Method and system for distinguishing American ginseng by integrating metabonomics and machine learning

Abstract

The invention discloses a method and a system for distinguishing American ginseng by integrating metabonomics and machine learning. The method comprises the following steps: preparing wild and cultivated American ginseng samples and a metabolic extract sample solution; carrying out ultra-high performance liquid chromatography-quadrupole rod-flight time high resolution mass spectrometry detection analysis on the metabolic extract sample solution to obtain LC-MS data of wild and cultivated American ginseng sample components; performing multivariate data analysis on the data based on a metabonomics technology and a molecular network strategy, and screening and annotating differential metabolites of wild and cultivated American ginseng; establishing and training a machine learning diagnosis model of differential metabolites of the wild and cultivated American ginseng to distinguish the wild and cultivated American ginseng; according to the method, metabonomics based on liquid chromatography-mass spectrometry is combined with machine learning to distinguish wild American ginseng and cultivated varieties, the difference of metabolites between the wild American ginseng and the cultivated American ginseng is determined, and classification is performed through an artificial intelligence diagnosis model, so that the origin and the cultivation method of the American ginseng are accurately identified.

Description

Technical Field

[0001] The present invention relates to the technical field of American ginseng identification, and more specifically to a method and system for distinguishing American ginseng by integrating metabolomics and machine learning. Background Art

[0002] American ginseng is a globally recognized medicinal and dietary plant, known for its important clinical uses and health-promoting properties. Wild American ginseng is mainly distributed in North America, especially in the United States and Canada, between 40° and 50° north latitude and 67° and 125° west longitude. However, wild populations are scarce and difficult to maintain, and advanced cultivation techniques are needed to ensure a stable supply.

[0003] At present, morphological characteristics and microscopic observation are still the main methods for distinguishing American ginseng varieties. Practice has shown that naked eye observation has poor reliability and it is difficult to distinguish the origin of American ginseng. Although genomics and other technologies have been applied to the identification of American ginseng in recent years, the procedures are cumbersome, the technology is difficult, and it is time-consuming and labor-intensive. In addition, American ginseng of different varieties and origins differs in quality and clinical efficacy, and its metabolic differences remain largely unexplored. Accurately distinguishing wild, cultivated and different sources of American ginseng is crucial for variety identification, geographical origin tracking and quality control.

[0004] Non-targeted metabolomics uses advanced liquid chromatography-mass spectrometry (LC-MS) analysis to obtain metabolite profiles in analytical samples, compare metabolite content differences, and systematically screen some specific biomarkers from the differential metabolites, which is helpful for the identification of plant phenotypes and the analysis of metabolic variation.

[0005] Therefore, how to determine the differences in metabolites between different varieties of American ginseng and identify American ginseng varieties efficiently, accurately and quickly is a problem that technical personnel in this field urgently need to solve. Summary of the invention

[0006] In view of this, the present invention provides a method and system for distinguishing American ginseng by integrating metabolomics and machine learning to solve the technical problems mentioned in the background technology.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] A method for distinguishing American ginseng by integrating metabolomics and machine learning, comprising the following steps:

[0009] S1. Prepare wild American ginseng and cultivated American ginseng samples, and prepare metabolic extract sample solutions;

[0010] S2. The metabolic extract sample solution was analyzed by ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry to obtain LC-MS data of the components of wild and cultivated American ginseng samples;

[0011] S3. Based on metabolomics technology and molecular network strategy, multivariate data analysis was performed on the LC-MS data of the components of wild and cultivated American ginseng samples, and the differential metabolites of wild and cultivated American ginseng were screened and annotated;

[0012] S4. Establish and train a machine learning diagnostic model of differential metabolites between wild and cultivated American ginseng to distinguish between wild and cultivated American ginseng.

[0013] Preferably, the method for preparing wild American ginseng and cultivated American ginseng samples is:

[0014] Take wild and cultivated American ginseng samples, grind them and pass them through No. 3 sieve, accurately weigh 1g of powder, put it in a stoppered conical flask, accurately add 50ml of water-saturated n-butanol, weigh the weight, heat reflux in a water bath to extract for 1.5 hours, cool, weigh again, make up the lost weight with water-saturated n-butanol, shake well, and filter.

[0015] Preferably, the method for preparing the metabolic extract sample solution is:

[0016] Accurately measure 25 ml of the filtrate, place it in an evaporating dish, evaporate it to dryness, add an appropriate amount of 50% methanol to the residue to dissolve it, transfer it to a 10 ml volumetric flask, add 50% methanol to the scale, shake well, pass it through a 0.22 μm microporous filter membrane, and take the filtrate.

[0017] Preferably, ultra-high performance liquid chromatography-quadrupole-time of flight high resolution mass spectrometry detection analysis includes:

[0018] S21. Optimize and evaluate the chromatographic columns, mobile phase types, and gradient elution procedures for chromatographic testing;

[0019] S22. Electrospray ionization was used for ionization for mass spectrometry detection, including ESI+ positive ion mode and ESI- negative ion mode.

[0020] Preferably, the optimized chromatographic conditions of step S21 are: chromatographic column C18, mobile phase is organic solvent acetonitrile and aqueous phase, the aqueous phase is 0.05mmol / L formic acid-water, gradient elution program is: 0-5min, 5-15% acetonitrile; 5-20min, 15-25% acetonitrile; 20-25min, 25-35% acetonitrile; 25-35min, 35-45% acetonitrile; 35-40min, 45-55% acetonitrile; flow rate 0.4mL / min, injection volume 2μL;

[0021] The contents of mass spectrometry detection were as follows: before analysis, accurate mass calibration was performed using Agilent standard conditioning mixture, the primary mass spectrometry scanning range was set to 100-1700 m / z, the drying gas was nitrogen, the temperature was 325°C, the flow rate was 6.8 L / min, the sheath gas temperature was 350°C, the capillary voltage was 4.0 kV, and the primary collision voltage was 150 V; differential metabolites were scanned in targeted mode for targeted MS analysis, the secondary mass spectrometry scanning range was 50-1000 m / z, and the secondary fragmentation voltages were 10 V, 25 V, and 40 V.

[0022] Preferably, the specific content of step S3 is:

[0023] S31. Preprocessing the raw LC-MS data, including peak extraction, identification, matching, alignment and normalization, to generate a metabolite expression matrix;

[0024] S32. Two unsupervised analysis models, principal component analysis (PCA) and hierarchical cluster analysis (HCA), were used to identify the chemical phenotypes of wild and cultivated American ginseng based on the metabolite expression matrix.

[0025] S33. + Positive ion mode and ESI - In the negative ion mode, the orthogonal partial least squares discriminant analysis model was used to screen the differential components among different American ginseng varieties;

[0026] S34. R2, Q2 and cross-validation tests were used to evaluate the model performance, and the differences in metabolites among American ginseng varieties were determined based on the variable weight values ​​and the p-values ​​of the t-test;

[0027] S35. Differential metabolites in American ginseng samples were identified by accurate molecular weight determination of molecular ion peaks and secondary mass spectrometry fragmentation analysis, and the main metabolites were determined by comparison with standards and mass spectrometry databases.

[0028] Preferably, the eight significantly different metabolites are: ginsenoside Rg6, ginsenoside Rg2, ginsenoside F1, ginsenoside Rk2, ginsenoside R2 and ginsenoside Rg10, as well as oleanolic acid-28-O-β-D-glucopyranoside and oleanolic acid-12-ene-3,11-dione.

[0029] Preferably, the method for establishing the machine learning diagnostic model in step S4 is specifically as follows:

[0030] Python was used to develop and evaluate machine learning models. Multiple classifiers including Gaussian Bayesian, logistic regression, neural network, random forest, and support vector machine were used to classify American ginseng. The ginseng samples were divided into a training set of 25 samples and a test set of 10 samples in a ratio of 5:2. The stability of the algorithm was verified by the 10-fold cross-validation method, and the model performance was evaluated using precision, recall, F1 score, and confusion matrix indicators.

[0031] Preferably, in step S4, the machine learning diagnostic model for differential metabolites between wild and cultivated American ginseng is a random forest classifier.

[0032] A system for distinguishing American ginseng by integrating metabolomics and machine learning, based on the method for distinguishing American ginseng by integrating metabolomics and machine learning, comprising a metabolite extract preparation device, an LC-MS detection module, a differential metabolite characterization module and a diagnosis module;

[0033] A metabolite extract preparation device, used for preparing samples of wild American ginseng and cultivated American ginseng, and preparing a metabolic extract sample solution;

[0034] LC-MS detection module, used to analyze the metabolic extract sample solution by ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry detection, to obtain LC-MS data of the components of wild and cultivated American ginseng samples;

[0035] The differential metabolite characterization module is used to perform multivariate data analysis on the LC-MS data of wild and cultivated American ginseng sample components based on metabolomics technology and molecular network strategies, and to screen and annotate the differential metabolites of wild and cultivated American ginseng;

[0036] The diagnostic module is used to establish and train a machine learning diagnostic model of differential metabolites between wild and cultivated American ginseng to distinguish between wild and cultivated American ginseng.

[0037] It can be seen from the above technical solution that compared with the prior art, the present invention discloses a method and system for distinguishing American ginseng by integrating metabolomics and machine learning, and proposes the use of metabolomics based on liquid chromatography-mass spectrometry LC-MS combined with machine learning to distinguish wild American ginseng from cultivated varieties, and determines the differences in metabolites between wild and cultivated American ginseng. It also constructs and trains an artificial intelligence diagnostic model for classification, so as to accurately identify the origin and cultivation method of American ginseng, which is of great significance for the identification of American ginseng varieties, the traceability of origin, and the promotion of market supervision. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0039] Figure 1 A schematic diagram of a method for distinguishing American ginseng by integrating metabolomics and machine learning provided by the present invention;

[0040] Figure 2 The chromatographic mass spectrum of American ginseng in positive ion detection provided by the present invention;

[0041] Figure 3 The chromatographic mass spectrum of the American ginseng in negative ion detection provided by the present invention;

[0042] Figure 4 The metabolic phenotype differentiation of American ginseng by principal component analysis and hierarchical cluster analysis under ESI+ and ESI- modes provided by the present invention;

[0043] Figure 5 A schematic diagram of the OPLS-DA provided by the present invention for distinguishing the metabolic phenotypes of American ginseng;

[0044] Figure 6 A schematic diagram of the differential metabolites between American ginseng cultivated in China and wild in North America from different origins provided by the present invention;

[0045] Figure 7 A schematic diagram of the differential metabolites between cultivated and wild American ginseng from different origins in North America provided by the present invention;

[0046] Figure 8 A schematic diagram of the differential metabolites between American ginseng cultivated in China and cultivated in North America from different origins provided by the present invention;

[0047] Fig. 9 A schematic diagram of screening differential metabolites of American ginseng from three different origins in a positive ion state mode provided by the present invention;

[0048] Fig.10 A schematic diagram of screening differential metabolites of American ginseng from three different origins in the negative ion state mode provided by the present invention;

[0049] Fig.11 A schematic diagram of five machine learning methods for distinguishing American ginseng from different origins in the positive ion state mode provided by the present invention;

[0050] Fig.12A schematic diagram of five machine learning methods for distinguishing American ginseng from different origins in the positive ion state mode provided by the present invention. DETAILED DESCRIPTION

[0051] 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.

[0052] The embodiment of the present invention discloses a method for distinguishing American ginseng by integrating metabolomics and machine learning, such as Figure 1 , including the following steps:

[0053] S1. Prepare wild American ginseng and cultivated American ginseng samples, and prepare metabolic extract sample solutions;

[0054] S2. The metabolic extract sample solution was analyzed by ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry to obtain LC-MS data of the components of wild and cultivated American ginseng samples;

[0055] S3. Based on metabolomics technology and molecular network strategy, multivariate data analysis was performed on the LC-MS data of the components of wild and cultivated American ginseng samples, and the differential metabolites of wild and cultivated American ginseng were screened and annotated;

[0056] S4. Establish and train a machine learning diagnostic model of differential metabolites between wild and cultivated American ginseng to distinguish between wild and cultivated American ginseng.

[0057] In order to further implement the above technical solution, the method for preparing wild American ginseng and cultivated American ginseng samples is as follows:

[0058] Take wild and cultivated American ginseng samples, grind them and pass them through No. 3 sieve, accurately weigh 1g of powder, put it in a stoppered conical flask, accurately add 50ml of water-saturated n-butanol, weigh the weight, heat reflux in a water bath to extract for 1.5 hours, cool, weigh again, make up the lost weight with water-saturated n-butanol, shake well, and filter.

[0059] In order to further implement the above technical solution, the method for preparing the metabolic extract sample solution is:

[0060] Accurately measure 25 ml of the filtrate, place it in an evaporating dish, evaporate it to dryness, add an appropriate amount of 50% methanol to the residue to dissolve it, transfer it to a 10 ml volumetric flask, add 50% methanol to the scale, shake well, pass it through a 0.22 μm microporous filter membrane, and take the filtrate.

[0061] To further implement the above technical solution, ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry detection and analysis includes:

[0062] S21. Optimize and evaluate the chromatographic columns, mobile phase types, and gradient elution procedures for chromatographic testing;

[0063] S22. Ionization using electrospray ionization for mass spectrometry detection, including ESI + positive ion mode and ESI-negative ion mode.

[0064] In order to further implement the above technical scheme, the optimized chromatographic conditions of step S21 are: chromatographic column C18, mobile phase is organic solvent acetonitrile and aqueous phase, the aqueous phase is 0.05mmol / L formic acid (mass spectrometry grade)-water, gradient elution program is: 0-5min, 5-15% acetonitrile; 5-20min, 15-25% acetonitrile; 20-25min, 25-35% acetonitrile; 25-35min, 35-45% acetonitrile; 35-40min, 45-55% acetonitrile; flow rate 0.4mL / min, injection volume 2μL;

[0065] The contents of mass spectrometry detection were as follows: before analysis, accurate mass calibration was performed using Agilent standard conditioning mixture, the primary mass spectrometry scanning range was set to 100-1700 m / z, the drying gas was nitrogen, the temperature was 325°C, the flow rate was 6.8 L / min, the sheath gas temperature was 350°C, the capillary voltage was 4.0 kV, and the primary collision voltage was 150 V; differential metabolites were scanned in targeted mode for targeted MS analysis, the secondary mass spectrometry scanning range was 50-1000 m / z, and the secondary fragmentation voltages were 10 V, 25 V, and 40 V;

[0066] The results are as follows Figure 2 For ESI + Ion chromatogram of American ginseng LC-MS in positive ion mode, Figure 3 For ESI - Ion chromatogram of American ginseng LC-MS in negative ion mode.

[0067] In order to further implement the above technical solution, the specific content of step S3 is:

[0068] S31. Preprocessing the raw LC-MS data, including peak extraction, identification, matching, alignment and normalization, to generate a metabolite expression matrix;

[0069] S32. Two unsupervised analysis models, principal component analysis (PCA) and hierarchical cluster analysis (HCA), were used to identify the chemical phenotypes of wild and cultivated American ginseng based on the metabolite expression matrix.

[0070] S33. +Positive ion mode and ESI - In the negative ion mode, the orthogonal partial least squares discriminant analysis model was used to screen the differential components among different American ginseng varieties;

[0071] S34. R2, Q2 and cross-validation tests were used to evaluate the model performance, and the differences in metabolites among American ginseng varieties were determined based on the variable weight values ​​and the p-values ​​of the t-test;

[0072] S35. Identify the differential metabolites in American ginseng samples by accurate molecular weight determination of molecular ion peaks and secondary mass spectrometry fragmentation analysis, and determine the main metabolites by comparison with standards and mass spectrometry databases;

[0073] In this embodiment, if Figure 4 For ESI + and ESI - The metabolic phenotype characteristics of wild, North American and Chinese cultivated American ginseng were determined by principal component analysis (PCA) and hierarchical cluster analysis (HCA). + Mode Figure 3 Figure A shows the principal component analysis in the positive ion state to distinguish the metabolic phenotypes of American ginseng from different origins. PCA shows that the quality control (QC) samples are concentrated in the scatter plot, indicating the robustness and reliability of the metabolomics analysis adopted; Figure B shows the principal component analysis to distinguish the metabolic phenotypes of American ginseng from different origins, including wild American ginseng, North American cultivated American ginseng, and Chinese cultivated American ginseng; Figure C shows the hierarchical clustering analysis to distinguish the metabolic phenotypes of American ginseng from different origins. HCA further shows that all wild American ginseng samples formed a unique cluster, while North American and Chinese cultivated American ginseng were classified into another chemical subtype; + Metabolomics data are consistent, ESI - Pattern analysis also revealed different metabolic subtypes among wild, North American, and Chinese cultivated American ginseng, which was confirmed by PCA and HCA. Specifically, in the ESI-pattern Figure 3 Figure A contains a PCA scatter plot of QC samples, showing the robust clustering and reliability of the LC-ESI-MS method, Figure B is a PCA scatter plot of QC samples removed in ESI-mode, and Figure C is a hierarchical clustering analysis of metabolic differences among the three types of American ginseng in ESI-mode; these analyses together highlight the obvious metabolic differences between wild and cultivated American ginseng.

[0074] Figure 5 To distinguish the chemotypes of wild, North American and Chinese cultivated American ginseng using partial least squares discriminant analysis (OPLS-DA), specifically: Figure 4 Figure A shows the comparison of metabolites from different origins under the OPLS-DA model in the positive ion state. +The orthogonal partial least squares discriminant analysis (OPLS-DA) under the same model showed that there was an obvious trend of inter-group separation and intra-group aggregation in wild and North American cultivated American ginseng samples; Figure 4 Figure B is the permutation test of the OPLS-DA model. 200 permutation tests confirmed the stability and reliability of the established OPLS-DA model. Figure 4 Figure C shows the differential components of American ginseng from different origins screened by SUS-PLOT, Figure D shows the differential components of American ginseng from different origins screened by scatter plot, and Figure E shows the differential components with the largest content fluctuations. Using the screening criteria of VIP greater than 1 and absolute p(corr) value greater than 0.5, 228 differential metabolites were identified, of which 40 were annotated and characterized.

[0075] Figure 6 For ESI + OPLS-DA analysis of the differences in metabolites between wild (C) and cultivated (N) American ginseng in China under the model. The analysis in Figure A shows that there is obvious separation and intra-group clustering between wild American ginseng (C) and cultivated (N) American ginseng in China; 200 permutation tests in Figure B confirm the robustness and reliability of the established OPLS-DA model; The loading diagram in Figure C depicts different metabolites, showing that metabolites at different levels are located in the periphery, and the metabolite identification criteria are VIP>1, p(corr)>0.5;

[0076] Figure 7 For ESI + OPLS-DA analysis of wild (C) and North American cultivated (W) American ginseng under the model, where A is the OPLS-DA model under the positive ion state to compare the metabolites from different origins, and the scatter plot shows the separation of wild ginseng and North American cultivated ginseng samples, with intra-group clustering; 200 permutation tests verified the robustness and reliability of the OPLS-DA model, and B is the permutation test of the OPLS-DA model, and the replacement test confirmed the stability and reliability of the model; C is a scatter plot to screen the differential components of American ginseng from different origins, and the loading map of differential metabolites highlights the metabolites with variable surrounding contents. The screening criteria for metabolites are VIP>1, p(corr)>0.5;

[0077] Figure 8 For ESI + OPLS-DA differentiation of North American (W) and Chinese (N) cultivated American ginseng in the model, where the analysis of Figure A shows that North American (W) and Chinese (N) cultivated American ginseng samples are separated into different groups, and clustering is observed within each group; 200 permutation tests in Figure B confirm the robustness and reliability of the OPLS-DA model; The loading diagram in Figure C shows different metabolites, which show different levels of metabolites located around. The screening criteria are VIP>1, pcorr>0.5;

[0078] Compared with wild American ginseng (C), 34 metabolites were decreased in North American cultivated American ginseng (W), and 6 metabolites were increased in North American cultivated American ginseng (W);

[0079] Similarly, in the ESI-mode, the OPLS-DA model identified 248 different metabolites in wild American ginseng (C) and Chinese cultivated American ginseng samples, of which 39 metabolites were annotated, and the levels of 34 metabolites were decreased and 5 metabolites were increased in Chinese cultivated American ginseng (N) compared with wild American ginseng (C);

[0080] In order to further implement the above technical scheme, 8 significantly different differential metabolites are: ginsenoside Rg6, ginsenoside Rg2, ginsenoside F1, ginsenoside Rk2, ginsenoside R2 and ginsenoside Rg10, as well as oleanolic acid-28-O-β-D-glucopyranoside and oleanolic acid-12-ene-3,11-dione.

[0081] like Fig. 9 From the comparison of ESI+ differential metabolites among the three groups, there were significant differences in 8 differential metabolites among wild ginseng, North American cultivated American ginseng and Chinese cultivated American ginseng: ginsenoside Rg6, ginsenoside Rg2, ginsenoside F1, ginsenoside Rk2, ginsenoside R2 and ginsenoside Rg10, as well as two oleanolic acid-type triterpenes (oleanolic acid-28-O-β-D-glucopyranoside and oleanolic acid-12-ene-3,11-dione). Compared with North American and Chinese cultivated ginseng, six ginsenosides were particularly abundant in wild American ginseng; the expression levels of ginsenoside Rk2, ginsenoside R2 and ginsenoside Rg6 were in the order of wild American ginseng > North American cultivated American ginseng > Chinese cultivated American ginseng.

[0082] like Fig.10 Figure 3 ESI-differential metabolite analysis of three groups. The comparative analysis of differential metabolites in Figure A identified three different compounds, including two ginsenosides (ginsenoside Rc and ginsenoside Re) and sucrose. Figure B shows that among the three groups, the Rd2 metabolite has the highest content in wild American ginseng, followed by ginsenoside Re, which is also higher in wild American ginseng, while fructose is higher in North American ginseng. Figure C shows that the expression trend of ginsenoside Rd2 is: wild American ginseng > Chinese cultivated American ginseng > North American cultivated American ginseng. Figure D is a structural description of the differential metabolites.

[0083] In order to further implement the above technical solution, in step S4, the machine learning diagnostic model for the differential metabolites between wild and cultivated American ginseng is a random forest classifier.

[0084] In this embodiment, machine learning model development and evaluation were performed using Python 3.11.4, and multiple classifiers such as Gaussian Bayes, logistic regression, neural network, random forest, and support vector machine were used to classify American ginseng, and they were classified according to statistically significant differential metabolites. The American ginseng samples were divided into a training set of 25 samples and a test set of 10 samples at a ratio of 5:2. The stability of the algorithm was verified by a 10-fold cross-validation method, and the model performance was evaluated using precision, recall, F1 score, and confusion matrix indicators:

[0085]

[0086] like Fig.11 , using ESI + The expression matrix of 8 differential metabolites identified in the model was used to construct multiple machine learning classifiers, aiming to identify the origin of American ginseng. The training set results showed that the recognition accuracy of random forest classifier and support vector machine both reached 100%, which was better than Gaussian naive Bayes, logistic regression and multi-layer perceptron classifiers, whose recognition accuracies were 90.91%, 72.73% and 63.63%, respectively; in the validation set, random forest classifier and support vector machine maintained the lead with 100% accuracy, followed by Gaussian naive Bayes (90.91%), logistic regression (72.73%) and multi-layer perceptron classifier (63.63%); misclassification was observed using Gaussian naive Bayes, logistic regression and multi-layer perceptron classifier algorithms; random forest classifier and support vector machine had the highest diagnostic performance for American ginseng, indicating that the combination of metabolomics and machine learning can effectively identify ginseng from different origins.

[0087] In addition, if Fig.12 , using ESI - The expression matrix of three differential metabolites in the model was used to establish multiple machine learning classifiers to diagnose the source of American ginseng. Among them, the random forest classifier had the highest accuracy in the training set (100.0%), followed by the logistic regression classifier and Gaussian naive Bayes classifier (95.83%), the multilayer perceptron classifier (87.5%) and the support vector machine (70.83%); in the test set, the accuracy of the random forest classifier, logistic regression and Gaussian naive Bayes classifier all reached 100%, among which the accuracy of the multilayer perceptron classifier was 90.91% and the accuracy of the support vector machine was 63.64%; except for the random forest classifier, all algorithms had misclassification problems; this shows that the random forest classifier has excellent diagnostic performance in the classification of American ginseng, which strengthens the practicality of ginseng source tracking combined with metabolomics and machine learning.

[0088] A system for differentiating American ginseng by integrating metabolomics and machine learning, based on a method for differentiating American ginseng by integrating metabolomics and machine learning, comprising a metabolite extract preparation device, an LC-MS detection module, a differential metabolite characterization module, and a diagnosis module;

[0089] A metabolite extract preparation device, used for preparing samples of wild American ginseng and cultivated American ginseng, and preparing a metabolic extract sample solution;

[0090] LC-MS detection module, used to analyze the metabolic extract sample solution by ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry detection, to obtain LC-MS data of the components of wild and cultivated American ginseng samples;

[0091] The differential metabolite characterization module is used to perform multivariate data analysis on the LC-MS data of wild and cultivated American ginseng sample components based on metabolomics technology and molecular network strategies, and to screen and annotate the differential metabolites of wild and cultivated American ginseng;

[0092] The diagnostic module is used to establish and train a machine learning diagnostic model of differential metabolites between wild and cultivated American ginseng to distinguish between wild and cultivated American ginseng.

[0093] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0094] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for distinguishing American ginseng by integrating metabolomics and machine learning, characterized in that: The following steps are involved: S1. Prepare wild American ginseng and cultivated American ginseng samples, and prepare metabolic extract sample solutions; S2. The metabolic extract sample solution was analyzed by ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry to obtain LC-MS data of the components of wild and cultivated American ginseng samples; S3. Based on metabolomics technology and molecular network strategy, multivariate data analysis was performed on the LC-MS data of the components of wild and cultivated American ginseng samples, and the differential metabolites of wild and cultivated American ginseng were screened and annotated; S4. Establish and train a machine learning diagnostic model of differential metabolites between wild and cultivated American ginseng to distinguish between wild and cultivated American ginseng.

2. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 1, characterized in that: The method for preparing wild American ginseng and cultivated American ginseng samples is as follows: Take wild and cultivated American ginseng samples, grind them and pass them through No. 3 sieve, accurately weigh 1g of powder, put it in a stoppered conical flask, accurately add 50ml of water-saturated n-butanol, weigh the weight, heat reflux in a water bath to extract for 1.5 hours, cool, weigh again, make up the lost weight with water-saturated n-butanol, shake well, and filter.

3. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 1, characterized in that: The method for preparing the metabolic extract sample solution is: Accurately measure 25 ml of the filtrate, place it in an evaporating dish, evaporate it to dryness, add an appropriate amount of 50% methanol to the residue to dissolve it, transfer it to a 10 ml volumetric flask, add 50% methanol to the scale, shake well, pass it through a 0.22 μm microporous filter membrane, and take the filtrate.

4. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 1, characterized in that: Ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry detection analysis includes: S21. Optimize and evaluate the chromatographic columns, mobile phase types, and gradient elution procedures for chromatographic testing; S22. Ionization using electrospray ionization for mass spectrometry detection, including ESI + positive ion mode and ESI-negative ion mode.

5. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 4, characterized in that: The optimized chromatographic conditions of step S21 are: chromatographic column C18, mobile phase is organic solvent acetonitrile and aqueous phase, the aqueous phase is 0.05mmol / L formic acid-water, gradient elution program is: 0-5min, 5-15% acetonitrile; 5-20min, 15-25% acetonitrile; 20-25min, 25-35% acetonitrile; 25-35min, 35-45% acetonitrile; 35-40min, 45-55% acetonitrile; flow rate 0.4mL / min, injection volume 2μL; The contents of mass spectrometry detection were as follows: before analysis, accurate mass calibration was performed using Agilent standard conditioning mixture, the primary mass spectrometry scanning range was set to 100-1700 m / z, the drying gas was nitrogen, the temperature was 325°C, the flow rate was 6.8 L / min, the sheath gas temperature was 350°C, the capillary voltage was 4.0 kV, and the primary collision voltage was 150 V; differential metabolites were scanned in targeted mode for targeted MS analysis, the secondary mass spectrometry scanning range was 50-1000 m / z, and the secondary fragmentation voltages were 10 V, 25 V, and 40 V.

6. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 1, characterized in that: The specific content of step S3 is: S31. Preprocessing the raw LC-MS data, including peak extraction, identification, matching, alignment and normalization, to generate a metabolite expression matrix; S32. Two unsupervised analysis models, principal component analysis (PCA) and hierarchical cluster analysis (HCA), were used to identify the chemical phenotypes of wild and cultivated American ginseng based on the metabolite expression matrix. S33. + In positive ion mode and ESI-negative ion mode, the orthogonal partial least squares discriminant analysis model was used to screen the differential components among different American ginseng varieties; S34. R2, Q2 and cross-validation tests were used to evaluate the model performance, and the differences in metabolites among American ginseng varieties were determined based on the variable weight values ​​and the p-values ​​of the t-test; S35. Differential metabolites in American ginseng samples were identified by accurate molecular weight determination of molecular ion peaks and secondary mass spectrometry fragmentation analysis, and the main metabolites were determined by comparison with standards and mass spectrometry databases.

7. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 1, characterized in that: The eight significantly different metabolites were ginsenoside Rg6, ginsenoside Rg2, ginsenoside F1, ginsenoside Rk2, ginsenoside R2 and ginsenoside Rg10, as well as oleanolic acid-28-O-β-D-glucopyranoside and oleanolic acid-12-ene-3,11-dione.

8. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 1, characterized in that: The method for establishing the machine learning diagnosis model in step S4 is specifically as follows: Python was used to develop and evaluate machine learning models. Multiple classifiers including Gaussian Bayesian, logistic regression, neural network, random forest, and support vector machine were used to classify American ginseng. The ginseng samples were divided into a training set of 25 samples and a test set of 10 samples in a ratio of 5:

2. The stability of the algorithm was verified by the 10-fold cross-validation method, and the model performance was evaluated using precision, recall, F1 score, and confusion matrix indicators.

9. The method for distinguishing American ginseng by integrating metabolomics and machine learning according to claim 1, characterized in that: In step S4, the machine learning diagnostic model for the differential metabolites between wild and cultivated American ginseng is a random forest classifier.

10. A system for distinguishing American ginseng by integrating metabolomics and machine learning, characterized in that: A method for distinguishing American ginseng by integrating metabolomics and machine learning based on any one of claims 1 to 9, comprising a metabolite extract preparation device, an LC-MS detection module, a differential metabolite characterization module, and a diagnosis module; A metabolite extract preparation device, used for preparing samples of wild American ginseng and cultivated American ginseng, and preparing a metabolic extract sample solution; LC-MS detection module, used to analyze the metabolic extract sample solution by ultra-high performance liquid chromatography-quadrupole-time of flight high-resolution mass spectrometry detection, to obtain LC-MS data of the components of wild and cultivated American ginseng samples; The differential metabolite characterization module is used to perform multivariate data analysis on the LC-MS data of wild and cultivated American ginseng sample components based on metabolomics technology and molecular network strategies, and to screen and annotate the differential metabolites of wild and cultivated American ginseng; The diagnostic module is used to establish and train a machine learning diagnostic model of differential metabolites between wild and cultivated American ginseng to distinguish between wild and cultivated American ginseng.

Images