Night Light Projector,Starry Night Projector,Galaxy Room Projector,Galaxy Starry Night Projector Zhongshan Lianwei Electrical Technology Co., Ltd. , https://www.zslwwaterpump.com
Establishment of Near-infrared Detection Model and Reliability Analysis of Crude Protein and Crude Fat Content in Soybean
Soybean grains contain approximately 40% crude protein and 20% crude fat, and are important plant protein and edible oil sources. Improving the protein and fat content of soybean grain is one of the important goals of quality improvement. Accurate determination of crude protein and crude fat content is a prerequisite. At present, the national crude protein standard determination method is based on the chemical analysis of the Kjeldahl method (GB290521982), and the crude fat standard method is based on the chemical analysis of the Soxhlet extraction method (GB290621982). Although it is generally considered that its reliability is high, the operation steps are tedious and the seeds are crushed so that the progeny cannot be continued. In the process of soybean quality improvement, a large number of germplasm resources, mutants, and hybrid offspring materials need to be identified, screened, and analyzed in a timely manner. At the same time, it is hoped that the grains with good quality can be kept intact after analysis so that they can be further propagated. Therefore, there is an urgent need for an accurate, rapid, simple and non-destructive method of detection of kernels.
Near Infrared Spectroscopy (NIS) is a physical testing technology developed in the late 1980s. It uses the vibration absorption of organic matter in the near-infrared spectral region to quickly determine the content of various chemical components in the sample. The frequency-doubling of various hydrogen-containing groups (CH, OH, NH, SH, etc.) contained in proteins, fats, sugars, etc. The coincidence spectrum band just falls in the near-infrared region to obtain the characteristic vibration information of these organic molecules containing hydrogen groups, thereby determining the content of its chemical components, and has become an important means for analyzing the quality of crops such as rice, wheat, and corn. Compared with chemical analysis, near-infrared spectroscopy is an indirect analytical method. The method for determining the content of a sample component is based on chemical analysis or other instrumental measurements. It is a kind of "regeneration" measurement method. Therefore, any NIR spectrometer calibrates each component or each parameter individually. Therefore, the key to using near-infrared to detect the content of crude protein and crude fat in soybean grains is to construct a correct model and verify the reliability of the model. In fact, the determination of crude protein and crude fat in soybeans is a simpler method. It is measured directly with the instrument. The SZF-06B Crude Fat Analyzer and SZF-06 Crude Fat Analyzer are two types of crude fat analyzers with different models and slightly different functions.
The aim of the study was to construct a near-infrared detection model using different modeling sample sets and use the chemical analysis results as a reference to explore the adaptation range of near-infrared detection instead of chemical analysis.
1 Materials and Methods
1.1 Establishment of Near Infrared Detection Model
Four modeling sample sets were used, and the chemical values ​​were obtained according to GB290521982 and GB290621982 by an authoritative testing center (indicated by M2). Near-infrared detection was performed using a Matrix2I FT-NIR spectrometer from Bruker GmbH, Germany. A suitable amount of soybean seeds was placed in a sample cup, and the near-infrared absorption spectra of soybean seeds were obtained by transmission scanning. Scanning spectra were from 4 000 to 12 600 cm. - 1 with a resolution of 16 cm-1, a scanning speed of 10 kHz, and a number of scans of 60 times. Each sample was reloaded and scanned 3 times. Using the Quant 2 method of Bruker's OPUS 412 software, a model of soybean powder, kernel near-infrared detection of crude protein and crude fat content was established by PLS, and the model was evaluated using cross-validation.
1.2 Comparison of Near Infrared Detection Model Reliability
1. 2. 1 Materials used 8 soybean varieties (lines), crude protein content of 35% ~ 45. 5%, crude fat content of 16. 2% ~ 22. 5%, of which, 1, 2, 3, 5, No. 8 and No. 8 materials are yellow seed coat materials, No. 4 is a seed coat material, and No. 6 is a seed coat material.
1. 2. 2 Detection method The seed of each test cultivar is fully mixed and divided into 3 equal parts (namely repeated 3 times), and randomly numbered as 24 sample numbers. Twenty-four samples were tested for the contents of crude protein and crude fat in the three NIR kernel models (indicated by M5, M6, and M7) and the NIR powder detection model (indicated by M4) established in the study; The authoritative testing department designated by Hebei Province was examined by chemical method (indicated by M1, M2, and M3).
1. 2. 3 Statistical Methods The stability was evaluated by the variance analysis to compare the differences in the coefficient of variation (CV) between the results. Variance analysis and correlation analysis were used to measure the consistency of trends in different test results. The correlation analysis is based on the average number. Analysis of variance, correlation analysis completed by SAS6. 12 software.
2 Results and Analysis
2.1 Near-infrared model establishment
Near-infrared detection model modeling The number of sample sets was 30, 77, 415, and 161, respectively (Table 1). 439,粗脂肪0。 Among them, the powder model sample coverage is relatively narrow, crude protein determination coefficient of 0.9788, crude fat determination coefficient 0. 9601, the largest in four models, cross-validation rms were crude protein 0.439, crude fat 0 . 287, smallest in 4 models. The sample set of the kernel model M6 includes 415 samples, and the sample range of crude protein and crude fat content is the widest among the 4 models. Its crude protein and crude fat determination coefficients are the largest among the three grain models, and cross-validation rms is the smallest among the three grain models.
2.2 Stability Analysis of Different Test Results
The coefficient of variation, CV, of repeated observations was used as a measure of the stability of the test results (Table 2). Crude protein content of three kinds of grain model test results CV average of 0.008, powder model CV is 0. 019, 3 chemical analysis results of CV average 0. 014; crude fat content of 3 kinds of grain model test results of CV average 0. 008, the powder model CV was 0.024, three chemical analysis CV averaged 0.013. The results showed that the three kinds of grain model test results have higher stability when detecting crude protein and crude fat content.
2.3 Consistency Analysis of Different Test Results
2.3.1 Analysis of Variance From the overall results of the eight tested varieties (Table 3), there were significant differences in the results of crude protein content detection among the three chemical testing centers, and there was no significant difference in the crude fat content test results. Near-infrared grain models M5, M6 and M7 and their modeling data sources chemical analysis of M2 crude protein content and crude fat content were not significantly different at 0.01 extremely significant levels. The analysis of variance of the crude protein and crude fat of the 8 tested varieties was analyzed by 7 methods. The results of the analysis (Table 4 and Table 5) showed that there were differences in the results of the 3 chemical tests; M5 and M7 measured higher levels of crude protein in the tested black seed coat 6 and 3 of the NIR models compared to 7 other tested varieties. There was no significant difference in the results of the test, and both were consistent with the results of the chemical analysis M2 of its modeled sample data source.
2.3.2 Correlation analysis The consistency of the detection results of different test methods for different species is a concern of researchers. In order to visually observe whether the trends of different test results are the same, the results of crude protein and crude fat of M2 to 8 test cultivars are plotted on the abscissa, and M5, M6 and M7 are used to detect crude protein and crude fat of 8 cultivars. The results are plotted on the ordinate and plotted separately (Figure 1). It can be seen from the above that the crude protein detection results of M5, M6, and M7 on the eight tested varieties are more linear than the crude fat detection results and the chemical analysis result M2. The linear relationship between M6 and M2 is stronger than M5 and M7. In the results of M5 and M7, it was clearly seen that the detection results of crude protein of variety 6 deviated from the linear relationship, and the near-infrared result was high. The correlation analysis showed that the crude protein detection results of the three grain models were significantly positively correlated with the M2 detection results, and the correlation coefficients of the M6 ​​and M2 chemical analysis were the highest. From the crude fat content test results, the three grain models were significantly positively correlated with M2, and the correlation coefficients of M6 and M2 chemical analysis were also the highest.
3 Conclusions and discussions
The precondition for using near-infrared technology to detect the content of crude protein and crude fat is to build a near-infrared detection model by using as many samples as possible, covering a wide range of samples and rich sample types. The M6 ​​model with the largest sample volume in the model set has the highest reliability. This is consistent with the results of a near-infrared detection model for rice protein content established by Wu et al. [7]. Therefore, when establishing a near-infrared detection model, a wide range of samples with a wide coverage should be used as a sample set.
The model was determined to have a minimum coefficient of 0. 8168, a maximum of 019,788, and an average of 0.9231. According to Wu Jinhong et al. [11], the determination coefficients of the three near-infrared detection models for protein content in rice were 01805, with a maximum of 0.909 and an average of 0.871.方。 Yan Yan et al established corn kernel oil near-infrared detection model determination coefficient reached 0. 9176. Although the near-infrared model has a high coefficient of determination, people are generally more willing to accept the results of the chemical analysis and are skeptical about the NIR results. The analysis results show that the coefficient of variation of the near-infrared grain model is less than that of the chemical analysis, and the quality of the soybean is analyzed by using near-infrared detection techniques such as Yao Xinxuan. The standard deviation of the near-infrared results is smaller than the standard deviation of the chemical analysis results. Tests and t-tests showed no significant differences between the two methods. The number of manually operated steps in the chemical analysis process is far more than the manual operation steps in the detection of the near-infrared grain model, which is a possible reason for reducing the stability of the chemical analysis.
From the result of the consistency of results, the near-infrared results are consistent with the chemical analysis results of its modeling sample data sources, and the consistency of the three grain models and M2 on the detection results of different species is better than three chemical analysis (M1, M2, M3) The consistency of results, especially the highest correlation coefficient between M6 and M2. It was shown that using the sample set with a large number of samples, a wide range of coverage, and a rich sample type set up a near-infrared detection model to detect the crude protein and crude fat content of soybean, and the results consistent with the chemical analysis can be obtained, which has good stability and reliability. NIR detection is an effective alternative to chemical analysis for the quality studies of low-generation selection, QTL s/gene mapping, introgression screening, and mutant screening in seed materials that need to be preserved. Near-infrared spectral characteristics in addition to protein, fat, grain size, grain color is also an important factor. Through analysis of variance within all species, it was shown that near-infrared detection may make the crude protein detection result of black seed coat higher, which may be due to the fact that dark seed coat has a certain influence on the absorption of near-infrared spectrum, and the sample set up to establish the model is in turn Mainly yellow seed coat materials. Therefore, when building a near-infrared model, a special model should be constructed based on the color of the seed coat.