Method for rapidly determining cannabidiol content based on mid-infrared spectrum prediction model

Abstract

The invention discloses a method for rapidly determining the content of cannabidiol based on an intermediate infrared spectrum prediction model, and belongs to the technical field of detection. The method comprises the following steps: pretreating different industrial hemp products, collecting an infrared spectrum curve of a sample, actually measuring the content of cannabidiol in the sample by liquid chromatography, pretreating and analyzing infrared spectrum data, establishing a model, and verifying the model. The method is based on the mid-infrared spectrum technology and is combined with the predictive capacity of the PLS model and the distinguishing effect of principal component analysis, and the content of cannabidiol in industrial hemp extract products and floral leaves can be rapidly detected. The method is simple and rich in information content, simple to operate, green, environment-friendly, reliable in performance and suitable for popularization and application, adopts nondestructive identification, has lower related consumables and maintenance cost, reduces the operation cost and the environmental protection risk, and saves a large amount of manpower, material resources and financial resources. The invention provides a novel method for rapidly determining the cannabidiol content of different industrial hemp products.

Description

Technical Field

[0001] The invention relates to a method for quickly determining cannabidiol content based on a mid-infrared spectrum prediction model, and belongs to the technical field of detection. Background Art

[0002] Industrial hemp is cultivated on the basis of the traditionally planted "hemp" or "Hanma". The content of △9-tetrahydrocannabinol in the dried flowers and leaves of the plant is less than 0.3% Wt. It can be used for fiber, seed or flower and leaf extraction and other industrial applications. Cannabinoid chemicals represented by cannabidiol have been found to have a variety of pharmacological activities such as antidepressant, antipsychotic, anti-inflammatory, and nervous system protection. In recent years, there have been many extractions, processing and application developments of medicinal ingredients in industrial hemp flowers and leaves. Therefore, domestic and foreign germplasm research and development institutions have studied and produced a variety of industrial hemp products with high cannabidiol and low △9-tetrahydrocannabinol content. Among them, full-spectrum oil is a typical industrial hemp extract product. It contains a variety of active ingredients. These ingredients work together to produce the unique pharmacological activity of full-spectrum oil. It is favored in the pharmaceutical and health care products industries due to its wide range of health benefits, so many counterfeit and inferior full-spectrum oil products have been spawned on the market. The purpose of the patent of this invention is to study the infrared spectral characteristics of industrial hemp and its derivatives, and to establish a fast, accurate and simple method for identifying industrial hemp products and varieties.

[0003] As a rapid detection technology, mid-infrared spectroscopy analysis technology has the advantages of short analysis time, simple operation, and low analysis cost. In recent years, it has been more and more widely used in food and drug quality analysis, authenticity identification, and origin identification. The present invention studies the infrared spectral characteristics of industrial hemp full-spectrum oil extraction, flower and leaf wine, and industrial hemp flower and leaf, and uses infrared spectroscopy technology and data processing software to predict the cannabidiol content and perform principal component analysis. The infrared rapid detection technology of this study can quickly distinguish different samples and quickly quantify the CBD content, which has good practical application value.

[0004] At present, the commonly used methods for detecting cannabidiol content are usually liquid chromatography (HPLC) or liquid chromatography-mass spectrometry (LC-MS) analysis. These methods have the problems of high cost (high consumption of mobile phase, environmental pollution, and high maintenance cost), long analysis time, and small detection volume. However, the method provided by the present invention can just solve these problems. The present invention uses a portable infrared spectrometer for determination, the equipment maintenance is simple, the detection method is fast and simple, and the pollution caused by the use of mobile phase to the environment during liquid chromatography or mass spectrometry detection is effectively avoided. Summary of the invention

[0005] In order to solve the problems existing in the detection process of cannabidiol in the prior art, the present invention provides another method for quickly determining the content of cannabidiol based on a mid-infrared spectroscopy prediction model, and the steps of the method include:

[0006] (1) Collect the infrared spectrum curve of the sample.

[0007] (2) Preprocessing the infrared spectrum curve obtained in step (1) to obtain mid-infrared spectrum data without noise interference.

[0008] (3) Determine the cannabidiol content in the sample.

[0009] (4) Regression modeling is performed on the cannabidiol content data obtained in step (3) and the mid-infrared data obtained in step (2) to obtain a mid-infrared prediction model for cannabidiol content.

[0010] (5) Detect the infrared spectrum curve of the sample to be measured, substitute it into the mid-infrared prediction model of cannabidiol content obtained in step (4), obtain the cannabidiol content of the sample to be tested, and complete all tests.

[0011] Preferably, in step (1), when the sample is a liquid, 0.1 g is dissolved in 2 mL of 75% ethanol solvent, ultrasonically treated for 15 min, 50 uL is placed on a 2 mm thick potassium bromide tablet, and another tablet is pressed on it as a background, and then placed in the detection window to be measured to obtain the infrared spectrum curve of the sample.

[0012] Preferably, in step (1), when the sample is solid, the sample is washed, dried, crushed, and sieved, and 0.1 g of the sieved powder sample is weighed and evenly spread on a test bench to obtain an infrared spectrum curve of the sample.

[0013] Preferably, the acquisition conditions of the mid-infrared spectrum in step (1) are: scanning range 700~3800cm -1 Each sample was collected 3 times. The mid-infrared spectrometer was Agilent ATR-Cary 630FTIR portable spectrometer, and the data acquisition software was Agilent MicroLab PC Software.

[0014] Preferably, the data processing software in step (2) is Agilent MicroLab Expert Software, version 1.1.0.1, and after preprocessing, the abnormal sample data is removed to obtain the mid-infrared spectrum curves of different industrial hemp samples excluding noise interference. The infrared spectrum preprocessing method includes multiplicative scatter correction (MSC), deviation correction, smoothing, variable standardization, first-order derivative processing, and second-order derivative processing.

[0015] Preferably, the cannabidiol content in step (3) is measured by ultra-high performance liquid chromatography Agilent Technologies 1290 Infinity.

[0016] Preferably, the modeling method described in step (4) refers to modeling using data measured by two high-precision instruments, a mid-infrared spectrometer and an ultra-high performance liquid chromatography. First, the mid-infrared spectral data of step (2) is used as the predicted value of the partial least squares regression analysis prediction model, and then the cannabidiol content of the same batch of samples is detected by an ultra-high performance liquid chromatograph. The data measured by the liquid chromatography is used as the true value of the partial least squares regression analysis prediction model. After that, the two sets of data are regressed and modeled. The modeling software is Agilent MicroLab Expert Software, version 1.1.0.1, and a mid-infrared spectral prediction model for the cannabidiol content in industrial hemp can be obtained.

[0017] Beneficial effects of the present invention

[0018] (1) This method is simple to operate. After the model is established, only simple training is required for the operators. The spectral data can be directly input into the model to obtain the results. No professional personnel are required. The detection speed is fast and the detection cost is low, which is very suitable for promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is the original mid-infrared spectrum of industrial hemp.

[0020] Figure 2 This is a spectrum preprocessing diagram, where (1) is multiplicative scatter correction (MSC); (2) is deviation correction (Deviation correction); (3) is smoothing processing (SG); (4) is variable standardization (SNV); (5) is first-order derivative processing; 6. Second-order derivative processing.

[0021] Figure 3The principal component analysis diagrams of industrial hemp are shown in Figure 1, where (1) is the principal component analysis diagram of two industrial hemp products and flowers and leaves (full-spectrum oil extracted from flowers and leaves, wine soaked from flowers and leaves, and industrial hemp flowers and leaves); (2) is the principal component analysis diagram of three full-spectrum oil products; and (3) is the principal component analysis diagram of three flowers and leaves.

[0022] Figure 4 This is a cluster analysis diagram of industrial hemp in different regions.

[0023] Figure 5 A mid-infrared spectroscopy prediction model for cannabidiol content in industrial hemp. DETAILED DESCRIPTION

[0024] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and through specific implementation methods. However, the following embodiments are only simple examples of the present invention and do not represent or limit the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

[0025] Instruments used in the examples: Agilent ATR-Cary 630 FTIR portable spectrometer; ultra-high performance liquid chromatography Agilent Technologies 1290 Infinity; data processing software: Agilent MicroLab Expert Software, version 1.1.0.1; MetaboAnalyst 5.0.

[0026] In the following implementation, the model is established based on mid-infrared scanning data and liquid chromatography measurement data. The samples used in the experiment include industrial hemp full-spectrum oil, flower and leaf wine, and industrial hemp flowers and leaves, all of which are provided by Yunnan Industrial Hemp Planting and Processing Enterprises and Yunnan Industrial Hemp Association.

[0027] Example 1

[0028] A mid-infrared spectrum prediction model is established, and the specific steps are as follows:

[0029] (1) Collect the mid-infrared spectrum curve of the test sample:

[0030] A. Dissolve 0.1g of industrial hemp full-spectrum oil and industrial hemp flower and leaf wine in 2mL of 75% ethanol solvent, ultrasonically treat for 15min, and set aside.

[0031] B. Use a pipette to drop 50uL of the industrial hemp full-spectrum oil extracted from step (A) and the industrial hemp flower and leaf wine samples on the potassium bromide tablet, then press another background tablet on it, and then place it in the detection window for testing. Repeat three times for each sample.

[0032] C. The pretreatment method of industrial hemp flower and leaf samples is cleaning, drying, crushing and sieving. Weigh 0.1 grams of flower and leaf powder sample obtained by sieving, evenly spread on the test bench for testing, the test time is 30 seconds, each sample is repeated three times, and the detection window is wiped with 95% alcohol after each test.

[0033] The obtained mid-infrared original spectrum is as follows Figure 1 shown.

[0034] (2) Preprocessing the original mid-infrared spectrum obtained in step (1) to obtain mid-infrared spectrum data without noise interference, specifically, performing scatter correction and preprocessing on the three repeated original mid-infrared spectra of each sample obtained in step (1), including multiplicative scatter correction (MSC), deviation correction (Deviation correction), smoothing (SG), variable standardization (SNV), first-order derivative processing, and second-order derivative processing. The processing results are as follows: Figure 2 As shown, Figure 2 (1) is the result of multiplicative scatter correction (MSC); Figure 2 (2) Deviation correction processing result; Figure 2 (3) is the result of smoothing; Figure 2 (4) Standardize the results of the variables; Figure 2 (5) First-order derivative processing; Figure 2 (6) Second-order derivative processing.

[0035] (3) Determination of cannabidiol content in the sample: Ultra-high performance liquid chromatography was used to determine the cannabidiol content in the sample. The specific steps are as follows:

[0036] A. Preparation of Standard Curve

[0037] Weigh 1 ml of 1000 ug / ml cannabidiol standard solution into a 10 ml volumetric flask and dilute to the mark to prepare a 200 ug / ml standard solution. Place the solution in a -20°C refrigerator away from light for later use. For subsequent standard working solution analysis, set 5 different gradient injections (0.2 μL, 0.5 μL, 1 μL, 2 μL, 5 μL).

[0038] B. Sample Analysis

[0039] Weigh 0.2g of industrial hemp flower and leaf powder sample, extract with 95% methanol-water solution, fix to 10mL, ultrasonicate for 30min, centrifuge at 12000r / min for 5min, take out the supernatant, filter through 0.22μm organic filter membrane, and transfer the filtrate to the chromatographic injection bottle for analysis. HPLC conditions: Waters Acquity BEH-C18 column (2.1mm×50mm, 1.7μm); mobile phase A is 0.1% phosphoric acid aqueous solution, B is methanol solution, isocratic elution: 75% B, flow rate 0.30ml / min, column temperature: 30℃; injection volume is set to 1μL. And check whether the concentration of the sample treatment solution is within the range of the standard curve. If not, the sample treatment solution needs to be diluted or concentrated.

[0040] C. Sample content calculation formula

[0041] The peak area of ​​the standard working solution is taken as y, and regression analysis is performed with the corresponding standard concentration to obtain the standard curve: y = 6.9806x-3.6129, R 2 =0.9997. Then, according to the ratio of the peak area to the concentration, the actual content of cannabidiol can be calculated. The actual content of cannabidiol in all samples is shown in Table 1.

[0042] Table 1 Measured results of cannabidiol content in samples

[0043]

[0044]

[0045] (4) The mid-infrared spectral data obtained in step (2) is used as the predicted value of the partial least squares regression analysis prediction model, and then the cannabidiol content of the same batch of samples is detected by ultra-high performance liquid chromatography. The data measured by liquid chromatography is used as the true value of the partial least squares regression analysis prediction model. The two sets of data are then regressed and modeled. The modeling software is Agilent MicroLab Expert Software, version 1.1.0.1, and the mid-infrared spectral prediction model of the cannabidiol content in industrial hemp can be obtained.

[0046] Compare the models established under different conditions, and select the model with small cross-validation bias (SEC), small root mean square error (RMSE) of the calibration model, and large cross-validation correlation coefficient R. Compare the models established under different conditions, and select the model with small cross-validation bias (SEC), small root mean square error (RMSE) of the calibration model, and large cross-validation correlation coefficient R.

[0047] The model with small root mean square error (RMSE) and large cross-validation correlation coefficient R. The comparison of different treatment conditions is shown in Table 2. Finally, MSC+SG+SNV+2nd order derivative was selected, and the linear equation obtained was y=0.97647612*x+0.01719390.

[0048] Table 2 Different treatment methods

[0049] Spectral processing SEC RMSE <![CDATA[R 2 ]]> MSC+SG+1st order derivative 0.3314 0.3292 0.82 MSC+SG+2nd order derivative 0.2918 0.2986 0.90 Bias correction + SG + 2nd order derivative 0.3142 0.3126 0.89 MSC+SG+SNV+2nd order derivative 0.2959 0.2914 0.98

[0050] Example 2

[0051] Model accuracy prediction

[0052] Import the infrared spectrum data of the industrial hemp sample to be tested into the mid-infrared prediction model, such as Figure 5 As shown in the figure, the results of industrial hemp from different origins were screened for verification, and these data were divided into a calibration set and a validation set. The prediction ability of the model was verified by internal cross-validation, and the accuracy and relative standard deviation of each calibration set and validation set, as well as the cross-validation correlation coefficient (R 2 ), residuals, and finally the average value is calculated. The results are shown in Table 3. The higher the accuracy and the larger the correlation coefficient, the better the prediction effect of the model. The accuracy of all calibration sets and validation sets is above 80%, and the relative standard deviation is less than 28%, indicating that the model can be used successfully.

[0053] The relative standard deviation (RSD) is calculated as:

[0054] The accuracy (A) of the validation set and calibration set is calculated as follows:

[0055] Import the predicted content of each calibration set and validation set calculated by the prediction model into the EXCEL table, use the STDEV function to calculate the standard deviation, and substitute it into the RSD calculation formula to obtain the RSD value.

[0056] In the accuracy calculation formula, |V| is the absolute value of the residual of each sample point. |P| is the predicted value of each sample point. The residual is equal to the true value measured by liquid chromatography minus the predicted value measured by the prediction model. After calculating the accuracy of each validation set and calibration set, the average value is calculated, and the final result is used as the basis for judging the model accuracy.

[0057] Table 3 Accuracy and RSD value of PLS ​​model (industrial hemp flower and leaf samples)

[0058] area Calibration set Validation set Residual Predicted value Calibration set accuracy (%) Validation set accuracy (%) RSD(%) Qujing 12 2 0.0172 0.2362 84.38 85.97 7.74～10.03 Xundian-A 4 1 0.0201 0.1569 86.91 99.73 13.86～21.89 Xundian-B 4 1 0.0195 0.1649 90.22 89.09 18.22～27.36 Yuxi 4 1 0.0785 0.0913 79.35 83.87 10.29～12.84 Kunming 10 2 0.0145 0.0916 83.38 85.97 18.09～27.79 Lufeng-A 8 2 0.0083 0.8684 99.88 89.26 4.71～5.44 Lufeng-B 8 2 0.0432 0.5036 91.59 89.43 3.17～3.60 Lufeng-C 8 2 0.0899 0.7520 88.40 95.04 3.64～1.01 Lufeng-D 4 1 0.0224 0.3818 94.19 82.20 4.20～4.77 Chuxiong 4 1 0.0183 0.1222 86.10 92.66 15.76～23.47

[0059] To further verify the accuracy of infrared spectroscopy detection, we directly imported the original spectral data obtained in step (1) of Example 1 into the principal component analysis model (PCA) to analyze whether it can effectively distinguish different varieties. The results show that the model has a good effect on variety differentiation and can effectively distinguish different varieties of industrial hemp, but its disadvantage is that it cannot quantify the cannabidiol content. The results are as follows Figure 3 As shown, Figure 3 (1) Distinguished between two types of industrial hemp products and flowers and leaves (full spectrum oil, industrial hemp flowers and leaves for wine brewing, and industrial hemp flowers and leaves). Figure 3 (2) The differences between the same full-spectrum oils but with different contents were distinguished. Figure 3 (3) The differences between different varieties of industrial hemp (Yunma No. 7, Yunma No. 8, Yunma No. 10) were distinguished. This model can prove the accuracy of infrared spectrum detection and provide a basis for the authenticity identification and variety differentiation of industrial hemp products.

[0060] According to the original spectral data obtained in step (1) of Example 1, the origin of the industrial hemp can also be verified by establishing a cluster analysis model, which can clearly distinguish the differences between the six origins of industrial hemp (Chuxiong, Qujing, Xundian, Yuxi, Baoshan, Kunming, and Tangdian). The distinction effect is obvious and the result is accurate. At the same time, it can provide a basis for origin tracing. The model is as follows Figure 4 shown.

[0061] Example 3 Comparison of actual value and predicted value

[0062] The cannabidiol content of the sample detected by ultra-high performance liquid chromatography was used as the true value, and the mid-infrared spectroscopy prediction model was used as the predicted value. The standard deviation of the CBD content prediction of the full-spectrum oil sample was 4.94, and the standard error was 1.71. The standard deviation of the CBD content prediction of the flower wine was 0.02, and the standard error was 0.01. The standard deviation of the CBD content prediction of the industrial hemp flower and leaf was 0.29, and the standard error was 0.016. The cross-validation coefficients of all samples were greater than 0.9, indicating that the model can be used successfully. The comparison of the true value and the predicted value of the cannabidiol content in the industrial hemp samples is shown in Table 4.

[0063] Table 4 shows the detailed information of the 22 samples.

[0064] No. Sample type Origin variety CBD actual content (%) quantity CBD predicted value content (%) 1 Flower leaf extract full spectrum oil Qujing City, Yunnan 68～71 5 70～73 2 Flower leaf extract full spectrum oil Qujing City, Yunnan 73～76 5 70～81 3 Flower leaf extract full spectrum oil Qujing City, Yunnan 58～61 5 59～63 4 Flower leaves soaked in wine Kunming, Yunnan Yunma No.7 0.0852 1 0.075 5 Flower leaves soaked in wine Kunming, Yunnan Yunma No.7 0.0649 1 0.070 6 Flower leaves soaked in wine Kunming, Yunnan Yunma No.7 0.0435 1 0.043 7 Industrial hemp flowers and leaves Kunming, Yunnan Yunma No.7 0.0610～0.0723 3 0.058～0.072 8 Industrial hemp flowers and leaves Kunming, Yunnan Yunma No.7 0.2500～0.2655 3 0.240～0.258 9 Industrial hemp flowers and leaves Kunming, Yunnan Yunma No.8 0.1930～0.2019 3 0.190～0.250 10 Industrial hemp flowers and leaves Kunming, Yunnan Yunma No.8 0.1700～0.1811 3 0.165～0.180 12 Industrial hemp flowers and leaves Kunming, Yunnan Yunma No.10 0.2810～0.3120 3 0.270～0.30 13 Industrial hemp flowers and leaves Kunming, Yunnan Yunma No.10 1.1110～1.1250 3 1.050～1.115 14 Industrial hemp flowers and leaves Qujing City, Yunnan Yunma No.7 0.2414～0.2523 3 0.220～0.250 15 Industrial hemp flowers and leaves Qujing City, Yunnan Yunma No.7 0.4600～0.5012 3 0.45～0.48 16 Industrial hemp flowers and leaves Xundian County, Yunnan Yunma No.7 0.0820～0.0980 3 0.080～0.101 17 Industrial hemp flowers and leaves Xundian County, Yunnan Yunma No.7 0.1275～0.1354 3 0.130～0.144 18 Industrial hemp flowers and leaves Yuxi City, Yunnan Yunma No.7 0.1400～0.1543 3 0.140～0.157 19 Industrial hemp flowers and leaves Yuxi City, Yunnan Yunma No.7 0.0491～0.0560 3 0.047～0.058 20 Industrial hemp flowers and leaves Lufeng City, Yunnan Foreign introduction 0.8770～0.912 3 0.801～0.950 21 Industrial hemp flowers and leaves Lufeng City, Yunnan Foreign introduction 1.1360～1.1400 3 1.135～1.145 22 Industrial hemp flowers and leaves Lufeng City, Yunnan Foreign introduction 1.0320～1.0421 3 1.025～1.055

[0065] Model Application

[0066] The samples to be tested are tested, and the correlation diagram between the experimental value and the predicted value is shown in the figure below. Figure 5 As shown, the R of the predicted value and the experimental value of the whole sample both reached above 0.9, with high prediction accuracy, indicating that the model can be used for the rapid determination and preliminary screening of the cannabidiol content in cannabis flowers and leaves.

Claims

1. A method for rapidly determining the content of cannabidiol based on a mid-infrared spectroscopy prediction model, characterized in that: The method steps include: (1) Collect infrared spectrum curve of sample; (2) preprocessing the infrared spectrum curve obtained in step (1) to obtain mid-infrared spectrum data without noise interference; (3) determining the cannabidiol content in the sample; (4) performing regression modeling on the cannabidiol content data obtained in step (3) and the mid-infrared data obtained in step (2) to obtain a mid-infrared prediction model for cannabidiol content; (5) Detect the infrared spectrum curve of the sample to be measured, substitute it into the mid-infrared prediction model of cannabidiol content obtained in step (4), obtain the cannabidiol content of the sample to be tested, and complete all tests.

2. The method for rapidly determining the content of cannabidiol based on the mid-infrared spectroscopy prediction model according to claim 1, characterized in that: The preprocessing methods in step (2) include multivariate scattering correction, bias correction, smoothing, variable standardization, first-order derivative processing, and second-order derivative processing.

3. The method for rapidly determining the content of cannabidiol based on the mid-infrared spectroscopy prediction model according to claim 1, characterized in that: The modeling method described in step (4) refers to modeling using data measured by two high-precision instruments, a mid-infrared spectrometer and an ultra-high performance liquid chromatography. First, the mid-infrared spectral data of step (2) is used as the predicted value of the partial least squares regression analysis prediction model, and then the cannabidiol content of the same batch of samples is detected by an ultra-high performance liquid chromatograph. The data measured by the liquid chromatography is used as the true value of the partial least squares regression analysis prediction model. After that, the two sets of data are regressed and modeled to obtain a mid-infrared spectral prediction model for the cannabidiol content in industrial hemp.

Images