JP2017026558A - Cocoa mass quality prediction method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for easily predicting quality of cocoa mass which has been difficult conventionally.SOLUTION: A cocoa mass quality prediction method comprises: a pretreatment step for performing pretreatment to pieces of cocoa mass for acquiring analysis samples; an appliance analysis step for subjecting the analysis samples to appliance analysis for acquiring pieces of appliance analysis data; and a multivariable analysis step for, performing multivariable analysis using pieces of appliance analysis data about pieces of cocoa mass and pieces of quality data indicating respective quality of pieces of cocoa mass, for creating a cocoa mass quality prediction model which indicates a relationship between the pieces of appliance analysis data and the qualities predicted by the appliance analysis data, thereby easily predicting the qualities of pieces of cacao mass.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a cacao mass quality prediction method. More particularly, the present invention relates to a method for predicting quality by instrumental analysis of a metabolite or degradation product of cacao mass.

  Cocoa beans are used as a raw material for chocolate, cocoa and the like by taking advantage of their unique flavor. The flavor of cocoa beans is formed by fermentation and drying in the production area and subsequent roasting in the raw material processing step. It is known that the quality of flavor and the like formed in these processes varies depending on the cocoa bean variety, production area, producer, and the like.

  Currently, the quality of cacao mass is determined by sensory tests based on the aroma, color, and taste of cacao mass by experts. It takes years of experience to acquire these skills. In addition, there is a disadvantage that the objectivity and reproducibility are opaque in some cases. Therefore, a simple mechanized quality prediction method is necessary for manufacturing process management and product quality management. Under such circumstances, many studies have been conducted on changes in specific components during the production process of cacao mass, and research focusing on changes in proteins and free amino acids due to fermentation, drying, and roasting (Non-patent Document 1). And research focusing on polyphenols and saccharides (Non-Patent Document 2). However, these techniques focus on specific components and are limited in data, so it was difficult to say that they represent the quality of the entire cacao mass.

In recent years, methods for quality control or characterization have been proposed by combining instrumental analysis methods and powerful computer-driven pattern recognition techniques.
For example, Patent Document 1 discloses a method of converting an analysis result obtained by instrumental analysis of green tea into numerical data, performing multivariate analysis, and predicting the quality (ranking) of green tea from the obtained analysis result. Yes.

JP 2009-14700 A

Journal of the Science of Food and Agriculture, 81, 281-288 2000 Journal of Japan Society for Food Science and Technology, Vol. 48, 848-851 2001

The taste of cocoa mass is mainly attributed to organic acids, fatty acids, amino acids, sugars, polyphenols and the like. Moreover, a volatile compound mainly contributes to the difference in the scent of cocoa mass.
Depending on the origin of the raw cocoa beans, fermentation method, drying method, roasting method, etc., the apparent steady state amount of the intermediate pathway metabolites and / or terminal accumulation of the final metabolites may vary, affecting the flavor of the cocoa mass.

  An object of the present invention is to provide a method for easily predicting the quality of cacao mass, which has been difficult in the past.

  The present invention is a method for predicting the quality of cocoa mass, a pretreatment step for pre-processing cocoa mass to obtain an analysis sample; an instrument analysis step for subjecting the analysis sample to instrument analysis to obtain instrument analysis data; The relationship between the device analysis data and the quality predicted from the device analysis data is represented by multivariate analysis using the device analysis data about the cocoa mass and the quality data representing the quality of each cocoa mass. And a multivariate analysis step of creating a cacao mass quality prediction model.

The multivariate analysis is preferably a PLS regression analysis.
The instrumental analysis is preferably a combination of gas chromatography and mass spectrometry.
As the instrumental analysis data, it is preferable to use retention time and mass spectrometry signal intensity for multivariate analysis.

It is preferable that the pretreatment includes a step of extracting a hydrophilic compound from the cocoa mass to obtain an extract.
Preferably, the instrumental analysis is a combination of gas chromatography and mass spectrometry, and the pretreatment further includes a step of silylating the extract.

  The instrumental analysis is a combination of gas chromatography and mass spectrometry, and among the analytical results of the instrumental analysis, succinic acid, uridine, alanine, adenine, glycine, leucine, serine, fumaric acid, threonine, pyroglutamic acid , Phenylalanine, tyrosine, shikimic acid, tryptophan, valine, isoleucine, xylitol, asparagine, aconite, glyceric acid, arabitol, sucrose, lactic acid, aspartic acid, arabinose, guanosine, proline, homocysteic acid, 2-dehydro-D-gluconic acid , Glutamic acid, gluconic acid, glycerol, trehalose, pyruvic acid, oxaloacetic acid, threonic acid, and sarcosine data are preferably used for multivariate analysis.

  According to the present invention, it is possible to predict the quality of cacao mass, which has been difficult in the past, with a more accurate and simple method.

It is a figure which shows the relationship between the predicted value of sensory evaluation by the PLS model of a training set, and the measured value by sensory evaluation. It is a figure which shows the relationship between the predicted value of sensory evaluation by a PLS model, and the measured value by sensory evaluation about all the 15 types of analysis samples of a test set (triangle plot) and a training set (square plot). It is a figure which shows the relationship between the predicted value of sensory evaluation by the PLS model of a training set, and the measured value by sensory evaluation. It is a figure which shows the relationship between the predicted value of sensory evaluation by a PLS model, and the measured value by sensory evaluation about all 15 types of analysis samples of a test set and a training set.

  The quality prediction target in the present invention is cacao mass. The cocoa mass in the present invention is a cocoa mass produced through fermentation, drying and roasting.

  In the present invention, the cocoa mass is appropriately pretreated according to the instrumental analysis step described below, and the analysis sample obtained by the pretreatment is subjected to various instrumental analyzes.

<Instrument analysis process>
In the present invention, the instrument analysis step refers to a step of performing analysis / measurement using an analysis instrument. Analysis methods include gas chromatography (GC), liquid chromatography (LC) (for example, high performance liquid chromatography (HPLC), ultra high performance liquid chromatography (UPLC)), mass spectrometry (MS), infrared spectroscopy. Analysis (IR) (for example, Fourier transform infrared spectroscopy (FT-IR)), near infrared spectroscopy (NIR) (for example, Fourier transform near infrared spectroscopy (FT-NIR)), nuclear magnetic resonance analysis ( NMR) (for example, Fourier transform nuclear magnetic resonance analysis (FT-NMR)) and the like.
The detector for gas chromatography (GC) is not particularly limited, and a known one can be used as appropriate. For example, a flame ionization detector (FID) can be suitably used.
These instrumental analysis methods may be combined, and examples thereof include GC / MS, LC / MS (particularly HPLC / MS, UPLC / MS) and the like.
The apparatus used for the instrumental analysis process is not particularly limited as long as it is capable of measuring metabolites (for example, amino acids, organic acids, sugars, etc.) contained in cocoa mass. Can be used. Moreover, the measurement conditions can be appropriately set as appropriate for the measurement of these substances.
In particular, GC / MS (a combination of gas chromatography and mass spectrometry) is preferable.
For example, quantitative analysis is performed by GC / MS on an extract containing a hydrophilic low-molecular-weight metabolite obtained by extracting cocoa mass or on a thermal decomposition product of cocoa mass.

<Pretreatment process>
The pretreatment of cocoa mass is performed according to the instrumental analysis used in order to make the substance to be analyzed in the cocoa mass suitable for instrumental analysis. Examples of the pretreatment include treatments such as drying, cutting, pulverization, and extraction. For example, the pulverization can be performed using an appropriate instrument such as a blender or a ball mill. The extraction can be performed using water, an organic solvent, or a mixture of these solvents. Examples of the organic solvent that can be used for extraction include methanol, ethanol, n-propanol, isopropanol, acetone, chloroform, and the like. These unit operations are singly or combined to set appropriate preprocessing conditions.
In the present invention, it is preferable that the pretreatment includes a step of extracting a hydrophilic compound from cacao mass to obtain an extract.

  Moreover, when performing GC as an instrumental analysis, Preferably, pre-processing includes the process of silylating the extract from cocoa mass. The silylation can be carried out using a silylating reagent (derivatizing agent) for GC which is usually used by those skilled in the art.

<Multivariate analysis process / quality prediction process>
The pre-treated cocoa mass sample (analysis sample) is subjected to instrumental analysis to obtain an analysis result. The obtained analytical result may be a fingerprint (chemical fingerprint) of the cocoa mass sample.
In the present invention, the analysis result is converted into numerical data to perform multivariate analysis. Examples of results (variables) obtained by analysis include retention time, wavelength (or wave number), and signal data (or ionic strength), spectral data such as absorbance. When instrumental analysis is GC / MS, it is preferable to use retention time and mass spectrometry signal intensity as variables.
In the multivariate analysis for creating the quality prediction model, the sensory evaluation score of cacao mass is further preferable as a variable.

As the multivariate analysis, an analysis tool usually used in chemometrics is employed for analysis of instrumental analysis data.
For example, PCA (principal component analysis), HCA (hierarchical cluster analysis), PLS regression analysis (Projection to Latent Structure), discriminate analysis, etc. Multivariate tools.
In addition, PLS regression analysis with partial least squares (Partial least square projection to Latent Structure) is used to confirm the relationship between two groups of related variables; for example, the relationship between a cocoa mass metabolite and its flavor quality. Is done. If necessary, multivariate analysis may be performed in combination with spectral filtering methods, eg, orthogonal signal correction (OSC) to remove interfering components.
Many of these analysis tools are commercially available as software, and arbitrary ones are available. Such a commercially available tool is generally provided with an operation manual so that multivariate analysis can be performed without knowledge of difficult mathematics and statistics.

  The PLS regression analysis suitably employed in the present invention is an effective technique for creating a calibration curve from spectrum data having a correlation between variables (for example, wave number and wavelength). Normally, if there is a correlation between variables, the regression accuracy will be significantly reduced depending on the combination of variables used. To avoid this, PLS converts the variables into variables that are uncorrelated with each other (latent variables). To do regression. That is, PLS is an analysis method in which data variables are orthogonally transformed and (multiple) regression analysis is performed using the new variables.

In the present invention, a plurality of cacao masses of known quality are pre-processed to obtain individual analysis samples, the individual analysis samples are subjected to instrument analysis to obtain individual instrument analysis data, and the individual instrument analysis data and Create a cocoa mass quality prediction model that represents the relationship between the device analysis data and the quality predicted from the device analysis data by performing multivariate analysis using the flavor quality (known) of each cocoa mass To do.
And, for the cacao mass whose quality is unknown, pre-processed to obtain an analytical sample, the analytical sample is subjected to instrumental analysis to obtain instrumental analysis data, and the instrumental analysis data is used to produce the cacao mass quality prediction model created above. The quality of the cacao mass whose quality is unknown can be predicted.

The multivariate analysis step and the quality prediction step in the present invention are performed using an arithmetic processing unit.
A suitable arithmetic processing unit includes a first storage unit that stores instrument analysis data output from the instrument analyzer, a second storage unit that stores input quality data (preferably a sensory evaluation score), Prediction model creation for creating a quality prediction model by performing multivariate analysis on device analysis data stored in the first storage unit and quality data stored in the second storage unit based on a preset program A third storage unit that stores the quality prediction model, a fourth storage unit that stores device analysis data for the measurement control, the device analysis data stored in the fourth storage unit, A predicted value calculation unit that compares the quality prediction model stored in the third storage unit based on a preset program and calculates the quality (preferably the value of the sensory evaluation score) predicted from the device analysis data And be prepared That.

For example, when the instrumental analysis is a GC / MS analysis, when a plurality of cacao mass samples with known sensory evaluation scores are input to the arithmetic processing unit, the instrumental analysis data obtained by the respective GC / MS analysis is input to the first processing unit. Stored in the department. Separately, when the sensory evaluation scores of the plurality of cocoa mass samples are input to the arithmetic processing device, they are stored in the second storage unit.
Then, in the prediction model creation unit of the arithmetic processing unit, multivariate analysis by the PLS method using instrument analysis data (for example, retention time and mass analysis signal intensity) of the plurality of cocoa mass samples and sensory evaluation scores of the plurality of cocoa mass samples as variables. And a sensory evaluation score prediction model (quality prediction model) is created and stored in the third storage unit. The instrumental analysis data is preprocessed by a known method as necessary before being used for multivariate analysis.
When quality prediction is performed, for the cacao mass sample whose sensory evaluation score is unknown, when device analysis data obtained by the same GC / MS analysis is input to the arithmetic processing device, this is stored in the fourth storage unit.
Then, in the predicted value calculation unit of the arithmetic processing unit, the input device analysis data is compared and collated with the sensory evaluation score prediction model (quality prediction model), and the input device analysis data (sensory evaluation score unknown) A sensory evaluation score (predicted value) corresponding to the cacao mass sample instrument analysis data) is calculated. Note that the instrumental analysis data is preprocessed by a known method as necessary before being used for comparison / collation with the quality prediction model.

For example, when an extract (analysis sample) containing a metabolite extracted from cacao mass is analyzed by GC / MS, the analysis results obtained include various metabolite retention times, mass spectra, ionic strength, and the like. .
In this case, among the components detected in the analysis, in particular, succinic acid, uridine, alanine, adenine, glycine, leucine, serine, fumaric acid, threonine, pyroglutamic acid, phenylalanine, tyrosine, shikimic acid, tryptophan, valine, Isoleucine, xylitol, asparagine, aconite, glyceric acid, arabitol, sucrose, lactic acid, aspartic acid, arabinose, guanosine, proline, homocysteic acid, 2-dehydro-D-gluconic acid, glutamic acid, gluconic acid, glycerol, trehalose, pyruvic acid Analysis results of oxaloacetate, threonic acid and sarcosine play an important role in predicting the quality of cocoa mass.
According to the present invention, in the cocoa mass sample with strong acidity, succinic acid, uridine, alanine, adenine, glycine, leucine, serine, threonine, pyroglutamic acid, phenylalanine, tyrosine, shikimic acid, tryptophan, valine, asparagine, In cocoa mass samples with relatively high amounts of lactic acid, aspartic acid, guanosine, and homocysteic acid, fumaric acid, isoleucine, xylitol, aconite, glyceric acid, arabitol, sucrose, arabinose, proline, 2-dehydro It has been shown that -D-gluconic acid, glutamic acid, gluconic acid, glycerol, pyruvic acid, oxaloacetic acid, threonic acid tend to exist as main metabolites.
In addition, in the cacao mass sample with strong astringency, succinic acid, glycine, alanine, threonine, leucine, serine, phenylalanine, adenine, tyrosine, lactic acid, asparagine, tryptophan, isoleucine, valine, glutamic acid, aspartic acid, Homocysteic acid, pyroglutamic acid, uridine, arabinose, sarcosine, guanosine, and 2-dehydro-D-gluconic acid have relatively high amounts of cocoa mass samples with weak astringency, such as xylitol, fumaric acid, glyceric acid, arabitol It has been found that sucrose, gluconic acid, glycerol and proline tend to exist as main metabolites.

In the present invention, the multivariate analysis may be performed by selecting a certain range of data important for flavor quality prediction instead of all the obtained data.
For example, among the analysis results, succinic acid, uridine, alanine, adenine, glycine, leucine, serine, fumaric acid, threonine, pyroglutamic acid, phenylalanine, tyrosine, shikimic acid, tryptophan, valine, isoleucine, xylitol, asparagine, aconite, Selected from glyceric acid, arabitol, sucrose, lactic acid, aspartic acid, arabinose, guanosine, proline, homocysteic acid, 2-dehydro-D-gluconic acid, glutamic acid, gluconic acid, glycerol, trehalose, pyruvic acid, oxaloacetic acid, threonic acid If one or more kinds of data, preferably all data is selected and the multivariate analysis is performed, the quality expressed by the sensory evaluation term “acidity” can be predicted well from the obtained analysis results. Can do.
For example, in GC / MS, each compound can be specified by the retention time.

  In the analysis results, succinic acid, glycine, alanine, threonine, leucine, serine, phenylalanine, adenine, tyrosine, lactic acid, asparagine, tryptophan, isoleucine, valine, glutamic acid, aspartic acid, homocysteic acid, pyroglutamic acid, uridine Select one or more data selected from arabinose, sarcosine, guanosine, 2-dehydro-D-gluconic acid, xylitol, fumaric acid, glyceric acid, arabitol, sucrose, gluconic acid, glycerol, proline, preferably all data If the multivariate analysis is performed, the quality expressed by the sensory evaluation term “astringency” can be predicted well from the obtained analysis result.

Furthermore, the accuracy of the prediction model obtained by the above method can be increased by accumulating data. Therefore, for example, if the tendency of the relative amount of metabolite in cocoa mass becomes clearer, the fingerprint is obtained from the analysis result of the relative amount of metabolite of cocoa mass of unknown quality, and the quality is predicted based on this fingerprint It is also possible to do.

  Hereinafter, the present invention will be described in more detail using examples, but the present invention is not limited to these examples.

<Cacao mass sample>
Fifteen species from Ghana, Ivory Coast, Indonesia, Vietnam, Ecuador, Dominica, and Venezuela were selected as representative cacao mass producing regions and used as samples (15 samples in total).
About these cacao masses, sensory evaluation was performed by the following method and used as cacao masses with known flavor quality.

<Sensory evaluation method>
The sensory evaluation regarding the flavor quality was determined by a sensory test score (sensory evaluation score) by a quantitative descriptive analysis method.
The sensory evaluation method was performed using a 15 cm line scale in which the instruction term (weak on the left side and strong on the right side) was written at 1 mm from both ends. It shows that predetermined | prescribed flavor or taste is so strong that the value of a sensory evaluation score is large.
The evaluation was performed using a trained panel consisting of 10 people aged 20 to 50 years (5 women and 5 men). Panels were selected based on their ability to detect five basic tastes (salt, sweet, sour, bitter and umami) and to recognize differences between cacao masses using a three-point comparison method.

Using the methodology of quantitative descriptive analysis, the evaluation terms for cocoa mass samples were developed by the panel itself. Quantitative descriptive analysis was performed on two sensory attributes of “acidity” and “astringency” defined as follows. Evaluation paper preparation and panelist training were established in three working sessions of 30 minutes each.
Acidity: One of the basic tastes. Typically taste of lactic acid or citric acid.
Astringency: Astringent taste. Typically taste of tannins.

For quantitative descriptive analysis, each of the 15 cocoa mass samples was mixed at 70% by weight cocoa mass, 20% by weight sugar, 10% by weight edible fat and oil, and provided at a temperature of about 55 ° C. Samples and panelists were placed according to the plan to avoid order effects. Three evaluations were performed on all samples. When distributing the samples to the panels, a three-digit random number was assigned. All evaluations were performed in a room that was fully air-conditioned and free from external odors, noise, and obstructions.
The value of the sensory evaluation score was the average of three evaluations of 10 panelists.

<Example 1: Preparation of analytical sample for GC / MS (pretreatment step)>
After the cocoa mass sample was ground in a mortar, 15 mg was placed in a 2 mL Eppendorf tube, and 3 mm zirconia beads were added and frozen in liquid nitrogen. Subsequently, it was pulverized with a ball mill (20 Hz, 1 minute, room temperature). 1000 μL of a mixed solution (MeOH: H 2 O: CHCl 3 = 2.5: 1: 1 (v / v / v)) was added to the pulverized cacao mass and mixed with a vortex mixer. To this mixture, 1.85 μL of apicic acid (8.04 mg / mL) was added as an internal standard. Next, centrifugation (3 minutes, 4 ° C., 16000 × g) was performed, and 800 μL of the supernatant was transferred to a 1.5 mL Eppendorf tube.
Thereto, 400 μL of milli-Q water was added, vortexed and centrifuged (3 minutes, 4 ° C., 16000 × g), and 500 μL of the supernatant was transferred to another 1.5 mL Eppendorf tube. It was concentrated by centrifugation for about 2 hours and then lyophilized overnight.

Next, 50 μL of a pyridine solution (20 mg / mL) of methoxyamine hydrochloride (manufactured by Sigma) was added to the extract as a first derivatizing agent. The mixture was incubated at 30 ° C. for 90 minutes.
Further, 50 μL of N-methyl-N- (trimethylsilyl) trifluoroacetamide (MSTFA: manufactured by GLscience) as a second derivatizing agent was added, and the extract was incubated at 37 ° C. for 30 minutes. Derivatization gave an analytical sample.

<Example 2: Analysis by GC / MS (instrument analysis process)>
1 μL of the analytical sample obtained in Example 1 above was injected into the GC / MS in split mode (25: 1, v / v). The GC / MS apparatus (GC / Q-MS analyzer) used in this example is GCMS-QP2010ultra (manufactured by Shimadzu Corporation), and 0.25 μm CP- connected with AOC-20i (manufactured by Shimadzu Corporation) as an autosampler. 30 m × 0.25 mm coated with SIL8CB low bleed (Varian) i. d. 689CN (manufactured by Agilent) equipped with a fused silica capillary column. The injection temperature was 230 ° C.
The flow rate of helium gas through the column was 1.12 mL / min. The column temperature was kept isothermal at 80 ° C. for 2 minutes, then increased to 330 ° C. at 15 ° C./minute and kept isothermal for 6 minutes. The transfer line and ion source temperatures were 250 ° C. and 200 ° C., respectively.
Ions were generated by 70 eV electron impact (EI) and 20 spectra per second were recorded over the mass range of m / z 85-650. The acceleration voltage was activated after a solvent delay of 250 seconds.

<Example 3: Analysis of data obtained by GC / MS>
[Data pre-processing]
In this example, multivariate analysis was performed after pre-processing instrument analysis data by the following method. That is, for the preprocessing of the data obtained by GC / MS in Example 2 above, the obtained raw chromatographic data was converted into an ANDI file (analysis data exchange protocol, * .cdf). Using the ANDI format, it was possible to convert and transfer data between different mass spectral data systems. The converted file (ANDI) was subjected to a data preprocessing procedure and the data points were reprocessed in detail. In addition, data conversion was also performed to obtain the best chromatographic data.
Total ion chromatographic data was then extracted and saved as an ANDI format without fragment data. These files were transferred to the software MetAlign (manufactured by Wageningen University) and aligned.

[Identification of components]
Identify prominent compounds in mass spectra obtained by GC / MS by comparing their mass spectra with the spectra in the library (internal library prepared from NIST library and certified standard chemicals) did.

[PLS regression analysis]
Subsequently, PLS regression analysis (projection to a latent structure by partial least squares method, hereinafter sometimes referred to as PLS) was performed using software AIoutput2 (ver 1.29) (manufactured by Osaka University), and a prediction model was created. PLS finds a relationship between two sets of variables (measurements and response values).

(Multivariate analysis process performed by the prediction model creation unit)
A GC / MS analysis result of each sample obtained was converted into matrix data, and then PLS regression analysis was performed to create a sensory evaluation score prediction model (also referred to as a PLS model in this specification).
That is, the retention time index obtained from the GC / MS chromatogram and the mass analysis signal intensity obtained from the mass spectrum are converted into matrix data (variable name of row: variable name of sample No. column: retention time), PLS regression coefficients as a sensory evaluation score prediction model (PLS model) were created by the PLS method using matrix data as explanatory variables, sensory evaluation scores of each sample as objective variables, and known samples as a training set.

(PLS model for “acidity”)
The case where the sensory evaluation score of “acidity” is used as the flavor quality will be described. First, the analysis samples were divided into training sets and test sets. Sample No. 13, Sample No. 14 and sample no. Fifteen samples were excluded as a test set for model validation.
That is, in the training set, sample No. 12 types of 1-12 were used.
There were no converted variables. All of the variables were aggregated and scaled to an average of 0 and a variance of 1 to reduce the chromatographic data noise effect.

FIG. 1 is a diagram showing a relationship between a predicted value of sensory evaluation by a PLS model of a training set and an actual measurement value by sensory evaluation, and FIG. 2 shows both a test set (triangle plot) and a training set (square plot). It is a figure which shows the relationship between the predicted value of sensory evaluation by a PLS model, and the measured value by sensory evaluation about all 15 types of analysis samples.
The completeness of the model, i.e. the number of potential factors in the PLS model, could be determined by cross-validation, and the optimal number was found in the balance between fit (to the model) and predictability. In addition, the PLS model was validated with a test set and the mean square error (RMSEP) of the predicted values was computed.
Two prominent components were extracted, and for the training set, a 96.6% variation in Y was described (R2 = 0.966), and an 83.1% variation in Y was predicted according to cross validation (Q2 = 0.831). R2 is an index indicating the fitness of the model. The closer this value is to 1, the higher the accuracy of the model. Q2 is an index indicating the prediction performance of the model. The closer this value is to 1, the higher the predictability of the model.
The test set was then predicted with the PLS model, and for the model evaluation based on the training set sample (RMSEE = 0.703), the prediction accuracy of the test sample (RMSEP = 0.835) was obtained. RMSEE is the square root of the mean square error by the training set, and RMSEP is the square root of the mean square error by the test set. The smaller these values, the higher the prediction accuracy.

(PLS model for “Astringency”)
A sensory evaluation score of “Astringency” was used as the flavor quality, and a PLS model was similarly created.
FIG. 3 is a diagram showing the relationship between the predicted value of sensory evaluation by the PLS model of the training set and the actually measured value by sensory evaluation, and FIG. It is a figure which shows the relationship between the predicted value of sensory evaluation by a model, and the measured value by sensory evaluation.
For the training set, a 98.7% variation in Y was described (R2 = 0.987) and a 93.8% variation in Y was predicted according to cross validation (Q2 = 0.938).
For model evaluation based on the training set sample (RMSEE = 0.298), the prediction accuracy of the test sample (RMSEP = 0.786) was obtained.

[Identification of compounds that contribute to prediction performance]
In the PLS model for “acidity” and the PLS model for “astringency”, compounds that have a large contribution to the prediction performance, that is, compounds that are important for predicting the respective sensory evaluation scores were identified.
As a result, in the sensory evaluation term “acidity”, as shown in Table 1, succinic acid, uridine, alanine, adenine, glycine, leucine, serine, threonine, pyroglutamic acid, phenylalanine, tyrosine, shikimic acid, tryptophan, valine, asparagine , Lactic acid, aspartic acid, guanosine, and homocysteic acid contributed positively to "sourness". That is, cocoa mass with a high sensory evaluation score includes succinic acid, uridine, alanine, adenine, glycine, leucine, serine, threonine, pyroglutamic acid, phenylalanine, tyrosine, shikimic acid, tryptophan, valine, asparagine, lactic acid, aspartic acid, guanosine, And there was a tendency to contain a lot of homocysteic acid. Conversely, cocoa mass having a low sensory evaluation score includes fumaric acid, isoleucine, xylitol, aconite, glyceric acid, arabitol, sucrose, arabinose, proline, 2-dehydro-D-gluconic acid, glutamic acid, gluconic acid, glycerol, pyruvic acid, It became clear that oxaloacetic acid and threonic acid contributed.
On the other hand, in the sensory evaluation term “astringency”, as shown in Table 2, cocoa mass having a high sensory evaluation score includes succinic acid, glycine, alanine, threonine, leucine, serine, phenylalanine, adenine, tyrosine, lactic acid, asparagine, Tryptophan, isoleucine, valine, glutamic acid, aspartic acid, homocysteic acid, pyroglutamic acid, uridine, arabinose, sarcosine, guanosine, and 2-dehydro-D-gluconic acid contribute to cocoa mass with a low sensory evaluation score. It was revealed that xylitol, fumaric acid, glyceric acid, arabitol, sucrose, gluconic acid, glycerol and proline contributed.

In Tables 1 and 2, “VIP” is a variable importance for the projection, and is a value that explains X in the PLS model and summarizes the importance of correlated variables with respect to Y.
“Contribution to“ acidity ”” and “contribution to“ astringency ”are related to whether each compound is positive (+) or negative (−) with respect to the flavor. Is determined by PLS loading weight analysis.

This demonstrates that the metabolomics obtained in Example 3 above using GC / MS in combination with chemometrics provides useful information in cocoa mass studies.
It has also been shown that metabolomics can be useful for cocoa mass quality measurement.

According to the present invention, it is possible to predict the quality of cacao mass, which has been difficult in the past, by a simple method. That is, quality can be predicted quickly and beneficially by screening cocoa mass by various instrumental analyzes such as chromatography.
Conventional instrumental analysis of cocoa mass has focused on a specific group of compounds such as amino acids and polyphenols, but according to the method of the present invention, the prediction is based on the results of analyzing multiple components at once. Therefore, more accurate quality prediction is possible, and further improvement of the cacao mass production process or cacao mass distribution storage process can be expected.

Claims (9)

A method for predicting the quality of cocoa mass,
A pretreatment step of pretreating cocoa mass to obtain an analytical sample;
An instrument analysis step of subjecting the analysis sample to instrument analysis to obtain instrument analysis data;
A relationship between the device analysis data and the quality predicted from the device analysis data by performing multivariate analysis using the device analysis data for the plurality of cocoa masses and the quality data representing the quality of each of the cocoa masses A multivariate analysis step for creating a cacao mass quality prediction model representing the cacao mass quality prediction method.   The method according to claim 1, further comprising: a quality prediction step of performing the preprocessing step and the device analysis step on cacao mass of unknown quality and collating the obtained device analysis data with the quality prediction model.   The method according to claim 1, wherein the multivariate analysis is a PLS regression analysis.   The method according to any one of claims 1 to 3, wherein the instrumental analysis is a combination of gas chromatography and mass spectrometry.   The method according to claim 4, wherein retention time and mass spectrometry signal intensity are used for multivariate analysis as the instrumental analysis data.   The method according to claim 1, wherein the pretreatment includes a step of extracting a hydrophilic compound from the cocoa mass to obtain an extract.   The method according to claim 6, wherein the instrumental analysis is a combination of gas chromatography and mass spectrometry, and the pretreatment further comprises silylating the extract.   The instrumental analysis is a combination of gas chromatography and mass spectrometry, and among the analytical results of the instrumental analysis, succinic acid, uridine, alanine, adenine, glycine, leucine, serine, fumaric acid, threonine, pyroglutamic acid , Phenylalanine, tyrosine, shikimic acid, tryptophan, valine, isoleucine, xylitol, asparagine, aconite, glyceric acid, arabitol, sucrose, lactic acid, aspartic acid, arabinose, guanosine, proline, homocysteic acid, 2-dehydro-D-gluconic acid The method according to claim 7 or 8, wherein one or more kinds of data selected from glutamic acid, gluconic acid, glycerol, trehalose, pyruvic acid, oxaloacetic acid, threonic acid, and sarcosine are used for multivariate analysis.   The method according to claim 1, wherein the quality is sour or astringent.

Images