART decision tree algorithm;

[0028] Calculate the variable importance score VIM (Variable Importance Measures) of each data in the original data set according to the Gini coefficient;

[0029] Data is screened according to the variable importance score VIM, data with a variable importance score VIM less than a preset threshold is eliminated, and a screened data set is obtained.

[0030] Preferably, step S3 comprises:

[0031] Divide the filtered data set into training set and test set;

[0032] Using the training set to train the random forest prediction model, to obtain a trained random forest prediction model;

[0033] The test set is used to evaluate the performance parameters of the trained random forest prediction model, and the trained random forest prediction model is iteratively optimized until the performance parameters meet the preset conditions to obtain the trained random forest prediction model.

[0034] Preferably, the screened data set is divided into a training set and a test set in a ratio of 9:1 using a ten-fold crossover method;

[0035] The performance parameters include coefficient of determination, root mean square error and mean absolute error;

[0036] The larger the determination coefficient is, and the smaller the root mean square error and mean absolute error are, the better the performance of the trained random forest prediction model is.

[0037] The present invention also provides a system for predicting the dynamic changes of a cold spring long-term enriched bacterial community, which uses the above-mentioned method for predicting the dynamic changes of a cold spring long-term enriched bacterial community, including:

[0038] Data acquisition unit: used to collect Raman spectra, environmental parameters, bacterial community α-diversity and relative abundance of key bacterial groups of bacterial culture samples at different enrichment stages;

[0039] Analyzing the Raman spectrum using the MCR-ALS algorithm to obtain metabolite data;

[0040] The metabolite data, environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups are jointly saved as an original data set;

[0041] Data screening unit: used to use the CART decision tree algorithm to perform importance calculation and data screening on the original data set to obtain the screened data set;

[0042] Model training unit: used to optimize and train the random forest prediction model used for prediction using the screened data set, and obtain the trained random forest prediction model;

[0043] Community dynamic change prediction unit: used to obtain metabolite data and environmental parameters of the bacterial culture sample to be predicted and input them into the trained random forest prediction model. The trained random forest prediction model outputs the prediction results of bacterial community alpha diversity and relative abundance of key bacterial groups.

[0044] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0045] The invention provides a method and system for predicting the dynamic changes of a cold spring long-term enriched bacterial community. First, Raman spectra, environmental parameters, α diversity of bacterial communities and relative abundance of key bacterial groups of long-term bacterial culture samples at different enrichment stages are collected; the Raman spectra are analyzed by using an MCR-ALS algorithm to obtain metabolite data; the metabolite data, environmental parameters, α diversity of bacterial communities and relative abundance of key bacterial groups are saved together as an original data set; then, a CART decision tree algorithm is used to perform importance calculation and data screening on the original data set to obtain a screened data set; then, a random forest prediction model is optimized and trained using the screened data set to obtain a trained random forest prediction model; finally, the metabolite data and environmental parameters of the bacterial culture samples to be predicted are obtained and input together into the trained random forest prediction model, and the trained random forest prediction model outputs prediction results of the α diversity of bacterial communities and the relative abundance of key bacterial groups.

[0046] The present invention has the following advantages:

[0047] 1) The present invention uses Raman spectroscopy and MCR-ALS algorithm to qualitatively and quantitatively analyze the compound components and their concentrations of culture samples. Compared with existing bacterial metabolite monitoring technologies, it can avoid sample pretreatment processes such as concentration and extraction, and achieve efficient, accurate and rapid detection of characteristic metabolites of different enriched samples;

[0048] 2) The present invention uses a variety of parameter data from the long-term enrichment stage of cold spring samples to construct a data set, and screens and trains the constructed random forest prediction model to achieve high-precision prediction of the diversity of enriched bacterial communities, which can indicate the dynamic changes of bacterial communities and effectively identify the "inflection point" of the long-term cultivation stage of cold springs;

[0049] 3) The present invention sets a random forest model to specifically predict the abundance of key functional bacteria in the cold spring enriched flora. Since the data set of the prediction model is integrated based on continuous culture monitoring of the same enrichment system, the functional bacteria of the enriched flora and the environmental parameters have continuity and periodicity compared to the non-continuous data set in the database. Adding the abundance of key functional bacteria as the prediction output can more accurately indicate the dynamic changes of the enriched flora over a long period, thereby formulating a more targeted sampling strategy and improving the accuracy and efficiency of the sampling operation;

[0050] 4) Compared with the existing deep-sea bacterial culture technology, the method for predicting the dynamic changes of cold spring long-term enriched bacterial communities of the present invention optimizes the problem of high sample consumption in the dynamic monitoring of bacterial communities during the long-term culture of deep-sea bacteria. At the same time, high-precision detection can be achieved without the need for pre-treatment processes such as DNA extraction and PCR library construction and sequencing, thereby efficiently, accurately and quickly indicating the dynamic changes of bacterial communities in samples at different enrichment stages under long periods. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of a method for predicting dynamic changes in a cold spring long-term enriched bacterial community provided in Example 1.

[0052] Figure 2 This is a diagram of the framework of a method for predicting dynamic changes in the community of cold spring long-term enriched bacteria provided in Example 2.

[0053] Figure 3 This is a structural diagram of a system for predicting dynamic changes in cold spring long-term enriched bacterial communities provided in Example 2. DETAILED DESCRIPTION

[0054] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0055] In order to better illustrate the present embodiment, some parts in the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product;

[0056] It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0057] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0058] Example 1

[0059] like Figure 1 As shown, this embodiment provides a method for predicting the dynamic changes of a cold spring long-term enriched bacterial community, comprising the following steps:

[0060] S1: Raman spectra, environmental parameters, α-diversity of bacterial communities and relative abundance of key bacterial groups of bacterial culture samples collected at different enrichment stages;

[0061] The Raman spectrum is analyzed using the MCR-ALS algorithm (multivariate curve resolution-alternating least squares method) to obtain metabolite data;

[0062] The metabolite data, environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups are jointly saved as an original data set;

[0063] S2: Use the CART decision tree algorithm to calculate the importance and filter the data of the original data set to obtain the filtered data set;

[0064] S3: Optimize and train the random forest prediction model used for prediction using the filtered data set to obtain a trained random forest prediction model;

[0065] S4: Obtain metabolite data and environmental parameters of the bacterial culture sample to be predicted and input them into a trained random forest prediction model. The trained random forest prediction model outputs prediction results of bacterial community alpha diversity and relative abundance of key bacterial groups.

[0066] In the specific implementation process, this embodiment predicts the dynamic changes of the community of the long-term enriched bacterial community in the cold spring based on machine learning. As an intelligent prediction algorithm, machine learning has accumulated a large number of algorithms and applications in the field of microbial community prediction. It can train and build an algorithm model based on the existing data set, and then predict the dynamic change law in the subsequent stage; the simulation prediction of machine learning can avoid the problems of large sampling volume and low frequency caused by conventional sampling sequencing monitoring. Therefore, machine learning has good adaptability for ultra-long-term enriched samples of cold springs, and can build a model through the data set sampled and collected in the early stage of enrichment culture, predict the change trend of the bacterial community in the subsequent enrichment stage to make reasonable decisions and achieve efficient sampling operations;

[0067] First, the Raman spectra, environmental parameters, bacterial community α-diversity, and relative abundance of key bacterial groups of long-term bacterial culture samples at different enrichment stages were collected;

[0068] Raman spectra were analyzed using the MCR-ALS (multivariate curve resolution-alternating least squares) algorithm to obtain metabolite data;

[0069] The MCR-ALS algorithm model is as follows:

[0070] D=CS T +E

[0071] Where D is the measurement data, i.e. the Raman spectrum data matrix of the characteristic metabolites; C is the concentration matrix, i.e. the concentration curves of different metabolites; S T is the pure substance response matrix, that is, the spectral response matrix of pure characteristic metabolites; E is the residual matrix of the model;

[0072] The MCR-ALS algorithm can decompose the unknown complex spectral data of mixed samples into pure components and concentrations through a bilinear model. It has good adaptability to culture samples with complex components, so it can compare the spectral differences between enriched samples to analyze the changes in bacterial metabolites. The collected Raman spectral data are divided into multiple parts, and the spectral signals and corresponding concentrations of various compounds in bacterial culture samples at different enrichment stages are calculated based on the MCR-ALS algorithm. 2 The robustness of the model was preliminarily evaluated by using the LOF value (>99%) and LOF value (<10%);

[0073] The MCR-ALS algorithm model is evaluated using the MCR-BANDS algorithm. The signal component contribution function f of the MCR-BANDS algorithm under specific constraints is n It is used to indicate the relative contribution of each component in the mixture, can indicate the degree of rotational blur, and evaluate the accuracy of the MCR-ALS algorithm model;

[0074] The MCR-BANDS algorithm model is as follows:

[0075]

[0076] Among them, f n It is a scalar value indicating the relative contribution of the signal value of a characteristic component to the overall signal of n components obtained in the mixed sample; represents the signal value of the characteristic component, ‖CS T ‖ represents all signals of the mixed sample;

[0077] When a characteristic component signal produces a certain ambiguity, the MCR-BANDS algorithm can obtain f by calculating the relative contribution of each component.n The maximum and minimum values ​​of and Based on the maximum and minimum relative contributions of the characteristic components, the difference ΔSCCF (the difference range is 0 to 1) is calculated to evaluate the rotational ambiguity of each characteristic component for characterization accuracy. n The maximum and minimum values ​​are equal, that is, ΔSCCF is 0, which indicates that the characteristic component has no remaining rotational ambiguity, that is, the characteristic component has obtained a unique solution through the iteration of the MCR-ALS algorithm model; the ΔSCCF value based on the MCR-BANDS algorithm is used to further evaluate the accuracy of the MCR-ALS algorithm model in analyzing the characteristic components of each part of the Raman spectral data, so as to achieve accurate qualitative and quantitative calculation of the metabolite composition and concentration of the enriched bacterial samples;

[0078] The metabolite difference data of each sample were formed based on the metabolite composition and concentration data of samples at different enrichment stages (adaptation, growth, reproduction and stability);

[0079] Metabolite data, environmental parameters, bacterial community alpha diversity, and relative abundance of key bacterial groups were co-saved as the original dataset;

[0080] Then, a random forest prediction model was established based on the random forest algorithm. The input of the prediction model is the Raman spectrum, environmental parameters and metabolite data of bacteria, and the output is the α diversity of bacterial communities and the relative abundance of key bacterial groups. The α diversity of bacterial communities is used to indicate the dynamic changes of bacterial communities.

[0081] The random forest algorithm can sort and screen the importance of multi-parameter data, and iteratively remove redundant parameters to improve the accuracy of the prediction model. The algorithm has a high tolerance for outliers in multiple environmental parameters and is not prone to overfitting. It has a good match in prediction models, especially in the prediction research of bacterial communities. This method proposes a specific model prediction output indicator (relative abundance of key functional bacteria) based on the periodicity and continuity of the cold spring enrichment cycle of the input data, which can more accurately identify the key turning points of the enrichment cycle.

[0082] Then, the CART decision tree algorithm is used to calculate the importance and filter the data of the original data set to obtain the filtered data set; the random forest model is constructed based on the CART decision tree algorithm, which divides the values ​​of the split attribute into two subsets, calculates the Gini index as a feature parameter based on the training data set, and divides the training data set using the minimum Gini index principle;

[0083] By calculating the importance of each input parameter and repeatedly iterating to screen the key parameters that affect the model output value, the optimal parameter data is obtained for subsequent model training, which can effectively improve the model training effect and the prediction accuracy after training is completed;

[0084] Then, the random forest prediction model is optimized and trained using the filtered data set, and the model parameters (number of sub-data sets, maximum number of features, etc.) are adjusted to improve the model accuracy, and a trained random forest prediction model is obtained;

[0085] Finally, during the long-term enrichment of cold spring samples, the samples are regularly monitored online in real time to obtain the metabolite data and environmental parameters of the bacterial culture samples to be predicted. The online monitoring data are then input into the trained random forest prediction model. The trained random forest prediction model outputs the prediction results of the α diversity of the bacterial community and the relative abundance of key bacterial groups, indicating the dynamic changes of the bacterial community of the enriched samples, so as to identify the turning point of the bacterial community change and formulate targeted sampling strategies to improve the sampling efficiency of cold spring enriched samples.

[0086] This method is based on the Raman spectral data of the enriched samples and the MCR-ALS algorithm for decomposition. It can analyze the pure compound components of complex enriched samples, achieve qualitative identification and quantitative solution of characteristic metabolites, characterize the metabolic differences between enriched samples, avoid the sample concentration and extraction operations required in the traditional metabolite detection process, and more efficiently, accurately and quickly form a metabolite data set for the subsequent construction of a random forest prediction model.

[0087] This method constructs a random forest prediction model based on multi-source data, realizes accurate prediction of bacterial community diversity index and abundance of key functional bacteria, indicates the dynamic changes of bacterial communities in long-term enrichment systems, and can be used to efficiently identify the "inflection point" of the long-term enrichment stage of cold springs, so as to formulate targeted sampling strategies to improve the sampling efficiency of enrichment samples.

[0088] This method constructs a prediction model through the random forest algorithm to achieve fast, efficient and accurate prediction of the community dynamics of enriched samples, optimizes the high sample consumption in the dynamic monitoring of bacterial communities during the long-term cultivation of deep-sea bacteria, and can be used to indicate the periodic succession and sampling decisions of samples in different enrichment stages (adaptation, growth, reproduction and stability) under long-term conditions.

[0089] Example 2

[0090] This embodiment provides a method for predicting the dynamic changes of a cold spring long-term enriched bacterial community, comprising the following steps:

[0091] S1: Raman spectra, environmental parameters, α-diversity of bacterial communities and relative abundance of key bacterial groups of bacterial culture samples collected at different enrichment stages;

[0092] The Raman spectrum is analyzed using the MCR-ALS algorithm (multivariate curve resolution-alternating least squares method) to obtain metabolite data;

[0093] The metabolite data, environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups are jointly saved as an original data set;

[0094] S2: Use the CART decision tree algorithm to calculate the importance and filter the data of the original data set to obtain the filtered data set;

[0095] S3: Optimize and train the random forest prediction model used for prediction using the filtered data set to obtain a trained random forest prediction model;

[0096] S4: Obtain metabolite data and environmental parameters of the bacterial culture sample to be predicted and input them into a trained random forest prediction model, wherein the trained random forest prediction model outputs prediction results of bacterial community alpha diversity and relative abundance of key bacterial groups;

[0097] In step S1, the bacterial enrichment stage includes: an adaptation stage, a growth stage, a reproduction stage and a stabilization stage;

[0098] In step S1, the environmental parameters include: water quality physical and chemical parameters, nutrient salt content and OD600 bacterial concentration parameters;

[0099] The bacterial community alpha diversity includes: Shannon index, Chao1 index, Ace index and Simpson index;

[0100] The relative abundance of the key bacterial groups includes: the relative abundance of anaerobic methane-oxidizing archaea ANME, sulfate-reducing bacteria SRB and methanogens;

[0101] The metabolite data include: the content of formic acid, acetic acid, methanol, formaldehyde and sulfate;

[0102] The water quality physical and chemical parameters include: dissolved methane and carbon dioxide content, temperature, salinity, pH, dissolved oxygen content;

[0103] The nutrient salt content includes: nitrate, dissolved organic matter DOM and total organic carbon TOC content;

[0104] In the step S1, a Raman spectrum is collected using a micro-confocal Raman spectroscopy probe;

[0105] Use multi-parameter water quality probes to collect water quality physical and chemical parameters;

[0106] The nutrient salt content and OD600 bacterial concentration parameters were collected using an ultraviolet spectrophotometer;

[0107] The bacterial culture samples at different enrichment stages are sequentially subjected to DNA extraction, PCR amplification, library construction and sequencing to obtain DNA sequencing data; and the alpha diversity of the bacterial community and the relative abundance of key bacterial groups are obtained according to the DNA sequencing data;

[0108] The step S1 further includes: evaluating the parsing accuracy of the MCR-ALS algorithm using the MCR-BANDS algorithm;

[0109] The step S2 comprises:

[0110] A decision tree is constructed using the original data set, and the Gini coefficient of each decision node in the decision tree is calculated using the CART decision tree algorithm;

[0111] Calculate the variable importance score VIM (Variable Importance Measures) of each data in the original data set according to the Gini coefficient;

[0112] Perform data screening according to the variable importance score VIM, eliminate data whose variable importance score VIM is less than a preset threshold, and obtain a screened data set;

[0113] The step S3 comprises:

[0114] The filtered data set was divided into training set and test set in a ratio of 9:1 using the ten-fold crossover method;

[0115] Using the training set to train the random forest prediction model, to obtain a trained random forest prediction model;

[0116] Use the test set to evaluate the performance parameters of the trained random forest prediction model, and iteratively optimize the trained random forest prediction model until the performance parameters meet the preset conditions, and obtain the trained random forest prediction model;

[0117] The performance parameters include coefficient of determination, root mean square error and mean absolute error;

[0118] The larger the determination coefficient is, and the smaller the root mean square error and mean absolute error are, the better the performance of the trained random forest prediction model is.

[0119] In the specific implementation process, Figure 2 As shown, firstly, a microscopic confocal Raman spectroscopy probe, a multi-parameter water quality probe, and an ultraviolet spectroscopy probe are used to regularly monitor the long-period enrichment samples of cold springs to collect raw data;

[0120] In this embodiment, the wavelength of the Raman spectrum is set to 532 nm to excite the Raman signal, the laser intensity is 10 mW, and the scanning wave number is 150-4200 cm -1 , the acquisition time is 10 s; the collected Raman spectral data are preprocessed by removing cosmic rays, denoising, baseline correction and normalization;

[0121] The multi-parameter water quality probe is used to monitor and collect dissolved gas content (methane and carbon dioxide), temperature, salinity, pH, and dissolved oxygen content of water quality;

[0122] The UV spectrophotometer probe is used to monitor and collect the nutrient salt content and OD600 parameters of the culture samples. The nutrient salt parameters include nitrate content and organic carbon content. The OD600 parameter is an indicator of the growth concentration of bacteria in the culture samples. The spectral range is 200-1000nm, with a step length of 1nm. The UV spectrum mainly collects the UV spectral absorbance of nitrate (220nm, 275nm), DOM (355nm, 700-800nm, 254nm, 250nm, 365nm, 440nm, 275-295nm), and TOC (255nm, 265nm and 275nm);

[0123] The principle of UV spectroscopic analysis of nitrate, DOM and TOC is based on the dual wavelength method for determining nitrate (220nm and 275nm), the UV absorption coefficient method for determining DOM types (colored dissolved organic matter CDOM, CDOM relative molecular weight, humus and fulvic acid DOM content ratio, aromatized DOM), and the UV absorption method for determining TOC (255nm, 265nm and 275nm);

[0124] The UV absorption coefficient for DOM monitoring is calculated as follows:

[0125] a(λ)=ln10A(λ) / l

[0126] Where a(λ) is the absorption coefficient at wavelength λ, in m -1 ; A(λ) is the absorbance at wavelength λ, and the average absorbance between 700-800nm ​​is deducted to remove errors such as instrument background noise; l is the light path of the cuvette, in meters;

[0127] In this example, cold spring long-term enrichment samples were collected regularly, bacterial DNA of the cultured samples was extracted, and 16S rDNA sequences were amplified by PCR using universal primers, library construction and sequencing were performed, and the α diversity (Shannon index, Chao1 index, Ace index, Simpson index) of the bacterial community and the relative abundance of key bacterial groups (methanotrophic bacteria, sulfate-reducing bacteria, methanogens, etc.) were calculated based on the sequencing results to indicate the dynamic changes of the bacterial community in the enriched samples;

[0128] By collecting the above Raman spectra, environmental parameters, bacterial community diversity and key bacterial group abundance data, the original collected data are formed;

[0129] Based on the original collected Raman spectral data, the MCR-ALS algorithm is used to analyze the types and concentrations of characteristic chemical components in the sample; Raman spectroscopy analysis of metabolites such as sulfate, formic acid, methanol, and acetic acid is based on the pre-experimental fitting concentration standard curve set up in a high-pressure environment;

[0130] The MCR-ALS algorithm can decompose the unknown complex spectral data of mixed samples into pure components and concentrations through a bilinear model; the Raman spectral data is divided into multiple parts for input by calculating multiple grayscale thresholds according to grayscale features using the multi-threshold segmentation method; the multi-threshold segmentation method has the characteristics of fast calculation and high efficiency; the cross least squares method ALS is used for iterative optimization until convergence, and the R 2 The robustness of the model was evaluated by the LOF value (>99%) and the robustness model was screened out;

[0131] The MCR-BANDS algorithm was used to evaluate the MCR-ALS algorithm model. The signal component contribution function (ΔSCCF value) of the MCR-BANDS algorithm can represent the degree of rotational blur. The closer the ΔSCCF value is to 0, the higher the accuracy of the model is, so as to evaluate the accuracy of the MCR-ALS algorithm model. Finally, the metabolite composition and concentration data of samples at different enrichment stages were obtained based on the output value of the MCR-ALS algorithm, forming a metabolite difference data set of samples at different enrichment stages.

[0132] Based on the metabolite data output by the MCR-ALS model, combined with the original data, an original data set including characteristic metabolite components and concentrations, environmental parameters, bacterial community diversity, and abundance data of key bacterial groups was formed;

[0133] The prediction model was constructed using the random forest algorithm to output the results of the alpha diversity (Shannon index, Chao1 index, Ace index, Simpson index) of the bacterial community of the enriched samples and the relative abundance of key functional bacterial groups (methanotrophic bacteria, sulfate-reducing bacteria, methanogens, etc.);

[0134] The CART decision tree algorithm of the random forest model is used to calculate the Gini index and divide the data set according to the minimum Gini index principle. At the same time, the variable importance score VIM of each environmental parameter is calculated based on the Gini index. The relative importance of each environmental parameter in the input data set to the output value is compared, and the environmental parameters with high importance are screened out, which are integrated and summarized to form the screened data set.

[0135] Among them, the key CART decision tree algorithm is based on the Gini coefficient calculation process of decision nodes and divided feature parameters as follows:

[0136] Gini index of decision tree nodes of data set S:

[0137]

[0138] Among them, k refers to the category of the decision tree, p k is the probability of sampling to category k, 1-p k is the probability of not sampling the k category;

[0139] When the data set S is divided into subset 1 and subset 2, the Gini coefficient of a single partition is calculated based on the Gini index of the node:

[0140]

[0141] Based on Gini split (S), follow the Gini index minimum principle to divide the CART decision tree in the random forest, and repeat the division until the following termination conditions are met: the number of samples in the current data set is less than a given value; the samples in the current data set all belong to the same class; the attribute variable of the current data set is 0; the depth of the decision tree is greater than a given value;

[0142] For each environmental parameter in the data set, the variable importance measures (VIM) of the environmental parameter are calculated through the Gini index of each node and partition:

[0143]

[0144] It refers to the characteristic variable X n The importance of the i-th decision tree is the set S of nodes where the feature variable appears in the i-th decision tree, and s is each node in the set S;

[0145] In the random forest operation, the variable importance scores of each feature in each decision tree are normalized to compare the importance of each parameter to the overall decision:

[0146]

[0147] It refers to the characteristic variable X n The relative importance of each variable, It refers to the characteristic variable X n The importance of the i-th decision tree;

[0148] Then, the screened data set was split into 90% training set for training and building the random forest model and 10% test set for testing the model output results according to the ten-fold cross-validation method. The training set was iterated to optimize the parameters of the random forest model (number of sub-datasets, maximum number of features, etc.) to improve the accuracy of model prediction. The evaluation parameters (coefficient of determination R 2 , root mean square error RMSE, mean absolute error MAE), R 2 The larger the value, the smaller the RMSE and MAE values, and the best random forest prediction model is selected;

[0149] Among them, the evaluation parameters of the ten-fold cross method are calculated as follows:

[0150]

[0151]

[0152]

[0153] Among them, R 2 is the coefficient of determination, RMSE is the root mean square error, and MAE is the mean absolute error, where y a is the ath measured value, y a1 is the predicted value of the ath model, y a2 is the arithmetic mean of the ath measured values. 2 The larger the value, the smaller the RMSE and MAE values, which means the better the test performance of the model;

[0154] During the long-term enrichment process of cold springs, microscopic confocal Raman spectroscopic probes, multi-parameter water quality probes, and ultraviolet spectroscopic probes are regularly used to collect Raman spectra of different enriched samples (the metabolite composition and concentration of the enriched samples are calculated based on the MCR-ALS algorithm), OD600, nutrient content (nitrate, sulfate, DOC), DOM types (aromatic, hydrophobic), dissolved gas content (methane and carbon dioxide), temperature, salinity, pH, dissolved oxygen content and other parameters to form online monitoring data;

[0155] The data obtained online are input into the optimal random forest prediction model, and the output is the predicted α diversity index (Shannon index, Chao1 index, Ace index, Simpson index) of the bacterial community and the relative abundance of key functional bacterial groups (methanotrophic bacteria, sulfate-reducing bacteria, methanogens, etc.). Compared with traditional sampling and DNA extraction for library construction and sequencing, it has the advantages of accuracy, high efficiency and low sample loss. The predicted data output by the random forest model can be used to indicate the dynamic changes of the bacterial community of the samples in the enrichment stage and the growth and reproduction of key bacteria, thereby identifying the turning point of the long-term enrichment stage of cold springs and formulating targeted sampling strategies to improve the sampling efficiency of enriched samples.

[0156] Example 3

[0157] like Figure 3 As shown, this embodiment provides a system for predicting the dynamic changes of a cold spring long-term enriched bacterial community, using a method for predicting the dynamic changes of a cold spring long-term enriched bacterial community described in Example 1 or 2, including:

[0158] Data acquisition unit 301: used to collect Raman spectra, environmental parameters, bacterial community α diversity and relative abundance of key bacterial groups of bacterial culture samples at different enrichment stages;

[0159] Analyzing the Raman spectrum using the MCR-ALS algorithm to obtain metabolite data;

[0160] The metabolite data, environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups are jointly saved as an original data set;

[0161] Data screening unit 302: used to perform importance calculation and data screening on the original data set using the CART decision tree algorithm to obtain a screened data set;

[0162] Model training unit 303: used to optimize and train the random forest prediction model used for prediction using the screened data set to obtain a trained random forest prediction model;

[0163] The community dynamic change prediction unit 304 is used to obtain the metabolite data and environmental parameters of the bacterial culture sample to be predicted and input them into the trained random forest prediction model. The trained random forest prediction model outputs the prediction results of the bacterial community α diversity and the relative abundance of key bacterial groups.

[0164] In the specific implementation process, first, the data acquisition unit 301 collects Raman spectra, environmental parameters, bacterial community α diversity and relative abundance of key bacterial groups of long-term bacterial culture samples at different enrichment stages; uses the MCR-ALS algorithm to analyze the Raman spectra to obtain metabolite data; and saves the metabolite data, environmental parameters, bacterial community α diversity and relative abundance of key bacterial groups together as an original data set;

[0165] Then the data screening unit 302 uses the CART decision tree algorithm to perform importance calculation and data screening on the original data set to obtain a screened data set;

[0166] Then, the model training unit 303 uses the screened data set to optimize the random forest prediction model and obtain a trained random forest prediction model;

[0167] Finally, the community dynamic change prediction unit 304 obtains the metabolite data and environmental parameters of the bacterial culture sample to be predicted and inputs them into the trained random forest prediction model, and the trained random forest prediction model outputs the prediction results of the bacterial community α diversity and the relative abundance of key bacterial groups;

[0168] This system forms a data set through routine sampling and collection of the complete bacterial community succession process (adaptation, growth, reproduction and stability) in the early stage of enrichment culture, and uses machine learning algorithms to build an accurate prediction model, thereby improving the sampling efficiency of long-term enrichment culture operations of cold spring flora.

[0169] The same or similar reference numerals correspond to the same or similar components;

[0170] The terms used in the drawings to describe positional relationships are only used for illustrative purposes and should not be construed as limiting this patent;

[0171] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not limitations on the implementation methods of the present invention. For ordinary technicians in the relevant field, other different forms of changes or modifications can be made on the basis of the above description. It is not necessary and impossible to list all the implementation methods here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for predicting the dynamic changes of cold spring long-term enriched bacterial communities, characterized in that: The following steps are involved: S1: Raman spectra, environmental parameters, α-diversity of bacterial communities and relative abundance of key bacterial groups of bacterial culture samples collected at different enrichment stages; The relative abundance of the key bacterial groups includes: the relative abundance of anaerobic methane-oxidizing archaea ANME, sulfate-reducing bacteria SRB and methanogens; Analyzing the Raman spectrum using the MCR-ALS algorithm to obtain metabolite data; The metabolite data, environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups are jointly saved as an original data set; S2: Use the CART decision tree algorithm to calculate the importance and filter the data of the original data set to obtain the filtered data set; S3: Optimize and train the random forest prediction model used for prediction using the filtered data set to obtain a trained random forest prediction model; S4: Obtain metabolite data and environmental parameters of the bacterial culture sample to be predicted and input them into a trained random forest prediction model. The trained random forest prediction model outputs prediction results of bacterial community alpha diversity and relative abundance of key bacterial groups.

2. The method for predicting the dynamic changes of a cold spring long-term enriched bacterial community according to claim 1, characterized in that: In step S1, the bacterial enrichment phase includes: an adaptation phase, a growth phase, a reproduction phase and a stabilization phase.

3. The method for predicting the dynamic changes of a cold spring long-term enriched bacterial community according to claim 1, characterized in that: In step S1, the environmental parameters include: water quality physical and chemical parameters, nutrient salt content and OD600 bacterial concentration parameters; The bacterial community alpha diversity includes: Shannon index, Chao1 index, Ace index and Simpson index; The metabolite data include: the contents of formic acid, acetic acid, methanol, formaldehyde and sulfate.

4. The method for predicting the dynamic changes of a cold spring long-term enriched bacterial community according to claim 3, characterized in that: The water quality physical and chemical parameters include: dissolved methane and carbon dioxide content, temperature, salinity, pH, dissolved oxygen content; The nutrient salt content includes: nitrate, dissolved organic matter DOM and total organic carbon TOC content.

5. The method for predicting the dynamic changes of a cold spring long-term enriched bacterial community according to claim 4, characterized in that: Raman spectra were collected using a micro-confocal Raman spectroscopy probe; Use multi-parameter water quality probes to collect water quality physical and chemical parameters; The nutrient salt content and OD600 bacterial concentration parameters were collected using an ultraviolet spectrophotometer; DNA extraction, PCR amplification, library construction and sequencing are performed on the bacterial culture samples at different enrichment stages in sequence to obtain DNA sequencing data; and the alpha diversity of the bacterial community and the relative abundance of key bacterial groups are obtained based on the DNA sequencing data.

6. The method for predicting the dynamic changes of a cold spring long-term enriched bacterial community according to claim 1 or 5, characterized in that: The step S1 further includes: using the MCR-BANDS algorithm to evaluate the parsing accuracy of the MCR-ALS algorithm.

7. The method for predicting the dynamic changes of a cold spring long-term enriched bacterial community according to claim 1, characterized in that: The step S2 comprises: A decision tree is constructed using the original data set, and the Gini coefficient of each decision node in the decision tree is calculated using the CART decision tree algorithm; Calculate the variable importance score VIM of each data in the original data set according to the Gini coefficient; Data is screened according to the variable importance score VIM, data with a variable importance score VIM less than a preset threshold is eliminated, and a screened data set is obtained.

8. The method for predicting the dynamic changes of cold spring long-term enriched bacterial communities according to claim 1, characterized in that: The step S3 comprises: Divide the filtered data set into training set and test set; Using the training set to train the random forest prediction model, to obtain a trained random forest prediction model; The test set is used to evaluate the performance parameters of the trained random forest prediction model, and the trained random forest prediction model is iteratively optimized until the performance parameters meet the preset conditions to obtain the trained random forest prediction model.

9. The method for predicting the dynamic changes of the community of cold spring long-term enriched bacteria according to claim 8, characterized in that: The filtered data set was divided into training set and test set in a ratio of 9:1 using the ten-fold crossover method; The performance parameters include coefficient of determination, root mean square error and mean absolute error; The larger the determination coefficient is, and the smaller the root mean square error and mean absolute error are, the better the performance of the trained random forest prediction model is.

10. A system for predicting dynamic changes of a cold spring long-term enriched bacterial community, using a method for predicting dynamic changes of a cold spring long-term enriched bacterial community as claimed in any one of claims 1 to 9, characterized in that: include: Data acquisition unit: used to collect Raman spectra, environmental parameters, bacterial community α-diversity and relative abundance of key bacterial groups of bacterial culture samples at different enrichment stages; Analyzing the Raman spectrum using the MCR-ALS algorithm to obtain metabolite data; The metabolite data, environmental parameters, bacterial community alpha diversity and relative abundance of key bacterial groups are jointly saved as an original data set; Data screening unit: used to use the CART decision tree algorithm to perform importance calculation and data screening on the original data set to obtain the screened data set; Model training unit: used to optimize and train the random forest prediction model used for prediction using the screened data set, and obtain the trained random forest prediction model; Community dynamic change prediction unit: used to obtain metabolite data and environmental parameters of the bacterial culture sample to be predicted and input them into the trained random forest prediction model. The trained random forest prediction model outputs the prediction results of bacterial community alpha diversity and relative abundance of key bacterial groups.

Images