CN102841072A - Method for identifying transgenic rice and non-transgenic rice based on NIR (Near Infrared Spectrum) - Google Patents

Abstract

The invention discloses a method for identifying transgenic rice and non-transgenic rice based on NIR (Near Infrared Spectrum), which comprises the following steps of: (1) transmitting the NIR to the rice seed samples, and collecting the diffuse reflection spectrum information of all the rice samples; (2) respectively preprocessing the diffuse reflection spectrum information for all the rice seed samples, extracting the spectrum information in characteristic spectrum regions after preprocessing through a principal component analysis method, selecting principal components, and acquiring the scores of the principal components; (3) building a model by the principal component scores corresponding to the rice seed sample spectrum information as an input and the rice seed type set values corresponding to the rice seed samples as an output; and (4) acquiring the scores of the principal components of the spectrum of rice seeds to be detected, taking the scores into the model in the step (3), and obtaining the types of the rice seeds to be detected. The invention is high in identification precision, simple to operate, low in cost, and is capable of realizing quickly and losslessly identifying the transgenic rice and the non-transgenic rice.

Description

A method for identifying transgenic rice and non-transgenic rice based on near-infrared spectroscopy

technical field

The invention belongs to the detection field of transgenic plants, and in particular relates to a method for discriminating transgenic rice and non-transgenic rice based on near-infrared spectroscopy.

Background technique

Rice is the largest grain crop in my country, and its annual output accounts for about 38% of the total grain output. Its production is related to the food security of the country. Transgenic technology breaks through the limitations of traditional rice breeding and provides a new way to ensure my country's food security. With the rapid development of rice genetic engineering technology, the research results of genetically modified rice are remarkable. In the past 20 years, the use of transgenic technology has successfully developed insect-resistant, disease-resistant, herbicide-resistant, stress-resistant, high-yield and high-quality transgenic rice, and obtained a large number of genetically modified rice lines with stable inheritance and expression of target traits and excellent agronomic traits.

As a large number of genetically modified crops gradually enter the market, the safety of genetically modified crops and food processed by genetically modified crops has also begun to attract people's attention. Essentially, GM crops are no different from conventionally bred varieties. Conventional breeding is generally achieved through sexual hybridization, while plant genetic engineering uses techniques such as Agrobacterium, gene gun, electric shock, and microinjection to introduce exogenous recombinant DNA into the plant genome. Traits and phenotypes should be more accurately predictable and safer in application, but it is still necessary to conduct safety tests on genetically modified crops.

The detection methods of genetically modified ingredients can be divided into qualitative detection methods and quantitative detection methods. At present, the commonly used detection methods of transgenic crops include PCR detection method, chemical tissue detection method, enzyme-linked immunosorbent assay, exogenous gene integration identification method, Westren hybridization method, biological assay detection method and so on. Conventional detection methods for some transgenic plants can no longer meet the needs of rapid and accurate detection. Some of these methods require membrane transfer and hybridization, which are cumbersome and expensive to operate, and some are not suitable for the detection of batch samples, which severely limits their application. The development trend of genetically modified detection technology is simple operation, low cost and strong applicability.

The invention patent application with the application publication number CN 102081075A discloses a method for identifying transgenic rice and non-transgenic rice, including: (1) separately measuring shikimic acid and The content of galacturonic acid; (2) if the content of shikimic acid and galacturonic acid of the rice to be detected is significantly lower than the content of shikimic acid and galacturonic acid of non-transgenic rice, the rice to be detected is transgenic Rice: As the detection methods of shikimic acid and galacturonic acid are relatively complicated, the operation steps of this invention are cumbersome and time-consuming.

Contents of the invention

The invention provides a method for identifying transgenic rice and non-transgenic rice based on near-infrared spectroscopy, and the method can quickly, non-destructively and simply identify transgenic rice and non-transgenic rice.

A method for discriminating transgenic rice and non-transgenic rice based on near-infrared spectroscopy, comprising the following steps:

(1) Transmit near-infrared spectra with wavenumbers ranging from 4000 to 10000 cm ^-1 to rice seed samples, and collect diffuse reflection spectral information of all rice seed samples;

(2) Preprocess the diffuse reflectance spectral information of all rice seed samples respectively, use the principal component analysis method to extract the spectral characteristic information in the characteristic spectral section, select the principal components, and obtain the score of each principal component;

(3) With the respective principal component scores corresponding to the spectral information of all rice seed samples as input, the rice seed type setting value corresponding to the rice seed samples is used as output to establish a model;

(4) According to the operations of steps (1)-(2), obtain the scores of the principal components of the spectrum of the rice seeds to be tested, and bring them into the model described in (3) to obtain the types of rice seeds to be tested.

Diffuse reflection light is the light that returns to the detector after the near-infrared light enters the sample to be tested, undergoes multiple reflections, refraction, diffraction, and absorption, and interacts with the molecules inside the sample to be tested. The structure and composition information of the sample to be tested. Since the internal structure and composition of transgenic rice seeds and non-transgenic rice seeds are different, combined with spectral information processing technology, the difference information can be extracted, and the small differences between spectral image information can be further highlighted and transformed through effective data processing methods. It is an identifiable signal for the identification of transgenic rice and non-transgenic rice.

In step (1), the rice seed samples are transgenic rice seed samples and non-transgenic rice seed samples, and the transgenic rice is transprotein gene rice or transregulated gene rice.

In step (2), in order to remove the influence of factors such as high-frequency random noise, baseline drift, and sample inhomogeneity, the spectrum needs to be preprocessed.

The selection of spectral preprocessing method affects the preprocessing effect and the extraction of effective spectral information. Through comparative analysis, the preprocessing method is preferably the standard normal transformation method.

The characteristic spectrum section refers to a section with obvious differences in the spectral information of transgenic rice seeds and non-transgenic rice seeds, and the identification of transgenic rice seeds and non-transgenic rice seeds can be realized by digitally processing the spectral information of the spectral section ; The characteristic spectral section in the present invention is preferably a spectral section with a wavenumber ranging from 4000 to 10000 cm ^-1 or 4000 to 8000 cm ^-1 .

The principal component analysis (PCA) is used to extract spectral feature information. The purpose of principal component analysis (PCA) is to reduce the dimensionality of spectral data, transform the original variables into a linear combination of a set of new variables that are orthogonal to each other, and eliminate the overlapping of multiple variables. information, at the same time, the new variable can best characterize the data structure characteristics of the original variable. Because the number of new variables of principal component analysis is small and irrelevant to each other, it is more conducive to the analysis of spectral information. The principal component analysis method can be implemented on Unscrambler software (manufactured by CAMO, USA).

The selection of principal components is closely related to the effective extraction of original information. When selecting principal components, the cumulative contribution rate should be ≥ 85%, but the selection of principal components should not be too many, otherwise unnecessary noise will be introduced and overfitting will be caused; When the standard normal transformation is used for preprocessing, the principal component score is preferably 4-8 through comparative analysis.

Step (2) is preferably: use the standard normal transformation method to preprocess the diffuse reflectance spectra of all rice seed samples, and use the principal component analysis method to extract the pretreated wavenumber range from 4000 to 8000cm ^-1 in the spectral information The spectral feature information of , select 5 principal components, and obtain the score of each principal component.

Step (2) is preferably: using the standard normal transformation method to preprocess the diffuse reflectance spectra of all rice seed samples, and using the principal component analysis method to extract the preprocessed wavenumber range from 4000 to 10000cm ^-1 in the spectral information The spectral feature information of , select 6 principal components, and obtain the score of each principal component.

In step (3), the rice seed types corresponding to the rice seed samples are transgenic and non-transgenic, and the set values of transgenic and non-transgenic can be set to -1 and 1, respectively, in the model.

Described model is preferably partial least squares discriminant analysis (PLS-DA) model, and PLS-DA method is a kind of discriminant analysis method based on PLS regression, because of taking into account the class membership information that auxiliary matrix provides with code form when constructing factor , so it has efficient identification ability and can improve the accuracy of rice type identification.

Compared with the prior art, the beneficial effects of the present invention are:

(1) The present invention is simple to operate, saves time and effort, and only needs to obtain the near-infrared spectrum of rice seeds to identify transgenic and non-transgenic rice;

(2) The invention has high identification accuracy and reliable results, and can realize rapid and non-destructive identification of transgenic and non-transgenic rice.

Description of drawings

Fig. 1 is the regression chart of predicted value and actual value of transgenic rice and non-transgenic rice in verification set of Example 1.

Detailed ways

Example 1

1. Build a model

(1) Select 80 transgenic rice seeds and 40 non-transgenic rice seeds as rice seed samples, and use Nicolet Nexus870 (Thermo Corporation USA) Fourier transform near-infrared spectrometer to emit wavenumbers in the range of 4000-10000cm- ¹ to the rice seed samples The near-infrared spectrum was collected by using OMNIC 6.0 software to collect the diffuse reflectance spectra of all rice seed samples; the number of scans of the near-infrared spectrometer was set to 32, and the resolution was 4cm ^-1 . During collection, the room temperature was controlled at about 25°C, and the humidity was kept stable.

The above-mentioned transgenic rice is TCTP gene transgenic rice and Osmi166 gene transgenic rice, and its acquisition method can be: the transgenic material adopts the mature embryo of rice respectively as explant induction, cultures callus, selects embryogenic callus as transformation recipient, Transgenic rice was obtained by transforming Agrobacterium tumefaciens EHA105 containing plant expression vectors p1301-TCTP and p1301-mi166 into rice calluses and differentiated through a series of screening. The non-transgenic rice is Zhonghua 11 rice.

(2) Use the standard normal transformation method to preprocess the diffuse reflectance spectra of all rice seed samples to obtain the preprocessed spectral information; use the principal component analysis method in the Unscrambler software to extract the wave number range after each pretreatment For the spectral characteristic information in the 4000～10000cm ^-1 spectral information, select the number of principal components as 6, and obtain the scores of each principal component (as shown in Table 1);

(3) Take the principal component scores corresponding to the spectral information of all rice seed samples as input, and use the rice seed type setting value corresponding to the rice seed sample as the output to establish the PLS-DA model; the rice seed type setting corresponding to the rice seed sample The values were set as follows: TCTP-transgenic rice and mi166-transgenic rice were set to -1, and non-transgenic rice was set to 1. Due to space limitations, only the data of 20 rice seed samples are listed here, see Table 1.

Table 1 Part of the database used for model building

NO _{x1 x2 x3 x4 x5} X ₆ Y 1 1.3530 -0.7930 0.7700 1.0820 0.3880 -0.1170 1 2 1.1700 0.5890 0.7120 -0.8540 0.8270 -0.5790 1 3 2.0450 1.2560 -0.0391 -0.9710 0.5480 -0.7620 1 4 -0.4370 -0.1520 1.0850 0.4820 0.4580 -0.3600 1 5 -0.6610 0.1800 0.4670 0.6060 0.5070 -0.5080 1 6 -2.7930 0.7790 -0.2370 0.5790 0.4760 0.1470 1 7 2.0430 -0.8770 -0.3850 0.8900 0.1430 0.2080 1 8 -2.5240 0.3730 -0.0229 0.3550 0.6040 0.3070 1 9 -1.3080 -0.1710 0.5250 0.1620 0.6420 0.0927 1 10 -4.5830 1.6350 -0.3410 1.0520 -0.1640 -0.2080 1 11 -1.0420 -0.1430 -0.5020 -0.8720 -0.6650 -0.7770 -1 12 -2.1080 -0.8060 0.0816 -0.9740 -0.0098 -0.5140 -1 13 1.3600 -2.1210 -0.0541 -0.0623 -0.4610 -0.2890 -1 14 -4.7990 -0.3640 0.1410 -0.7450 -0.6880 -0.3830 -1 15 0.9920 -1.4920 -0.3650 1.3850 -0.7650 -0.1610 -1 16 -0.4150 -0.2960 0.2750 -1.3150 -0.0383 -0.7440 -1 17 0.6530 -1.8010 0.2070 -0.8100 -0.2300 -0.3810 -1 18 -2.5350 -1.3630 0.5500 -0.6420 0.0111 -0.0736 -1 19 -2.1330 -1.1560 0.6900 -0.5170 -0.1900 -0.7030 -1 20 1.3300 -3.3070 0.6430 0.1860 -0.1150 -0.5490 -1

Among them, NO refers to the serial number of the rice seed sample, X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , and X ₆ respectively correspond to the scores of each principal component; Y is the output value.

2. Use the model to predict the rice seed type of the rice seed samples in the correction set

After the PLS-DA model is established, the six principal component scores of the rice seed samples in the correction set obtained in step 1 according to steps (1)-(2) are brought into the established PLS-DA model in step 1 to obtain the output value (such as shown in Table 2); through the output value, the rice seed type of the rice seed sample is determined according to the following principles: when the predicted value is less than 0, it is transgenic rice; when the predicted value is greater than 0, it is non-transgenic rice. Due to space limitations, only the data of 20 rice seed samples are listed here, see Table 2.

Table 2 Predicted rice seed types and actual rice seed types of rice seed samples in the calibration set

NO _{x1 x2 x3 x4 x5} X ₆ Y S ₁ S ₂ 1 1.3530 -0.7930 0.7700 1.0820 0.3880 -0.1170 0.88 1 1 2 1.1700 0.5890 0.7120 -0.8540 0.8270 -0.5790 1.213 1 1 3 2.0450 1.2560 -0.0391 -0.9710 0.5480 -0.7620 1.162 1 1 4 -0.4370 -0.1520 1.0850 0.4820 0.4580 -0.3600 0.976 1 1 5 -0.6610 0.1800 0.4670 0.6060 0.5070 -0.5080 0.793 1 1 6 -2.7930 0.7790 -0.2370 0.5790 0.4760 0.1470 0.665 1 1 7 2.0430 -0.8770 -0.3850 0.8900 0.1430 0.2080 0.222 1 1 8 -2.5240 0.3730 -0.0229 0.3550 0.6040 0.3070 0.604 1 1 9 -1.3080 -0.1710 0.5250 0.1620 0.6420 0.0927 0.635 1 1 10 -4.5830 1.6350 -0.3410 1.0520 -0.1640 -0.2080 0.653 1 1 11 -1.0420 -0.1430 -0.5020 -0.8720 -0.6650 -0.7770 -1.046 -1 -1 12 -2.1080 -0.8060 0.0816 -0.9740 -0.0098 -0.5140 -1.015 -1 -1 13 1.3600 -2.1210 -0.0541 -0.0623 -0.4610 -0.2890 -1.253 -1 -1 14 -4.7990 -0.3640 0.1410 -0.7450 -0.6880 -0.3830 -1.264 -1 -1 15 0.9920 -1.4920 -0.3650 1.3850 -0.7650 -0.1610 -0.717 -1 -1 16 -0.4150 -0.2960 0.2750 -1.3150 -0.0383 -0.7440 -0.456 -1 -1 17 0.6530 -1.8010 0.2070 -0.8100 -0.2300 -0.3810 -1.153 -1 -1 18 -2.5350 -1.3630 0.5500 -0.6420 0.0111 -0.0736 -0.919 -1 -1 19 -2.1330 -1.1560 0.6900 -0.5170 -0.1900 -0.7030 -0.909 -1 -1 20 1.3300 -3.3070 0.6430 0.1860 -0.1150 -0.5490 -1.566 -1 -1

Among them, NO refers to the serial number of the rice seed sample, X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , and X ₆ respectively correspond to the scores of each principal component; Y is the output value, S ₁ is the type of rice seed predicted by the model, S ₂ is the actual rice seed type; in columns S ₁ and S ₂ , 1 represents non-transgenic rice, and -1 represents transgenic rice.

Through data analysis, it can be seen that the coefficient of determination R ² _c of the calibration set is 0.9183, and the root mean square error RMSECV of the calibration set is 0.2695.

3. Use the model to predict the rice seed type of the rice seeds to be tested in the verification set

Take 60 rice seeds to be tested as a verification set, obtain the six principal component scores of the rice seeds to be tested according to steps (1)-(2) in 1, and bring them into the established PLS-DA model in 1, The model output value Y is obtained; through the output value Y, the type of the rice seed to be tested is determined according to the judgment principle in 2. Due to space limitations, only the data of 25 rice seeds to be tested are listed here, as shown in Table 3.

Table 3 The predicted rice seed type and the actual rice seed type of the rice seeds to be tested in the verification set

NO _{x1 x2 x3 x4 x5} X ₆ Y S ₁ S ₂ 1 -2.9310 -0.3970 -0.6650 0.5860 -0.1070 0.3520 0.968 1 1 2 -0.6670 -0.2310 0.1920 0.0401 0.0825 0.2030 0.73 1 1 3 -2.2690 -1.0310 0.8860 -0.3120 -0.0803 0.0724 1.188 1 1 4 -0.5720 -0.3310 0.3430 -0.1160 -0.2660 -0.0198 1.178 1 1 5 -0.7930 -0.4970 0.0681 0.2100 -0.1360 0.5340 0.846 1 1 6 6.3560 0.4730 -0.8840 0.2470 -0.5010 0.1160 0.986 1 1 7 -2.6240 0.2530 -0.0890 -0.3410 0.1510 -0.3470 0.505 1 1 8 -3.1970 0.5690 -0.6570 -0.2880 -0.2240 -0.6270 0.935 1 1 9 -1.0460 1.7940 -0.7750 0.2840 -0.2470 -0.2890 0.978 1 1 10 -0.8210 1.0490 -0.2000 0.4340 -0.1160 -0.1670 0.293 1 1 11 -3.7120 0.7550 -0.0027 -0.1010 -0.3900 -0.4650 0.28 1 1 12 0.2270 0.6260 0.0185 -0.3020 0.2860 0.5710 0.871 1 1 13 -2.8670 -0.9860 0.9000 -0.4200 0.1780 0.3570 -0.722 -1 -1 14 -2.5370 -0.0343 0.8240 -0.5960 -0.1770 0.2700 -0.771 -1 -1 15 2.9340 1.1510 1.5360 -0.7920 -0.3610 0.0080 -0.886 -1 -1 16 3.2440 1.3420 1.3470 -1.1520 -0.3610 -0.3780 -1.093 -1 -1 17 -1.3670 0.9700 -0.4390 0.0960 0.4480 0.0888 -0.959 -1 -1 18 0.0870 0.3080 0.8900 -0.0888 0.5510 0.1600 -1.023 -1 -1

19 -2.4050 0.3110 -0.2150 0.1330 -0.0730 -0.3800 -0.959 -1 -1 20 3.1820 -0.0256 1.6990 0.8590 0.5650 -0.5930 -1.554 -1 -1 twenty one 0.2270 -2.7020 -0.3470 -0.1950 0.2150 -0.6300 -0.45 -1 -1 twenty two -3.4210 -0.4880 -0.1550 0.2390 0.2590 -0.3490 -0.878 -1 -1 twenty three -0.8670 0.3150 0.3190 0.0802 -0.1740 -0.7870 -1.316 -1 -1 twenty four -1.4390 0.1650 0.0678 0.4940 -0.4470 -1.2150 -0.944 -1 -1 25 -0.3240 1.0430 0.0300 -0.2890 -0.1670 0.4830 -0.877 -1 -1

Among them, NO refers to the serial number of the rice seed to be tested, X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , and X ₆ correspond to the scores of each principal component respectively; Y is the output value, and S ₁ is the type of rice seed predicted by the model , S ₂ is the actual rice seed type; in columns S ₁ and S ₂ , 1 represents non-transgenic rice, and -1 represents transgenic rice.

Through data analysis, it can be seen that the determination coefficient R ² _p of the verification set is 0.8979, and the root mean square error RMSEP of the verification set is 0.2878. The regression chart of the predicted value and actual value of transgenic rice and non-transgenic rice in the verification set is shown in Figure 1.

Example 2

1. Build a model

(1) With embodiment 1.

(2) Use the standard normal transformation method to preprocess the diffuse reflectance spectra of all rice seed samples to obtain the preprocessed spectral information; use the principal component analysis method in the Unscrambler software to extract the wave number range after each pretreatment For the spectral feature information in the 4000～8000cm ^-1 spectral segment, select the principal component number as 5, and obtain the scores of each principal component (as shown in Table 4);

(3) Take the principal component scores corresponding to the spectral information of all rice seed samples as input, and use the rice seed type setting value corresponding to the rice seed sample as the output to establish the PLS-DA model; the rice seed type setting corresponding to the rice seed sample The values were set as follows: -1 for TCTP-transgenic rice and mi166-transgenic rice, and 1 for non-transgenic rice. Due to space limitations, only the data of 20 rice seed samples are listed here, see Table 4.

Table 4 Part of the database used for model building

NO _{x1 x2 x3 x4 x5} X ₆ Y

1 1.5760 -1.2890 0.2420 0.7620 0.1610 0.0249 1 2 0.7060 1.4910 -0.1330 0.4010 0.2350 -0.2130 1 3 2.2380 1.8980 -0.1730 0.3890 0.0969 -0.5140 1 4 -1.4210 -0.5080 0.3290 0.9520 -0.0171 -0.1600 1 5 -1.3610 -0.4000 0.2890 0.5180 0.0336 -0.3000 1 6 -3.8860 -0.1460 0.0375 0.2370 0.1880 -0.2860 1 7 3.0420 -1.4200 0.1070 -0.1660 0.1130 0.1830 1 8 -3.5180 -0.0658 -0.0167 0.2420 0.3000 0.0238 1 9 -2.0770 0.1050 0.1180 -0.2780 0.2160 0.0316 1 10 -6.2470 0.3450 0.8330 0.4870 -0.5970 -0.1830 1 11 -1.0590 0.2430 -0.6420 -0.0786 -0.2050 0.1910 -1 12 -3.0900 0.0187 -0.5990 0.2100 -0.0894 0.0323 -1 13 2.2590 -1.2940 -0.2860 -0.2610 -0.1850 0.2650 -1 14 -6.5120 0.3220 -0.6920 -0.2590 0.1780 0.4370 -1 15 2.2500 -2.4760 -0.0601 -0.2320 -0.3630 0.2610 -1 16 -0.9040 0.9580 -0.5490 -0.2850 0.0365 -0.1090 -1 17 1.1180 -0.1560 -0.3270 -0.7240 0.2460 -0.2080 -1 18 -3.4160 -0.1470 -0.2280 -0.2390 -0.0240 -0.0630 -1 19 -3.2530 -0.4290 -0.2160 -0.4260 0.0084 -0.0983 -1 20 2.0300 -1.7920 -0.1390 -0.9330 -0.0915 -0.5240 -1

Among them, NO refers to the serial number of the rice seed sample, X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , and X ₆ respectively correspond to the scores of each principal component; Y is the output value.

2. Use the model to predict the rice seed type of the rice seed samples in the correction set

After the PLS-DA model is established, the six principal component scores of the rice seed samples in the correction set obtained in step 1 according to steps (1)-(2) are brought into the established PLS-DA model in step 1 to obtain the output value (such as shown in Table 5); through the output value, the rice seed type of the rice seed sample is determined according to the following principles: when the predicted value is less than 0, it is transgenic rice; when the predicted value is greater than 0, it is non-transgenic rice. Due to space limitations, only the data of 20 rice seed samples are listed here, see Table 5.

Table 5 Predicted rice seed type and actual rice seed type of rice seed samples in the calibration set

NO _{x1 x2 x3 x4 x5} X ₆ Y S ₁ S ₂ 1 1.5760 -1.2890 0.2420 0.7620 0.1610 0.0249 1.069 1 1 2 0.7060 1.4910 -0.1330 0.4010 0.2350 -0.2130 1.235 1 1 3 2.2380 1.8980 -0.1730 0.3890 0.0969 -0.5140 1.336 1 1 4 -1.4210 -0.5080 0.3290 0.9520 -0.0171 -0.1600 1.159 1 1 5 -1.3610 -0.4000 0.2890 0.5180 0.0336 -0.3000 0.895 1 1 6 -3.8860 -0.1460 0.0375 0.2370 0.1880 -0.2860 0.29 1 1 7 3.0420 -1.4200 0.1070 -0.1660 0.1130 0.1830 0.275 1 1 8 -3.5180 -0.0658 -0.0167 0.2420 0.3000 0.0238 0.379 1 1 9 -2.0770 0.1050 0.1180 -0.2780 0.2160 0.0316 0.369 1 1 10 -6.2470 0.3450 0.8330 0.4870 -0.5970 -0.1830 0.938 1 1 11 -1.0590 0.2430 -0.6420 -0.0786 -0.2050 0.1910 -0.951 -1 -1 12 -3.0900 0.0187 -0.5990 0.2100 -0.0894 0.0323 -0.862 -1 -1 13 2.2590 -1.2940 -0.2860 -0.2610 -0.1850 0.2650 -0.727 -1 -1 14 -6.5120 0.3220 -0.6920 -0.2590 0.1780 0.4370 -1.306 -1 -1 15 2.2500 -2.4760 -0.0601 -0.2320 -0.3630 0.2610 -0.951 -1 -1 16 -0.9040 0.9580 -0.5490 -0.2850 0.0365 -0.1090 -0.438 -1 -1 17 1.1180 -0.1560 -0.3270 -0.7240 0.2460 -0.2080 -0.395 -1 -1 18 -3.4160 -0.1470 -0.2280 -0.2390 -0.0240 -0.0630 -0.608 -1 -1 19 -3.2530 -0.4290 -0.2160 -0.4260 0.0084 -0.0983 -0.773 -1 -1 20 2.0300 -1.7920 -0.1390 -0.9330 -0.0915 -0.5240 -1.068 -1 -1

Among them, NO refers to the serial number of the rice seed sample, X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , and X ₆ correspond to the scores of each principal component respectively; Y is the output value, S ₁ is the type of rice seed predicted by the model, S ₂ is the actual rice seed type; in columns S ₁ and S ₂ , 1 represents non-transgenic rice, and -1 represents transgenic rice.

Through data analysis, it can be seen that the coefficient of determination R ² _c of the calibration set is 0.8578, and the root mean square error RMSECV of the calibration set is 0.3555.

3. Use the model to predict the rice seed type of the rice seeds to be tested in the verification set

Take 60 rice seeds to be tested as a verification set, obtain the six principal component scores of the rice seeds to be tested according to steps (1) to (3) in 1, and bring them into the established PLS-DA model to obtain the model Output value Y; through the output value Y, determine the type of the rice seed to be tested according to the judgment principle in 2. Due to space limitations, only the data of 25 rice seeds to be tested are listed here, see Table 6.

Table 6 The predicted rice seed type and the actual rice seed type of the rice seeds to be tested in the validation set

NO _{x1 x2 x3 x4 x5} X ₆ Y S ₁ S ₂ 1 -3.4170 -1.0060 -0.7280 0.6610 0.0677 -0.2440 0.858 1 1 2 -0.9130 -0.0728 -0.0108 0.1660 -0.0995 0.0692 0.787 1 1 3 -3.6300 -0.3270 -0.0092 -0.0052 -0.1700 -0.1260 0.994 1 1 4 -0.9640 -0.0178 -0.1990 -0.2020 -0.0221 -0.0748 0.713 1 1 5 -1.0860 -0.5690 -0.2700 0.6380 -0.3260 0.0469 0.616 1 1 6 8.0500 -0.3310 -0.3900 -0.0776 -0.2580 0.0654 1.431 1 1 7 -3.0820 0.2200 0.5230 -0.1130 -0.0029 0.4140 0.375 1 1 8 -3.4340 0.1150 0.0832 -0.3410 -0.2630 0.2880 1.216 1 1 9 0.0422 1.4380 -0.3540 -0.0101 -0.5670 0.4380 0.998 1 1 10 -0.2760 0.9760 -0.5780 0.1790 -0.1710 0.3150 0.953 1 1 11 -4.3810 0.9250 -0.0026 -0.9750 -0.4250 0.2840 0.880 1 1 12 0.5170 0.8000 0.6040 0.7250 0.0724 0.1790 0.324 1 1 13 -4.3450 -0.3570 0.4730 0.3960 0.1150 -0.0927 -0.537 -1 -1 14 -3.5570 0.6740 0.4930 -0.0930 0.0196 -0.3320 -1.026 -1 -1 15 3.2640 2.4540 0.1720 0.1120 -0.0863 -0.3330 -0.754 -1 -1 16 3.7650 2.4540 0.8320 -1.1410 0.1980 -0.5470 -1.203 -1 -1 17 -0.9060 0.8170 0.2280 1.0120 0.1660 0.5320 -1.083 -1 -1 18 -0.1260 1.1780 0.2870 0.9450 0.2910 0.4300 -1.299 -1 -1 19 -2.6800 0.2390 -0.3510 0.1930 0.1640 0.4000 -1.048 -1 -1 20 3.0620 1.2720 -1.0030 0.1120 0.6080 0.3440 -1.242 -1 -1 twenty one -0.6180 -3.1850 -0.0892 -0.3880 0.7230 0.4070 0.045 1 1 twenty two -4.3620 -0.6470 -0.0510 -0.2630 0.4750 0.3470 -0.709 -1 -1 twenty three -0.9680 0.6720 -0.3570 -0.7200 0.0012 0.5530 -0.550 -1 -1 twenty four -1.5620 0.3000 -1.0520 -1.3440 0.4130 0.6250 -0.768 -1 -1 25 0.1140 1.1700 0.6600 -0.1340 0.0222 -0.3180 -1.524 -1 -1

Among them, NO refers to the serial number of the rice seed to be tested, X ₁ , X ₂ , X ₃ , X ₄ , X ₅ , and X ₆ correspond to the scores of each principal component; Y is the output value, and S ₁ is the type of rice seed predicted by the model , S ₂ is the actual rice seed type; in columns S ₁ and S ₂ , 1 represents non-transgenic rice, and -1 represents transgenic rice.

Through data analysis, it can be seen that the determination coefficient R ² _p of the verification set is 0.8344, and the root mean square error RMSEP of the verification set is 0.3543.

Claims (7)

1. the method based near infrared spectrum discriminating transgenic paddy rice and non-transgenic paddy rice is characterized in that, may further comprise the steps:

(1) be 4000～10000cm to rice paddy seed sample emission wave-number range ^-1Near infrared spectrum, and gather the spectrum information that diffuses of all rice paddy seed samples;

(2) respectively the spectrum information that diffuses of all rice paddy seed samples is carried out pre-service, utilize PCA to extract the spectral signature information in the pretreated characteristic spectrum section, choose major component, and obtain the score of each major component;

(3) with corresponding each principal component scores of all rice paddy seed sample light spectrum informations as input, as output, set up model with the corresponding rice paddy seed type setting value of rice paddy seed sample;

(4) obtain each principal component scores of rice paddy seed spectrum to be measured according to the operation of step (1)～(2), carry it into model described in (3), obtain rice paddy seed type to be measured.

2. the method based near infrared spectrum discriminating transgenic paddy rice and non-transgenic paddy rice as claimed in claim 1 is characterized in that in the step (2), preprocess method is the standard normal converter technique.

3. the method based near infrared spectrum discriminating transgenic paddy rice and non-transgenic paddy rice as claimed in claim 1 is characterized in that in the step (2), said characteristic spectrum section is that wave-number range is 4000～10000cm ^-1The spectrum section.

4. the method based near infrared spectrum discriminating transgenic paddy rice and non-transgenic paddy rice as claimed in claim 1 is characterized in that in the step (2), said characteristic spectrum section is that wave-number range is 4000～8000cm ^-1The spectrum section.

5. the method based near infrared spectrum discriminating transgenic paddy rice and non-transgenic paddy rice as claimed in claim 1 is characterized in that in the step (2), number of principal components is 4～8.

6. like claim 3 or 4 described methods, it is characterized in that in the step (2), number of principal components is 5 based near infrared spectrum discriminating transgenic paddy rice and non-transgenic paddy rice.

7. the method based near infrared spectrum discriminating transgenic paddy rice and non-transgenic paddy rice as claimed in claim 1 is characterized in that in the step (3), said model is the partial least squares discriminant analysis model.

Images