Determination of Total Saccharide Content in Auricularia auricula Based on Near-Infrared Spectroscopy

In this paper, the content of total saccharide in Auricularia auricula from different regions was determined. ,en, near-infrared (NIR) technology was used to collect the spectral information of the samples.,e sample data were divided into calibration set and validation set.,e best quantitativemodel of the total saccharide content ofA. auriculawas established by selecting the parameters such as spectral range, pretreatment method, and partial least square method (PLS) main factor number of the calibration set data. ,e validation set data were used to verify the reliability of this model. In this model, the original spectrum was used to preprocess by standard normal variate (SNV) + second derivative (SD) to eliminate the scattering effect caused by uneven particle distribution and the influence of noise on spectral data. ,e spectrum range was 4000–10000 cm, and the final choice of PLS main factor number was 11. Under this condition, the calibration set Rc of the model was 0.9092, the root mean square error of calibration (RMSEC) was 1.405, the root mean square error of prediction (RMSEP) was 1.507, and the residual predictive deviation (RPD) was 3.32. ,e validation samples were used to test the model, and the result showed that Rv � 0.9048 of the validation set. ,e result proved that the predicted value of the validation samples had a good linear relationship with the measured value. According to the T-test of the two sets of data in the validation set, the difference between the predicted value and the chemical value was not significant (P≥ 0.05). ,e results were in line with the expected objectives. ,e established NIR quantitative model can be used to predict the total saccharide content of the black fungus sample to be tested.


Introduction
Auricularia auricula is the fruiting body of edible fungus used as medicine and food [1,2], which contains active ingredients such as cellulose, polysaccharide, and minerals such as calcium, iron, and phosphorus [3]. A. auricula polysaccharide (AAP) is one of the main effective components and has biological activities such as antitumor, lowering blood glucose, anticoagulant, antiaging, and enhancing immunity [4][5][6][7]. erefore, A. auricula has great potential for medicinal development. At present, the methods of AAP extraction include water extraction, acid-base extraction, ultrasonic-assisted extraction, and enzyme extraction, but these methods usually take a lot of time and manpower and damage the original medicine to a certain extent [8]. Because of the complex composition of polysaccharides, it is difficult to determine them directly. In practice, the content of total saccharide is often used to replace the content of polysaccharide to evaluate the quality of food. So, it is necessary to find a rapid and accurate method to determine the total saccharide content of A. auricula.
In recent years, near-infrared (NIR) spectroscopy has attracted more and more attention because of its fast detection speed, environmental protection, and no loss of samples. Especially since the 1990s, NIR had been widely used in industrial fields, such as applied physics, textile, architecture, history, pharmacy, and food [9][10][11][12][13][14][15]. In the near-infrared spectrum analysis, the commonly used modeling methods include multiple linear regression (MLR), principal component regression (PCR), partial least squares (PLS), and other linear correction methods, as well as artificial neural network (ANN) and support vector machine (SVM) and other nonlinear correction methods [16][17][18].
In this paper, the PLS method was used as the modeling method to establish the near-infrared total saccharide prediction model of A. auricula.

Materials and Methods
2.1. Materials. 166 batches of A. auricula were collected from different regions in China during 2018-2019, and the specific information was shown in Table 1. All samples were sealed and stored in a cool and dry place.

Preparation of Test Solution.
e sample of A. auricula was ground, passed through a 20 mesh sieve, and then 0.25 g of A. auricula powder was precisely weighed and placed in a 250 mL conical flask. 50 mL water and 15 mL concentrated hydrochloric acid were added, heated, and refluxed in a 100°C water bath for 3 h. After cooling to room temperature, the filter residue was washed with distilled water, the filtrate and washing solution were combined, and the volume of water was fixed to 250 mL to prepare the test solution.

Drawing of Standard
Curve. 0.1, 0.2, 0.4, 0.6, 0.8, and 1.0 mL of 100 mg/L glucose standard solution were accurately taken to 10 mL test tube with plug, and then distilled water was added to make up to 10 mL. 1.0 mL of 5% phenol solution was added into the test solution, and then 5.0 mL of concentrated sulfuric acid was added. e reaction solution was mixed by using a rapid mixer. en, the test tube was placed in a water bath at 30°C for 20 min, and an appropriate amount of reaction solution was taken to measure the absorbance at 490 nm. e standard curve was drawn, and the standard curve equation was calculated with the absorbance value as the ordinate and the glucose concentration (mg/mL) as the abscissa.

Determination of Sugar Content in
Samples. 0.2 ml of the test solution was accurately piped into a 10 mL tube with plug, and distilled water was added to make up to 1.0 mL. e absorbance was measured according to the above steps. e total saccharide content w was calculated by the standard curve, and the value was expressed in percentage (%). e calculation formula is as follows: where v 1 is the constant volume of sample, V 2 is the volume of the sample solution taken during colorimetric determination, m 1 is the sugar content in the sample solution which can be obtained from the standard curve, m 2 is the sample quality, and ω is the water content of the sample.

e Method of Near-Infrared Spectrum
Collection. 166 batches of A. auricula powder samples (80 mesh) were spread separately on the integrating sphere of the NIR spectrometer, and the samples should be completely covered with the integrating sphere. e scanning spectrum range was 4000-10000 cm −1 , the scanning times were 64, and the resolution was 8 cm −1 . e original spectral data of sample were shown in Figure 1.

Method and Verification of Establishing Quantitative
Model.
e spectral data of samples were analyzed by Spectrum Quant software, and partial least square (PLS) method was selected as the modeling method. Four indexes, RMSEC (root mean square error of calibration), RMSEP (root mean square error of prediction), R 2 (coefficient of determination), and RPD (residual predictive deviation), were used to evaluate the performance of the model. Generally, the larger the R 2 , the smaller the RMSEC and RMSEP, and the better the model prediction effect. And, the smaller the difference between RMSEC and RMSEP, the better the generalization ability of the prediction model. Generally, the ratio of RMSEC/RMSEP needs to be controlled between 0.8 and 1.2. RPD is the final evaluation index of the model, and its evaluation standard adopts the threshold segmentation method [19][20][21]. When RPD <2.0, it indicates that the model is very poor and cannot be applied; when RPD is 2.0-2.5, it indicates that the model can make a rough prediction and correlation evaluation of the samples; when RPD is 2.5-3.0, it indicates that the model is better and can be used for quantitative analysis of samples; when RPD> 3.0, it indicates that the model has excellent predictive ability and can be used for model analysis [22]. e calculation formula of RPD is as follows:

Determination Results of Total Saccharide Content of
A. auricula. According to formula (1), the measurement results of the total saccharide content in 166 batches were shown in Table 2.

Removal of Model Abnormal Samples.
In this study, software was used to identify the abnormal samples automatically. e PCA score map was used to cluster analyse the spectral data of A. auricula samples, and the differences among the samples were counted. Diagnosis of abnormal data by test of studentized residual and leverage values method [23]. Finally, 47 batches of abnormal samples were removed from the model. e modeling process was shown in Figures 2 and 3.

Division of Calibration Set and Verification Set.
e calibration set and verification set were selected from all samples randomly. e following two conditions must be met during the selection process: ① ensure that the total saccharide content of the verification set samples covers the content gradient of all samples. ② Ensure that the content range of the selected verification set samples does not exceed the calibration set.
In Table 3, the samples of the verification set were completely distributed within the range of the calibration set, indicating that the division of the sample set meets the modeling requirements.

Selection of Spectral Range for Modeling.
e calibration model is established by choosing appropriate spectral range which has strong prediction ability, and good robustness can often be obtained [24]. In Figure 1, the sample had obvious characteristic peak absorption at 5100 cm −1 and 7000 cm −1 . Meanwhile, R 2 , RMSEC, RMSEP, and RPD were used as model evaluation indicators to establish the model in four spectral intervals. e appropriate spectral range was selected by comparing the modeling results of different spectral intervals. e results are shown in Table 4. In Table 4, when the spectral range was 4000-10000 cm −1 , the values of RMSEC and RMSEP were the smallest, and the model effect was the best.

Selection of Spectral Pretreatment Methods.
It is necessary to preprocess the original spectrogram. e robustness and accuracy of the model can be improved by preprocessing the spectrogram. Common preprocessing methods include S-G convolution smoothing, FD (first derivative), SD (second derivative), SNV (standard normal variable transformation), and MSC (multivariate scattering correction). In this study, R 2 , RMSEC, and RMSEP were used as model evaluation indicators to compare the modeling results of different pretreatment methods and selected the appropriate spectral pretreatment method.
In Table 5, the best pretreatment method of total saccharide model of A. auricula was "SNV + SD." e spectrum of A. auricula after pretreatment was shown in Figure 4.

Determination of the Main Factor of PLS.
When the method of partial least squares was used to establish the model, the appropriate number of PLS main factors was related to the quality of the model directly. If the number of PLS main factors was too small, it would cause the loss of useful information in the original spectrum, which would often reduce the predictive ability of the model. If the number of PLS main factors was too much, some information that was not related to the sample would be included.
is situation would lead to a larger prediction error of the model [25]. According to the comprehensive evaluation index R 2 and RMSEP, the number of main factors used in PLS modeling was 11.

External Verification.
e near-infrared quantitative model was used to predict the total saccharide content of the validation set samples, and the predicted value of the quantitative model and the measured chemical value were shown in Table 6 and Figure 5. In Figure 5, R 2 � 0.9048 was higher than 0.9 of the verification set, which indicated that there was a good linear relationship between the predicted value and the measured value of the model. en, the statistical software of SPSS was used to test the predicted results with paired T-test. e set confidence was 95%. e results showed that there was no significant difference between the predicted value and the measured chemical value of the quantitative model (P � 0.182 ≥ 0.05).

Evaluation of the Model.
e parameters of the quantitative model were shown in Table 7, and the linear scatter diagram of the chemical value and the predicted value of the model was shown in Figure 6. e best near-infrared model of total saccharide content of A. auricula was obtained by the screening of modeling parameters.

Discussion
e quantitative analysis of edible mushroom mainly focused on the determination of water, protein, polysaccharides, triterpene, and other components. Although the chemical standard analysis method can detect the total saccharide content, its operation steps are tedious and the efficiency is low; there are some limitations in the practical application process. In the research on the detection of effective components in Cordyceps militaris, Wang Di [26] proved that it was feasible to use near-infrared technology combined with PLS algorithm to achieve rapid prediction of effective components in edible mushroom. Li Junshan [27] directly measured the water content in Poria cocos by nearinfrared technology combined with PLS method and established a prediction model. e optimal number of the latent variables for the model was 8. e coefficient of determination (R 2 ) for the model was 0.998. e RMSECV was 0.453, and RMSEP was 0.366.
e results of the model showed that the method of determining the water content of Poria cocos by near-infrared technology was faster and simpler than other methods. Wu Lun [28] determined the saponins content of the Honey-Fried Processing of Rhizoma Cimicifugae by near-infrared technology. rough the study of the band range, the spectrum used for modeling was finally determined, and the intervals are 5200-6700 cm −1   According to the documentation, the combination of near-infrared spectroscopy and chemometrics to establish a prediction model can quickly and accurately predict the content of the tested substance, and it is suitable for the rapid detection of nutrients in edible mushroom.

Conclusion
In this study, the near-infrared model of the total saccharide of A. auricula established had 104 calibration sets and 15 validation sets, the best pretreatment method was SNV +SD, the best spectral range was 4000-10000 cm −1 , and the best PLS main factor number was 11. R 2 , RMSEC, RMSEP, and RPD of the quantitative model were 0.9092, 1.405, 1.507, and 3.32, respectively. e predicted value of the model had a good linear relationship with the measured value. According to the T-test of the validation set, the difference between the predicted value and the measured chemical value was not significant (P ≥ 0.05). e results showed that the quantitative model had good fitting ability, and it is feasible to determine the total saccharide content of A. auricula by near-infrared spectroscopy. is method is simple and accurate, which can effectively determine the total saccharide content of A. auricula. It can provide technical reference for further expanding the application of near-infrared spectroscopy in the determination of total saccharide content, and it is expected to be used in the detection and analysis of the quality of traditional Chinese medicine.

Data Availability
e data used to support the findings of this study are included within the article.

Conflicts of Interest
All authors declare that there are no conflicts of interest regarding this study. Chemical value (%) Figure 6: e predicted vs. measured value of the optimal model for total saccharide.