WO2020031447A1 - Sample evaluation/estimation method by fluorescence fingerprint analysis, program, and device - Google Patents
Sample evaluation/estimation method by fluorescence fingerprint analysis, program, and device Download PDFInfo
- Publication number
- WO2020031447A1 WO2020031447A1 PCT/JP2019/018739 JP2019018739W WO2020031447A1 WO 2020031447 A1 WO2020031447 A1 WO 2020031447A1 JP 2019018739 W JP2019018739 W JP 2019018739W WO 2020031447 A1 WO2020031447 A1 WO 2020031447A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- fluorescence
- sample
- test sample
- fingerprint information
- fluorescent fingerprint
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/62—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light
- G01N21/63—Systems in which the material investigated is excited whereby it emits light or causes a change in wavelength of the incident light optically excited
- G01N21/64—Fluorescence; Phosphorescence
Definitions
- the present invention relates to a method, a program, and an apparatus for evaluating and estimating a sample using fluorescent fingerprint analysis, and more particularly, to an evaluation and estimating method, a program, and an apparatus that can be suitably used for evaluating and estimating a trace component in a sample. .
- the excitation wavelength ( ⁇ Ex) and the fluorescence wavelength Corresponding points are plotted in a three-dimensional space having (measurement wavelength) ( ⁇ Em) and fluorescence intensity (IEx, Em) as three orthogonal axes.
- a visualization of a set of these points is called a fluorescent fingerprint or an Excitation Emission Matrix (EEM).
- the fluorescent fingerprint can be represented as a three-dimensional graph by displaying the fluorescence intensity of each point in a contour shape, a color distribution, or the like (see FIG. 3), or can be represented as a two-dimensional graph (see FIG. 4). ).
- Non-Patent Document 1 introduces a technique for estimating the mixing ratio of buckwheat flour using a fluorescent fingerprint.
- such analysis using fluorescent fingerprints allows characterization of test samples without performing pretreatment such as fluorescent staining on the test samples, and is easier to operate. Measurement can be performed in a short time, the amount of information is large and quantification can be performed relatively easily, nondestructive and non-contact measurement is possible, and the device is relatively It has such advantages as being inexpensive.
- fluorescence spectrophotometer that scans the excitation wavelength and continuously measures each fluorescence spectrum is required, but a fluorescence spectrophotometer that has such a function is also commercially available. (Hitachi High-Tech Science F-7000, etc.).
- JP 2017-36991 A JP-A-2017-51162
- a fluorescent fingerprint information obtaining step of obtaining fluorescent fingerprint information comprising data of excitation wavelength, fluorescent wavelength, and fluorescent intensity for the test sample, and an axis such that a peak value of the fluorescent intensity unique to the test sample appears by a second derivative process.
- a pre-processing step including at least a secondary differentiation process of the fluorescent fingerprint information along the axis, and at least the fluorescent fingerprint information on which the secondary differentiation process has been performed is used as an explanatory variable.
- Aspect 2 2.
- Aspect 5 The method for evaluating a test sample by fluorescent fingerprint analysis according to aspects 1 to 4, wherein the test sample contains chlorogenic acid. (Aspect 6) 5.
- (Aspect 7) 5. The method for evaluating a test sample by fluorescent fingerprint analysis according to aspects 1 to 4, wherein the test sample contains rutin.
- (Aspect 8) Any one of the excitation wavelength / fluorescence wavelengths in which the excitation wavelength is near 285 nm and the fluorescence wavelength is at least one of the values near 410, 415, 420, 425, 430, 435, 440, 450 and 460 nm.
- Aspect 7 The method for evaluating a test sample by fluorescence fingerprint analysis according to aspect 7, wherein a calibration curve of rutin is created by a combination of wavelengths.
- the test material is a raw material of a tobacco product
- the harmony condition is a condition of a room at a temperature of 22 ° C. and a humidity of 60%
- the predetermined time is 24 hours or more.
- (Aspect 15) Estimating the content of a specific component contained in a sample based on the calibration curve obtained by the method for evaluating a test sample by the fluorescent fingerprint analysis described in Aspects 1 to 14 and the fluorescent fingerprint information of an unknown sample. A method for estimating the characteristic amount of a component.
- the excitation wavelength is a value near 285 nm and the fluorescence wavelengths are all values near 410, 415, 420, 425, 430, 435, 440, 450 and 460 nm. 18.
- Preprocessing means including at least a second derivative process of the fluorescent fingerprint information along the at least the fluorescent fingerprint information on which the second derivative process has been performed is used as an explanatory variable, and a known quantitative value of the test sample is used as a target variable.
- An apparatus for estimating a component amount to be estimated. (Aspect 21) 21. The apparatus according to aspect 20, wherein the fluorescent fingerprint information of the unknown sample is processed by the preprocessing unit, and the processed fluorescent fingerprint information is input to the component amount estimating unit.
- the term “nearby value” also refers to a wavelength existing in an error range of the excitation wavelength or the fluorescence (measurement) wavelength because the excitation wavelength or the fluorescence (measurement) wavelength necessarily involves an error. It has been added to clarify inclusion.
- the error range can vary depending on the measuring equipment, the measuring conditions, and the like.
- a" program is a data processing method described based on an arbitrary language or description method, and does not ask the form of a source code, a binary code, or the like.
- the “program” may be configured in a single form, but may be configured in a distributed manner as a plurality of modules or libraries, and may be configured to achieve its function in cooperation with other existing programs. It may be configured.
- the “apparatus” may be configured as hardware, or may be configured as a combination of function realizing means for realizing various functions by computer software.
- the function realizing means may include, for example, a program module.
- the present invention it is possible to create a highly accurate estimation model (calibration curve). In particular, it is possible to accurately determine a specific component from a sample containing components having similar chemical structures.
- FIG. 1 is a flowchart for explaining an outline of an embodiment of the present invention.
- FIG. 2 is an explanatory diagram illustrating an outline of a spectrum of fluorescence emitted from the measurement target when the measurement target is irradiated with excitation light.
- FIG. 3 is a contour graph showing a three-dimensional example of a fluorescent fingerprint.
- FIG. 4 is a contour graph showing a two-dimensional example of a fluorescent fingerprint.
- FIG. 5 is a diagram illustrating an example of a fluorescent fingerprint of a mixed sample containing scopoletin, chlorogenic acid, and rutin.
- FIG. 6A is a diagram illustrating an example of a fluorescent fingerprint of a sample that includes a single substance.
- FIG. 6A is a diagram illustrating an example of a fluorescent fingerprint of a sample that includes a single substance.
- FIG. 6B is a diagram illustrating an example of a fluorescent fingerprint of a sample containing chlorogenic acid alone.
- FIG. 6C is a diagram illustrating an example of a fluorescent fingerprint of a sample containing rutin alone.
- FIG. 7A is a graph plotting points defined by a known content of scopoletin and an estimated value based on fluorescent fingerprint information for a plurality of samples.
- FIG. 7B is a graph plotting points defined by a known content of chlorogenic acid and an estimated value based on fluorescent fingerprint information for a plurality of samples.
- FIG. 7C is a graph plotting points defined by a known content of rutin and an estimated value based on fluorescent fingerprint information for a plurality of samples.
- FIG. 7A is a graph plotting points defined by a known content of scopoletin and an estimated value based on fluorescent fingerprint information for a plurality of samples.
- FIG. 7B is a graph plotting points defined by a known content of chlorogenic acid and an estimated value based on
- FIG. 8A is a graph plotting points defined by an actual measurement value (chemical analysis value) of rutin content and an estimated value based on pre-processed fluorescent fingerprint information for a plurality of samples.
- FIG. 8B is a graph plotting points defined by measured values (chemical analysis values) of chlorogenic acid contents and estimated values based on pre-processed fluorescent fingerprint information for a plurality of samples.
- FIG. 9 is a graph in which, for a plurality of samples, points defined by actual measured values (chemical analysis values) of rutin content and estimated values based on fluorescent fingerprint information without preprocessing are plotted.
- FIG. 10 is a block diagram for explaining an outline of another embodiment of the present invention.
- FIG. 1 is a flowchart for explaining an outline of an embodiment of the present invention.
- a test sample with a known type and content of components is prepared (S01), and a fluorescent fingerprint is measured on such a known test sample to acquire fluorescent fingerprint information (S02).
- preprocessing is performed on the acquired fluorescent fingerprint information (S03).
- One embodiment of the present invention is particularly characterized by this preprocessing, and the details of the preprocessing will be described later.
- an estimation model (calibration curve) is created (S04). Specifically, this modeling is performed by using various multivariate analysis methods and data mining methods to construct an estimation expression using preprocessed fluorescent fingerprint information as an explanatory variable and a known content as an objective variable.
- the calibration curve (regression equation) for estimating the content of the specific component in the test sample from the fluorescent fingerprint information is created.
- the algorithm used for constructing the estimation formula may be a machine learning algorithm that is more general and supports nonlinear phenomena, such as support vector machine (SVM), random forest (RF), and neural network. An example of a multivariate analysis method used for modeling will be described later.
- the content of the specific component contained in the unknown sample is estimated based on the fluorescent fingerprint information of the unknown sample (S06).
- FIG. 10 is a block diagram for explaining an outline of another embodiment of the present invention.
- the apparatus 100 for evaluating and estimating a sample by fluorescence fingerprint analysis inputs fluorescence fingerprint information including data of excitation wavelength, fluorescence wavelength, and fluorescence intensity of the sample, and performs peak differentiation of the fluorescence intensity peculiar to the sample by a second derivative process.
- An axis is set so that a value appears, and a preprocessing unit 110 including at least a second differentiation process of the fluorescent fingerprint information along the axis, and an output of the preprocessing unit 110 as an input, at least the second differentiation process Is used as an explanatory variable, and a known quantitative value of the test sample is used as a target variable, and an estimation model creating means 120 for obtaining a calibration curve, and the calibration obtained by the estimation model creating means 120 are used.
- the obtained fluorescent fingerprint information is input to the preprocessing means 110, and preprocessing is performed on the input fluorescent fingerprint information.
- This aspect also has a feature in this preprocessing, similarly to the above-described embodiment of the present invention, and the details of the preprocessing will be described later.
- the relationship between the preprocessed fluorescent fingerprint information and the known content of the specific component is modeled by the estimation model creation means 120 to create an estimation model (calibration curve).
- This modeling is similar to the above-described embodiment of the present invention.
- the estimation model (calibration curve) thus constructed is verified, its validity is confirmed, and the estimation model (calibration curve) whose validity is confirmed is stored in a memory (not shown) or the like.
- the component amount estimating means 130 estimates the content of the specific component contained in the unknown sample based on the fluorescent fingerprint information of the unknown sample using the estimation model (calibration curve) whose effectiveness has been confirmed. It is desirable to perform pre-processing by the pre-processing unit 110 also on the fluorescent fingerprint information of the unknown sample (the sample evaluation / estimation device 100 based on the fluorescent fingerprint analysis of FIG. 10 adopts such a configuration. However, such preprocessing can be omitted as necessary.
- the measured value of the fluorescent fingerprint (fluorescence spectrum for each excitation wavelength) can be used as it is, but it is necessary to perform various pretreatments as necessary.
- One embodiment of the present invention aims to improve measurement accuracy by adopting a novel method (secondary differential processing unique to the present invention) as described in detail below, particularly in preprocessing. is there.
- the three orthogonal axes of the excitation wavelength ( ⁇ Ex), the fluorescence wavelength (measurement wavelength) ( ⁇ Em), and the fluorescence intensity (IEx, Em) in the three-dimensional space are set to the y-axis, the x-axis, and the z-axis, respectively, to measure the test sample.
- the coordinates of each point P i (1 ⁇ i ⁇ N) of the EEM (fluorescent fingerprint) obtained in (1) are represented by (x i , y i , z i ).
- a plane determined by the x axis and the y axis is an xy plane
- a plane determined by the x axis and the z axis is an xz plane
- a plane determined by the y axis and the z axis is a yz plane.
- a wz ′ plane orthogonal to the xy plane is obtained.
- z ′ f (w).
- the steepest change point of the gradient is obtained as a peak.
- the number of peaks is not limited to one, and there may be a plurality of peaks. In this case, information corresponding to the plurality of peaks is obtained.
- an optimal w-axis (hereinafter, referred to as an “optimal axis”) for a test sample, a technique exemplified below is effective.
- a technique exemplified below is effective.
- the w axis determined by this is set as the optimal axis.
- the points P i (1 ⁇ i ⁇ N) of the fluorescent fingerprint the sum of squares of distances from points near the bottom surface (points near the xy plane) of the fluorescent fingerprint is minimized.
- a straight line (least square line) is obtained, a wz 'plane including the least square line is determined. Therefore, the w axis determined by this is set as the optimal axis.
- the setting of the optimal axis is not limited to the above-described method.
- the w-axis obtained by the above-described method is set as a temporary optimal axis, and the temporary optimal axis is appropriately rotated on the xy plane. Then, the w-axis may be set so that the peak value of the fluorescence intensity unique to the test sample appears. Further, depending on the test sample, a simplified method of setting the optimum axis as an axis parallel to the x-axis or the y-axis may be employed.
- Patent Document 2 exemplifies a second derivative as a signal processing operation for a two-dimensionally developed fluorescent fingerprint (see paragraph [0026], etc.). Note that neither is described nor suggested.
- non-fluorescent component removal processing in order to remove noise from the measured fluorescent fingerprint and obtain effective fluorescent fingerprint information, non-fluorescent component removal processing, scattered light removal processing, low sensitivity
- One or a combination of the area deletion processing can be adopted as the preprocessing.
- one or more combinations of arithmetic processing such as centering, standardization, standardization, baseline correction, smoothing, auto scaling, logarithmic conversion (Log10) are pre-processed on the acquired fluorescent fingerprint information. It can also be adopted as.
- the secondary differential processing that has been used before can also be used together.
- the order of application can be set as appropriate.
- non-fluorescent component removal processing is performed. It is desirable to precede processing such as scattered light removal processing and low-sensitivity area deletion processing.
- Multivariate analysis method used for modeling As a multivariate analysis method used in modeling, various analysis methods such as PLS (Partial Least Squares) regression analysis, multiple regression analysis, principal component regression analysis, and least square method can be used.
- PLS Partial Least Squares
- PLS regression analysis is a method of extracting a principal component so that the covariance between the principal component and the objective variable is maximized, and is effective when there is a strong correlation between explanatory variables (when multicollinearity occurs).
- Principal component regression analysis is a method of extracting principal components so that the variance of the principal components is maximized.Principal component analysis is performed using only explanatory variables, and the minimum between the obtained principal components and the objective variable is calculated. This is to perform multiple regression analysis by the square method. The multiple regression analysis applies the least squares method between the explanatory variable and the objective variable, and has different characteristics from the principal component regression analysis.
- trace components are often present in the sample as polyphenols having similar chemical structures (for example, tobacco raw materials).
- polyphenols having similar chemical structures have similar fluorescent fingerprints, even if the fluorescent fingerprint method is applied, it is expected that it will be difficult to extract only a specific trace component and accurately quantify it. .
- FIG. 5 shows a fluorescent fingerprint of a mixed sample containing chlorogenic acid (10 ppm), rutin (1 ppm), and scopoletin (0.1 ppm).
- FIG. 6A, FIG. 6B, and FIG. 6C show that each of scopoletin (1 ppm: (A)), chlorogenic acid (1 ppm: (B)), and rutin (1 ppm: (C)) contained in the mixed sample was used alone.
- 5 shows a fluorescent fingerprint of a test sample containing the sample.
- Table 1 shows the mixed sample, the wavelength (excitation wavelength / fluorescence wavelength) near the peak position of the fluorescent fingerprint of scopoletin (1 ppm), chlorogenic acid (1 ppm), and rutin (1 ppm), and the corresponding fluorescence intensity. It is a summary of.
- test sample ⁇ When using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a sample> Forty samples (solvent: 50% ethanol / water solution) prepared by mixing rutin / chlorogenic acid / scopoletin samples at various ratios were prepared. Each sample was prepared as a test sample by pulverizing to a particle size of 1 mm or less and sufficiently mixed. By such pulverization and mixing, the accuracy of measurement is ensured even in a case where rutin, chlorogenic acid, and scopoletin are not uniformly distributed but localized in the raw material.
- test samples used were stored in advance to stabilize the water content.
- Each sample having a known rutin content was pulverized to a particle size of 1 mm or less and sufficiently mixed to prepare a test sample.
- the sample is a tobacco raw material
- rutin is not uniformly present but localized in the tobacco raw material.
- the sample is reduced to a certain particle size (1 mm diameter) or less. It is preferable to obtain a fluorescent fingerprint after pulverizing the mixture and thoroughly mixing the mixture.
- the rutin content of each sample was determined in advance by high performance liquid chromatography (HPLC).
- test samples used were stored in advance to stabilize the water content.
- the measurement conditions were as follows: excitation light 200-600 nm, fluorescence 200-700 nm, resolution 5 nm, slit width 5 nm, photomultiplier sensitivity 700 V. Considering a resolution of 5 nm, the measurement wavelength allows at least an error of about 5 nm.
- Preprocessing for fluorescent fingerprint information ⁇ When using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a sample> Preprocessing is performed on fluorescent fingerprint information obtained from test samples in which samples of rutin, chlorogenic acid, and scopoletin are mixed at various ratios.
- preprocessing for example, dedicated software such as Matlab or PLS_toolbox is used.
- the pre-processing in addition to the second-order differentiation processing specific to the present invention described in detail in the above-mentioned ⁇ Pre-processing for fluorescent fingerprint information>, preferably, the second-order differentiation processing previously used for each spectrum and the components are preferably used. There is a process of removing a wavelength that does not contribute to information.
- VIP Variable important projection
- iPLS interval PLS
- G Genetic algorithms
- e Jack-knife analysis
- F Forward interval PLS
- BiPLS Backward interval PLS
- G Synergy interval PLS (siPLS)
- H LASSO type method
- the method described in the above-mentioned ⁇ Preprocessing for fluorescent fingerprint information> is applied.
- the fluorescent fingerprints of rutin, chlorogenic acid, and scopoletin are used.
- An effective result can be obtained by performing a second derivative process on the information along an axis parallel to the x-axis (fluorescence wavelength axis).
- the calibration curve uses the acquired fluorescent fingerprint information as an explanatory variable and the content of known components (rutin, chlorogenic acid, scopoletin) as an objective variable, and performs PLS regression analysis (hereinafter simply referred to as “PLS”). There is also).
- PLS PLS regression analysis
- the PLS does not directly use the information of the explanatory variable X for modeling the objective variable y, but converts part of the information of the explanatory variable X into a latent variable t, and models the objective variable y using the latent variable t. Is what you do.
- the number of latent variables can be determined, for example, using a predictive explanatory variance value obtained by cross-validation as an index.
- Latent variables may also be called principal components.
- the above (1) and (2) are represented by the following (3) and (4).
- X t 1 p 1 T + E (3)
- y t 1 q 1 + f (4)
- t 1 is a latent variable (vector)
- p 1 is loading (vector)
- q 1 is a coefficient (scalar).
- t 1 Xw 1 (5)
- w 1 is a standardized weight vector.
- PLS is a covariance y T t 1 between y and t 1
- the norm of w 1 is 1 (
- 1 ) is intended to determine the t 1 that maximizes under the condition that, t 1 May be calculated using the so-called Lagrange undetermined multiplier method. Since the calculation method using the Lagrange's undetermined multiplier method is well known, the details of the calculation are omitted, and only the calculation results regarding w 1 , p 1 , and q 1 are shown as (6) to (8) below.
- t 1 in the equations (7) and (8) is a vector calculated by substituting w 1 obtained in the equation (6) into the equation (5).
- the above-described PLS regression analysis is applied to the calibration sample group, and a calibration curve for estimating the content of each component from the acquired fluorescent fingerprint information is created.
- a calibration curve for estimating the content of each component from the acquired fluorescent fingerprint information is created.
- pre-processing for the acquired fluorescent fingerprints.
- (1) second-order differentiation processing unique to the present invention (2) removal processing for wavelengths that do not contribute to component information, and the like. It is desirable to perform pre-processing.
- the pre-processing of (2) for example, the following processing can be adopted.
- the content of each component is estimated from the acquired fluorescent fingerprint information using the calibration curve, and the calibration curve is verified.
- FIG. 7A to 7C are graphs of data of a validation sample group for each component of rutin, chlorogenic acid, and scopoletin.
- FIG. 7A shows scopoletin
- FIG. 7B shows chlorogenic acid
- FIG. 7C shows rutin. ing.
- FIG. 8A shows the measured value (chemical analysis value) by high-performance liquid chromatography (HPLC) on the horizontal axis and the estimated value by the fluorescent fingerprint on the vertical axis, and the pretreatment was performed for each sample belonging to the validation sample group. It is the graph which plotted the corresponding point of the case.
- FIG. 8B shows the measured value (chemical analysis value) by high-performance liquid chromatography (HPLC) on the horizontal axis and the estimated value by the fluorescent fingerprint on the vertical axis, and the pretreatment was performed for each sample belonging to the validation sample group. It is the graph which plotted the corresponding point of the case.
- the content of each component of rutin, chlorogenic acid, and scopoletin is based on the fluorescent fingerprint information of the unknown sample, and the content of each component contained in the sample is determined.
- the amount can be estimated effectively.
- This simplified embodiment uses only nine wavelengths of 285/410, 415, 420, 425, 430, 435, 440, 450, and 460 (nm) as excitation wavelength / fluorescence wavelength (measurement wavelength) to obtain fluorescent fingerprint information. Is obtained to determine the amount of rutin.
- the method for quantification is basically the same as the case where all the fluorescent fingerprint information is used, and details are omitted.
- the wavelength is limited as described above (generally less than 10 wavelengths)
- multiple regression analysis (MLR) can be used to generate a calibration curve.
- MLR multiple regression analysis
- the nine wavelengths of 285/410, 415, 420, 425, 430, 435, 440, 450, and 460 are specific excitation / fluorescence wavelengths at which the fluorescence intensity takes a maximum value based on the chemical structure of rutin. It is equivalent.
- the simplified embodiment is not limited to this, and the excitation wavelength / fluorescence wavelength (measurement wavelength) is 285/410, 415, 420, 425, 430, 435, 440, 450, and 460 (nm).
- An embodiment including at least one can be adopted. In this case, although the measurement accuracy is inferior to the case where only 9 wavelengths are used, an improvement in speed can be realized.
Abstract
In this sample evaluation/estimation method by fluorescence fingerprint analysis, a highly accurate estimation model (a calibration curve) is obtained by performing a novel form of secondary differential processing on acquired fluorescence fingerprint information. A test sample with a known content of a specific component is prepared (S01) and fluorescence fingerprint information is acquired (S02). Pre-processing including the novel secondary differential processing is performed (S03) on the fluorescence fingerprint information, and an estimation model (a calibration curve) for estimating the content of the specific component in a test sample from fluorescence fingerprint information is created (S04). After verification (S05) of the calibration curve, said calibration curve is applied to an unknown sample and the content of the specific component included in the unknown sample is estimated (S06).
Description
本発明は、蛍光指紋分析を利用した試料の評価・推定方法、プログラム、及び装置に関し、特に、試料中の微量成分の評価・推定に好適に利用し得る評価・推定方法、プログラム、及び装置に関する。
The present invention relates to a method, a program, and an apparatus for evaluating and estimating a sample using fluorescent fingerprint analysis, and more particularly, to an evaluation and estimating method, a program, and an apparatus that can be suitably used for evaluating and estimating a trace component in a sample. .
図２に示すように、蛍光物質を含む試験試料に、段階的に波長を変化させながら励起光を照射し、試験試料から発せられる光（蛍光）を測定すると、励起波長（λＥｘ）、蛍光波長（測定波長）（λＥｍ）、蛍光強度（ＩＥｘ，Ｅｍ）を３直交軸とする3次元空間において、対応するポイントがプロットされる。
これらのポイントの集合を可視化したものを、蛍光指紋、または、励起蛍光マトリクス（Excitation Emission Matrix；ＥＥＭ）と呼ぶ。
蛍光指紋は、各ポイントの蛍光強度を等高線形状や色分布等で表示することにより、３次元グラフとして表すことができ（図３参照）、また、２次元グラフとして表すこともできる（図４参照）。 As shown in FIG. 2, when a test sample containing a fluorescent substance is irradiated with excitation light while changing the wavelength stepwise, and light (fluorescence) emitted from the test sample is measured, the excitation wavelength (λEx) and the fluorescence wavelength Corresponding points are plotted in a three-dimensional space having (measurement wavelength) (λEm) and fluorescence intensity (IEx, Em) as three orthogonal axes.
A visualization of a set of these points is called a fluorescent fingerprint or an Excitation Emission Matrix (EEM).
The fluorescent fingerprint can be represented as a three-dimensional graph by displaying the fluorescence intensity of each point in a contour shape, a color distribution, or the like (see FIG. 3), or can be represented as a two-dimensional graph (see FIG. 4). ).
これらのポイントの集合を可視化したものを、蛍光指紋、または、励起蛍光マトリクス（Excitation Emission Matrix；ＥＥＭ）と呼ぶ。
蛍光指紋は、各ポイントの蛍光強度を等高線形状や色分布等で表示することにより、３次元グラフとして表すことができ（図３参照）、また、２次元グラフとして表すこともできる（図４参照）。 As shown in FIG. 2, when a test sample containing a fluorescent substance is irradiated with excitation light while changing the wavelength stepwise, and light (fluorescence) emitted from the test sample is measured, the excitation wavelength (λEx) and the fluorescence wavelength Corresponding points are plotted in a three-dimensional space having (measurement wavelength) (λEm) and fluorescence intensity (IEx, Em) as three orthogonal axes.
A visualization of a set of these points is called a fluorescent fingerprint or an Excitation Emission Matrix (EEM).
The fluorescent fingerprint can be represented as a three-dimensional graph by displaying the fluorescence intensity of each point in a contour shape, a color distribution, or the like (see FIG. 3), or can be represented as a two-dimensional graph (see FIG. 4). ).
このような蛍光指紋は、３次元の膨大な情報を有する試験試料固有のパターンを示しており、各種の鑑別や定量等に使用できる。例えば、非特許文献１には、蛍光指紋を利用した蕎麦粉混合割合の推定手法等が紹介されている。
Such a fluorescent fingerprint shows a unique pattern of a test sample having a large amount of three-dimensional information, and can be used for various discriminations, quantification, and the like. For example, Non-Patent Document 1 introduces a technique for estimating the mixing ratio of buckwheat flour using a fluorescent fingerprint.
このような蛍光指紋を用いた分析は、他の分光分析手法と比較して、試験試料に対して蛍光染色などの前処理を行うことなく試験試料のキャラクタリゼーションが可能であること、操作が容易で計測も短時間で行えること、情報量が多く定量も比較的容易に行えること、非破壊・非接触での計測が可能であること、紫外～可視光を用いていることから装置が比較的安価であること、などの長所を有している。
Compared to other spectroscopic analysis methods, such analysis using fluorescent fingerprints allows characterization of test samples without performing pretreatment such as fluorescent staining on the test samples, and is easier to operate. Measurement can be performed in a short time, the amount of information is large and quantification can be performed relatively easily, nondestructive and non-contact measurement is possible, and the device is relatively It has such advantages as being inexpensive.
蛍光指紋を計測するには、励起波長をスキャンして、それぞれの蛍光スペクトルを連続的に計測する機能を有する蛍光分光光度計が必要であるが、このような機能を有する蛍光分光光度計も市販されている（日立ハイテクサイエンス社製Ｆ－7000等）。
To measure fluorescence fingerprints, a fluorescence spectrophotometer that scans the excitation wavelength and continuously measures each fluorescence spectrum is required, but a fluorescence spectrophotometer that has such a function is also commercially available. (Hitachi High-Tech Science F-7000, etc.).
蛍光指紋分析を利用した既知の試料の評価・推定手法においては、前処理を行っても、精度の高い推定モデル（検量線）の作成が困難な場合があった。特に、化学構造が類似した成分を含む試料中から特定の成分の定量を行う場合に困難性が高かった。
に お い て In the method of evaluating and estimating known samples using fluorescence fingerprint analysis, it was sometimes difficult to create a highly accurate estimation model (calibration curve) even with preprocessing. In particular, it is difficult to quantify a specific component from a sample containing components having similar chemical structures.
本発明は、このような課題を解決するために提案されたものであり、実施の態様を例示すれば、以下のとおりである。
（態様１）
試験試料について励起波長・蛍光波長・蛍光強度のデータからなる蛍光指紋情報を取得する蛍光指紋情報取得工程と、２次微分処理により前記試験試料に特有な前記蛍光強度のピーク値が現れるように軸を設定し、前記軸に沿った前記蛍光指紋情報の２次微分処理を少なくとも含む前処理工程と、少なくとも前記２次微分処理が実施された蛍光指紋情報を説明変数とし、前記試験試料についての既知の定量値を目標変数として、検量線を取得する推定モデル作成工程と、を含む、蛍光指紋分析による試験試料の評価方法。
（態様２）
前記推定モデル作成工程において、多変量解析によって上記検量線を作成することを特徴とする態様１に記載の蛍光指紋分析による試験試料の評価方法。
（態様３）
前記多変量解析は、PLS回帰分析であることを特徴とする態様２に記載の蛍光指紋分析による試験試料の評価方法。
（態様４）
前記前処理工程において、前記蛍光指紋情報に対して低感度領域の削除処理を行うことを特徴とする態様１～３に記載の蛍光指紋分析による試験試料の評価方法。
（態様５）
前記試験試料が、クロロゲン酸を含む、態様１～４に記載の蛍光指紋分析による試験試料の評価方法。
（態様６）
前記試験試料が、スコポレチンを含む、態様１～４に記載の蛍光指紋分析による試験試料の評価方法。
（態様７）
前記試験試料が、ルチンを含む、態様１～４に記載の蛍光指紋分析による試験試料の評価方法。
（態様８）
前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値のうちの少なくとも１つである波長のいずれかの励起波長／蛍光波長の組み合わせにより、ルチンの検量線を作成することを特徴とする態様７に記載の蛍光指紋分析による試験試料の評価方法。
（態様９）
前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値の全てである波長の励起波長／蛍光波長の組み合わせの両方により、ルチンの検量線を作成することを特徴とする態様８に記載の蛍光指紋分析による試験試料の評価方法。
（態様１０）
前記試験試料は、たばこ製品の原料であることを特徴とする態様１～９に記載の蛍光指紋分析による試験試料の評価方法。
（態様１１）
前記試験試料は、励起光の照射前に粉末状に粉砕・混合されることを特徴とする態様１～１０に記載の蛍光指紋分析による試験試料の評価方法。
（態様１２）
前記試験粉砕によって、試料が１ｍｍ以下の粒径とされることを特徴とする態様９に記載の蛍光指紋分析による試験試料の評価方法。
（態様１３）
前記試験材料は、事前に水分量を一定化するために、所定の調和条件で所定時間蔵置されることを特徴とする態様１～１２に記載の蛍光指紋分析による試験試料の評価方法。
（態様１４）
前記試験材料は、たばこ製品の原料であり、前記調和条件は、温度２２℃、湿度６０％の室内という条件であり、前記所定時間は２４時間以上であることを特徴とする態様１３に記載の蛍光指紋分析による試験試料の評価方法。
（態様１５）
態様１～１４に記載の蛍光指紋分析による試験試料の評価方法により得られた検量線と未知の試料の蛍光指紋情報とに基づき、前記試料に含有される特定成分の含有量を推定することを特徴とする成分量推定方法。
（態様１６）
前記試料に含有される特定成分の含有量を推定する際に、前記蛍光指紋分析による試験試料の評価方法における前処理と同一の処理を行うことを特徴とする態様１５に記載の成分量推定方法。
（態様１７）
前記未知の試料の蛍光指紋情報を取得するために、励起波長が２８５nm近傍値であって蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値のうち少なくとも１つである波長のいずれかの励起波長／蛍光波長の組み合わせを用いることを特徴とする態様１５、１６に記載の成分量推定方法。
（態様１８）
前記未知の資料の蛍光指紋情報を取得するために、前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０，４５０，４６０nm近傍値の全てである波長の励起波長／蛍光波長の組み合わせの両方を用いることを特徴とする態様１７に記載の成分量推定方法。
（態様１９）
コンピュータに態様１～１８に記載の方法を実行させるためのプログラム。
（態様２０）
試料についての励起波長・蛍光波長・蛍光強度のデータからなる蛍光指紋情報を入力し、２次微分処理により前記試料に特有な前記蛍光強度のピーク値が現れるように軸を設定し、前記軸に沿った前記蛍光指紋情報の２次微分処理を少なくとも含む前処理手段と、少なくとも前記２次微分処理が実施された蛍光指紋情報を説明変数とし、前記試験試料についての既知の定量値を目標変数として、検量線を取得する推定モデル作成手段と、前記推定モデル作成手段により取得された前記検量線と未知の試料の蛍光指紋情報とに基づき、前記未知の試料に含有される特定成分の含有量を推定する成分量推定手段と、を具備することを特徴とする装置。
（態様２１）
前記未知の試料の蛍光指紋情報は前記前処理手段で処理され、処理後の蛍光指紋情報が前記成分量推定手段に入力されることを特徴とする態様２０に記載の装置。 The present invention has been proposed to solve such a problem, and the following is an example of an embodiment.
(Aspect 1)
A fluorescent fingerprint information obtaining step of obtaining fluorescent fingerprint information comprising data of excitation wavelength, fluorescent wavelength, and fluorescent intensity for the test sample, and an axis such that a peak value of the fluorescent intensity unique to the test sample appears by a second derivative process. And a pre-processing step including at least a secondary differentiation process of the fluorescent fingerprint information along the axis, and at least the fluorescent fingerprint information on which the secondary differentiation process has been performed is used as an explanatory variable. And estimating a calibration curve using the quantitative value of the target variable as a target variable.
(Aspect 2)
2. The method for evaluating a test sample by fluorescence fingerprint analysis according toaspect 1, wherein the calibration curve is created by multivariate analysis in the estimation model creation step.
(Aspect 3)
The method for evaluating a test sample by fluorescent fingerprint analysis according toaspect 2, wherein the multivariate analysis is a PLS regression analysis.
(Aspect 4)
The method for evaluating a test sample by fluorescent fingerprint analysis according to any one ofaspects 1 to 3, wherein in the preprocessing step, a process of deleting a low-sensitivity region is performed on the fluorescent fingerprint information.
(Aspect 5)
5. The method for evaluating a test sample by fluorescent fingerprint analysis according toaspects 1 to 4, wherein the test sample contains chlorogenic acid.
(Aspect 6)
5. The method for evaluating a test sample by fluorescent fingerprint analysis according toaspects 1 to 4, wherein the test sample contains scopoletin.
(Aspect 7)
5. The method for evaluating a test sample by fluorescent fingerprint analysis according toaspects 1 to 4, wherein the test sample contains rutin.
(Aspect 8)
Any one of the excitation wavelength / fluorescence wavelengths in which the excitation wavelength is near 285 nm and the fluorescence wavelength is at least one of the values near 410, 415, 420, 425, 430, 435, 440, 450 and 460 nm. Aspect 7. The method for evaluating a test sample by fluorescence fingerprint analysis according to aspect 7, wherein a calibration curve of rutin is created by a combination of wavelengths.
(Aspect 9)
Both the excitation wavelength / fluorescence wavelength combination where the excitation wavelength is near 285 nm and the fluorescence wavelength is all of the values near 410, 415, 420, 425, 430, 435, 440, 450, and 460 nm, The method for evaluating a test sample by fluorescent fingerprint analysis according toaspect 8, wherein a calibration curve of rutin is prepared.
(Aspect 10)
The method for evaluating a test sample by fluorescent fingerprint analysis according toaspects 1 to 9, wherein the test sample is a raw material of a tobacco product.
(Aspect 11)
The method for evaluating a test sample by fluorescence fingerprint analysis according toaspects 1 to 10, wherein the test sample is pulverized and mixed into a powder before irradiation with excitation light.
(Aspect 12)
10. The method for evaluating a test sample by fluorescent fingerprint analysis according toaspect 9, wherein the sample is reduced to a particle size of 1 mm or less by the test pulverization.
(Aspect 13)
13. The method for evaluating a test sample by fluorescence fingerprint analysis according toaspects 1 to 12, wherein the test material is stored for a predetermined period of time under a predetermined condition in order to stabilize the water content in advance.
(Aspect 14)
The test material is a raw material of a tobacco product, the harmony condition is a condition of a room at a temperature of 22 ° C. and a humidity of 60%, and the predetermined time is 24 hours or more. Evaluation method of test sample by fluorescent fingerprint analysis.
(Aspect 15)
Estimating the content of a specific component contained in a sample based on the calibration curve obtained by the method for evaluating a test sample by the fluorescent fingerprint analysis described inAspects 1 to 14 and the fluorescent fingerprint information of an unknown sample. A method for estimating the characteristic amount of a component.
(Aspect 16)
The method for estimating a component amount according toaspect 15, wherein when estimating the content of the specific component contained in the sample, the same processing as the preprocessing in the test sample evaluation method by the fluorescent fingerprint analysis is performed. .
(Aspect 17)
In order to acquire the fluorescence fingerprint information of the unknown sample, at least one of the excitation wavelength is a value near 285 nm and the fluorescence wavelength is a value near 410, 415, 420, 425, 430, 435, 440, 450, 460 nm. 17. The component amount estimating method according to aspects 15 and 16, wherein any one of the excitation wavelength / fluorescence wavelength combination of the following wavelengths is used.
(Aspect 18)
In order to obtain the fluorescence fingerprint information of the unknown material, the excitation wavelength is a value near 285 nm and the fluorescence wavelengths are all values near 410, 415, 420, 425, 430, 435, 440, 450 and 460 nm. 18. The component amount estimation method according to aspect 17, wherein both the combination of the excitation wavelength and the fluorescence wavelength of a certain wavelength are used.
(Aspect 19)
A program for causing a computer to execute the method according toaspects 1 to 18.
(Aspect 20)
Fluorescent fingerprint information composed of data of excitation wavelength, fluorescence wavelength, and fluorescence intensity of the sample is input, and an axis is set so that a peak value of the fluorescence intensity unique to the sample appears by a second derivative process. Preprocessing means including at least a second derivative process of the fluorescent fingerprint information along the at least the fluorescent fingerprint information on which the second derivative process has been performed is used as an explanatory variable, and a known quantitative value of the test sample is used as a target variable. Based on the calibration curve and the fingerprint information of the unknown sample acquired by the estimation model creation unit, and obtains the content of the specific component contained in the unknown sample. An apparatus for estimating a component amount to be estimated.
(Aspect 21)
21. The apparatus according toaspect 20, wherein the fluorescent fingerprint information of the unknown sample is processed by the preprocessing unit, and the processed fluorescent fingerprint information is input to the component amount estimating unit.
（態様１）
試験試料について励起波長・蛍光波長・蛍光強度のデータからなる蛍光指紋情報を取得する蛍光指紋情報取得工程と、２次微分処理により前記試験試料に特有な前記蛍光強度のピーク値が現れるように軸を設定し、前記軸に沿った前記蛍光指紋情報の２次微分処理を少なくとも含む前処理工程と、少なくとも前記２次微分処理が実施された蛍光指紋情報を説明変数とし、前記試験試料についての既知の定量値を目標変数として、検量線を取得する推定モデル作成工程と、を含む、蛍光指紋分析による試験試料の評価方法。
（態様２）
前記推定モデル作成工程において、多変量解析によって上記検量線を作成することを特徴とする態様１に記載の蛍光指紋分析による試験試料の評価方法。
（態様３）
前記多変量解析は、PLS回帰分析であることを特徴とする態様２に記載の蛍光指紋分析による試験試料の評価方法。
（態様４）
前記前処理工程において、前記蛍光指紋情報に対して低感度領域の削除処理を行うことを特徴とする態様１～３に記載の蛍光指紋分析による試験試料の評価方法。
（態様５）
前記試験試料が、クロロゲン酸を含む、態様１～４に記載の蛍光指紋分析による試験試料の評価方法。
（態様６）
前記試験試料が、スコポレチンを含む、態様１～４に記載の蛍光指紋分析による試験試料の評価方法。
（態様７）
前記試験試料が、ルチンを含む、態様１～４に記載の蛍光指紋分析による試験試料の評価方法。
（態様８）
前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値のうちの少なくとも１つである波長のいずれかの励起波長／蛍光波長の組み合わせにより、ルチンの検量線を作成することを特徴とする態様７に記載の蛍光指紋分析による試験試料の評価方法。
（態様９）
前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値の全てである波長の励起波長／蛍光波長の組み合わせの両方により、ルチンの検量線を作成することを特徴とする態様８に記載の蛍光指紋分析による試験試料の評価方法。
（態様１０）
前記試験試料は、たばこ製品の原料であることを特徴とする態様１～９に記載の蛍光指紋分析による試験試料の評価方法。
（態様１１）
前記試験試料は、励起光の照射前に粉末状に粉砕・混合されることを特徴とする態様１～１０に記載の蛍光指紋分析による試験試料の評価方法。
（態様１２）
前記試験粉砕によって、試料が１ｍｍ以下の粒径とされることを特徴とする態様９に記載の蛍光指紋分析による試験試料の評価方法。
（態様１３）
前記試験材料は、事前に水分量を一定化するために、所定の調和条件で所定時間蔵置されることを特徴とする態様１～１２に記載の蛍光指紋分析による試験試料の評価方法。
（態様１４）
前記試験材料は、たばこ製品の原料であり、前記調和条件は、温度２２℃、湿度６０％の室内という条件であり、前記所定時間は２４時間以上であることを特徴とする態様１３に記載の蛍光指紋分析による試験試料の評価方法。
（態様１５）
態様１～１４に記載の蛍光指紋分析による試験試料の評価方法により得られた検量線と未知の試料の蛍光指紋情報とに基づき、前記試料に含有される特定成分の含有量を推定することを特徴とする成分量推定方法。
（態様１６）
前記試料に含有される特定成分の含有量を推定する際に、前記蛍光指紋分析による試験試料の評価方法における前処理と同一の処理を行うことを特徴とする態様１５に記載の成分量推定方法。
（態様１７）
前記未知の試料の蛍光指紋情報を取得するために、励起波長が２８５nm近傍値であって蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値のうち少なくとも１つである波長のいずれかの励起波長／蛍光波長の組み合わせを用いることを特徴とする態様１５、１６に記載の成分量推定方法。
（態様１８）
前記未知の資料の蛍光指紋情報を取得するために、前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０，４５０，４６０nm近傍値の全てである波長の励起波長／蛍光波長の組み合わせの両方を用いることを特徴とする態様１７に記載の成分量推定方法。
（態様１９）
コンピュータに態様１～１８に記載の方法を実行させるためのプログラム。
（態様２０）
試料についての励起波長・蛍光波長・蛍光強度のデータからなる蛍光指紋情報を入力し、２次微分処理により前記試料に特有な前記蛍光強度のピーク値が現れるように軸を設定し、前記軸に沿った前記蛍光指紋情報の２次微分処理を少なくとも含む前処理手段と、少なくとも前記２次微分処理が実施された蛍光指紋情報を説明変数とし、前記試験試料についての既知の定量値を目標変数として、検量線を取得する推定モデル作成手段と、前記推定モデル作成手段により取得された前記検量線と未知の試料の蛍光指紋情報とに基づき、前記未知の試料に含有される特定成分の含有量を推定する成分量推定手段と、を具備することを特徴とする装置。
（態様２１）
前記未知の試料の蛍光指紋情報は前記前処理手段で処理され、処理後の蛍光指紋情報が前記成分量推定手段に入力されることを特徴とする態様２０に記載の装置。 The present invention has been proposed to solve such a problem, and the following is an example of an embodiment.
(Aspect 1)
A fluorescent fingerprint information obtaining step of obtaining fluorescent fingerprint information comprising data of excitation wavelength, fluorescent wavelength, and fluorescent intensity for the test sample, and an axis such that a peak value of the fluorescent intensity unique to the test sample appears by a second derivative process. And a pre-processing step including at least a secondary differentiation process of the fluorescent fingerprint information along the axis, and at least the fluorescent fingerprint information on which the secondary differentiation process has been performed is used as an explanatory variable. And estimating a calibration curve using the quantitative value of the target variable as a target variable.
(Aspect 2)
2. The method for evaluating a test sample by fluorescence fingerprint analysis according to
(Aspect 3)
The method for evaluating a test sample by fluorescent fingerprint analysis according to
(Aspect 4)
The method for evaluating a test sample by fluorescent fingerprint analysis according to any one of
(Aspect 5)
5. The method for evaluating a test sample by fluorescent fingerprint analysis according to
(Aspect 6)
5. The method for evaluating a test sample by fluorescent fingerprint analysis according to
(Aspect 7)
5. The method for evaluating a test sample by fluorescent fingerprint analysis according to
(Aspect 8)
Any one of the excitation wavelength / fluorescence wavelengths in which the excitation wavelength is near 285 nm and the fluorescence wavelength is at least one of the values near 410, 415, 420, 425, 430, 435, 440, 450 and 460 nm. Aspect 7. The method for evaluating a test sample by fluorescence fingerprint analysis according to aspect 7, wherein a calibration curve of rutin is created by a combination of wavelengths.
(Aspect 9)
Both the excitation wavelength / fluorescence wavelength combination where the excitation wavelength is near 285 nm and the fluorescence wavelength is all of the values near 410, 415, 420, 425, 430, 435, 440, 450, and 460 nm, The method for evaluating a test sample by fluorescent fingerprint analysis according to
(Aspect 10)
The method for evaluating a test sample by fluorescent fingerprint analysis according to
(Aspect 11)
The method for evaluating a test sample by fluorescence fingerprint analysis according to
(Aspect 12)
10. The method for evaluating a test sample by fluorescent fingerprint analysis according to
(Aspect 13)
13. The method for evaluating a test sample by fluorescence fingerprint analysis according to
(Aspect 14)
The test material is a raw material of a tobacco product, the harmony condition is a condition of a room at a temperature of 22 ° C. and a humidity of 60%, and the predetermined time is 24 hours or more. Evaluation method of test sample by fluorescent fingerprint analysis.
(Aspect 15)
Estimating the content of a specific component contained in a sample based on the calibration curve obtained by the method for evaluating a test sample by the fluorescent fingerprint analysis described in
(Aspect 16)
The method for estimating a component amount according to
(Aspect 17)
In order to acquire the fluorescence fingerprint information of the unknown sample, at least one of the excitation wavelength is a value near 285 nm and the fluorescence wavelength is a value near 410, 415, 420, 425, 430, 435, 440, 450, 460 nm. 17. The component amount estimating method according to
(Aspect 18)
In order to obtain the fluorescence fingerprint information of the unknown material, the excitation wavelength is a value near 285 nm and the fluorescence wavelengths are all values near 410, 415, 420, 425, 430, 435, 440, 450 and 460 nm. 18. The component amount estimation method according to aspect 17, wherein both the combination of the excitation wavelength and the fluorescence wavelength of a certain wavelength are used.
(Aspect 19)
A program for causing a computer to execute the method according to
(Aspect 20)
Fluorescent fingerprint information composed of data of excitation wavelength, fluorescence wavelength, and fluorescence intensity of the sample is input, and an axis is set so that a peak value of the fluorescence intensity unique to the sample appears by a second derivative process. Preprocessing means including at least a second derivative process of the fluorescent fingerprint information along the at least the fluorescent fingerprint information on which the second derivative process has been performed is used as an explanatory variable, and a known quantitative value of the test sample is used as a target variable. Based on the calibration curve and the fingerprint information of the unknown sample acquired by the estimation model creation unit, and obtains the content of the specific component contained in the unknown sample. An apparatus for estimating a component amount to be estimated.
(Aspect 21)
21. The apparatus according to
なお、上述の態様において、「近傍値」という語句は、励起波長や蛍光（測定）波長が誤差を必然的に伴うことから、励起波長や蛍光（測定）波長の誤差範囲に存在する波長をも包含することを明確化するために付したものである。因みに、誤差範囲は測定機器や測定条件等により変動し得るものである。
In the above-described embodiment, the term “nearby value” also refers to a wavelength existing in an error range of the excitation wavelength or the fluorescence (measurement) wavelength because the excitation wavelength or the fluorescence (measurement) wavelength necessarily involves an error. It has been added to clarify inclusion. Incidentally, the error range can vary depending on the measuring equipment, the measuring conditions, and the like.
また、「プログラム」とは、任意の言語や記述方法に基づき記述されたデータ処理方法であり、ソースコードやバイナリコード等の形式を問うものではない。また、「プログラム」は単一の形で構成されてもよいが、複数のモジュールやライブラリとして分散構成されてもよく、また、他の既存のプログラムと協働してその機能を達成するように構成されたものであってもよい。
「In addition, a" program "is a data processing method described based on an arbitrary language or description method, and does not ask the form of a source code, a binary code, or the like. Further, the “program” may be configured in a single form, but may be configured in a distributed manner as a plurality of modules or libraries, and may be configured to achieve its function in cooperation with other existing programs. It may be configured.
また、「装置」は、ハードウエアとして構成されてもよいが、コンピュータのソフトウエアによって各種機能を実現する機能実現手段の組合せとして構成されてもよい。機能実現手段には、例えば、プログラムモジュールが含まれ得る。
The “apparatus” may be configured as hardware, or may be configured as a combination of function realizing means for realizing various functions by computer software. The function realizing means may include, for example, a program module.
本発明によれば、精度の高い推定モデル（検量線）の作成が可能になる。また、特に、化学構造が類似した成分を含む試料中から特定の成分の定量を的確に行うことができる。
According to the present invention, it is possible to create a highly accurate estimation model (calibration curve). In particular, it is possible to accurately determine a specific component from a sample containing components having similar chemical structures.
以下、本発明の実施の一態様を説明するとともに、試料中に微量成分として含まれるクロロゲン酸、スコポレチン、ルチンの含有量の本発明に基づく評価・推定方法について説明する。また、特に、タバコ原料を試料とした場合について、たばこ原料中のルチン含有量の本発明に基づく評価・推定方法について説明する。
なお、この実施の一態様により、本発明が限定されるものではないことに留意されたい。 Hereinafter, an embodiment of the present invention will be described, and a method for evaluating and estimating the contents of chlorogenic acid, scopoletin, and rutin contained as trace components in a sample according to the present invention will be described. In addition, a method for evaluating and estimating the content of rutin in tobacco raw materials based on the present invention, particularly when a tobacco raw material is used as a sample, will be described.
Note that the present invention is not limited by one embodiment of this embodiment.
なお、この実施の一態様により、本発明が限定されるものではないことに留意されたい。 Hereinafter, an embodiment of the present invention will be described, and a method for evaluating and estimating the contents of chlorogenic acid, scopoletin, and rutin contained as trace components in a sample according to the present invention will be described. In addition, a method for evaluating and estimating the content of rutin in tobacco raw materials based on the present invention, particularly when a tobacco raw material is used as a sample, will be described.
Note that the present invention is not limited by one embodiment of this embodiment.
＜本発明の実施の一態様の概要＞
図１は、本発明の実施の一態様の概要を説明するためのフローチャートである。 <Overview of One Embodiment of the Present Invention>
FIG. 1 is a flowchart for explaining an outline of an embodiment of the present invention.
図１は、本発明の実施の一態様の概要を説明するためのフローチャートである。 <Overview of One Embodiment of the Present Invention>
FIG. 1 is a flowchart for explaining an outline of an embodiment of the present invention.
最初に、成分の種類や含有量が既知の試験試料を準備し（Ｓ０１）、このような既知の試験試料に対して蛍光指紋を測定し、蛍光指紋情報を取得する（Ｓ０２）。
First, a test sample with a known type and content of components is prepared (S01), and a fluorescent fingerprint is measured on such a known test sample to acquire fluorescent fingerprint information (S02).
次に、取得された蛍光指紋情報に対する前処理を行う（Ｓ０３）。本発明の実施の一態様は、特に、この前処理に特徴を有しており、前処理の詳細は後述する。
Next, preprocessing is performed on the acquired fluorescent fingerprint information (S03). One embodiment of the present invention is particularly characterized by this preprocessing, and the details of the preprocessing will be described later.
次に、前処理済みの蛍光指紋情報と特定成分の既知の含有量との関係をモデル化し、推定モデル（検量線）を作成する（Ｓ０４）。このモデル化は、具体的には、前処理済みの蛍光指紋情報を説明変数、既知の含有量を目的変数とする推定式を、様々な多変量解析手法やデータマイニング手法を使用して構築することにより行われ、蛍光指紋情報から試験試料中の特定成分の含有量を推定する検量線（回帰式）が作成される。なお、推定式の構築に用いるアルゴリズムはより汎用的かつ非線形現象にも対応する機械学習アルゴリズム、例えばsupport vector machine (SVM), random forest (RF), neural networkなどでも良い。モデル化の際に用いられる多変量解析手法の例は後述する。
Next, the relationship between the preprocessed fluorescent fingerprint information and the known content of the specific component is modeled, and an estimation model (calibration curve) is created (S04). Specifically, this modeling is performed by using various multivariate analysis methods and data mining methods to construct an estimation expression using preprocessed fluorescent fingerprint information as an explanatory variable and a known content as an objective variable. The calibration curve (regression equation) for estimating the content of the specific component in the test sample from the fluorescent fingerprint information is created. Note that the algorithm used for constructing the estimation formula may be a machine learning algorithm that is more general and supports nonlinear phenomena, such as support vector machine (SVM), random forest (RF), and neural network. An example of a multivariate analysis method used for modeling will be described later.
このようにして構築された推定モデル（検量線）を検証し、その有効性を確認する（Ｓ０５）。
(4) The estimation model (calibration curve) thus constructed is verified and its effectiveness is confirmed (S05).
有効性が確認された推定モデル（検量線）を用いて、未知の試料の蛍光指紋情報に基づき、前記未知の試料に含有される特定成分の含有量を推定する（Ｓ０６）。
(4) Using the estimated model (calibration curve) whose effectiveness has been confirmed, the content of the specific component contained in the unknown sample is estimated based on the fluorescent fingerprint information of the unknown sample (S06).
＜本発明の実施の別の一態様の概要＞
図１０は、本発明の実施の別の一態様の概要を説明するためのブロック図である。 <Overview of Another Embodiment of the Present Invention>
FIG. 10 is a block diagram for explaining an outline of another embodiment of the present invention.
図１０は、本発明の実施の別の一態様の概要を説明するためのブロック図である。 <Overview of Another Embodiment of the Present Invention>
FIG. 10 is a block diagram for explaining an outline of another embodiment of the present invention.
蛍光指紋分析による試料の評価・推定装置１００は、試料についての励起波長・蛍光波長・蛍光強度のデータからなる蛍光指紋情報を入力とし、２次微分処理により前記試料に特有な前記蛍光強度のピーク値が現れるように軸を設定し、前記軸に沿った前記蛍光指紋情報の２次微分処理を少なくとも含む前処理手段１１０と、前記前処理手段１１０の出力を入力とし、少なくとも前記２次微分処理が実施された蛍光指紋情報を説明変数とし、前記試験試料についての既知の定量値を目標変数として、検量線を取得する推定モデル作成手段１２０と、前記推定モデル作成手段１２０により取得された前記検量線と未知の試料の蛍光指紋情報とに基づき、前記未知の試料に含有される特定成分の含有量を推定する成分量推定手段１３０とを具備している。
The apparatus 100 for evaluating and estimating a sample by fluorescence fingerprint analysis inputs fluorescence fingerprint information including data of excitation wavelength, fluorescence wavelength, and fluorescence intensity of the sample, and performs peak differentiation of the fluorescence intensity peculiar to the sample by a second derivative process. An axis is set so that a value appears, and a preprocessing unit 110 including at least a second differentiation process of the fluorescent fingerprint information along the axis, and an output of the preprocessing unit 110 as an input, at least the second differentiation process Is used as an explanatory variable, and a known quantitative value of the test sample is used as a target variable, and an estimation model creating means 120 for obtaining a calibration curve, and the calibration obtained by the estimation model creating means 120 are used. Component amount estimating means 130 for estimating the content of a specific component contained in the unknown sample based on the line and the fluorescent fingerprint information of the unknown sample. To have.
まず、既知の蛍光分光光度計等を用いて、成分の種類や含有量が既知の試験試料の蛍光指紋情報を取得する。
First, using a known fluorescence spectrophotometer or the like, obtain the fluorescence fingerprint information of the test sample whose component type and content are known.
次に、取得された蛍光指紋情報を前記前処理手段１１０に入力し、入力された蛍光指紋情報に対する前処理を行う。本態様も、前述の本発明の実施の一態様と同様、特に、この前処理に特徴を有しており、前処理の詳細は後述する。
Next, the obtained fluorescent fingerprint information is input to the preprocessing means 110, and preprocessing is performed on the input fluorescent fingerprint information. This aspect also has a feature in this preprocessing, similarly to the above-described embodiment of the present invention, and the details of the preprocessing will be described later.
次に、推定モデル作成手段１２０により、前処理済みの蛍光指紋情報と特定成分の既知の含有量との関係をモデル化し、推定モデル（検量線）を作成する。このモデル化は、前述の本発明の実施の一態様と同様である。そして、このようにして構築された推定モデル（検量線）を検証し、その有効性を確認し、有効性が確認された推定モデル（検量線）を不図示のメモリ等に記憶しておく。
Next, the relationship between the preprocessed fluorescent fingerprint information and the known content of the specific component is modeled by the estimation model creation means 120 to create an estimation model (calibration curve). This modeling is similar to the above-described embodiment of the present invention. Then, the estimation model (calibration curve) thus constructed is verified, its validity is confirmed, and the estimation model (calibration curve) whose validity is confirmed is stored in a memory (not shown) or the like.
成分量推定手段１３０は、有効性が確認された推定モデル（検量線）を用いて、未知の試料の蛍光指紋情報に基づき、前記未知の試料に含有される特定成分の含有量を推定する。なお、未知の試料の蛍光指紋情報に対しても、前記前処理手段１１０による前処理を行うことが望ましい（図１０の蛍光指紋分析による試料の評価・推定装置１００は、そのような構成を採用している）が、必要に応じてこのような前処理を省略することもできる。
The component amount estimating means 130 estimates the content of the specific component contained in the unknown sample based on the fluorescent fingerprint information of the unknown sample using the estimation model (calibration curve) whose effectiveness has been confirmed. It is desirable to perform pre-processing by the pre-processing unit 110 also on the fluorescent fingerprint information of the unknown sample (the sample evaluation / estimation device 100 based on the fluorescent fingerprint analysis of FIG. 10 adopts such a configuration. However, such preprocessing can be omitted as necessary.
以下、蛍光指紋情報に対する前処理、モデル化の際に用いられる多変量解析手法、試料中に微量成分として含まれるクロロゲン酸、スコポレチン、ルチンの含有量の本発明に基づく評価・推定方法について説明する。なた、特に、タバコ原料を試料とした場合について、たばこ原料中のルチン含有量の本発明に基づく評価・推定方法について説明する。
Hereinafter, a description will be given of the preprocessing of the fluorescent fingerprint information, a multivariate analysis method used in modeling, and a method for evaluating and estimating the contents of chlorogenic acid, scopoletin, and rutin contained as trace components in a sample based on the present invention. . A method for estimating and estimating the content of rutin in a tobacco raw material based on the present invention, particularly when a tobacco raw material is used as a sample, will be described.
＜蛍光指紋情報に対する前処理＞
試験試料の蛍光指紋情報を取得する際には、蛍光指紋の計測値（励起波長毎の蛍光スペクトル）をそのまま用いることもできるが、必要に応じて、各種の前処理を行う必要がある。 <Preprocessing for fluorescent fingerprint information>
When acquiring the fluorescent fingerprint information of the test sample, the measured value of the fluorescent fingerprint (fluorescence spectrum for each excitation wavelength) can be used as it is, but it is necessary to perform various pretreatments as necessary.
試験試料の蛍光指紋情報を取得する際には、蛍光指紋の計測値（励起波長毎の蛍光スペクトル）をそのまま用いることもできるが、必要に応じて、各種の前処理を行う必要がある。 <Preprocessing for fluorescent fingerprint information>
When acquiring the fluorescent fingerprint information of the test sample, the measured value of the fluorescent fingerprint (fluorescence spectrum for each excitation wavelength) can be used as it is, but it is necessary to perform various pretreatments as necessary.
本発明の実施の一態様は、特に、前処理において、以下に詳述するような新規な手法（本発明に特有な２次微分処理）を採用することにより、測定精度の向上を図るものである。
One embodiment of the present invention aims to improve measurement accuracy by adopting a novel method (secondary differential processing unique to the present invention) as described in detail below, particularly in preprocessing. is there.
３次元空間における、励起波長（λＥｘ）、蛍光波長（測定波長）（λＥｍ）、蛍光強度（ＩＥｘ，Ｅｍ）の３直交軸を、それぞれ、ｙ軸、ｘ軸、ｚ軸とし、試験試料の測定で得られたＥＥＭ（蛍光指紋）の各ポイントＰi（１≦ｉ≦Ｎ）の座標を（ｘi，ｙi，ｚi）で表す。また、ｘ軸及びｙ軸で決定される平面をｘ－ｙ平面、ｘ軸及びｚ軸で決定される平面をｘ－ｚ平面、ｙ軸及びｚ軸で決定される平面をｙ－ｚ平面と呼ぶ。
The three orthogonal axes of the excitation wavelength (λEx), the fluorescence wavelength (measurement wavelength) (λEm), and the fluorescence intensity (IEx, Em) in the three-dimensional space are set to the y-axis, the x-axis, and the z-axis, respectively, to measure the test sample. The coordinates of each point P i (1 ≦ i ≦ N) of the EEM (fluorescent fingerprint) obtained in (1) are represented by (x i , y i , z i ). A plane determined by the x axis and the y axis is an xy plane, a plane determined by the x axis and the z axis is an xz plane, and a plane determined by the y axis and the z axis is a yz plane. Call.
ｘ－ｙ平面上にｗ軸を取り、ｗ軸と交わりｚ軸と平行な軸をｚ’軸とすると、ｘ－ｙ平面と直交するｗ－ｚ’平面が得られる、このｗ－ｚ’平面を平行移動させ、前記蛍光指紋を切って得られた各断面は、ｚ’＝ｆ（ｗ）という関数で表すことができる。この関数の２次微分値（ｄ2ｚ’/ｄｗ2＝（ｄ2/ｄｗ2）ｆ（ｗ））を求める。求めた２次微分値から、勾配の一番急激な変化点をピークとして求める。
なお、ピークは1つとは限らず、複数存在する可能性があり、この場合には、複数のピークに対応する情報を取得する。 If the w-axis is taken on the xy plane, and the axis that intersects the w-axis and is parallel to the z-axis is the z′-axis, a wz ′ plane orthogonal to the xy plane is obtained. Can be translated and each section obtained by cutting the fluorescent fingerprint can be represented by a function of z ′ = f (w). The second derivative (d 2 z ′ / dw 2 = (d 2 / dw 2 ) f (w)) of this function is obtained. From the obtained secondary differential value, the steepest change point of the gradient is obtained as a peak.
The number of peaks is not limited to one, and there may be a plurality of peaks. In this case, information corresponding to the plurality of peaks is obtained.
なお、ピークは1つとは限らず、複数存在する可能性があり、この場合には、複数のピークに対応する情報を取得する。 If the w-axis is taken on the xy plane, and the axis that intersects the w-axis and is parallel to the z-axis is the z′-axis, a wz ′ plane orthogonal to the xy plane is obtained. Can be translated and each section obtained by cutting the fluorescent fingerprint can be represented by a function of z ′ = f (w). The second derivative (d 2 z ′ / dw 2 = (d 2 / dw 2 ) f (w)) of this function is obtained. From the obtained secondary differential value, the steepest change point of the gradient is obtained as a peak.
The number of peaks is not limited to one, and there may be a plurality of peaks. In this case, information corresponding to the plurality of peaks is obtained.
そして、このような処理を行うことにより、１次微分では検出が困難な、埋もれ易い微かなピークであっても有効に抽出することができる。また１次微分処理と異なり、スペクトル取得時に問題となる波長依存のベースラインの補正を容易に行うことができる。
By performing such a process, it is possible to effectively extract even a faint peak that is difficult to detect by the first derivative and is easily buried. Further, unlike the first-order differentiation processing, it is possible to easily correct a wavelength-dependent baseline which becomes a problem when acquiring a spectrum.
そこで、上述のようなｗ軸の設定をどのように行うべきかが問題となる。試験試料に最適なｗ軸（以下、「最適軸」という）を設定するために、以下に例示するような手法が有効である。
（１）ｗ軸の方向を固定し、ｗ－ｚ’平面を平行移動させると、当該ｗ軸について前記蛍光指紋の断面積（∫ｆ（ｗ）ｄｗ）の最大値が得られる。ｗ軸の方向を変えて、この断面積の最大値が最大となるｗ軸（ｗSmax）を求め、このｗSmaxを最適軸とする。
（２）ｗ軸の方向を固定し、ｗ－ｚ’平面を平行移動させると、当該ｗ軸について前記蛍光指紋の断面の底辺（ｘ－ｙ平面に接する辺）近傍の長さの最大値が得られる。ｗ軸の方向を変えて、この長さの最大値が最大となるｗ軸（ｗLmax）を求め、このｗLmaxを最適軸とする。
（３）蛍光指紋の各ポイントＰi（１≦ｉ≦Ｎ）からの距離の二乗の総和が最小となるような直線（最小二乗直線）を求めると、当該最小二乗直線を含むｗ－ｚ’平面が決定される。そこで、これにより決定されるｗ軸を最適軸とする。
（４）蛍光指紋の各ポイントＰi（１≦ｉ≦Ｎ）の中、前記蛍光指紋の底面近傍のポイント（ｘ－ｙ平面近傍のポイント）からの距離の二乗の総和が最小となるような直線（最小二乗直線）を求めると、当該最小二乗直線を含むｗ－ｚ’平面が決定される。そこで、これにより決定されるｗ軸を最適軸とする。 Therefore, how to set the w-axis as described above becomes a problem. In order to set an optimal w-axis (hereinafter, referred to as an “optimal axis”) for a test sample, a technique exemplified below is effective.
(1) When the direction of the w-axis is fixed and the wz ′ plane is moved in parallel, the maximum value of the cross-sectional area (（f (w) dw) of the fluorescent fingerprint is obtained for the w-axis. By changing the direction of the w-axis, a w-axis (w Smax ) at which the maximum value of the cross-sectional area becomes maximum is determined, and this w Smax is set as the optimum axis.
(2) When the direction of the w-axis is fixed and the wz ′ plane is moved in parallel, the maximum value of the length near the bottom (side in contact with the xy plane) of the cross section of the fluorescent fingerprint with respect to the w-axis becomes can get. By changing the direction of the w-axis, a w-axis (w Lmax ) at which the maximum value of the length is maximum is determined, and this w Lmax is set as the optimum axis.
(3) When a straight line (least square line) that minimizes the sum of squares of distances from each point P i (1 ≦ i ≦ N) of the fluorescent fingerprint is obtained, wz ′ including the least square line is obtained. The plane is determined. Therefore, the w axis determined by this is set as the optimal axis.
(4) Among the points P i (1 ≦ i ≦ N) of the fluorescent fingerprint, the sum of squares of distances from points near the bottom surface (points near the xy plane) of the fluorescent fingerprint is minimized. When a straight line (least square line) is obtained, a wz 'plane including the least square line is determined. Therefore, the w axis determined by this is set as the optimal axis.
（１）ｗ軸の方向を固定し、ｗ－ｚ’平面を平行移動させると、当該ｗ軸について前記蛍光指紋の断面積（∫ｆ（ｗ）ｄｗ）の最大値が得られる。ｗ軸の方向を変えて、この断面積の最大値が最大となるｗ軸（ｗSmax）を求め、このｗSmaxを最適軸とする。
（２）ｗ軸の方向を固定し、ｗ－ｚ’平面を平行移動させると、当該ｗ軸について前記蛍光指紋の断面の底辺（ｘ－ｙ平面に接する辺）近傍の長さの最大値が得られる。ｗ軸の方向を変えて、この長さの最大値が最大となるｗ軸（ｗLmax）を求め、このｗLmaxを最適軸とする。
（３）蛍光指紋の各ポイントＰi（１≦ｉ≦Ｎ）からの距離の二乗の総和が最小となるような直線（最小二乗直線）を求めると、当該最小二乗直線を含むｗ－ｚ’平面が決定される。そこで、これにより決定されるｗ軸を最適軸とする。
（４）蛍光指紋の各ポイントＰi（１≦ｉ≦Ｎ）の中、前記蛍光指紋の底面近傍のポイント（ｘ－ｙ平面近傍のポイント）からの距離の二乗の総和が最小となるような直線（最小二乗直線）を求めると、当該最小二乗直線を含むｗ－ｚ’平面が決定される。そこで、これにより決定されるｗ軸を最適軸とする。 Therefore, how to set the w-axis as described above becomes a problem. In order to set an optimal w-axis (hereinafter, referred to as an “optimal axis”) for a test sample, a technique exemplified below is effective.
(1) When the direction of the w-axis is fixed and the wz ′ plane is moved in parallel, the maximum value of the cross-sectional area (（f (w) dw) of the fluorescent fingerprint is obtained for the w-axis. By changing the direction of the w-axis, a w-axis (w Smax ) at which the maximum value of the cross-sectional area becomes maximum is determined, and this w Smax is set as the optimum axis.
(2) When the direction of the w-axis is fixed and the wz ′ plane is moved in parallel, the maximum value of the length near the bottom (side in contact with the xy plane) of the cross section of the fluorescent fingerprint with respect to the w-axis becomes can get. By changing the direction of the w-axis, a w-axis (w Lmax ) at which the maximum value of the length is maximum is determined, and this w Lmax is set as the optimum axis.
(3) When a straight line (least square line) that minimizes the sum of squares of distances from each point P i (1 ≦ i ≦ N) of the fluorescent fingerprint is obtained, wz ′ including the least square line is obtained. The plane is determined. Therefore, the w axis determined by this is set as the optimal axis.
(4) Among the points P i (1 ≦ i ≦ N) of the fluorescent fingerprint, the sum of squares of distances from points near the bottom surface (points near the xy plane) of the fluorescent fingerprint is minimized. When a straight line (least square line) is obtained, a wz 'plane including the least square line is determined. Therefore, the w axis determined by this is set as the optimal axis.
なお、最適軸の設定は上記手法に限られるものではなく、例えば、上記手法により得られたｗ軸を暫定的な最適軸として、この暫定的な最適軸をｘ－ｙ平面上で適宜回転して、前記試験試料に特有な前記蛍光強度のピーク値が現れるようにｗ軸を設定するようにしてもよい。
また、試験試料によっては、最適軸をｘ軸、又は、ｙ軸に平行な軸として設定する簡素化した手法を採用することもできる。 The setting of the optimal axis is not limited to the above-described method. For example, the w-axis obtained by the above-described method is set as a temporary optimal axis, and the temporary optimal axis is appropriately rotated on the xy plane. Then, the w-axis may be set so that the peak value of the fluorescence intensity unique to the test sample appears.
Further, depending on the test sample, a simplified method of setting the optimum axis as an axis parallel to the x-axis or the y-axis may be employed.
また、試験試料によっては、最適軸をｘ軸、又は、ｙ軸に平行な軸として設定する簡素化した手法を採用することもできる。 The setting of the optimal axis is not limited to the above-described method. For example, the w-axis obtained by the above-described method is set as a temporary optimal axis, and the temporary optimal axis is appropriately rotated on the xy plane. Then, the w-axis may be set so that the peak value of the fluorescence intensity unique to the test sample appears.
Further, depending on the test sample, a simplified method of setting the optimum axis as an axis parallel to the x-axis or the y-axis may be employed.
なお、特許文献２には、２次元展開された蛍光指紋に対する信号処理演算として２次微分が例示されているが（段落〔００２６〕等参照）、本発明の上記最適軸に係る技術思想については、記載も示唆もされていないことに留意されたい。
Note that Patent Document 2 exemplifies a second derivative as a signal processing operation for a two-dimensionally developed fluorescent fingerprint (see paragraph [0026], etc.). Note that neither is described nor suggested.
上記本発明に特有な２次微分処理の外に、計測された蛍光指紋からノイズを除去して有効な蛍光指紋情報を得るために、非蛍光成分の除去処理、散乱光の除去処理、低感度領域の削除処理の中の１つ又は複数の組合せを前処理として採用することができる。また、取得された蛍光指紋情報に対し、中心化、規格化、標準化、ベースライン補正、平滑化、オートスケーリング、対数変換（Log10）等の演算処理の中の１つ又は複数の組合せを前処理として採用することもできる。さらに、従前使用されている２次微分処理を併せて使用することもできる。
In addition to the above-described second differentiation processing unique to the present invention, in order to remove noise from the measured fluorescent fingerprint and obtain effective fluorescent fingerprint information, non-fluorescent component removal processing, scattered light removal processing, low sensitivity One or a combination of the area deletion processing can be adopted as the preprocessing. In addition, one or more combinations of arithmetic processing such as centering, standardization, standardization, baseline correction, smoothing, auto scaling, logarithmic conversion (Log10) are pre-processed on the acquired fluorescent fingerprint information. It can also be adopted as. Furthermore, the secondary differential processing that has been used before can also be used together.
なお、上述の本発明に特有な２次微分処理とそれ以外の前処理を併用する場合に、その適用順序は適宜設定可能であるが、処理の効率化の観点から、非蛍光成分の除去処理、散乱光の除去処理、低感度領域の削除処理等の処理を先行させることが望ましい。
In the case where the above-described second derivative processing unique to the present invention and other pre-processing are used together, the order of application can be set as appropriate. However, from the viewpoint of processing efficiency, non-fluorescent component removal processing is performed. It is desirable to precede processing such as scattered light removal processing and low-sensitivity area deletion processing.
＜モデル化の際に用いられる多変量解析手法＞
モデル化の際に用いられる多変量解析手法として、ＰＬＳ（Partial Least Squares）回帰分析、重回帰分析、主成分回帰分析、最小二乗法等の各種の解析手法を用いることができる。 <Multivariate analysis method used for modeling>
As a multivariate analysis method used in modeling, various analysis methods such as PLS (Partial Least Squares) regression analysis, multiple regression analysis, principal component regression analysis, and least square method can be used.
モデル化の際に用いられる多変量解析手法として、ＰＬＳ（Partial Least Squares）回帰分析、重回帰分析、主成分回帰分析、最小二乗法等の各種の解析手法を用いることができる。 <Multivariate analysis method used for modeling>
As a multivariate analysis method used in modeling, various analysis methods such as PLS (Partial Least Squares) regression analysis, multiple regression analysis, principal component regression analysis, and least square method can be used.
ＰＬＳ回帰分析は、主成分と目的変数との共分散が最大になるように主成分を抽出する手法であり、説明変数の間に強い相関がある場合（多重共線性を生ずる場合）に有効な手法である。
PLS regression analysis is a method of extracting a principal component so that the covariance between the principal component and the objective variable is maximized, and is effective when there is a strong correlation between explanatory variables (when multicollinearity occurs). Method.
主成分回帰分析は、主成分の分散が最大になるように主成分を抽出する手法であり、説明変数のみを用いて主成分分析を行い、得られた主成分と目的変数との間で最小二乗法による重回帰分析を行うものである。
重回帰分析は、説明変数と目的変数との間で最小二乗法を適用するものであり、主成分回帰分析とは異なる特徴を有するものである。 Principal component regression analysis is a method of extracting principal components so that the variance of the principal components is maximized.Principal component analysis is performed using only explanatory variables, and the minimum between the obtained principal components and the objective variable is calculated. This is to perform multiple regression analysis by the square method.
The multiple regression analysis applies the least squares method between the explanatory variable and the objective variable, and has different characteristics from the principal component regression analysis.
重回帰分析は、説明変数と目的変数との間で最小二乗法を適用するものであり、主成分回帰分析とは異なる特徴を有するものである。 Principal component regression analysis is a method of extracting principal components so that the variance of the principal components is maximized.Principal component analysis is performed using only explanatory variables, and the minimum between the obtained principal components and the objective variable is calculated. This is to perform multiple regression analysis by the square method.
The multiple regression analysis applies the least squares method between the explanatory variable and the objective variable, and has different characteristics from the principal component regression analysis.
なお、上述の各解析手法自体は周知であり、本発明のモデル化の際に特殊な処理を要請するものでもないので処理内容の詳細は省略するが、ＰＬＳに関しては、検量線の作成との関連で後述する。
The above-described analysis methods are well-known, and do not require any special processing when modeling the present invention. Therefore, details of the processing contents are omitted. It will be described later in connection with the present invention.
＜試料中のクロロゲン酸、スコポルチン、ルチンの含有量の推定への適用例＞
次に、本発明の手法を試料中の微量成分（クロロゲン酸、スコポルチン、ルチン）の含有量の推定に適用した適用例について詳述する。 <Example of application to estimation of the contents of chlorogenic acid, scoportin, and rutin in a sample>
Next, application examples in which the method of the present invention is applied to estimating the content of trace components (chlorogenic acid, scoportin, rutin) in a sample will be described in detail.
次に、本発明の手法を試料中の微量成分（クロロゲン酸、スコポルチン、ルチン）の含有量の推定に適用した適用例について詳述する。 <Example of application to estimation of the contents of chlorogenic acid, scoportin, and rutin in a sample>
Next, application examples in which the method of the present invention is applied to estimating the content of trace components (chlorogenic acid, scoportin, rutin) in a sample will be described in detail.
これらの微量成分（クロロゲン酸、スコポルチン、ルチン）はポリフェノールの一種であり、蕎麦や果皮に多く含まれる栄養素の１つとして知られており、抗酸化作用等を有することから、食品や化粧品等様々な業界から注目されている。
These trace components (chlorogenic acid, scoportin, rutin) are a kind of polyphenols, and are known as one of the nutrients often contained in buckwheat and pericarp. Industry has attracted attention.
これらの微量成分の定量方法としては、試料中のこれらの微量成分を抽出液で抽出し、これを高速液体クロマトグラフ（HPLC）で定量する方法が知られているが、この方法では分析結果を得るまでに労力と時間を要するという問題がある。
As a method for quantifying these trace components, a method is known in which these trace components in a sample are extracted with an extract and quantified by high performance liquid chromatography (HPLC). There is a problem that it takes labor and time to obtain.
これらの微量成分は、試料中に化学構造が類似したポリフェノールとして混在して存在する場合が多い（例えば、たばこ原料等）。このような場合、化学構造が類似したポリフェノールは類似した蛍光指紋を有するため、蛍光指紋法を適用しても、特定の微量成分だけを抜き出して正確に定量することには困難性が予想される。
These trace components are often present in the sample as polyphenols having similar chemical structures (for example, tobacco raw materials). In such a case, since polyphenols having similar chemical structures have similar fluorescent fingerprints, even if the fluorescent fingerprint method is applied, it is expected that it will be difficult to extract only a specific trace component and accurately quantify it. .
上記のような化学構造が類似した微量成分の定量に関する困難性に係る事情等を、図５、図６Ａ～図６Ｃに基づき説明する。
(5) The circumstances related to the difficulty in quantifying a trace component having a similar chemical structure as described above will be described with reference to FIGS. 5 and 6A to 6C.
図５は、クロロゲン酸（１０ｐｐｍ）、ルチン（１ｐｐｍ）、スコポレチン（０．１ｐｐｍ）を含む混合サンプルの蛍光指紋を表している。他方、図６Ａ、図６Ｂ、図６Ｃは、混合サンプルに含まれる、スコポレチン（１ｐｐｍ：（Ａ））、クロロゲン酸（１ｐｐｍ：（Ｂ））、ルチン（１ｐｐｍ：（Ｃ））のそれぞれを単独で含む試験試料の蛍光指紋を表している。以下の表１は、前記混合サンプル、並びに、スコポレチン（１ｐｐｍ）、クロロゲン酸（１ｐｐｍ）、ルチン（１ｐｐｍ）の蛍光指紋のピーク位置近傍の波長（励起波長／蛍光波長）、及び、対応する蛍光強度を纏めたものである。
FIG. 5 shows a fluorescent fingerprint of a mixed sample containing chlorogenic acid (10 ppm), rutin (1 ppm), and scopoletin (0.1 ppm). On the other hand, FIG. 6A, FIG. 6B, and FIG. 6C show that each of scopoletin (1 ppm: (A)), chlorogenic acid (1 ppm: (B)), and rutin (1 ppm: (C)) contained in the mixed sample was used alone. 5 shows a fluorescent fingerprint of a test sample containing the sample. Table 1 below shows the mixed sample, the wavelength (excitation wavelength / fluorescence wavelength) near the peak position of the fluorescent fingerprint of scopoletin (1 ppm), chlorogenic acid (1 ppm), and rutin (1 ppm), and the corresponding fluorescence intensity. It is a summary of.
図５及び図６Ａ～図６Ｃ並びに表１から、概略、以下のような知見が得られる。
・混合物の蛍光指紋において、ルチンのピーク位置近傍の波長はスコポレチンと一致している（図５の符号「Ａ」を付した箇所参照）。
・混合物の蛍光指紋において、クロロゲン酸にみられるピーク位置（図５の符号「Ｂ」を付した箇所参照）が現れている。
・混合物の蛍光指紋において、ルチンに関する特徴は、一見したところでは視認できない。
・試料中の含有量が１ｐｐｍの場合、蛍光強度は、スコポレチンとルチンで３桁相違する。
・但し、蛍光指紋のパターンは、スコポレチン、クロロゲン酸、ルチンでわずかに異なる。 From FIG. 5 and FIGS. 6A to 6C and Table 1, the following findings are obtained.
-In the fluorescent fingerprint of the mixture, the wavelength near the peak position of rutin coincides with that of scopoletin (see the portion indicated by the symbol "A" in Fig. 5).
-In the fluorescent fingerprint of the mixture, the peak position (see the portion indicated by the symbol "B" in Fig. 5) observed in chlorogenic acid appears.
-In the fluorescent fingerprint of the mixture, the features relating to rutin are not visible at first glance.
-When the content in the sample is 1 ppm, the fluorescence intensity differs between scopoletin and rutin by three orders of magnitude.
-However, the pattern of the fluorescent fingerprint is slightly different for scopoletin, chlorogenic acid, and rutin.
・混合物の蛍光指紋において、ルチンのピーク位置近傍の波長はスコポレチンと一致している（図５の符号「Ａ」を付した箇所参照）。
・混合物の蛍光指紋において、クロロゲン酸にみられるピーク位置（図５の符号「Ｂ」を付した箇所参照）が現れている。
・混合物の蛍光指紋において、ルチンに関する特徴は、一見したところでは視認できない。
・試料中の含有量が１ｐｐｍの場合、蛍光強度は、スコポレチンとルチンで３桁相違する。
・但し、蛍光指紋のパターンは、スコポレチン、クロロゲン酸、ルチンでわずかに異なる。 From FIG. 5 and FIGS. 6A to 6C and Table 1, the following findings are obtained.
-In the fluorescent fingerprint of the mixture, the wavelength near the peak position of rutin coincides with that of scopoletin (see the portion indicated by the symbol "A" in Fig. 5).
-In the fluorescent fingerprint of the mixture, the peak position (see the portion indicated by the symbol "B" in Fig. 5) observed in chlorogenic acid appears.
-In the fluorescent fingerprint of the mixture, the features relating to rutin are not visible at first glance.
-When the content in the sample is 1 ppm, the fluorescence intensity differs between scopoletin and rutin by three orders of magnitude.
-However, the pattern of the fluorescent fingerprint is slightly different for scopoletin, chlorogenic acid, and rutin.
そこで、スコポレチン、クロロゲン酸、ルチン間でわずかに異なる蛍光指紋のパターンに着目して、試料中のそれぞれの含有量を推定することを可能にすることが考えられる。本適用例は、取得された蛍光指紋の前処理において、前述のような新規な手法を採用することにより、簡便かつ迅速に、試料中のスコポレチン、クロロゲン酸、ルチンの含有量を推定することができるものである。
Therefore, it is conceivable that it is possible to estimate the content of each sample in the sample by focusing on the pattern of the fluorescent fingerprint slightly different among scopoletin, chlorogenic acid, and rutin. In this application example, in the pre-processing of the obtained fluorescent fingerprints, the content of scopoletin, chlorogenic acid, and rutin in the sample can be easily and quickly estimated by employing the above-described novel method. You can do it.
以下、本発明の実施の一態様の各工程について説明する。
〔試験試料の準備〕
＜ルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合＞
ルチン・クロロゲン酸・スコポレチンの標品を様々な割合で混合した４０サンプル（溶媒は５０％エタノール／水溶液）を準備した。各サンプルについては、粒径１ｍｍ以下に粉砕し十分に混合したものを試験試料として用意した。このような粉砕・混合により、ルチン・クロロゲン酸・スコポレチンが原料中に一様に分布せず局在しているようなケースであっても、測定の精度を担保するようにした。 Hereinafter, each step of one embodiment of the present invention will be described.
[Preparation of test sample]
<When using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a sample>
Forty samples (solvent: 50% ethanol / water solution) prepared by mixing rutin / chlorogenic acid / scopoletin samples at various ratios were prepared. Each sample was prepared as a test sample by pulverizing to a particle size of 1 mm or less and sufficiently mixed. By such pulverization and mixing, the accuracy of measurement is ensured even in a case where rutin, chlorogenic acid, and scopoletin are not uniformly distributed but localized in the raw material.
〔試験試料の準備〕
＜ルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合＞
ルチン・クロロゲン酸・スコポレチンの標品を様々な割合で混合した４０サンプル（溶媒は５０％エタノール／水溶液）を準備した。各サンプルについては、粒径１ｍｍ以下に粉砕し十分に混合したものを試験試料として用意した。このような粉砕・混合により、ルチン・クロロゲン酸・スコポレチンが原料中に一様に分布せず局在しているようなケースであっても、測定の精度を担保するようにした。 Hereinafter, each step of one embodiment of the present invention will be described.
[Preparation of test sample]
<When using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a sample>
Forty samples (solvent: 50% ethanol / water solution) prepared by mixing rutin / chlorogenic acid / scopoletin samples at various ratios were prepared. Each sample was prepared as a test sample by pulverizing to a particle size of 1 mm or less and sufficiently mixed. By such pulverization and mixing, the accuracy of measurement is ensured even in a case where rutin, chlorogenic acid, and scopoletin are not uniformly distributed but localized in the raw material.
また、試験試料は事前に水分量を一定化するために蔵置されたものを用いた。例えばたばこ原料の場合、調和条件（22度60%の室内）で２４時間以上蔵置することが好ましい。このように測定前に水分量を一定化させておくことで蛍光のピークシフトが起こりにくくなる。
試 験 In addition, the test samples used were stored in advance to stabilize the water content. For example, in the case of tobacco raw materials, it is preferable to store them in harmony conditions (room at 22 degrees 60%) for 24 hours or more. By keeping the water content constant before the measurement, the peak shift of the fluorescence hardly occurs.
＜ルチン含有量の測定にたばこ原料を試料として用いる場合＞
ルチン含有量が既知の各サンプルについては、粒径１ｍｍ以下に粉砕し十分に混合したものを試験試料として用意した。 <When tobacco raw material is used as a sample for measurement of rutin content>
Each sample having a known rutin content was pulverized to a particle size of 1 mm or less and sufficiently mixed to prepare a test sample.
ルチン含有量が既知の各サンプルについては、粒径１ｍｍ以下に粉砕し十分に混合したものを試験試料として用意した。 <When tobacco raw material is used as a sample for measurement of rutin content>
Each sample having a known rutin content was pulverized to a particle size of 1 mm or less and sufficiently mixed to prepare a test sample.
試料がたばこ原料であった場合、ルチンはたばこ原料中に均一に存在するのではなく局在することがわかっていることから、このように、測定前に試料を一定粒径（１ｍｍ径）以下に粉砕し、十分に混合してから蛍光指紋を取得することが好ましいものである。なお、各サンプルのルチン含有量については、事前に高速液体クロマトグラフ（HPLC）により定量しておいた。
If the sample is a tobacco raw material, it is known that rutin is not uniformly present but localized in the tobacco raw material. Thus, before the measurement, the sample is reduced to a certain particle size (1 mm diameter) or less. It is preferable to obtain a fluorescent fingerprint after pulverizing the mixture and thoroughly mixing the mixture. In addition, the rutin content of each sample was determined in advance by high performance liquid chromatography (HPLC).
また、試験試料は事前に水分量を一定化するために蔵置されたものを用いた。たばこ原料の場合は、調和条件（22度60%の室内）で２４時間以上蔵置することが好ましい。このように測定前に水分量を一定化させておくことで蛍光のピークシフトが起こりにくくなる。
試 験 In addition, the test samples used were stored in advance to stabilize the water content. In the case of tobacco raw materials, it is preferable to store them in harmony conditions (room at 22 degrees 60%) for 24 hours or more. By keeping the water content constant before the measurement, the peak shift of the fluorescence hardly occurs.
但し、ルチンに関しては溶かせる溶媒が限られているため、固体のまま測定するのが望ましい場合もあり、たばこ原料中のルチンの含有量を測定（推定）するために固体のままで蛍光指紋情報を取得する場合もある。
However, it is sometimes desirable to measure rutin in a solid state because the solvent in which it can be dissolved is limited. In order to measure (estimate) the content of rutin in tobacco raw materials, it is necessary to measure fluorescent fingerprint information in a solid state. You may also get
〔蛍光指紋情報の取得〕
試験試料の蛍光指紋情報を取得するために、蛍光指紋測定装置として、日立ハイテクサイエンス社製F-7000を用い、反射法（FrontFace）により測定を行った。 [Acquisition of fluorescent fingerprint information]
In order to acquire the fluorescent fingerprint information of the test sample, measurement was performed by a reflection method (FrontFace) using F-7000 manufactured by Hitachi High-Tech Science as a fluorescent fingerprint measuring device.
試験試料の蛍光指紋情報を取得するために、蛍光指紋測定装置として、日立ハイテクサイエンス社製F-7000を用い、反射法（FrontFace）により測定を行った。 [Acquisition of fluorescent fingerprint information]
In order to acquire the fluorescent fingerprint information of the test sample, measurement was performed by a reflection method (FrontFace) using F-7000 manufactured by Hitachi High-Tech Science as a fluorescent fingerprint measuring device.
測定条件は、励起光２００－６００nm, 蛍光２００－７００nm, 分解能５nm, スリット幅５nm，フォトマル感度７００Vであった。なお、分解能５nmを考慮すれば、測定波長は少なくとも５nm程度の誤差を許容するものである。
The measurement conditions were as follows: excitation light 200-600 nm, fluorescence 200-700 nm, resolution 5 nm, slit width 5 nm, photomultiplier sensitivity 700 V. Considering a resolution of 5 nm, the measurement wavelength allows at least an error of about 5 nm.
〔蛍光指紋情報に対する前処理〕
＜ルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合＞
ルチン・クロロゲン酸・スコポレチンの標品を様々な割合で混合した試験試料から取得された蛍光指紋情報に対して前処理を行う。この前処理には、例えば、MatlabやPLS_toolbox等の専用ソフトウエアが使用される。前処理としては、前述の＜蛍光指紋情報に対する前処理＞において詳述した本発明に特有な２次微分処理の外に、好ましくは、各スペクトルに対する従前使用されている２次微分処理や、成分情報に寄与しない波長を除去する処理が挙げられる。因みに、成分情報に寄与しない波長を除去する処理として、例えば、以下のような手法を採用し得るが、各処理手法自体は既知であり、その詳細についての説明は省く。
（ａ）Variable important projection(VIP)
（ｂ）interval PLS(iPLS)
（ｃ）Genetic algorithms(GA)
（ｄ）Jack-knife分析
（ｅ）Forward interval PLS
（ｆ）Backward interval PLS(biPLS)
（ｇ）Synergy interval PLS(siPLS)
（ｈ）LASSO type method [Preprocessing for fluorescent fingerprint information]
<When using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a sample>
Preprocessing is performed on fluorescent fingerprint information obtained from test samples in which samples of rutin, chlorogenic acid, and scopoletin are mixed at various ratios. For this preprocessing, for example, dedicated software such as Matlab or PLS_toolbox is used. As the pre-processing, in addition to the second-order differentiation processing specific to the present invention described in detail in the above-mentioned <Pre-processing for fluorescent fingerprint information>, preferably, the second-order differentiation processing previously used for each spectrum and the components are preferably used. There is a process of removing a wavelength that does not contribute to information. Incidentally, as a process for removing a wavelength that does not contribute to the component information, for example, the following method may be adopted. However, each processing method itself is known, and a detailed description thereof will be omitted.
(A) Variable important projection (VIP)
(B) interval PLS (iPLS)
(C) Genetic algorithms (GA)
(D) Jack-knife analysis (e) Forward interval PLS
(F) Backward interval PLS (biPLS)
(G) Synergy interval PLS (siPLS)
(H) LASSO type method
＜ルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合＞
ルチン・クロロゲン酸・スコポレチンの標品を様々な割合で混合した試験試料から取得された蛍光指紋情報に対して前処理を行う。この前処理には、例えば、MatlabやPLS_toolbox等の専用ソフトウエアが使用される。前処理としては、前述の＜蛍光指紋情報に対する前処理＞において詳述した本発明に特有な２次微分処理の外に、好ましくは、各スペクトルに対する従前使用されている２次微分処理や、成分情報に寄与しない波長を除去する処理が挙げられる。因みに、成分情報に寄与しない波長を除去する処理として、例えば、以下のような手法を採用し得るが、各処理手法自体は既知であり、その詳細についての説明は省く。
（ａ）Variable important projection(VIP)
（ｂ）interval PLS(iPLS)
（ｃ）Genetic algorithms(GA)
（ｄ）Jack-knife分析
（ｅ）Forward interval PLS
（ｆ）Backward interval PLS(biPLS)
（ｇ）Synergy interval PLS(siPLS)
（ｈ）LASSO type method [Preprocessing for fluorescent fingerprint information]
<When using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a sample>
Preprocessing is performed on fluorescent fingerprint information obtained from test samples in which samples of rutin, chlorogenic acid, and scopoletin are mixed at various ratios. For this preprocessing, for example, dedicated software such as Matlab or PLS_toolbox is used. As the pre-processing, in addition to the second-order differentiation processing specific to the present invention described in detail in the above-mentioned <Pre-processing for fluorescent fingerprint information>, preferably, the second-order differentiation processing previously used for each spectrum and the components are preferably used. There is a process of removing a wavelength that does not contribute to information. Incidentally, as a process for removing a wavelength that does not contribute to the component information, for example, the following method may be adopted. However, each processing method itself is known, and a detailed description thereof will be omitted.
(A) Variable important projection (VIP)
(B) interval PLS (iPLS)
(C) Genetic algorithms (GA)
(D) Jack-knife analysis (e) Forward interval PLS
(F) Backward interval PLS (biPLS)
(G) Synergy interval PLS (siPLS)
(H) LASSO type method
本発明に特有な２次微分処理に関しては、前述の＜蛍光指紋情報に対する前処理＞において詳述した手法を適用するが、本混合試料のケースでみれば、ルチン、クロロゲン酸、スコポレチンの蛍光指紋情報に対して、ｘ軸（蛍光波長軸）に平行な軸に沿った２次微分処理を行うことにより、有効な結果を得ることができる。
For the second derivative processing unique to the present invention, the method described in the above-mentioned <Preprocessing for fluorescent fingerprint information> is applied. However, in the case of the present mixed sample, the fluorescent fingerprints of rutin, chlorogenic acid, and scopoletin are used. An effective result can be obtained by performing a second derivative process on the information along an axis parallel to the x-axis (fluorescence wavelength axis).
＜ルチン含有量の測定のためにたばこ原料を試料として用いる場合＞
基本的に、前述したルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合と同様の前処理を行えばよい。 <When tobacco raw material is used as a sample for measurement of rutin content>
Basically, the same pretreatment as in the case of using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a standard as described above may be performed.
基本的に、前述したルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合と同様の前処理を行えばよい。 <When tobacco raw material is used as a sample for measurement of rutin content>
Basically, the same pretreatment as in the case of using a sample in which rutin, chlorogenic acid, and scopoletin are mixed as a standard as described above may be performed.
〔検量線の作成・検証〕
検量線は、具体的には、取得された蛍光指紋情報を説明変数、既知の成分（ルチン、クロロゲン酸、スコポレチン）の含有量を目的変数とし、ＰＬＳ回帰分析（以下、単に「ＰＬＳ」ということもある）を使用して作成する。 [Creation and verification of calibration curve]
Specifically, the calibration curve uses the acquired fluorescent fingerprint information as an explanatory variable and the content of known components (rutin, chlorogenic acid, scopoletin) as an objective variable, and performs PLS regression analysis (hereinafter simply referred to as “PLS”). There is also).
検量線は、具体的には、取得された蛍光指紋情報を説明変数、既知の成分（ルチン、クロロゲン酸、スコポレチン）の含有量を目的変数とし、ＰＬＳ回帰分析（以下、単に「ＰＬＳ」ということもある）を使用して作成する。 [Creation and verification of calibration curve]
Specifically, the calibration curve uses the acquired fluorescent fingerprint information as an explanatory variable and the content of known components (rutin, chlorogenic acid, scopoletin) as an objective variable, and performs PLS regression analysis (hereinafter simply referred to as “PLS”). There is also).
ＰＬＳ回帰分析に関し、その概略を簡潔に説明しておく。
ＰＬＳでは、説明変数Ｘ（行列）と目的変数ｙ（ベクトル）は、以下の二つの基本式（１）、（２）を満たしている。
Ｘ＝ＴＰT＋Ｅ （１）
ｙ＝Ｔｑ＋ｆ （２）
ここで、Ｔは潜在変数（行列）、Ｐはローディング（行列）、Ｅは説明変数Ｘの残差（行列）、ｑは係数（ベクトル）、ｆは目的変数の残差（ベクトル）、ＰTはＰの転置行列である。 The outline of the PLS regression analysis will be briefly described.
In PLS, an explanatory variable X (matrix) and an objective variable y (vector) satisfy the following two basic expressions (1) and (2).
X = TP T + E (1)
y = Tq + f (2)
Here, T is a latent variable (matrix), P is loading (matrix), E is a residual (matrix) of the explanatory variable X, q is a coefficient (vector), f is a residual of the objective variable (vector), P T Is the transposed matrix of P.
ＰＬＳでは、説明変数Ｘ（行列）と目的変数ｙ（ベクトル）は、以下の二つの基本式（１）、（２）を満たしている。
Ｘ＝ＴＰT＋Ｅ （１）
ｙ＝Ｔｑ＋ｆ （２）
ここで、Ｔは潜在変数（行列）、Ｐはローディング（行列）、Ｅは説明変数Ｘの残差（行列）、ｑは係数（ベクトル）、ｆは目的変数の残差（ベクトル）、ＰTはＰの転置行列である。 The outline of the PLS regression analysis will be briefly described.
In PLS, an explanatory variable X (matrix) and an objective variable y (vector) satisfy the following two basic expressions (1) and (2).
X = TP T + E (1)
y = Tq + f (2)
Here, T is a latent variable (matrix), P is loading (matrix), E is a residual (matrix) of the explanatory variable X, q is a coefficient (vector), f is a residual of the objective variable (vector), P T Is the transposed matrix of P.
因みに、ＰＬＳは、説明変数Ｘの情報を目的変数ｙのモデリングに直接用いるのではなく、説明変数Ｘの情報の一部を潜在変数ｔに変換し、潜在変数ｔを用いて目的変数ｙをモデリングするものである。なお、潜在変数の数は、例えば、クロスバリデーションによる予測的説明分散値を指標として決定することができる。また、潜在変数は、主成分と呼ばれることもある。
Incidentally, the PLS does not directly use the information of the explanatory variable X for modeling the objective variable y, but converts part of the information of the explanatory variable X into a latent variable t, and models the objective variable y using the latent variable t. Is what you do. The number of latent variables can be determined, for example, using a predictive explanatory variance value obtained by cross-validation as an index. Latent variables may also be called principal components.
特に、１成分モデルの場合には、上記の（１）、（２）は、以下の（３）、（４）で表される。
Ｘ＝ｔ1ｐ1 T＋Ｅ （３）
ｙ＝ｔ1ｑ1＋ｆ （４）
ここで、ｔ1は潜在変数（ベクトル）、ｐ1はローディング（ベクトル）、ｑ1は係数（スカラー）である。 In particular, in the case of the one-component model, the above (1) and (2) are represented by the following (3) and (4).
X = t 1 p 1 T + E (3)
y = t 1 q 1 + f (4)
Here, t 1 is a latent variable (vector), p 1 is loading (vector), and q 1 is a coefficient (scalar).
Ｘ＝ｔ1ｐ1 T＋Ｅ （３）
ｙ＝ｔ1ｑ1＋ｆ （４）
ここで、ｔ1は潜在変数（ベクトル）、ｐ1はローディング（ベクトル）、ｑ1は係数（スカラー）である。 In particular, in the case of the one-component model, the above (1) and (2) are represented by the following (3) and (4).
X = t 1 p 1 T + E (3)
y = t 1 q 1 + f (4)
Here, t 1 is a latent variable (vector), p 1 is loading (vector), and q 1 is a coefficient (scalar).
今、ｔ1がＸの線形結合で表されると仮定すると、以下の（５）が成立する。
ｔ1＝Ｘｗ1 （５）
ここで、ｗ1は規格化された重みベクトルである。 Now, assuming that t 1 is represented by a linear combination of X, the following (5) holds.
t 1 = Xw 1 (5)
Here, w 1 is a standardized weight vector.
ｔ1＝Ｘｗ1 （５）
ここで、ｗ1は規格化された重みベクトルである。 Now, assuming that t 1 is represented by a linear combination of X, the following (5) holds.
t 1 = Xw 1 (5)
Here, w 1 is a standardized weight vector.
ＰＬＳは、ｙとｔ1との共分散ｙTｔ1を、ｗ1のノルムが１（｜ｗ1｜＝１）という条件下で最大化するようなｔ1を求めるものであり、ｔ1の算出には、所謂ラグランジュの未定乗数法を用いればよい。ラグランジュの未定乗数法を用いた計算手法は周知であるから、計算の詳細は省略し、ｗ1、ｐ1、ｑ1に関する計算結果のみ、以下の（６）～（８）として示す。
ｗ1＝ＸTｙ／｜ＸTｙ｜ （６）
ｐ1＝ＸTｔ1／ｔ1 Tｔ1 （７）
ｑ1＝ｙＴｔ１／ｔ1 Tｔ1 （８）
なお、（７）、（８）式のｔ1は、（６）式で求めたｗ1を（５）式に代入することにより算出されたベクトルである。 PLS is a covariance y T t 1 between y and t 1, the norm of w 1 is 1 (| w 1 | = 1 ) is intended to determine the t 1 that maximizes under the condition that, t 1 May be calculated using the so-called Lagrange undetermined multiplier method. Since the calculation method using the Lagrange's undetermined multiplier method is well known, the details of the calculation are omitted, and only the calculation results regarding w 1 , p 1 , and q 1 are shown as (6) to (8) below.
w 1 = X T y / | X T y | (6)
p 1 = X T t 1 / t 1 T t 1 (7)
q 1 = yTt1 / t 1 T t 1 (8)
Note that t 1 in the equations (7) and (8) is a vector calculated by substituting w 1 obtained in the equation (6) into the equation (5).
ｗ1＝ＸTｙ／｜ＸTｙ｜ （６）
ｐ1＝ＸTｔ1／ｔ1 Tｔ1 （７）
ｑ1＝ｙＴｔ１／ｔ1 Tｔ1 （８）
なお、（７）、（８）式のｔ1は、（６）式で求めたｗ1を（５）式に代入することにより算出されたベクトルである。 PLS is a covariance y T t 1 between y and t 1, the norm of w 1 is 1 (| w 1 | = 1 ) is intended to determine the t 1 that maximizes under the condition that, t 1 May be calculated using the so-called Lagrange undetermined multiplier method. Since the calculation method using the Lagrange's undetermined multiplier method is well known, the details of the calculation are omitted, and only the calculation results regarding w 1 , p 1 , and q 1 are shown as (6) to (8) below.
w 1 = X T y / | X T y | (6)
p 1 = X T t 1 / t 1 T t 1 (7)
q 1 = yTt1 / t 1 T t 1 (8)
Note that t 1 in the equations (7) and (8) is a vector calculated by substituting w 1 obtained in the equation (6) into the equation (5).
多成分モデルについても、同様の手法で計算できるが、計算手法は周知であるから詳細は省略する。
The same method can be used to calculate the multi-component model, but the calculation method is well-known, and thus the details are omitted.
まず、ルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合について、検量線の作成・検証の詳細を説明する。
(1) First, details of preparation and verification of a calibration curve for the case of using a sample in which rutin, chlorogenic acid and scopoletin are mixed as a standard will be described.
上述のような手法で得られた検量線を、ルチン、クロロゲン酸、スコポレチンの各成分について作成・検証するために、ルチン、クロロゲン酸、スコポレチンの各成分の含有量が既知の複数のサンプルを、検量線の作成に使用するためのキャリブレーション用サンプル群と、検量線を検証して有効性を確認するためのバリデーション用サンプル群に分けて用意する。
Calibration curve obtained by the above-mentioned method, rutin, chlorogenic acid, in order to create and verify each component of scopoletin, rutin, chlorogenic acid, a plurality of samples whose content of each component of scopoletin is known, A calibration sample group for use in creating a calibration curve and a validation sample group for verifying the calibration curve to confirm its effectiveness are prepared.
キャリブレーション用サンプル群に対して上述のＰＬＳ回帰分析を適用し、取得された蛍光指紋情報から各成分について含有量を推定する検量線を作成する。なお、検量線の作成に当たり、取得された蛍光指紋に対する前処理を省くこともできるが、（１）本発明に特有な２次微分処理、（２）成分情報に寄与しない波長に対する除去処理等の前処理を行うことが望ましい。（２）の前処理に関しては、例えば、以下のような処理を採用することができる。
・非蛍光成分の除去，散乱光の除去，低感度領域の除去
・対数変換（Log10）→従前の二次微分→規格化（normalize）→オートスケーリング（autoscale）
・VIPによる波長限定
因みに、上記（１）、（２）の前処理の順序は、適宜決めることができるが、（２）を先行させることが望ましい。 The above-described PLS regression analysis is applied to the calibration sample group, and a calibration curve for estimating the content of each component from the acquired fluorescent fingerprint information is created. In preparing the calibration curve, it is possible to omit the pre-processing for the acquired fluorescent fingerprints. However, (1) second-order differentiation processing unique to the present invention, (2) removal processing for wavelengths that do not contribute to component information, and the like. It is desirable to perform pre-processing. For the pre-processing of (2), for example, the following processing can be adopted.
・ Removal of non-fluorescent components, removal of scattered light, removal of low-sensitivity region ・ Logarithmic transformation (Log10) → Previous second derivative → Normalize → Auto scaling (autoscale)
-Wavelength limitation by VIP The order of the pre-processing of the above (1) and (2) can be determined as appropriate, but it is desirable to precede the (2).
・非蛍光成分の除去，散乱光の除去，低感度領域の除去
・対数変換（Log10）→従前の二次微分→規格化（normalize）→オートスケーリング（autoscale）
・VIPによる波長限定
因みに、上記（１）、（２）の前処理の順序は、適宜決めることができるが、（２）を先行させることが望ましい。 The above-described PLS regression analysis is applied to the calibration sample group, and a calibration curve for estimating the content of each component from the acquired fluorescent fingerprint information is created. In preparing the calibration curve, it is possible to omit the pre-processing for the acquired fluorescent fingerprints. However, (1) second-order differentiation processing unique to the present invention, (2) removal processing for wavelengths that do not contribute to component information, and the like. It is desirable to perform pre-processing. For the pre-processing of (2), for example, the following processing can be adopted.
・ Removal of non-fluorescent components, removal of scattered light, removal of low-sensitivity region ・ Logarithmic transformation (Log10) → Previous second derivative → Normalize → Auto scaling (autoscale)
-Wavelength limitation by VIP The order of the pre-processing of the above (1) and (2) can be determined as appropriate, but it is desirable to precede the (2).
次に、バリデーション用サンプル群について、取得された蛍光指紋情報から前記検量線を用いて各成分の含有量を推定し、検量線の検証を行う。
Next, for the validation sample group, the content of each component is estimated from the acquired fluorescent fingerprint information using the calibration curve, and the calibration curve is verified.
表２は、ルチン、クロロゲン酸、スコポレチンの各成分について、キャリブレーション用サンプル群の決定係数Ｒ2及び検量線作成時の標準誤差ＳＥＣ、並びに、バリデーション用サンプル群の決定係数Ｒ2及び検量線評価（検証）時の標準誤差ＳＥＰを纏めたものである。
Table 2, rutin, chlorogenic acid, each component of scopoletin, coefficient of determination of the sample group for calibration R 2 and standard error SEC when creating the calibration curve, and the determination of the validation sample group coefficient R 2 and calibration evaluation It is a compilation of the standard error SEP at the time of (verification).
図７Ａ～図7Ｃは、ルチン、クロロゲン酸、スコポレチンの各成分について、バリデーション用サンプル群のデータをグラフ化したものであり、図７Ａはスコポレチン、図７Ｂはクロロゲン酸、図７Ｃはルチンをそれぞれ表している。
7A to 7C are graphs of data of a validation sample group for each component of rutin, chlorogenic acid, and scopoletin. FIG. 7A shows scopoletin, FIG. 7B shows chlorogenic acid, and FIG. 7C shows rutin. ing.
上述の表２及び図７Ａ～図7Ｃからも明らかなように、ルチン、クロロゲン酸、スコポレチンの検量線は、良好な推定精度を有しており、当該検量線の有効性が確認されている。
明 ら か As is clear from Table 2 and FIGS. 7A to 7C, the calibration curves of rutin, chlorogenic acid, and scopoletin have good estimation accuracy, and the validity of the calibration curves has been confirmed.
次に、ルチン含有量の測定のためにたばこ原料を試料として用いる場合について、キャリブレーション及びバリデーションの推定精度について詳述する。
Next, the accuracy of calibration and validation estimation when using tobacco raw materials as samples for the measurement of rutin content will be described in detail.
図８Ａは、横軸に高速液体クロマトグラフ（HPLC）による実測値（化学分析値）、縦軸に蛍光指紋による推定値を取り、バリデーション用サンプル群に属する各サンプルについて、前記前処理を行った場合の対応する点をプロットしたグラフである。
FIG. 8A shows the measured value (chemical analysis value) by high-performance liquid chromatography (HPLC) on the horizontal axis and the estimated value by the fluorescent fingerprint on the vertical axis, and the pretreatment was performed for each sample belonging to the validation sample group. It is the graph which plotted the corresponding point of the case.
前述のような、ルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合と同様の前処理を行った場合のキャリブレーション用サンプル群において、決定係数Ｒ2＝０．９７（ＳＥＣ＝０．０８６％）であり、化学分析値と検量線による推定値との間に高い相関を有し、良好な推定精度であることが確認されている。また、図８Ａによれば、バリデーション用サンプル群における推定精度は、決定係数Ｒ2＝０．９１（ＳＥＰ＝０．１９％）あり、当該検量線の有効性が確認されている。
As described above, in the calibration sample group where the same pretreatment was performed as in the case of using a sample in which rutin / chlorogenic acid / scopoletin was mixed as a standard, the coefficient of determination R 2 = 0.97 (SEC = 0) 0.086%), which shows a high correlation between the chemical analysis value and the estimated value based on the calibration curve, confirming that the estimation accuracy is good. According to FIG. 8A, the estimation accuracy in the validation sample group has a coefficient of determination R 2 = 0.91 (SEP = 0.19%), and the validity of the calibration curve has been confirmed.
なお、前述のような前処理を省いた場合には、キャリブレーション用サンプル群における推定精度は、決定係数Ｒ2＝０．９３（ＳＥＣ＝０．１６％）、バリデーション用サンプル群における推定精度は、図９に示されるように、決定係数Ｒ2＝０．８７（ＳＥＰ＝０．２２％）であり、化学分析値と検量線による推定値との間に高い相関を有してはいるものの、前述のような前処理を行った場合と比較すると、推定精度について若干の低下が認められる。
When the pre-processing described above is omitted, the estimation accuracy in the calibration sample group is determined by the coefficient of determination R 2 = 0.93 (SEC = 0.16%), and the estimation accuracy in the validation sample group is As shown in FIG. 9, the coefficient of determination R 2 = 0.87 (SEP = 0.22%), and although there is a high correlation between the chemical analysis value and the value estimated by the calibration curve, However, as compared with the case where the above-described preprocessing is performed, a slight decrease in the estimation accuracy is recognized.
次に、クロロゲン酸含有量の測定のためにたばこ原料を試料として用いる場合について、キャリブレーション及びバリデーションの推定精度について詳述する。
Next, the accuracy of calibration and validation estimation in the case where tobacco raw materials are used as samples for measuring the chlorogenic acid content will be described in detail.
図８Ｂは、横軸に高速液体クロマトグラフ（HPLC）による実測値（化学分析値）、縦軸に蛍光指紋による推定値を取り、バリデーション用サンプル群に属する各サンプルについて、前記前処理を行った場合の対応する点をプロットしたグラフである。
FIG. 8B shows the measured value (chemical analysis value) by high-performance liquid chromatography (HPLC) on the horizontal axis and the estimated value by the fluorescent fingerprint on the vertical axis, and the pretreatment was performed for each sample belonging to the validation sample group. It is the graph which plotted the corresponding point of the case.
前述のような、ルチン・クロロゲン酸・スコポレチンを標品で混合した試料を用いる場合と同様の前処理を行った場合のキャリブレーション用サンプル群において、決定係数Ｒ2＝０．８７（ＳＥＣ＝０．１９％）であり、化学分析値と検量線による推定値との間に高い相関を有し、良好な推定精度であることが確認されている。また、図８Ｂによれば、バリデーション用サンプル群における推定精度は、決定係数Ｒ2＝０．８８（ＳＥＰ＝０．２０％）あり、当該検量線の有効性が確認されている。
As described above, in the calibration sample group in which the same pretreatment as in the case of using a sample in which rutin / chlorogenic acid / scopoletin was mixed as a standard, the coefficient of determination R 2 = 0.87 (SEC = 0) .19%), which shows a high correlation between the chemical analysis value and the estimated value based on the calibration curve, confirming that the estimation accuracy is good. According to FIG. 8B, the estimation accuracy in the validation sample group has a determination coefficient R 2 = 0.88 (SEP = 0.20%), and the validity of the calibration curve has been confirmed.
〔未知試料における各成分の含有量の推定〕
有効性が確認された検量線を用いて、ルチン、クロロゲン酸、スコポレチンの各成分の含有量が未知の試料の蛍光指紋情報に基づき、前記試料に含有される各成分の含有量を推定する。 [Estimation of content of each component in unknown sample]
Using the calibration curve whose effectiveness has been confirmed, the content of each component contained in the sample is estimated based on the fluorescent fingerprint information of the sample whose content of rutin, chlorogenic acid, and scopoletin is unknown.
有効性が確認された検量線を用いて、ルチン、クロロゲン酸、スコポレチンの各成分の含有量が未知の試料の蛍光指紋情報に基づき、前記試料に含有される各成分の含有量を推定する。 [Estimation of content of each component in unknown sample]
Using the calibration curve whose effectiveness has been confirmed, the content of each component contained in the sample is estimated based on the fluorescent fingerprint information of the sample whose content of rutin, chlorogenic acid, and scopoletin is unknown.
なお、各成分の含有量が未知の試料について、取得された蛍光指紋に対する前処理を省くこともできるが、検量線を取得したときと同様、（１）本発明に特有な２次微分処理、（２）成分情報に寄与しない波長に対する除去処理等の前処理を行い、次に、処理後の蛍光指紋から検量線に基づいて未知試料の各成分の含有量を推定することが望ましい。なお、上記（１）、（２）の前処理の順序は、適宜決めることができるが、（２）を先行させることが望ましい。
Note that, for a sample whose content of each component is unknown, it is possible to omit the pre-processing for the obtained fluorescent fingerprint, but as in the case where the calibration curve is obtained, (1) the secondary differentiation processing unique to the present invention, (2) It is desirable to perform preprocessing such as removal processing for wavelengths that do not contribute to the component information, and then estimate the content of each component of the unknown sample based on the calibration curve from the processed fluorescent fingerprint. The order of the pre-processing of (1) and (2) can be determined as appropriate, but it is desirable that (2) precede.
各成分以外のポリフェノール成分をノイズとして有効に除去し、各成分に特徴的に強い蛍光を示す特定の励起／蛍光波長（各成分の化学構造に基づいて決まる蛍光強度が最大になる波長条件）を使用することにより、蛍光指紋法による各成分の的確な定量が実現できる。
Effectively remove polyphenol components other than each component as noise, and set a specific excitation / fluorescence wavelength (a wavelength condition that maximizes the fluorescence intensity determined based on the chemical structure of each component) that gives each component a characteristically strong fluorescence. By using, accurate quantification of each component by the fluorescent fingerprint method can be realized.
そして、このように有効性が確認された検量線を用いて、ルチン、クロロゲン酸、スコポレチンの各成分の含有量が未知の試料の蛍光指紋情報に基づき、前記試料に含有される各成分の含有量を有効に推定できる。
なお、たばこ原料等のルチンを含有する未知試料中のルチンの定量に関しては、以下に詳述するような簡便化した手法を有効に利用できる。 Then, using the calibration curve thus validated, the content of each component of rutin, chlorogenic acid, and scopoletin is based on the fluorescent fingerprint information of the unknown sample, and the content of each component contained in the sample is determined. The amount can be estimated effectively.
For the determination of rutin in an unknown sample containing rutin such as a tobacco raw material, a simplified method described in detail below can be effectively used.
なお、たばこ原料等のルチンを含有する未知試料中のルチンの定量に関しては、以下に詳述するような簡便化した手法を有効に利用できる。 Then, using the calibration curve thus validated, the content of each component of rutin, chlorogenic acid, and scopoletin is based on the fluorescent fingerprint information of the unknown sample, and the content of each component contained in the sample is determined. The amount can be estimated effectively.
For the determination of rutin in an unknown sample containing rutin such as a tobacco raw material, a simplified method described in detail below can be effectively used.
＜ルチンに関する簡便化した適用例＞
これまでの説明においては、基本的に、全ての蛍光指紋情報を使用してルチン含有量を推定する態様を説明してきたが、精度は多少犠牲にしても、安価なハンディタイプのデバイスへの適用や簡素化された測定工程による測定の迅速化等に対する要請が存在する。
そこで、このような簡易測定という要請に応えるための簡便化した態様を以下説明する。 <Simplified application example of rutin>
In the above description, basically, the mode of estimating the rutin content using all the fluorescent fingerprint information has been described. However, even if accuracy is somewhat sacrificed, application to an inexpensive hand-held device is performed. There is a demand for speeding up the measurement by a simplified measurement process and the like.
Therefore, a simplified mode for responding to such a demand for simple measurement will be described below.
これまでの説明においては、基本的に、全ての蛍光指紋情報を使用してルチン含有量を推定する態様を説明してきたが、精度は多少犠牲にしても、安価なハンディタイプのデバイスへの適用や簡素化された測定工程による測定の迅速化等に対する要請が存在する。
そこで、このような簡易測定という要請に応えるための簡便化した態様を以下説明する。 <Simplified application example of rutin>
In the above description, basically, the mode of estimating the rutin content using all the fluorescent fingerprint information has been described. However, even if accuracy is somewhat sacrificed, application to an inexpensive hand-held device is performed. There is a demand for speeding up the measurement by a simplified measurement process and the like.
Therefore, a simplified mode for responding to such a demand for simple measurement will be described below.
この簡素化した態様は、励起波長／蛍光波長（測定波長）として、２８５／４１０，４１５，４２０，４２５，４３０，４３５，４４０，４５０，４６０（nm）の９波長のみを用いて蛍光指紋情報を取得し、ルチンの定量を行うものである。定量のための手法は、基本的に、全ての蛍光指紋情報を使用するケースと同様であり、詳細は省略することとするが、このように波長を限定した（概ね１０波長未満の）場合には、検量線の作成に際し、PLS回帰以外に、重回帰分析（MLR）も利用可能である。なお、波長を限定しない場合にはPLS回帰を使用することが望ましい。
This simplified embodiment uses only nine wavelengths of 285/410, 415, 420, 425, 430, 435, 440, 450, and 460 (nm) as excitation wavelength / fluorescence wavelength (measurement wavelength) to obtain fluorescent fingerprint information. Is obtained to determine the amount of rutin. The method for quantification is basically the same as the case where all the fluorescent fingerprint information is used, and details are omitted. However, when the wavelength is limited as described above (generally less than 10 wavelengths), In addition to the PLS regression, multiple regression analysis (MLR) can be used to generate a calibration curve. When the wavelength is not limited, it is desirable to use PLS regression.
なお、２８５／４１０，４１５，４２０，４２５，４３０，４３５，４４０，４５０，４６０（nm）の９波長は、ルチンの化学構造に基づいて蛍光強度が最大値を取る特定の励起／蛍光波長に相当するものである。
The nine wavelengths of 285/410, 415, 420, 425, 430, 435, 440, 450, and 460 (nm) are specific excitation / fluorescence wavelengths at which the fluorescence intensity takes a maximum value based on the chemical structure of rutin. It is equivalent.
このような特定数の波長を使用する簡便化した態様については、キャリブレーション用サンプル群における推定精度は、決定係数Ｒ2＝０．７３（ＳＥＣ＝０．２５％）であった。また、バリデーション用サンプル群における推定精度は、決定係数Ｒ2＝０．８２（ＳＥＰ＝０．２６％）であり、前述の態様と比較すると、推定精度の低下が認められるものの、簡易測定という要請に十分応え得るレベル範囲にあるものといえる。
For such a simplified embodiment using a specific number of wavelengths, the estimation accuracy in the calibration sample group was a coefficient of determination R 2 = 0.73 (SEC = 0.25%). In addition, the estimation accuracy of the validation sample group is a coefficient of determination R 2 = 0.82 (SEP = 0.26%). It can be said that it is within the level range that can sufficiently respond to.
また、簡便化した態様はこれに限られるものではなく、励起波長／蛍光波長（測定波長）として、２８５／４１０，４１５，４２０，４２５，４３０，４３５，４４０，４５０，４６０（nm）のうち少なくとも１つを含むような態様を採用することができる。この場合には、９波長のみを用いるケースと比較して、測定精度の点では劣るものの、迅速性の向上を実現できる。
Further, the simplified embodiment is not limited to this, and the excitation wavelength / fluorescence wavelength (measurement wavelength) is 285/410, 415, 420, 425, 430, 435, 440, 450, and 460 (nm). An embodiment including at least one can be adopted. In this case, although the measurement accuracy is inferior to the case where only 9 wavelengths are used, an improvement in speed can be realized.
そして、上述のような簡便化した態様を、ハンディタイプのデバイスや、簡素化された測定工程として実現することにより、安価且つ迅速に測定結果を取得することができる。
By implementing the above-described simplified embodiment as a handy-type device or a simplified measurement process, a measurement result can be obtained quickly and inexpensively.
なお、上述のような簡便化した適用例に関しては、迅速な定量を行うために、取得された蛍光指紋に対する前処理を省くこともできる。
In addition, in the above-described simplified application example, in order to perform rapid quantification, it is also possible to omit the preprocessing for the acquired fluorescent fingerprint.
また、上述のような簡便化した適用例は、たばこ原料中のルチンの含有量の推定のみに有効なものではなく、ルチンを含有する他の材料に対しても有効に適用することができる。
Further, the simplified application example described above is not only effective for estimating the content of rutin in tobacco raw materials, but also can be effectively applied to other materials containing rutin.
本発明は、上述した実施の態様以外にも、特許請求の範囲に記載した技術的思想の範囲内において種々の異なる実施の態様を採用し得ることに留意されたい。
留意 It should be noted that the present invention can adopt various different embodiments within the scope of the technical idea described in the claims, in addition to the above-described embodiments.
１００：蛍光指紋分析による試料の評価・推定装置
１１０：前処理手段
１２０：推定モデル作成手段
１３０：成分量推定手段 100: Apparatus for evaluating and estimating a sample by fluorescent fingerprint analysis 110: Preprocessing means 120: Estimation model creation means 130: Component amount estimation means
１１０：前処理手段
１２０：推定モデル作成手段
１３０：成分量推定手段 100: Apparatus for evaluating and estimating a sample by fluorescent fingerprint analysis 110: Preprocessing means 120: Estimation model creation means 130: Component amount estimation means
Claims (21)
- 試験試料について励起波長・蛍光波長・蛍光強度のデータからなる蛍光指紋情報を取得する蛍光指紋情報取得工程と、
２次微分処理により前記試験試料に特有な前記蛍光強度のピーク値が現れるように軸を設定し、前記軸に沿った前記蛍光指紋情報の２次微分処理を少なくとも含む前処理工程と、
少なくとも前記２次微分処理が実施された蛍光指紋情報を説明変数とし、前記試験試料についての既知の定量値を目標変数として、検量線を取得する推定モデル作成工程と、
を含む、蛍光指紋分析による試験試料の評価方法。 A fluorescent fingerprint information acquiring step of acquiring fluorescent fingerprint information comprising data of excitation wavelength, fluorescence wavelength, and fluorescence intensity for the test sample;
A preprocessing step that sets an axis so that a peak value of the fluorescence intensity unique to the test sample appears by the second derivative processing, and includes at least a second derivative processing of the fluorescent fingerprint information along the axis,
As an explanatory variable at least the fluorescent fingerprint information on which the second derivative processing has been performed, a known quantitative value of the test sample as a target variable, and an estimation model creating step of obtaining a calibration curve;
And a method for evaluating a test sample by fluorescence fingerprint analysis. - 前記推定モデル作成工程において、多変量解析によって上記検量線を作成することを特徴とする請求項１に記載の蛍光指紋分析による試験試料の評価方法。 The method for evaluating a test sample by fluorescence fingerprint analysis according to claim 1, wherein in the estimation model creation step, the calibration curve is created by multivariate analysis.
- 前記多変量解析は、PLS回帰分析であることを特徴とする請求項２に記載の蛍光指紋分析による試験試料の評価方法。 The method according to claim 2, wherein the multivariate analysis is a PLS regression analysis.
- 前記前処理工程において、前記蛍光指紋情報に対して低感度領域の削除処理を行うことを特徴とする請求項１～３の何れか１項に記載の蛍光指紋分析による試験試料の評価方法。 4. The method for evaluating a test sample by fluorescent fingerprint analysis according to any one of claims 1 to 3, wherein in the preprocessing step, a process of deleting a low-sensitivity region is performed on the fluorescent fingerprint information.
- 前記試験試料が、クロロゲン酸を含む、請求項１～４の何れか１項に記載の蛍光指紋分析による試験試料の評価方法。 (5) The method for evaluating a test sample by fluorescence fingerprint analysis according to any one of (1) to (4), wherein the test sample contains chlorogenic acid.
- 前記試験試料が、スコポレチンを含む、請求項１～４の何れか１項に記載の蛍光指紋分析による試験試料の評価方法。 (5) The method for evaluating a test sample by fluorescence fingerprint analysis according to any one of (1) to (4), wherein the test sample contains scopoletin.
- 前記試験試料が、ルチンを含む、請求項１～４の何れか１項に記載の蛍光指紋分析による試験試料の評価方法。 (5) The method for evaluating a test sample by fluorescence fingerprint analysis according to any one of (1) to (4), wherein the test sample contains rutin.
- 前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値のうちの少なくとも１つである波長のいずれかの励起波長／蛍光波長の組み合わせにより、ルチンの検量線を作成することを特徴とする請求項７に記載の蛍光指紋分析による試験試料の評価方法。 Any one of the excitation wavelength / fluorescence wavelengths where the excitation wavelength is near 285 nm and the fluorescence wavelength is at least one of the values near 410, 415, 420, 425, 430, # 435, 440, # 450, and 460 nm 8. The method for evaluating a test sample by fluorescence fingerprint analysis according to claim 7, wherein a calibration curve of rutin is created by a combination of wavelengths.
- 前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値の全てである波長の励起波長／蛍光波長の組み合わせの両方により、ルチンの検量線を作成することを特徴とする請求項８に記載の蛍光指紋分析による試験試料の評価方法。 Both the excitation wavelength / fluorescence wavelength combination in which the excitation wavelength is a value near 285 nm and the fluorescence wavelength is all values near 410, 415, 420, 425, 430, # 435, 440, # 450, and 460 nm, The method for evaluating a test sample by fluorescence fingerprint analysis according to claim 8, wherein a calibration curve of rutin is prepared.
- 前記試験試料は、たばこ製品の原料であることを特徴とする請求項１～９の何れか１項に記載の蛍光指紋分析による試験試料の評価方法。 The method for evaluating a test sample by fluorescent fingerprint analysis according to any one of claims 1 to 9, wherein the test sample is a raw material of a tobacco product.
- 前記試験試料は、励起光の照射前に粉末状に粉砕・混合されることを特徴とする請求項１～１０の何れか１項に記載の蛍光指紋分析による試験試料の評価方法。 The method according to any one of claims 1 to 10, wherein the test sample is pulverized and mixed into a powder before irradiation with excitation light.
- 前記試験粉砕によって、試料が１ｍｍ以下の粒径とされることを特徴とする請求項９に記載の蛍光指紋分析による試験試料の評価方法。 10. The method for evaluating a test sample by fluorescent fingerprint analysis according to claim 9, wherein the sample is reduced to a particle size of 1 mm or less by the test pulverization.
- 前記試験材料は、事前に水分量を一定化するために、所定の調和条件で所定時間蔵置されることを特徴とする請求項１～１２の何れか１項に記載の蛍光指紋分析による試験試料の評価方法。 The test sample according to any one of claims 1 to 12, wherein the test material is stored under predetermined harmonic conditions for a predetermined time in order to stabilize a water content in advance. Evaluation method.
- 前記試験材料は、たばこ製品の原料であり、前記調和条件は、温度２２℃、湿度６０％の室内という条件であり、前記所定時間は２４時間以上であることを特徴とする請求項１３に記載の蛍光指紋分析による試験試料の評価方法。 14. The test material according to claim 13, wherein the test material is a raw material of a tobacco product, the harmony condition is a condition of a room at a temperature of 22 ° C and a humidity of 60%, and the predetermined time is 24 hours or more. Method for evaluating test samples by fluorescence fingerprint analysis.
- 請求項１～１４の何れか１項に記載の蛍光指紋分析による試験試料の評価方法により得られた検量線と未知の試料の蛍光指紋情報とに基づき、前記試料に含有される特定成分の含有量を推定することを特徴とする成分量推定方法。 A specific component contained in the sample based on the calibration curve obtained by the method for evaluating a test sample by the fluorescent fingerprint analysis according to any one of claims 1 to 14 and fluorescent fingerprint information of an unknown sample. A component amount estimation method characterized by estimating an amount.
- 前記試料に含有される特定成分の含有量を推定する際に、前記蛍光指紋分析による試験試料の評価方法における前処理と同一の処理を行うことを特徴とする請求項１５に記載の成分量推定方法。 The component amount estimation according to claim 15, wherein when estimating the content of the specific component contained in the sample, the same process as the pre-processing in the test sample evaluation method by the fluorescent fingerprint analysis is performed. Method.
- 前記未知の試料の蛍光指紋情報を取得するために、励起波長が２８５nm近傍値であって蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０, ４５０，４６０nm近傍値のうち少なくとも１つである波長のいずれかの励起波長／蛍光波長の組み合わせを用いることを特徴とする請求項１５又は１６に記載の成分量推定方法。 In order to acquire the fluorescence fingerprint information of the unknown sample, at least one of the excitation wavelengths near 285 nm and the fluorescence wavelengths near 410, 415, 420, 425, 430, # 435, 440, # 450, and 460 nm is used. 17. The component amount estimating method according to claim 15, wherein a combination of any one of the following excitation wavelengths / fluorescence wavelengths is used.
- 前記未知の資料の蛍光指紋情報を取得するために、前記励起波長が２８５nm近傍値であって前記蛍光波長が４１０，４１５，４２０，４２５，４３０, ４３５，４４０，４５０，４６０nm近傍値の全てである波長の励起波長／蛍光波長の組み合わせの両方を用いることを特徴とする請求項１７に記載の成分量推定方法。 In order to obtain the fluorescent fingerprint information of the unknown material, the excitation wavelength is near 285 nm and the fluorescence wavelength is 410, 415, 420, 425, 430, 435, 440, 450, and 460 nm. 18. The component amount estimating method according to claim 17, wherein both the combination of the excitation wavelength and the fluorescence wavelength of a certain wavelength are used.
- コンピュータに請求項１～１８の何れか１項に記載の方法を実行させるためのプログラム。 A program for causing a computer to execute the method according to any one of claims 1 to 18.
- 試料についての励起波長・蛍光波長・蛍光強度のデータからなる蛍光指紋情報を入力し、２次微分処理により前記試料に特有な前記蛍光強度のピーク値が現れるように軸を設定し、前記軸に沿った前記蛍光指紋情報の２次微分処理を少なくとも含む前処理手段と、
少なくとも前記２次微分処理が実施された蛍光指紋情報を説明変数とし、前記試験試料についての既知の定量値を目標変数として、検量線を取得する推定モデル作成手段と、
前記推定モデル作成手段により取得された前記検量線と未知の試料の蛍光指紋情報とに基づき、前記未知の試料に含有される特定成分の含有量を推定する成分量推定手段と、
を具備することを特徴とする装置。 Fluorescent fingerprint information consisting of excitation wavelength / fluorescence wavelength / fluorescence intensity data for the sample is input, and an axis is set so that a peak value of the fluorescence intensity unique to the sample appears by a second derivative process. Preprocessing means including at least a second derivative processing of the fluorescent fingerprint information along
Estimation model creating means for acquiring a calibration curve, using at least the fluorescent fingerprint information on which the second derivative processing has been performed as an explanatory variable, and a known quantitative value for the test sample as a target variable,
Component amount estimation means for estimating the content of a specific component contained in the unknown sample, based on the calibration curve and the fluorescence fingerprint information of the unknown sample obtained by the estimation model creation means,
An apparatus comprising: - 前記未知の試料の蛍光指紋情報は前記前処理手段で処理され、処理後の蛍光指紋情報が前記成分量推定手段に入力されることを特徴とする請求項２０に記載の装置。 21. The apparatus according to claim 20, wherein the fluorescent fingerprint information of the unknown sample is processed by the preprocessing unit, and the processed fluorescent fingerprint information is input to the component amount estimating unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2020536325A JP7021755B2 (en) | 2018-08-10 | 2019-05-10 | Sample evaluation / estimation methods, programs, and equipment by fluorescent fingerprint analysis |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2018151556 | 2018-08-10 | ||
| JP2018-151556 | 2018-08-10 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020031447A1 true WO2020031447A1 (en) | 2020-02-13 |
Family
ID=69415438
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2019/018739 WO2020031447A1 (en) | 2018-08-10 | 2019-05-10 | Sample evaluation/estimation method by fluorescence fingerprint analysis, program, and device |
Country Status (3)
| Country | Link |
|---|---|
| JP (1) | JP7021755B2 (en) |
| TW (1) | TW202009471A (en) |
| WO (1) | WO2020031447A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112085061A (en) * | 2020-08-10 | 2020-12-15 | 杭州凯曼健康科技有限公司 | Quantum dot immunofluorescence curve feature identification method |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112816446B (en) * | 2020-12-24 | 2022-02-01 | 四川长虹电器股份有限公司 | Method for detecting powder decay of fluorescent wheel based on fluorescence spectrum |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002520005A (en) * | 1998-07-08 | 2002-07-09 | ノボザイムス アクティーゼルスカブ | Use of phenol oxidase in tobacco processing. |
| JP2010185719A (en) * | 2009-02-10 | 2010-08-26 | National Agriculture & Food Research Organization | Method and apparatus for discriminating grain flour |
| JP2014035255A (en) * | 2012-08-08 | 2014-02-24 | National Agriculture & Food Research Organization | Organism-count model generation device, organism-count model generation method, cytometer, and program |
| JP2016045179A (en) * | 2014-08-26 | 2016-04-04 | 国立研究開発法人農業・食品産業技術総合研究機構 | Cheese maturation degree indicator estimating method, cheese maturation degree indicator estimating device, computer-executable program, natural cheese, processed cheese, and manufacturing method for processed cheese |
| JP2016176830A (en) * | 2015-03-20 | 2016-10-06 | 雪印種苗株式会社 | Method of estimating lactic acid buffering capacity of fermented product materials |
| JP2017036991A (en) * | 2015-08-10 | 2017-02-16 | ハウス食品グループ本社株式会社 | Method for discriminating production region of spice by excitation-emission matrix analysis |
| JP2017051162A (en) * | 2015-09-11 | 2017-03-16 | 国立研究開発法人農業・食品産業技術総合研究機構 | Method and device for estimating viable cell count on sample surface, and program incorporated into the device |
| JP2018004571A (en) * | 2016-07-07 | 2018-01-11 | 国立研究開発法人農業・食品産業技術総合研究機構 | Heating history estimation method, heating history estimation device, program, and manufacturing method |
-
2019
- 2019-05-10 JP JP2020536325A patent/JP7021755B2/en active Active
- 2019-05-10 WO PCT/JP2019/018739 patent/WO2020031447A1/en active Application Filing
- 2019-05-24 TW TW108118071A patent/TW202009471A/en unknown
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002520005A (en) * | 1998-07-08 | 2002-07-09 | ノボザイムス アクティーゼルスカブ | Use of phenol oxidase in tobacco processing. |
| JP2010185719A (en) * | 2009-02-10 | 2010-08-26 | National Agriculture & Food Research Organization | Method and apparatus for discriminating grain flour |
| JP2014035255A (en) * | 2012-08-08 | 2014-02-24 | National Agriculture & Food Research Organization | Organism-count model generation device, organism-count model generation method, cytometer, and program |
| JP2016045179A (en) * | 2014-08-26 | 2016-04-04 | 国立研究開発法人農業・食品産業技術総合研究機構 | Cheese maturation degree indicator estimating method, cheese maturation degree indicator estimating device, computer-executable program, natural cheese, processed cheese, and manufacturing method for processed cheese |
| JP2016176830A (en) * | 2015-03-20 | 2016-10-06 | 雪印種苗株式会社 | Method of estimating lactic acid buffering capacity of fermented product materials |
| JP2017036991A (en) * | 2015-08-10 | 2017-02-16 | ハウス食品グループ本社株式会社 | Method for discriminating production region of spice by excitation-emission matrix analysis |
| JP2017051162A (en) * | 2015-09-11 | 2017-03-16 | 国立研究開発法人農業・食品産業技術総合研究機構 | Method and device for estimating viable cell count on sample surface, and program incorporated into the device |
| JP2018004571A (en) * | 2016-07-07 | 2018-01-11 | 国立研究開発法人農業・食品産業技術総合研究機構 | Heating history estimation method, heating history estimation device, program, and manufacturing method |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112085061A (en) * | 2020-08-10 | 2020-12-15 | 杭州凯曼健康科技有限公司 | Quantum dot immunofluorescence curve feature identification method |
| CN112085061B (en) * | 2020-08-10 | 2022-04-26 | 杭州凯曼健康科技有限公司 | Quantum dot immunofluorescence curve feature identification method |
Also Published As
| Publication number | Publication date |
|---|---|
| TW202009471A (en) | 2020-03-01 |
| JPWO2020031447A1 (en) | 2021-05-13 |
| JP7021755B2 (en) | 2022-02-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2425336C2 (en) | Method of spectral analysis | |
| JP3245157B2 (en) | Measurement and correction of spectral data | |
| JP6803750B2 (en) | Equipment and methods for adjusted hair color | |
| JP6089345B2 (en) | Multicomponent regression / multicomponent analysis of temporal and / or spatial series files | |
| US10557792B2 (en) | Spectral modeling for complex absorption spectrum interpretation | |
| KR101939887B1 (en) | Texture analysis of a coated surface using pivot-normalization | |
| EP3290908B1 (en) | Unknown sample determining method | |
| WO2020031447A1 (en) | Sample evaluation/estimation method by fluorescence fingerprint analysis, program, and device | |
| JP2019503490A (en) | Spectroscopic analysis method and apparatus using multiple processing of infrared and fluorescent spectral data | |
| EP2366991A1 (en) | Probabilistic scoring for components of a mixture | |
| CN111157115B (en) | Underwater Brillouin scattering spectrum acquisition method and device | |
| US11860102B2 (en) | Fluorescence fingerprint analysis for quantifying TSNAs in processed tobacco raw material | |
| Muñoz de la Peña et al. | Resolution of ternary mixtures of salicylic, salicyluric and gentisic acids by partial least squares and principal component regression: Optimization of the scanning path in the excitation-emission matrices | |
| Chen et al. | Adaptive multiscale regression for reliable Raman quantitative analysis | |
| JP2016206060A (en) | Spectroscopic measurement device and spectroscopic measurement method | |
| CN113933275B (en) | Quantitative analysis method, separation method, device and equipment based on biological imaging | |
| CN116399836A (en) | Cross-talk fluorescence spectrum decomposition method based on alternating gradient descent algorithm | |
| Baldo et al. | Proposition and experimental evaluation of a point-based compensation approach to reduce systematic errors in CT measurements | |
| CN110836878B (en) | Convolution interpolation coupling Gaussian mixture model rapid three-dimensional fluorescence peak searching method | |
| JP2014038043A (en) | Spectral measurement method | |
| JP2022527850A (en) | Methods for configuring a spectroscopic measuring device | |
| WO2015125311A1 (en) | Spectroscopic measurement method | |
| Wang et al. | A multimodal spectrometer for Raman scattering and near-infrared absorption measurement | |
| CN111157116B (en) | Underwater Brillouin scattering spectrum test system | |
| Zolfonoun | Determination of 6 Li and 7 Li abundances by inductively coupled plasma-optical emission spectrometry combined with derivative spectroscopy, multivariate curve resolution and multivariate calibration methods |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19847092 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 2020536325 Country of ref document: JP Kind code of ref document: A |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 19847092 Country of ref document: EP Kind code of ref document: A1 |