Establishment method and application of ephedra root producing area identification / procyanidine quantitative model based on near infrared spectrum
Through near-infrared spectroscopy technology and data modeling methods, a model of ephedra root origin identification and proanthocyanin quantitative analysis was established, solving the cumbersome and time-consuming problems of traditional methods, and achieving rapid and accurate detection results.
Patent Information
- Application Number
- CN202411676783.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-22
- Publication Date
- 2025-02-28
AI Technical Summary
The prior art is difficult to effectively identify the ephedra roots and their proanthocyanins content in different origins. The traditional methods are cumbersome, time-consuming and damage to the samples.
By collecting ephedra root samples from different origins, detecting their near-infrared spectral data, and dividing the samples using the KS algorithm to perform data preprocessing. Then, a place of origin identification model and a quantitative model of proanthocyanin were established, and feature variables and PLS-DA models were extracted using CARS, and LS-SVM modeling were modeled.
The non-destructive, quick and efficient identification of the origin and proanthocyanin content of ephedra roots is achieved, and the accuracy and efficiency of detection are improved.
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy, and belongs to the field of medicinal material detection. Background Art
[0002] Ephedra root is the dried rhizome of Ephedra sinica Stapf and Ephedraintremedia Schrenket CAMey., plants of the Ephedraceae family. It is sweet, astringent, and neutral in nature, and is regulated by the heart and lung meridians. Its traditional efficacy is to strengthen the exterior and stop sweating, and it is used for spontaneous sweating and night sweats. Modern pharmacological studies have found that ephedra root contains a variety of chemical components such as alkaloids, flavonoids, and organic acids, and has pharmacological effects such as anti-inflammatory, antioxidant, anti-tumor, and hypotensive. Proanthocyanidins are flavonoid dimer or polymer compounds. Studies have found that ephedra root is rich in proanthocyanidins. Tao Huaming isolated a variety of proanthocyanidin dimer compounds from ephedra root, and other studies have found that the proanthocyanidin components of ephedra root have good antioxidant and anti-ulcerative colitis effects.
[0003] Ephedra root is mainly produced in Shanxi, Hubei, Gansu, Xinjiang, Inner Mongolia, Yunnan and other regions. The geographical conditions and ecological environment of different origins will lead to certain differences in the content of active ingredients and present different quality characteristics. The inspection of ephedra root in the "Chinese Pharmacopoeia 2020 Edition" only includes the determination of moisture and ash content, and does not determine the index of its chemical composition content, which cannot effectively detect the quality of ephedra root. Therefore, it is of great significance to identify ephedra roots from different origins and determine the content of their active ingredients.
[0004] Traditional methods for qualitative analysis and quantitative identification of the origin of medicinal materials mainly include microscopic identification, fingerprint, gas chromatography, liquid chromatography, etc. These methods are cumbersome, time-consuming, and will cause certain damage to the sample. Therefore, it is necessary to establish a simple, fast and reliable qualitative and quantitative analysis method for ephedra root. Near-infrared spectroscopy technology is widely used in food, medicine, agriculture and other fields with its outstanding advantages of non-destructive, fast and efficient. In recent years, near-infrared has also been used in the analysis and quality control of traditional Chinese medicines, such as Schisandra chinensis, saffron, and Poria cocos. However, due to its weak absorption intensity, wide spectral band, and serious spectral band overlap, it cannot be used alone as a means of qualitative and quantitative analysis. It is usually combined with chemometrics to obtain qualitative or quantitative information to establish an analysis model. There is currently no report on the near-infrared spectroscopy detection method for ephedra root. Summary of the invention
[0005] Purpose of the invention: The purpose of the present invention is to provide a method for constructing an ephedra root origin identification / proanthocyanidin quantification model based on near-infrared spectroscopy and its application.
[0006] Technical solution: The present invention provides a method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy, comprising the following steps:
[0007] (1) Collect ephedra root samples from different origins and detect their near-infrared spectral data;
[0008] (2) Divide the samples using KS algorithm;
[0009] (3) Perform data preprocessing;
[0010] (4) Model building:
[0011] (4.1) Establishment of origin identification model: Carry out CARS to extract characteristic variables from the preprocessed data, determine the optimal variable subset, and establish the PLS-DA model, which is the origin identification model;
[0012] (4.2) Establishment of the quantitative model of proanthocyanidins: The proanthocyanidin content in the sample was determined, and LS-SVM modeling was performed together with the preprocessed data to obtain the proanthocyanidin quantitative model.
[0013] Furthermore, the near infrared spectroscopy data detection conditions described in step (1) are: scanning range 12500-4000cm -1 , scan times 64 times, resolution 8cm -1 .
[0014] Furthermore, the production areas described in step (1) include Gansu, Inner Mongolia, Yunnan and Hebei.
[0015] Furthermore, when the established model is an origin identification model, the sample division method in step (2) is to use two-thirds of the samples from each region as a calibration set and one-third as a validation set; when the established model is an anthocyanin quantification model, the sample division method in step (2) is to use two-thirds of the total number of samples as a calibration set and one-third as a validation set.
[0016] Furthermore, when the established model is an origin identification model, the data preprocessing method in step (3) is the first-order derivative method.
[0017] Furthermore, when the established model is a proanthocyanidin quantitative model, the data preprocessing method in step (3) is the standard normal variable method.
[0018] Furthermore, the 48th variable subset selected in step (4.1) is the optimal variable subset, containing 83 variables.
[0019] Furthermore, the method for determining the proanthocyanidin content in step (4.2) is the DMAC method.
[0020] Furthermore, the specific steps of the DMAC method are: using ethanol: water: HCl = 6: 1: 1 as a solvent, diluting DMAC powder to obtain a DMAC solution; weighing a proanthocyanidin standard in a volumetric flask, adding ethanol to dissolve, and preparing a proanthocyanidin standard solution with a concentration of 0 to 100 μg / mL; adding the DMAC solution to the proanthocyanidin standard solution, reading the absorbance at 640 nm every 1 minute at 25° C. for 30 minutes, drawing a proanthocyanidin calibration curve, measuring the absorbance of the sample under the same conditions, and inserting it into the calibration curve.
[0021] The present invention also provides the application of the ephedra root origin identification / proanthocyanidin quantification model constructed by the above method in identifying the ephedra root origin or proanthocyanidin content, and the origins include Gansu, Inner Mongolia, Yunnan and Hebei.
[0022] Furthermore, the near-infrared spectral data of the ephedra roots to be tested from the above four origins are detected and brought into the origin identification model / proanthocyanidin quantification model to obtain the ephedra root origin / proanthocyanidin content information.
[0023] Beneficial effects: Compared with the prior art, the present invention has the following outstanding significant advantages: the ephedra root origin identification / proanthocyanidin quantification model constructed by the present invention can realize the ephedra root origin identification and proanthocyanidin quantification analysis of the above four origins in a non-destructive, fast and efficient manner, only by measuring near-infrared spectral data. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 : Flow chart of the model constructed by the present invention.
[0025] Figure 2 : Original near-infrared spectra of ephedra root and near-infrared spectra after pretreatment.
[0026] Figure 3 : Principal component analysis diagram of Ephedra root and principal component analysis diagram after pretreatment.
[0027] Figure 4 : CARS screening variable process diagram.
[0028] Figure 5 : MCCV selection latent variable map.
[0029] Figure 6 : Correlation diagram of validation set and calibration set for LSSVM quantitative model establishment. DETAILED DESCRIPTION
[0030] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.
[0031] Example 1
[0032] 1. Experimental Materials
[0033] Proanthocyanidin A1 standard: Chengdu Purifa Technology Development Co., Ltd.; 4-(Dimethylamino)cinnamaldehyde (DMAC) solution: Beijing Bailingwei Technology Co., Ltd.; Anhydrous ethanol: Sinopharm Chemical Reagent Co., Ltd. 47 batches of ephedra roots were collected from different origins, including 14 batches in Gansu, a total of 70 samples, 17 batches in Inner Mongolia, a total of 85 samples, 14 batches in Yunnan, a total of 70 samples, and 2 batches in Hebei, a total of 10 samples.
[0034] 2. Near-infrared spectroscopy acquisition
[0035] Use a high-speed grinder to crush the ephedra root, pass it through a 60-mesh sieve, pack it in a ziplock bag, and store it dry. Put the ephedra root powder into a quartz cup and mix it evenly. Use air as a reference and use a Fourier transform near-infrared spectrometer in the integrating sphere diffuse reflectance mode to collect the sample spectrum. The scanning range is 12500-4000cm -1 , scan times 64 times, resolution 8cm -1 ,Each sample was collected three times in parallel and the average spectral value was taken.
[0036] The results are as follows Figure 2 As shown, Figure 2 The original near-infrared spectra of ephedra roots from different origins all have similar variation trends. -1 The absorption peak at 7000cm is mainly attributed to the first harmonic frequency of OH and the second harmonic frequency of CO. -1 There are also absorption peaks of varying degrees, mainly the near-infrared absorption band of R-OH. However, there is no obvious difference between the original spectra of ephedra roots from different origins, and the spectral information overlaps seriously. It is difficult to conduct qualitative analysis and quantitative identification of ephedra roots from different origins using the original near-infrared spectra. In addition, near-infrared spectral data may also be affected by irrelevant noise, baseline drift, overlapping bands, etc. In order to improve the stability and accuracy of subsequent modeling, it is necessary to eliminate various interference information in the original spectral data through preprocessing methods.
[0037] 3. Model Building
[0038] 3.1 Establishment of qualitative model
[0039] (1) The original near-infrared spectral data were divided using the KS algorithm. Two-thirds of the samples from the four origins were taken as the calibration set (158 sample data) and one-third was taken as the validation set (77 sample data).
[0040] (2) The raw near-infrared spectral data of ephedra root were preprocessed using Savitzky-Golay smoothing (SG9), first-order derivative (1D), second-order derivative (2D), and standard normal variable transformation (SNV).
[0041] (3) Perform principal component analysis (PCA) on the original spectral data and the preprocessed spectral data respectively.
[0042] The results are as follows Figure 3 As shown, Figure 3 a is the PCA result of original spectra from different origins. The distributions of samples from the four origins overlap and cannot be distinguished. Figure 3 b is the PCA result after 1D processing. The spatial distribution of each origin is separated to a certain extent, but qualitative identification between the origins is still not possible. The same is true when combined with the optimization of several other preprocessing methods. Therefore, the principal component analysis method is not selected.
[0043] (4) Partial least squares discriminant analysis (PLS-DA) was performed on the original spectral data and the preprocessed spectral data. The results are shown in Table 1.
[0044] Table 1
[0045] Preprocessing Latent variables Calibration set prediction accuracy Validation set prediction accuracy Raw 10 85.06 86.42 SG(9) 10 81.17 83.95 1D 9 91.56 87.65 2D 8 90.26 82.72 SNV 11 85.71 83.95
[0046] The data after 1D preprocessing has the best effect. In order to further improve the prediction accuracy of the model, the sample data is further subjected to competitive adaptive reweighted sampling (CARS) to screen variables. The number of sampling runs (N) is set to 100. The 1D preprocessed data is selected as an example in the CARS key variable extraction process. Figure 4 , A, B, and C are the changes in the number of variables, RMSECV, and variable regression coefficients in the 1D preprocessed data as the number of samplings increases. Figure A shows that in the early stage of sampling, the variables decrease rapidly, and as the number of samplings increases, the rate of variable reduction becomes gentle. Figure B shows that with the elimination of irrelevant variables, the RMSECV value becomes smaller, and the minimum is at 48 sampling times. Then, as the number of samplings increases, the RMSECV value begins to rise, indicating that some key information is eliminated, resulting in poor model prediction performance. Figure c shows the change path of the regression coefficient in each sampling. Given that the RMSECV of the PLS model established with the variable subset obtained in the 48th sampling is the smallest, this subset is taken as the key variable subset related to origin identification, with a total of 83 variables. The results of the PLS-DA analysis after screening the variables are shown in Table 2.
[0047] Table 2
[0048] Preprocessing Latent variables Calibration set prediction accuracy Validation set prediction accuracy Raw 10 85.06 75.32 SG(9) 10 85.06 77.78 1D 9 94.16 95.06 2D 8 90.91 74.07 SNV 11 89.61 79.01
[0049] It can be seen that after the CARS variable screening, the prediction accuracy of the calibration set under SG treatment did not change, and the prediction accuracy of the validation set decreased; although the prediction accuracy of the calibration set under 2D and SNV treatments was improved, the prediction accuracy of the validation set was also reduced; only the accuracy of the calibration set and validation set under 1D treatment was improved compared with before, the calibration set prediction accuracy was 94.16, and the validation set prediction accuracy was 95.06, which was more stable. Therefore, 1D was selected as the best pretreatment method. After the model was established, the near-infrared spectrum of the sample to be tested was collected, and the data was brought into the above model to obtain the origin information of Ephedra root.
[0050] 3.2 Establishment of quantitative model
[0051] (1) Determination of the content of proanthocyanidins in each sample
[0052] 1) Prepare standard stock solution and proanthocyanidin A1 concentration gradient working solution: dilute DMAC powder (4-(dimethylamino)cinnamaldehyde) with ethanol: water: HCl = 6: 1: 1 as solvent to obtain 0.1% (w / v) DMAC solution. Accurately weigh 10 mg of proanthocyanidin A1 standard in a volumetric flask, add ethanol to dissolve and dilute to the mark to obtain a standard stock solution with a concentration of 1.0 mg / mL. The solution is prepared and used immediately. Concentration configuration: 0μg / mL, 3.125μg / mL, 6.25μg / mL, 12.5μg / mL, 25μg / mL, 50μg / mL, 100μg / mL, to obtain a standard series of proanthocyanidin A1 working solutions.
[0053] Preparation of test samples: Crush the ephedra root and pass it through a 60-mesh sieve to obtain the ephedra root powder. Accurately weigh 2.00 g of each ephedra root powder in a 50 mL centrifuge tube, add 95% ethanol, the solid-liquid ratio is 15:1, and after ultrasonication for 90 min, the mixture is centrifuged at 3000× for 15 min at 4°C, and the supernatant is filtered through filter paper to obtain the test sample.
[0054] 2) Determine absorbance and draw a standard curve: After adding 70 μL of standard solution to a 96-well plate, add 210 μL of DMAC solution and read the absorbance at 640 nm every 1 minute at 25°C for 30 minutes. Calculate at maximum absorbance, draw a proanthocyanidin A1 calibration curve with absorbance as the ordinate and proanthocyanidin A1 concentration as the abscissa; the results are expressed as mg proanthocyanidin A1 / g sample. Each result is repeated three times in parallel. The calibration curve is: Y=0.021X+0.0554(R 2 =0.9935).
[0055] 3) Determine the proanthocyanidin content of each sample: Determine the absorbance of each sample to be tested and bring it into the calibration curve to calculate the content.
[0056] (2) The original near-infrared spectral data were divided into samples using the KS algorithm. Two-thirds of the total number of samples were used as the calibration set (158 sample data), and one-third was used as the validation set (77 sample data), regardless of origin.
[0057] (3) The grouped data were preprocessed using Savitzky-Golay smoothing (SG9), first-order derivative (1D), second-order derivative (2D), and standard normal variable transformation (SNV).
[0058] (4) The MCCV method was used to compare the optimal latent variables under different preprocessing conditions. The results are as follows: Figure 5 On this basis, the PLS model was established, and the results are shown in Table 3. The results under SG treatment were relatively good, but the prediction performance of each index was average.
[0059] Table 3
[0060] Preprocessing Latent variables RMSEC <![CDATA[Calibration set R 2 > RMSEP <![CDATA[Validation set R 2 > RPD Raw 10 1.6212 0.6610 1.6743 0.5188 1.4038 SG 10 1.6939 0.6298 1.6487 0.5127 1.4256 1D 7 1.6477 0.6514 1.7939 0.4597 1.3102 2D 6 1.8963 0.5470 1.9503 0.3519 1.2051 SNV 10 1.6540 0.6471 1.7258 0.4811 1.3663
[0061] (5) The modeling results of the preprocessed data using the least squares support vector machine (LS-SVM) are shown in Table 4, RMSEC (calibration set) and RMSEP (validation set) and their corresponding R 2 Compared with PLS, the modeling results are significantly improved, among which the modeling results under SNV pretreatment are the best, with RMSEC of 0.001 and R 2 is 1.0000, RMSEP is 0.5844, and its R 2 It is 0.9399.
[0062] Table 4
[0063] γ <![CDATA[δ 2 ]]> RMSEC <![CDATA[Calibration set R 2 > RMSEP <![CDATA[Validation set R 2 > RPD Raw 7233756.268 2.502683993 0.0012 1.0000 0.6447 0.9264 3.6457 SG 6854025.521 2.662632643 0.0020 1.0000 0.7228 0.9088 3.2518 1D 21841.741 3.04E-05 0.0002 1.0000 0.6164 0.9384 3.8132 2D 8350.8408 4.18E-06 0.0005 1.0000 1.3541 0.6735 1.7358 SNV 3392091.389 17.67178211 0.0001 1.0000 0.5844 0.9399 4.0222
[0064] The correlation between the calibration set and the validation set under this processing method is as follows Figure 6 , proving that the accuracy of proanthocyanidins in this method is high. Therefore, SNV pretreatment + LS-SVM was selected to establish a proanthocyanidin quantitative model. After the model was established, near-infrared spectra of the samples to be tested were collected, and the data were brought into the above model to obtain the proanthocyanidin content information in ephedra root.
Claims
1. A method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy, characterized in that: The following steps are involved: (1) Collect ephedra root samples from different origins and detect their near-infrared spectral data; (2) Divide the samples using KS algorithm; (3) Perform data preprocessing; (4) Model building: (4.1) Establishment of origin identification model: Carry out CARS to extract characteristic variables from the preprocessed data, determine the optimal variable subset, and establish the PLS-DA model, which is the origin identification model; (4.2) Establishment of the quantitative model of proanthocyanidins: The proanthocyanidin content in the sample was determined, and LS-SVM modeling was performed together with the preprocessed data to obtain the proanthocyanidin quantitative model.
2. The method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy according to claim 1, characterized in that: The near infrared spectroscopy data detection conditions described in step (1) are: scanning range 12500-4000cm -1 , scan times 64 times, resolution 8cm -1 .
3. The method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy according to claim 1, characterized in that: The origins described in step (1) include Gansu, Inner Mongolia, Yunnan and Hebei.
4. The method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy according to claim 1, characterized in that: When the established model is an origin identification model, the sample division method in step (2) is to use two-thirds of the samples from each region as a calibration set and one-third as a validation set; when the established model is an anthocyanin quantification model, the sample division method in step (2) is to use two-thirds of the total number of samples as a calibration set and one-third as a validation set.
5. The method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy according to claim 1, characterized in that: When the established model is an origin identification model, the data preprocessing method in step (3) is the first-order derivative method; when the established model is an anthocyanin quantitative model, the data preprocessing method in step (3) is the standard normal variable method.
6. The method for constructing an ephedra root origin identification / proanthocyanidin quantification model based on near infrared spectroscopy according to claim 1, characterized in that: The optimal variable subset described in step (4.1) is the variable subset screened for the 48th time, which contains 83 variables.
7. The method for constructing an ephedra root origin identification / proanthocyanidin quantitative model based on near infrared spectroscopy according to claim 1, characterized in that: The method for determining the proanthocyanidin content in step (4.2) is the DMAC method.
8. The method for constructing an ephedra root origin identification / proanthocyanidin quantification model based on near infrared spectroscopy according to claim 7, characterized in that: The specific steps of the DMAC method are: using ethanol: water: HCl = 6: 1: 1 as a solvent, diluting DMAC powder to obtain a DMAC solution; weighing a proanthocyanidin standard in a volumetric flask, adding ethanol to dissolve, and preparing a proanthocyanidin standard solution with a concentration of 0 to 100 μg / mL; Add DMAC solution to the proanthocyanidin standard solution, read the absorbance at 640 nm every 1 minute for 30 minutes at 25°C, draw a proanthocyanidin calibration curve, measure the absorbance of the sample under the same conditions, and substitute it into the calibration curve.
9. Use of the ephedra root origin identification / proanthocyanidin quantification model constructed by any one of claims 1 to 8 in identifying the origin of ephedra root or the proanthocyanidin content, characterized in that: The origins mentioned include Gansu, Inner Mongolia, Yunnan and Hebei.
10. The use according to claim 9, characterized in that: By detecting the near-infrared spectral data of the ephedra root to be tested and bringing it into the origin identification model / proanthocyanidin quantification model, the origin / proanthocyanidin content information of the ephedra root can be obtained.