Method for quantifying mycotoxins

Abstract

Provided is a method for quantifying mycotoxins belonging to type-B trichothecenes produced by mold of the genus Fusarium in wheat grains of common wheat, the method comprising: acquiring, for each of a plurality of types of wheat grains that are samples for creating a calibration curve, an optical spectrum through a transmission method using light of wavelengths including the near-infrared band of wavelengths of less than 1,000 nm; quantifying, by a known quantification method, mycotoxins for each of the plurality of types of wheat grains that are samples for creating the calibration curve; analyzing, through PLS regression analysis, the acquired information on the optical spectrum including the near-infrared band of wavelengths of less than 1,000 nm and the acquired quantification information on the mycotoxins, and establishing a regression equation indicating the relationship between the optical spectrum information and the quantification of the mycotoxins; acquiring, for a wheat grain sample to be measured and through a transmission method, the optical spectrum using light of wavelengths including the near-infrared band of wavelengths of less than 1,000 nm; and calculating the content of mycotoxins on the basis of the established regression equation.

Description

Method for quantifying mycotoxins

[0001] This invention relates to a method for quantifying deoxynivalenol and nivalenol, mycotoxins contained in wheat grains, by spectroscopy.

[0002] Some fungi belonging to the genus Fusarium are pathogens of Fusarium head blight, which affects wheat and other crops, and produce mycotoxins such as deoxynivalenol (hereinafter also known as DON) and nivalenol (hereinafter also known as NIV) that are harmful to the human body. Therefore, the Food Sanitation Act has established a standard for DON content, stipulating that it must not exceed 1.0 mg / kg.

[0003] Non-patent document 1 specifies that the deoxynivalenol test method for wheat uses a high-performance liquid chromatograph with an ultraviolet spectrophotometer (HPLC-UV), a liquid chromatograph-mass spectrometer (LC-MS), or a liquid chromatograph-tandem mass spectrometer (LC-MS / MS) as the quantitative method, and the ELISA (Enzyme-Linked Immunosorbent Assay) method as the screening method.

[0004] Non-patent document 2 describes measuring deoxynivalenol in durum wheat using near-infrared spectroscopy. Durum wheat is crushed into a powder and exposed to near-infrared light at 10,000 to 4,000 cm⁻¹. －１ Absorption was measured in the specified frequency band and analyzed using linear discriminant analysis (LDA). The same paper also attempts partial least squares regression analysis (PLS), but concludes that it is not suitable for the quantitative analysis of deoxynivalenol in durum wheat.

[0005] "On the Test Method for Deoxynivalenol in Wheat," *Rapid Analysis of Deoxynivalenol in Durum Wheat by FT-NIR Spectroscopy*, Annalisa De Girolamo Etal, Toxins 2014, 6, 3129-3143 (Report No. 0930-2)

[0006] The deoxynivalenol test method for wheat described in Non-Patent Literature 1 requires the wheat grain to be crushed into powder as a pretreatment, and the extraction and quantification of DON using reagents is time-consuming. The analytical method in Non-Patent Literature 2 is limited to durum wheat and is not applicable to common wheat. Furthermore, this analytical method also requires the sample to be crushed into powder, making it non-destructive and time-consuming. Although it is stated that analysis using linear discriminant analysis (LDA) was successful, it only involved dividing the sample into about three groups, making it closer to qualitative analysis than quantitative analysis, and moreover, there are many misjudgments depending on the group. From the perspective of food safety, there is an urgent need for a simple and rapid method for measuring mycotoxin content in wheat production sites.

[0007] The purpose of this invention is to provide a simple and highly accurate quantitative method for measuring deoxynivalenol and nivalenol, mycotoxins of red mold contained in common wheat grains.

[0008] To solve the above problems, the present invention provides a method for quantifying mycotoxins belonging to type B trichothecenes produced by Fusarium fungi in wheat grains of common wheat. This method involves obtaining the optical spectrum of multiple types of wheat grains, which are samples for calibration curve creation, using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm, by transmission analysis. The mycotoxin is quantified for each type of wheat grain in the calibration curve creation samples using a known quantitative method. The acquired optical spectrum information including the near-infrared band with wavelengths less than 1000 nm and the mycotoxin quantification information are analyzed using PLS regression analysis to set a regression equation showing the relationship between the optical spectrum information and the mycotoxin quantification. The optical spectrum of the wheat grains, which are the sample to be measured, is obtained using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm, by transmission analysis, and the mycotoxin content is calculated based on the set regression equation. As for mycotoxins, deoxynivalenol (DON) or nivalenol (NIV) may be quantified, or the total amount of both may be quantified.

[0009] According to the method for quantifying mycotoxins belonging to type B trichothecenes produced by Fusarium fungi contained in wheat grains, the mycotoxins contained in wheat grains can be quantified simply and quickly.

[0010] This is a flowchart outlining the process for setting the regression equation in the quantitative analysis of mycotoxins. This graph shows an example of the untreated spectrum (570-1100 nm) of a wheat sample. This graph shows an example of the spectrum of a wheat sample after applying first derivative, Detrend, and SNV processing, selecting the range from 860 nm to 1020 nm. This is a scatter plot comparing the calculated DON content based on a calibration curve obtained by performing PLS regression on all samples with the measured values ​​obtained by LC-MS. This graph shows the regression coefficients of the calibration curve obtained by performing PLS regression on all samples for the DON content obtained by LC-MS. This is a scatter plot comparing the calculated DON content after cross-validation with the measured values ​​obtained by LC-MS. This is a scatter plot comparing the calculated DON content based on a calibration curve obtained by performing PLS regression on a calibration curve creation set with the measured values ​​obtained by LC-MS. This graph shows the regression coefficients of the calibration curve obtained by performing PLS regression on a calibration curve creation set for DON content obtained by LC-MS. This is a scatter plot comparing the calculated DON content of a calibration curve validation set based on the calibration curve with the measured values ​​obtained by LC-MS. The spectra of wheat samples were subjected to first derivative, Detrend, and SNV processing, and the spectra from 860 nm to 1015 nm were selected. This is a scatter plot comparing the calculated NIV content based on a calibration curve obtained by performing PLS regression on all samples with the measured values ​​obtained by LC-MS. This graph shows the regression coefficients of the calibration curve obtained by performing PLS regression on all samples for NIV content obtained by LC-MS. This is a scatter plot comparing the calculated NIV content when cross-validation is performed with the measured values ​​obtained by LC-MS. This is a scatter plot comparing the calculated NIV content based on a calibration curve obtained by performing PLS regression on a calibration curve creation set with the measured values ​​obtained by LC-MS. This graph shows the regression coefficients of the calibration curve obtained by performing PLS regression on a calibration curve creation set for NIV content obtained by LC-MS. It also includes a scatter plot comparing the calculated NIV content of the calibration curve validation set based on the calibration curve with the measured values ​​obtained by LC-MS.The spectra of wheat samples were subjected to first derivative analysis, Detrend analysis, and SNV processing, and the spectra from 860 nm to 1010 nm were selected. This is a scatter plot comparing the calculated total DON and NIV content based on a calibration curve obtained by performing PLS regression on all samples with the measured values ​​obtained by LC-MS. This graph shows the regression coefficients of the calibration curve obtained by performing PLS regression on all samples for the total DON and NIV content obtained by LC-MS. This is a scatter plot comparing the calculated total DON and NIV content after cross-validation with the measured values ​​obtained by LC-MS. This is a scatter plot comparing the calculated total DON and NIV content based on a calibration curve obtained by performing PLS regression on the calibration curve creation set with the measured values ​​obtained by LC-MS. This graph shows the regression coefficients of the calibration curve obtained by performing PLS regression on the calibration curve creation set for the total DON and NIV content obtained by LC-MS. This is a scatter plot comparing the calculated total DON and NIV content of a calibration curve validation set based on a calibration curve with the measured values ​​obtained by LC-MS. The spectra of the wheat samples were subjected to first derivative, detrend, and SNV processing, and the spectra from 860 nm to 985 nm were selected. This is a scatter plot comparing the calculated DON content based on a calibration curve obtained by performing PLS regression on all samples with the measured values ​​obtained by the EZ-M system. This is a graph showing the regression coefficients of the calibration curve obtained by performing PLS regression on all samples for the DON content obtained by the EZ-M system. This is a scatter plot comparing the calculated DON content when cross-validation is performed with the measured values ​​obtained by the EZ-M system. This is a scatter plot comparing the calculated DON content based on a calibration curve obtained by performing PLS regression on a calibration curve creation set with the measured values ​​obtained by the EZ-M system. This graph shows the regression coefficients of the calibration curve obtained by performing PLS regression on a calibration curve creation set for the DON content obtained by the EZ-M system. It is a scatter plot comparing the calculated DON content values ​​from the calibration curve verification set based on the calibration curve with the measured values ​​obtained by the EZ-M system.

[0011] The embodiments for carrying out the present invention will now be described. In the quantitative method of the present invention, as will be described later, a regression equation is set that shows the relationship between optical spectrum information and the quantitative determination of mycotoxins, and calculations are performed based on that regression equation. Therefore, it is preferable to have an analytical device equipped with a computing device (CPU) and memory, and capable of installing and executing a program that performs calculations based on the determined regression equation. Such spectroscopic analyzers are commercially available. Here, an example will be given using the Inframatic 9500 (PerkinElmer) as the spectroscopic analyzer.

[0012] Figure 1 is a flowchart outlining the process of setting up a regression equation in the quantitative determination of mycotoxins. First, a sufficient number of samples, i.e., test materials, are prepared to create a calibration curve, i.e., to set up a regression equation. These samples are wheat grains.

[0013] Approximately 19 types of samples were prepared here. These samples will be used for creating calibration curves. Each type of sample needs to be in sufficient quantity to allow for optical spectrum acquisition and quantitative analysis other than spectroscopic analysis, but it is preferable that the wheat grains within each type of sample be as uniform in quality as possible. Furthermore, it is preferable that the mycotoxin content of the samples being measured is distributed across different types of samples to broadly cover the range to be measured.

[0014] In this example, in addition to the sample used to create the calibration curve, six samples were prepared as calibration curve verification samples.

[0015] The samples included different varieties of common wheat from both Japan and overseas, and 25 different types were prepared by varying the processing and preparation methods.

[0016] For each type of sample used to create a calibration curve, the optical spectrum is acquired, and quantitative analysis is performed using a known analytical method other than the near-infrared spectroscopy of the present invention (measurement of conventional analytical values). The optical spectrum information and the conventional analytical values ​​are then associated and stored. For each type of sample, the sample may be divided into two parts; the optical spectrum may be acquired using one part, and the conventional quantitative analysis may be performed using the other part. Furthermore, since the acquisition of the optical spectrum can be performed non-destructively, the conventional quantitative analysis may be performed using the same sample after the optical spectrum has been acquired.

[0017] Similarly, optical spectrum acquisition and conventional analytical value measurement are performed on the calibration curve verification samples. Alternatively, the calibration curve creation samples and calibration curve verification samples may be separated beforehand, and then optical spectrum acquisition and quantitative analysis using methods other than spectroscopic analysis may be performed. Alternatively, at this stage, both types of samples (25 in total in this example) may be treated the same way without distinction, and then the results may be used to separate the data for calibration curve creation and the data for calibration curve verification.

[0018] This section describes quantitative analysis that does not rely on near-infrared spectroscopy. The quantitative analysis described here uses established analytical methods. Any method capable of accurately measuring mycotoxin components such as deoxynivalenol or nivalenol is acceptable, even if it is time-consuming or costly.

[0019] Liquid chromatography-mass spectrometry (LC-MS) is a test method for deoxynivalenol or nivalenol. This method is described in the Food Safety Bureau Notification No. 0930-2, "Test Method for Deoxynivalenol in Wheat."

[0020] Next, an example of a spectroscopic analyzer used to acquire the optical spectrum will be described. The analyzer used is capable of measuring the absorbance of a wheat grain sample by transmission using light with wavelengths that include a particularly suitable band within the 570 nm to 1095 nm wavelength range. This band includes relatively short wavelengths of less than 1000 nm. Near-infrared light is light (electromagnetic waves) in the wavelength range between the visible and infrared regions, and is generally in the 800 to 2500 nm range. In this invention, at least the shortest wavelength range within the near-infrared region is used, but light in the visible region with even shorter wavelengths may also be used. Therefore, in the embodiments described later, an analyzer capable of acquiring the optical spectrum in these wavelength ranges is used. This is an analyzer capable of measuring the absorbance of a wheat grain sample by transmission.

[0021] The acquisition of the optical spectrum will be explained using the PerkinElmer Inframatic 9500 near-infrared spectrometer as an example. If you directly feed a brown rice sample into the sample input port of this spectrometer (for samples of approximately 200g or more), or if you pack the brown rice into a dedicated cuvette with an 18mm optical path length and insert it into the sample input port (for samples of only a few tens of grams), you can obtain near-infrared spectral data in about 30 seconds. No pretreatment such as milling or polishing of the wheat grains is required. The absorbance spectrum is collected when the sample is transmitted through it. After measurement, the wheat grains in the sample can be recovered as is. By performing this procedure for all types of samples, the optical spectrum for each can be obtained.

[0022] The optical spectrum is obtained by measuring absorbance using the transmission method. Therefore, information about the components of the entire wheat grain, including the core, not just the surface, can be obtained.

[0023] Next, we will explain how to determine the calibration curve. First, we select the data for creating the calibration curve. In this example, spectrum acquisition and quantitative analysis were performed on 25 samples. To ensure that the values ​​obtained from the quantitative analysis are widely distributed, we select the necessary number of samples (19 in this example) for calibration curve creation, and the remaining samples (6 in this example) for calibration curve verification. Figure 2 is a graph showing an example of the optical spectrum of wheat grain samples, showing all the absorbance spectra of the samples used for calibration curve creation and calibration curve verification. Here, the absorbance spectrum of the samples used for calibration curve creation is shown in lighter colors, and the absorbance spectrum of the samples used for calibration curve verification is shown in darker colors.

[0024] Next, regression analysis is performed. Here, raw spectrum information showing the relationship between wavelength and absorbance may be used, but preprocessing is preferable, and in particular, first derivative, detrend, and SNV processing is preferred. Figure 3 shows an example of the optical spectrum of wheat grains selected in the 860 nm to 1020 nm band after first derivative, detrend, and SNV processing. For the preprocessed optical spectrum of a certain sample, if the evaluation values ​​at wavelengths λ1, λ2, λ3... are x1, x2, x3... and the regression coefficients for them are a1, a2, a3..., then the DON content y for that sample is related by a linear combination as follows: y = a1・x1 + a2・x2 + a3・x3 + ... Partial least squares regression analysis is performed on the information of the sample selected for calibration curve creation to determine these regression coefficients a1, a2, a3.... This sets up the regression equation. Figure 5 is a graph showing the regression coefficients of the DON calibration curve in wheat. This calibration curve uses information from 860nm to 1020nm, which is a portion of the spectral information in the wavelength range of 570nm to 1100nm. In this example, PLS regression analysis is performed on the information from 19 samples selected for calibration curve creation. Therefore, if the sample data is split into calibration curve creation and calibration curve validation, the information from the 6 samples used for calibration curve validation is not involved in setting the regression equation. These calibration curve validation samples are used as the samples to be measured.

[0025] This section describes the case where all data is used without splitting the calibration curve creation set and the calibration curve validation set. Using the near-infrared spectrum of wheat collected with IM9500 and the DON content values ​​actually obtained by LC-MS, a calibration curve was created to determine the DON content from the spectrum, and evaluation was performed by cross-validation. After performing first derivative, detrend, and SNV processing on the spectrum obtained with IM9500, a calibration curve was created using PLS regression in the wavelength range of 860 nm to 1020 nm (Figure 3).

[0026] Figure 4 is a scatter plot comparing the calculated DON content (IM9500 predicted value) based on a calibration curve determined by PLS regression for all 25 samples without separating the calibration curve creation and calibration curve verification, with the measured value obtained by LC-MS. The accuracy during calibration curve creation is R 2 The coefficient of determination was 0.9971, and the standard error (SEC) was 0.030. Figure 5 is a graph showing the calibration curve at this time. It shows the regression coefficients (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0027] Figure 6 is a scatter plot comparing the calculated DON content (IM9500 predicted value) after cross-validation (tolerance verification) with the measured value obtained by LC-MS. The accuracy in cross-validation is R 2 The result was 0.9417, and the SECV (standard error) was 0.135.

[0028] Next, we will explain the process of dividing the samples into two groups: a calibration curve creation set (19 points) and a calibration curve verification set (6 points). First, a calibration curve is created using the calibration curve creation set, and then the results are evaluated using the calibration curve verification set.

[0029] Similar to the example shown in Figure 3, after performing first derivative, detrend, and SNV processing on the near-infrared spectrum, a calibration curve was created using PLS regression in the wavelength range of 860 nm to 1020 nm. Figure 7 is a scatter plot showing the correlation between IM9500 predicted values ​​(vertical axis) and measured values ​​(horizontal axis) when calibration curves were created for 19 samples used for calibration curve creation. The accuracy at this time was R2 The coefficient of determination was 0.9993 and the standard error (SEC) was 0.021. Figure 8 is a graph showing the calibration curve at this time. It shows the regression coefficients (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0030] Next, we will explain the results of evaluating the six calibration curve validation sets using the obtained calibration curve. The DON content was calculated by applying the regression equation set up according to the procedure described above to the calibration curve validation samples, and the validity of the regression equation was evaluated. Figure 9 is a scatter plot comparing the calculated DON content (IM9500 predicted value) based on the calibration curve with the measured value obtained from LC-MS. 2 The result was 0.9624, and the SEP (Standard Error) was 0.150.

[0031] Next, we will explain an example of quantifying nivalenol using the LC-MS method.

[0032] This section describes the case where all data is used without splitting the calibration curve creation set and the calibration curve validation set. Using the near-infrared spectra of wheat collected with IM9500 and the NIV content values ​​actually obtained by LC-MS, a calibration curve was created to determine the NIV content from the spectrum, and evaluation was performed by cross-validation. After performing first derivative, detrend, and SNV processing on the spectra obtained with IM9500, a calibration curve was created using PLS regression in the wavelength range of 860 nm to 1015 nm (Figure 10).

[0033] Figure 11 is a scatter plot comparing the calculated NIV content (IM9500 predicted value) based on a calibration curve determined by PLS regression for all 25 samples without separating the calibration curve creation and calibration curve verification, with the measured value obtained by LC-MS. The accuracy during calibration curve creation is R 2 The coefficient of determination was 0.9991 and the standard error (SEC) was 0.033. Figure 12 is a graph showing the calibration curve at this time. It shows the regression coefficients (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0034] Fig. 13 is a scatter plot comparing the calculated NIV content (IM9500 predicted value) obtained when cross-validation was performed with the actually measured values obtained by LC-MS. The accuracy in cross-validation is R 2 of 0.9563 and a SECV (standard error) of 0.231.

[0035] Next, a case where samples are divided into two groups: a calibration curve preparation set (19 points) and a calibration curve validation set (6 points), a calibration curve is first prepared with the calibration curve preparation set, and then evaluation is performed with the calibration curve validation set is described.

[0036] Similar to the example shown in Fig. 10, after performing first-order differentiation, Detrend and SNV processing on the near-infrared spectrum, a calibration curve was prepared by PLS regression using the wavelength range of 860 nm to 1015 nm. Fig. 14 is a scatter diagram showing the correlation between the IM9500 predicted value (vertical axis) and the actually measured value (horizontal axis) when a calibration curve was prepared using 19 samples for calibration curve preparation. The accuracy at this time is R 2 (coefficient of determination) of 0.9998 and a SEC (standard error) of 0.023. Fig. 15 is a graph showing the calibration curve at this time, showing the regression coefficient (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0037] Next, the results of evaluating 6 points of the calibration curve validation set using the obtained calibration curve are described. For the calibration curve validation samples, the regression formula set by the above procedure was applied to calculate the NIV content, and the validity of the regression formula was evaluated. Fig. 16 is a scatter plot comparing the calculated NIV content (IM9500 predicted value) based on the calibration curve with the actually measured values obtained by LC-MS. R 2 of 0.9890 and a SEP (standard error) of 0.304. Although nivalenol was not quantified in Non-Patent Document 2, nivalenol could also be quantified in the present example, and favorable results were obtained.

[0038] Next, quantification of the total content of DON and NIV will be described. First, a case where all data is used without dividing into a calibration curve creation set and a calibration curve verification set is described. Using the near-infrared spectrum of wheat collected with an IM9500 and the total content of DON and NIV actually obtained by LC-MS, a calibration curve for determining the total amount of DON and NIV from the spectrum was created, and evaluation by cross-validation was performed. After performing first-order differentiation, Detrend, and SNV processing on the spectrum obtained by the IM9500, a calibration curve was created by PLS regression using the wavelength range of 860 nm to 1010 nm (Fig. 17).

[0039] Fig. 18 is a scatter plot comparing the calculated values (IM9500 predicted values) of the total amount of DON and NIV, which are obtained by determining the calibration curve through performing PLS regression on all 25 samples without dividing into a calibration curve creation set and a calibration curve verification set, with the actually measured values obtained by LC-MS. The accuracy during calibration curve creation was expressed as R 2 (coefficient of determination) of 0.9984 and SEC (standard error) of 0.065. Fig. 19 is a graph showing the calibration curve obtained at this time, which shows the regression coefficient (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0040] Fig. 20 is a scatter plot comparing the calculated values (IM9500 predicted values) of the total amount of DON and NIV when cross-validation is performed with the actually measured values obtained by LC-MS. The accuracy in cross-validation was R 2 of 0.9480 and SECV (standard error) of 0.376.

[0041] Next, a case where samples are divided into two groups: a calibration curve creation set (19 samples) and a calibration curve verification set (6 samples), a calibration curve is first created with the calibration curve creation set, and then evaluation is performed with the calibration curve verification set will be described.

[0042] Similar to the example shown in Figure 17, after performing first derivative, detrend, and SNV processing on the near-infrared spectrum, a calibration curve was created using PLS regression in the wavelength range of 860 nm to 1010 nm. Figure 21 is a scatter plot showing the correlation between IM9500 predicted values ​​(vertical axis) and measured values ​​(horizontal axis) when calibration curves were created for 19 samples used for calibration curve creation. The accuracy at this time was R 2 The coefficient of determination was 0.9996 and the standard error (SEC) was 0.046. Figure 22 is a graph showing the calibration curve at this time. It shows the regression coefficients (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0043] Next, we will explain the results of evaluating the six calibration curve validation sets using the obtained calibration curve. The total DON and NIV amounts were calculated by applying the regression equation set up according to the procedure described above to the calibration curve validation samples, and the validity of the regression equation was evaluated. Figure 23 is a scatter plot comparing the calculated values ​​of total DON and NIV based on the calibration curve (IM9500 predicted values) with the measured values ​​obtained from LC-MS. 2 The result was 0.9860, and the SEP (Standard Error) was 0.493.

[0044] Next, we will describe an example of quantifying deoxynivalenol using the EZ-M method. In this example, we used data obtained from Charm Science's EZ-M system, which is based on the principle of antigen-antibody reaction (immunochromatography).

[0045] This section describes the case where all data is used without splitting the calibration curve creation set and the calibration curve validation set. Using the near-infrared spectrum of wheat collected with IM9500 and the DON content values ​​actually obtained with the EZ-M system, a calibration curve was created to determine the DON content from the spectrum, and evaluation was performed by cross-validation. After performing first derivative, detrend, and SNV processing on the spectrum obtained with IM9500, a calibration curve was created using PLS regression in the wavelength range of 860 nm to 985 nm (Figure 24).

[0046] Figure 25 is a scatter plot comparing the calculated DON content (IM9500 predicted value) based on a calibration curve determined by performing PLS regression on all 25 samples without separating the calibration curve creation and calibration curve verification, with the measured value obtained by the EZ-M system. The accuracy during calibration curve creation is R 2 The coefficient of determination was 0.9718, and the standard error (SEC) was 0.113. Figure 26 is a graph showing the calibration curve at this time. It shows the regression coefficients (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0047] Figure 27 is a scatter plot comparing the calculated DON content (IM9500 predicted value) after cross-validation (tolerance verification) with the measured value obtained by the EZ-M system. The accuracy in cross-validation is R 2 The result was 0.9133, and the SECV (standard error) was 0.202.

[0048] Next, we will explain the process of dividing the samples into two groups: a calibration curve creation set (19 points) and a calibration curve verification set (6 points). First, a calibration curve is created using the calibration curve creation set, and then the results are evaluated using the calibration curve verification set.

[0049] Similar to the example shown in Figure 24, after performing first derivative, Detrend, and SNV processing on the near-infrared spectrum, a calibration curve was created using PLS regression in the wavelength range of 860 nm to 985 nm. Figure 29 is a scatter plot showing the correlation between IM9500 predicted values ​​(vertical axis) and measured values ​​(horizontal axis) when calibration curves were created for 19 samples used for calibration curve creation. The accuracy at this time was R 2 The coefficient of determination was 0.9601 and the standard error (SEC) was 0.071. Figure 29 is a graph showing the calibration curve at this time. It shows the regression coefficients (vertical axis) for each wavelength (horizontal axis) in the obtained calibration curve.

[0050] Next, we will explain the results of evaluating the six calibration curve validation sets using the obtained calibration curve. The DON content was calculated by applying the regression equation set up according to the procedure described above to the calibration curve validation samples, and the validity of the regression equation was evaluated. Figure 30 is a scatter plot comparing the calculated DON content (IM9500 predicted value) based on the calibration curve with the measured value obtained from the EZ-M system. 2 The result was 0.9676, and the SEP (Standard Error) was 0.133.

[0051] The above shows that the calculated levels of mycotoxins belonging to type B trichothecenes produced by Fusarium fungi, such as deoxynivalenol (DON) and nivalenol (NIV), in wheat grains, as determined by this invention, show a high correlation with measurements using liquid chromatography-mass spectrometry (LC-MS) and Charm Science's EZ-M system. Near-infrared spectroscopy allows for the quantitative determination of mycotoxin content in wheat grains in a short time and at low cost. Furthermore, it can be performed non-destructively, the procedure is simple, and no expertise in chemical analysis is required.

[0052] Furthermore, the evaluation process described above after acquiring near-infrared spectral information can be automated. The calibration curve data is stored, and a computer program is created to handle the acquisition of near-infrared spectral information for the input sample, pre-processing of the spectral information (such as first derivative, detrend, and SNV processing), selection of the wavelength band to be used, and calculation of mycotoxin content based on a regression equation. This program is then installed on the computer. The user can then import the near-infrared spectral information into the computer to obtain the final predicted value.

[0053] Furthermore, the program may be installed in a near-infrared spectrometer equipped with a processing unit (CPU) and memory. In this case, the user can obtain the final predicted value by simply feeding the sample into the analyzer once.

Claims

1. A method for quantifying trichothecene type B mycotoxins produced by Fusarium fungi, comprising: obtaining the optical spectrum of wheat grains of multiple types of common wheat, which are samples for calibration curve creation, using a transmission method with light wavelengths including the near-infrared band with wavelengths less than 1000 nm for each type; quantifying the mycotoxins belonging to trichothecene type B produced by Fusarium fungi for each type of wheat grain in the calibration curve creation samples using a known quantitative method; analyzing the obtained optical spectrum information including the near-infrared band with wavelengths less than 1000 nm and the mycotoxin quantification information using PLS regression analysis to set a regression equation showing the relationship between the optical spectrum information and the mycotoxin quantification; obtaining the optical spectrum of wheat grains, which are the sample to be measured, using a transmission method with light wavelengths including the near-infrared band with wavelengths less than 1000 nm; and calculating the mycotoxin content based on the set regression equation.

2. A method for quantifying deoxynivalenol in wheat grains, comprising: obtaining the optical spectrum of each type of wheat grain of multiple types of common wheat, which are samples for calibration curve creation, using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm by transmission method; quantifying deoxynivalenol (DON) for each type of wheat grain of the calibration curve creation sample using a known quantitative method; analyzing the obtained optical spectrum information including the near-infrared band with wavelengths less than 1000 nm and the quantification information of deoxynivalenol by PLS regression analysis to set a regression equation showing the relationship between the optical spectrum information and the quantification of deoxynivalenol; obtaining the optical spectrum of the wheat grains, which are the sample to be measured, using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm by transmission method; and calculating the deoxynivalenol content based on the set regression equation.

3. A method for quantifying nivalenol in wheat grains, comprising: obtaining the optical spectrum of each type of wheat grain of multiple types of common wheat, which are samples for calibration curve creation, using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm by transmission method; quantifying nivalenol (NIV) for each type of wheat grain of the calibration curve creation sample using a known quantitative method; analyzing the obtained optical spectrum information including the near-infrared band with wavelengths less than 1000 nm and the quantification information of nivalenol by PLS regression analysis to set a regression equation showing the relationship between the optical spectrum information and the quantification of nivalenol; obtaining the optical spectrum of the wheat grains, which are the sample to be measured, using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm by transmission method; and calculating the nivalenol content based on the set regression equation.

4. For each type of wheat grain of multiple types of common wheat, which are used as calibration curve samples, the optical spectrum was obtained by transmission using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm. Deoxynivalenol (DON) and nivalenol (NIV) were quantified for each type of wheat grain of the calibration curve samples using known quantitative methods. The obtained optical spectrum information including the near-infrared band with wavelengths less than 1000 nm and the quantitative information of deoxynivalenol and nivalenol were analyzed using PLS regression analysis to set up a regression equation that shows the relationship between the optical spectrum information and the quantitative values ​​of deoxynivalenol and nivalenol. A method for quantifying the total amount of deoxynivalenol and nivalenol in wheat grains, which involves obtaining the optical spectrum of wheat grains, a sample to be measured, using a transmission method with light at wavelengths including the near-infrared band with wavelengths less than 1000 nm, and calculating and summing the respective contents of deoxynivalenol and nivalenol based on a set regression equation.

5. For each type of wheat grain of multiple varieties of common wheat, which are used as calibration curve samples, the optical spectrum was obtained by transmission using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm. The total amounts of deoxynivalenol (DON) and nivalenol (NIV) were quantified for each type of wheat grain in the calibration curve samples using a known quantitative method. The acquired optical spectrum information including the near-infrared band with wavelengths less than 1000 nm and the quantification information of the total amounts of deoxynivalenol and nivalenol were analyzed using PLS regression analysis to set up a regression equation that shows the relationship between the optical spectrum information and the quantification of the total amounts of deoxynivalenol and nivalenol. A method for quantifying the total amount of deoxynivalenol and nivalenol in wheat grains, comprising obtaining the optical spectrum of wheat grains, which are the sample to be measured, using light with wavelengths including the near-infrared band with wavelengths less than 1000 nm, by transmission, and calculating and summing the total amounts of deoxynivalenol and nivalenol based on a set regression equation.

Images