WO2000071993A1

WO2000071993A1 - Near infrared spectroscopy system and method for the identification of genetically modified grain

Info

Publication number: WO2000071993A1
Application number: PCT/US2000/014262
Authority: WO
Inventors: Charles R. Hurburgh, Jr.; Craig A. Heithoff; Glen R. Rippke
Original assignee: Iowa State University Research Foundation, Inc.
Priority date: 1999-05-24
Filing date: 2000-05-23
Publication date: 2000-11-30
Also published as: AU5043000A

Abstract

Genetically modified (GMO) grain can be distinguished from non-GMO grain by a test procedure that makes use of near infrared (NIR) spectroscopy. Calibrations derived for individual GMO events enable positive identification of GMO and non-GMO grains. The procedure is particularly useful at grain handling or delivery points, such as local elevators. System software and hardware can be incorporated in the grain stream to permit analysis of intermittent grain samples. In many cases, the necessary calibrations can be implemented in existing NIR instrumentation.

Description

NEAR INFRARED SPECTROSCOPY SYSTEM AND METHOD FOR THE IDENTIPICAΗON OF GENEΗCALLY

MODIFIED GRAIN

TECHNICAL FIELD

5 The present invention relates to analysis of grain and, more particularly, to identification of genetically modified grain.

BACKGROUND

Identification of genetically modified ("GMO") grain products has become an important issue in agriculture for producers, grain processors and grain exporters. For

1 o example, some new crops are genetically modified to produce grain with a nutrient composition specially designed to benefit a particular animal species, such as chickens or pigs. Such grain can command a premium price, provided that its identity is preserved through the chain of commerce, from the producer to the end user. In addition, some countries have proposed labeling of foods containing genetically

15 modified grain, which may require grain handlers to distinguish between GMO and non-GMO grain.

Simple verification of labels and purchase receipts may not be sufficient to ensure the identity of a particular lot of grain as GMO or non-GMO grain. There is a possibility that non-GMO seed may be inadvertently mixed with non-GMO seed, or 20 vice versa. The problem is complicated by the fact that firms buying the grain at original point of delivery locations, such as local grain elevators, ordinarily are not equipped to identify grain as GMO or non-GMO on site. Therefore, grain purchasers and processors may seek to impose warranty conditions on grain producers.

Wet chemical tests for GMO and non-GMO grain identification are presently 25 available, and are reasonably effective. The tests are technically complicated, however, and can require an extended period of time to obtain the results. Consequently, use of wet chemical technologies for on-site verification at locations such as elevators is generally not feasible.

SUMMARY

30 The present invention is directed to methods and systems for distinguishing between genetically modified (GMO) and non-GMO grains using near infrared (NLR) spectroscopy, either reflectance, absorbance or transmittance. NLR spectroscopy can be used to differentiate GMO from non-GMO grain with a very high degree of accuracy upon calculation of NIR calibrations for individual GMO events. In this manner, NIR spectroscopy can be used to quickly verify the presence or absence of GMO grain.

Verification using NLR spectroscopy can reduce the difficulties associated with identity preservation of grain for seed companies, grain producers, grain processors, grain exporters and purchasers. Moreover, NLR spectroscopy requires little technical expertise and can provide results in a shorter period of time, compared to wet chemistry techniques.

NLR spectroscopy can be applied to analyze samples taken from a grain stream. Alternatively, individual samples can be taken or submitted manually and subjected to NLR spectroscopy. In either case, the use of NIR spectroscopy provides a rapid and accurate technique for identification of GMO grain, with little impact on grain throughput.

The NLR spectroscopy process can be automated, and can be implemented in part by adaptation of existing NLR spectroscopy equipment already used in many grain handling locations. Accordingly, this technique can be especially advantageous for high-throughput facilities, such as seed company production facilities or local grain elevators, where prompt verification, minimal cost, and limited technical intervention are desirable.

The NLR spectroscopy technique can be particularly useful for identification of GMO and non-GMO soybeans and corn. Other types of grains, amenable to differentiation between GMO and non-GMO grains, include small grains such as barley, oats, wheat, rye and rice. It is contemplated that NIR calibrations can be calculated for so-called "stacked" GMO events in grains, providing an indication for each event.

Users may dispense with lengthy, complicated wet chemistry techniques in favor of the more rapid results provided by the simple NLR spectroscopy technique. Indeed, the NLR spectroscopy device can be coupled to a computer that processes the data and generates a discrete positive or negative indication concerning the presence or absence of GMO grain. Users can make use of a catalog of calibrations between GMO and non-GMO products, which take into account differences in spectral signature between grain type, GMO event, and geographic regions. The catalog can be organized to allow selection of the desired calibrations by the user. The catalog can be stored by a computer associated with the NIR spectroscopy device, which permits calibrations to be selectively retrieved for use with spectral data from individual grain samples.

A different calibration ordinarily will be provided for each known GMO event and grain type. Accordingly, as new GMO grains are added to the market, the catalog can be updated, e.g., by shipment on physical media or downloading via the internet. Users may elect to obtain the entire catalog and subscribe to updates. Alternatively, a user may elect to obtain specific portions of the catalog. In this manner, a user can be provided with only those calibrations best suited for grain produced or processed at each location.

In one embodiment, the present invention provides a method for distinguishing between genetically modified (GMO) and non-GMO grain, the method comprising subjecting a grain sample to near infrared spectroscopy to produce spectral data, and determining whether the grain sample contains GMO grain based on the spectral data.

In another embodiment, the present invention provides a system for distinguishing between genetically modified (GMO) and non-GMO grain, the system comprising a near infrared spectroscopy device that subjects a grain sample to near infrared spectroscopy to produce spectral data, and a processor that determines whether the grain sample contains GMO grain based on the spectral data produced by the near infrared spectroscopy device. In another embodiment, the present invention provides a method for distinguishing between genetically modified (GMO) and non-GMO grain, the method comprising measuring the infrared spectral signature of a grain sample, and determining whether the grain sample contains GMO grain based on the analysis.

In a further embodiment, the present invention provides a method for identifying genetically modified (GMO) grain, the method comprising analyzing a grain sample without application of wet chemical assay, and determining whether the grain sample contains GMO grain based on the analysis. In another embodiment, the present invention provides a method for identifying genetically modified (GMO) soybeans, the method comprising subjecting a sample of soybean seed to near infrared spectroscopy to produce spectral data, and determining whether the sample contains GMO soybeans based on the spectral data. Other advantages, features, and embodiments of the present invention will become apparent from the following detailed description and claims.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram of a system for identification of GMO grain;

FIG. 2 is a flow diagram of a calibration and measurement process for GMO grain identification;

FIG. 3 is a graph illustrating NLR spectral data for GMO and non-GMO grain samples;

FIG. 4 is a histogram of calibration values for a first NIR spectroscopy device; and FIG. 5 is a histogram of calibration values for a second NLR spectroscopy device.

Like reference numbers and designations in the various drawings indicate like elements.

DETAILED DESCRIPTION FIG. 1 is a block diagram of a system 10 for identification of GMO grain. As shown in FIG. 1, system 10 includes an NLR spectroscopy device 12, a processor 14, a storage device 16, and a user input device 18. Processor 14 controls device 12 to subject a grain sample 20 to near infrared spectroscopy. Device 12 collects NLR spectral data from grain sample 20 and transmits the data to processor 14. Processor 14 determines whether the grain sample contains GMO grain based on the spectral data. Storage device 16 stores a catalog containing a plurality of calibration equations indicative of differences in spectral characteristics for particular GMO and non-GMO grains. Processor 14 applies the spectral data to a calibration equation retrieved from storage device 16 to calculate a value. The calculated value indicates whether grain sample 20 contains grain with a GMO event. Processor 14 can be realized by a general purpose computer, such as a personal computer or workstation, that executes program code to carry out the NIR spectroscopy process in combination with device 12. Alternatively, processor 14 can be integrated with device 12. Storage device 16, which stores a catalog of calibration equations, can be realized by a hard drive or removable media drive associated with the computer. In addition to code arranged to drive the spectroscopy process, processor 14 may execute statistical, analysis applications for generation of the calibration equations and application of spectral data to the calibration equations to identify GMO grain in grain sample 20. A commercially available software application suitable for analyzing spectral data generated by NIR device 12 is the

Unscrambler™ software marketed by Camo A/S, of Oslo, Norway. User input device 18 can be used by a user to select particular calibration equations or categories thereof, and activate the spectroscopy process. Alternatively, the spectroscopy process can be automated. NLR device 12 can take the form of a commercially available NIR device, such as the Foss Grainspec™ or Foss/Tecator Lnfratec™ NLR instruments, available from Foss North America, of Eden Prairie, Minnesota, USA. Both instruments are presently used in a number of grain elevators throughout the Midwest United States to detect a variety of grain characteristics such as moisture, protein, oil, and fiber content. Both instruments operate in reflectance mode. With appropriate calibration equations, GMO grain identification can be implemented using existing NLR instruments, including those already installed at grain elevators.

New calibrations can be loaded into NLR device 12 or a memory coupled to an associated processor 14, and used to identify GMO grains. A separate calibration is provided for each GMO event in each type of grain and for each type or model of NLR device 12. With adequate computing capacity in processor 14, all available calibrations for a product can be calculated almost simultaneously, providing quick feedback to a user concerning the genetic makeup of a grain sample and the presence of multiple GMO events. Typical GMO grains are obtained by introduction into of plant of a preselected

DNA segment, e.g., an exogenous or recombinant DNA sequence. The DNA is introduced into the genome of the plant by transformation. Each successful transformation event (GMO event) introduces a gene(s) for a desired plant attribute. Useful GMO events include, for example, genes conferring tolerance to herbicides such as glyphosate (Roundup Ready™), bromoxynil, sulfonylureas or phosphinothricin resistance genes, genes conferring insect resistance such as Bacillus thuringiensis endotoxin protein genes and genes for specialty traits such as improved amino acid composition, e.g., dihydrodipicolinic acid synthase genes, lOkD zein genes, 15kD zein genes, aspartate kinase genes, or soybean albumin seed protein genes, as well as antisense or ribozyme genes that inhibit expression of an endogenous gene. Device 12 can be realized by an NLR instrument commonly used in local grain elevators to collect spectral data in the 850-1050 nm range to detect the characteristics of the samples. If device 12 is a Foss Grainspec™, for example, spectral values are collected at 33 wavelengths in the 850-1050 nm range. If device 12 is a Foss/Tecator lnfratec™, 100 spectral values in the same wavelength range are collected. Other NLR devices, which collect reflectance data in the mid-infrared spectral ranges (about 600-2500 nm) can also be used.

NLR devices are currently used to analyze grain for, e.g., moisture, protein, oil, fiber, and saturated fats, and typically complete the analysis in approximately 60 seconds per sample. A wide range of samples is collected each harvest with distinct compositional values to make a range for the different calibrations. The calibrations, a set of multiple linear equations with spectral values, in Beer's Law form ln(l/T), as the independent variable, are statistically based upon the reference values of the samples collected.

For system 10 to detect the difference between the GMO and non-GMO grains, the two types of grains must have a structural or biochemical distinction that would affect NLR spectral properties. The existence of such differences was not previously known. Indeed, previous studies of genetically modified grains have indicated that there are structural and biochemical similarities between GMO and non-GMO grains. FIG. 2 is a flow diagram of a calibration and measurement process for GMO grain identification. As shown in FIG. 2, the calibration process, indicated generally by block 22, involves obtaining a representative product sample set, as indicated by block 24. The sample set may include, for example, one hundred or more different samples of a particular grain type, each sample known to be GMO or non-GMO grain. Each sample is subjected to NLR spectroscopy to collect spectral data, as indicated by block 26. The spectral data are typically collected at a plurality of different wavelengths. As indicated by block 28, the spectral data also can be collected at a number of different sample temperatures to permit identification of temperature- induced variations in the spectral signature. Reference values are then assigned to each sample such that a GMO event is a value of 1, whereas a non-GMO event is a value of 0, as indicated by block 30. Processor 14, for example, can be appropriately programmed to carry out the desired assignment of a reference value.

Next, the spectral data are subjected to pretreatment and statistical analysis, as indicated by block 32. As indicated by block 34, for example, the spectral data can be smoothed, normalized, and filtered prior to application of a statistical analysis technique such as partial least squares (PLS), multiple linear regression (MLR), or neural network (NN) analysis. Such pretreatment differs from known NIR spectral data pretreatments. The usual objective of NIR calibrations is to make the calibrations insensitive to shifts in spectral data across the entire wavelength range. This permits the usual calibrations to identify relative differences in spectra between a reference sample and a test sample that result from quantitative differences in the chemical constituents of the grain (e.g., protein, oil, starch, individual amino acids and the like). In contrast, the pretreatments applied herein include the predicted constituent values of bias-insensitive calibrations in the independent variables. This type of pretreatment allows the statistical model to identify average spectral differences attributable to the GMO event from those attributable to constituents alone.

Although unlike usual NLR pretreatments, the calibrations described herein for detecting GMO and non-GMO grain can be readily implemented in software. Each calibration equation may correspond to one GMO event and may take the form of a multi-term equation, as indicated by block 38. The statistical software calculates the best fit for the wavelength information to the reference value, i.e., discrete values of 1 or 0. An equation is produced with as many terms as independent variables, often in the form y = b0+blXl+ . . . bnXn, where b values are generated by the calibration software. The linear form is not necessary, however, and the inclusion of the previously predicted constituent levels generally tends to make the calibration equations non-linear.

Calibration equations can be prepared and generated at various local sites, if desired. It ordinarily will be more desirable, however, to generate the calibration equations at a central laboratory and then distribute them to users in the field for installation, as indicated by block 40. Calibration equations can be distributed to a virtually unlimited number of NIR systems 10. For example, distribution can be by physical media such as CD-ROM, DVD-ROM and tapes, or by electronic downloads via dial-up or internet connections. In either case, the calibration equations can be widely disseminated, particularly to local grain elevators, and updated from time-to- time when new GMO varieties are being grown in a region. Although a large number of calibration equations may be distributed, a user can select an appropriate category or categories of equations based on the type of grain, the type of GMO event, or the region from which the grain was produced. Once the calibration equations are loaded into the NLR system, the measurement process, generally indicated by block 44, can be applied to identify the presence of GMO grains. As indicated by block 46, it may be desirable to standardize individual NLR devices 12 by analyzing a number of samples having known spectral signatures, e.g., ten to twenty samples, both in the instruments to be used at other locations and in a master unit (the unit that was used to create the calibrations). There are generally small differences in the optical properties of individual instruments that must be accounted for if all instruments are to most effectively utilize the same calibration equations. Preferably, standardization is accomplished by applying a bias to the values calculated for the known samples by an individual unit. Such a bias forces the values calculated by an individual unit to correspond to those obtained on a master unit, and can be achieved in software or hardware. Alternatively, spectral data obtained from the spectroscopy device 12 can be modified in software or hardware to standardize the data to spectral data from the master unit, prior to use of the calibration in the GMO grain analysis process. As another alternative, procedures can be used to adjust, wavelength by wavelength, the spectral data of all individual units before the calibration equation is calculated. It is contemplated that any of the above approaches will achieve satisfactory GMO calibrations. However, the bias approach, which is in effect shifting the rounding point up or down from an initial value of 0.5, is expected to be the simplest and most adaptable method, since calibrations for the presence or absence of GMO grain require a discrete, qualitative classification.

Following standardization, spectral data is collected, as indicated by block 48, for the desired grain samples, indicated by block 50. The grain samples can be manually prepared and placed within the NLR device by known techniques. Alternatively, the measurement process can be readily automated, e.g., using a conveyor that intermittently removes grain samples from a grain stream, prepares them for analysis and delivers them to the NLR device. Upon collection of spectral data, the processor associated with the NLR device computes event calibration values using the calibration equations, as indicated by block 52. The processor may compute the event calibration values using any number of calibrations, as indicated by block 54. In this manner, a sample can be tested against calibrations for multiple GMO events, providing an indication of the presence or absence of each event to the user.

With reference values of 1 and 0, the calibration values can provide a simple "yes" or "no" indication of the presence of a GMO event, as indicated by blocks 56 and 58. By rounding values to 1 or 0, the value is always an integer, as indicated by block 60. In the embodiment of FIG. 2, if the calibration value is greater than or equal to 1, the sample is classified as a GMO grain of the respective event, as indicated by block 62. If the calibration is less than or equal to 0, the grain sample is classified as a non-GMO grain for the respective event, as indicated by block 64. The values could, of course, be reversed, i.e., 1 could be indicative of a non-GMO grain and 0 could be indicative of a GMO event. The objective is to formulate the calibration equations to produce, from the spectral data, a simple indication of the presence or absence of a GMO event. In this manner, the output of the measurement process can be a discrete "yes" or "no," or positive or negative result to be received by the user. An indicator device such as a display, status light, or audible alarm can be coupled to the processor to facilitate notification of a user. With multiple calibration equations, however, it should be noted that the user may receive more than one "yes" or "no" indications if multiple GMO event grain is analysed. Raw measurement values can also be displayed on the indicator device, if so desired by a user. EXAMPLE

The following example illustrates the application of a GMO grain identification technique in accordance with an embodiment of the present invention. Experiments were conducted using Foss Grainspec™ and Foss/Tecator lnfratec™ NIR instruments, available from Foss North America, of Eden Prairie, Minnesota, USA. The spectral data were analyzed with Unscrambler™ software, available from Camo A/S, of Oslo, Norway.

Grain samples were collected from strip-plots located in Bremer and Tama/Grundy counties in Iowa, USA, that contained soybeans known to be either GMO or non-GMO. The GMO soybeans contained a gene conferring tolerance to glyphosate (Roundup™), and are referred to herein as RR soybeans. The non-GMO soybeans did not contain the glyphosate tolerance gene or any other transgene. Of the non-RR soybeans, 35 samples came from the Bremer county plot and 28 came from the Tama/Grundy county plot. Of the RR soybeans, 24 samples came from the Bremer county plot and 27 came from the Tama/Grundy county plot.

Samples from each plot were divided into non-RR and RR categories. The samples were prepared run in the lnfratec and Grainspec devices successively. The NLR data were collected, along with estimated protein, oil, fiber, and saturated fat percentages, calculated from existing calibrations. The Grainspec device collects optical values at 33 wavelengths in the 850-1050 nm range. The lnfratec device collects 100 values in the same wavelength range. The calibration and subsequent user-operated measurement process is illustrated in the block diagram of FIG. 2.

The composition results showed a very strong correlation between the two instruments, as indicated in Table 1 below, and little compositional difference between the two types of soybeans. The nutritional composition of the samples was similar to that of the overall 1998 soybean harvest in Iowa.

FIG. 3 is a graph illustrating NIR spectral data for GMO and non-GMO grain samples. In particular, the graph of FIG. 3 shows the difference in average spectral data between RR and non-RR samples. Spectral data for non-RR samples is indicated by reference numeral 66, whereas spectral data for RR samples is indicated by reference numeral 68. The nearly constant difference means that the constituent calibration could be use to model this, but also that a specific regression intended to find differences would also be successful.

For RR identification, a reference value of zero was input for the non-RR soybeans and a reference value of one was input for the RR soybeans. The ones and zeros replaced chemistry values and were then regressed against the spectral values. Through this process, an equation was derived for each instrument. This equation gave a score for each sample. A histogram of the scores can be found in FIG. 4 (Grainspec) and FIG. 5 (lnfratec). The shape of the histogram shows that the Grainspec had a slight advantage at separating the classes. The two-peak shape in FIG. 4 reflects concentration around two values, 1.0 and 0.0. Inspection of the histogram in FLG. 5 suggested that a lower cutoff value (about 0.4) may be more suitable for classifying grain when using the Lnfratec instrument.

Samples lying on the extreme outer limits of the set (optically) were considered outliers and were not used in the determination of the line. Samples were outliers for a number of reasons. There were several damaged or incomplete scans. Samples were also at the optical extremes to the point where they could not be considered statistically part of the group. If the sample is on the outer edge of a grouping, it is considered not part of the family of centralized samples, which causes it to be recognized as an outlier. If the computed result for a non-RR sample was rounded to one or greater, the sample was identified incorrectly. If the computed result for a RR sample was rounded to zero or less, the sample was identified incorrectly. Overall, the Grainspec and Lnfratec test instruments correctly identified 99.0% and 89.7% of the samples, respectively, as indicated by Table 2 below. The Grainspec device correctly identified a somewhat higher percentage, and both instruments did better at identifying non-RR than RR. A larger database would likely improve accuracy of both instruments. TABLE 2

The lnfratec calibration was used to test 39 additional soybean samples, 20 non-RR samples from Linn County, Iowa, and 19 RR samples from Clay County, Iowa. The results are shown in Table 3 below. With a -0.1 bias, 95% of the non-RR and 84% of the RR samples were identified correctly. The bias meant that the cutoff between non-RR and RR classification, i.e., the point at which there is the least overlap between types, was 0.4 instead of 0.5. The histogram of FIG. 5 had suggested that a lower cutoff value might be preferable. There were no outliers in the validation set. The bias improved the accuracy of identifying RR, but did not affect identification of non-RR soybeans. These results show that calibration equations could be developed to accurately distinguish between RR and non-RR soybeans.

TABLE 3

The foregoing detailed description has been provided for a better understanding of the invention and is for exemplary purposes only. Modifications may be apparent to those skilled in the art without deviating from the spirit and scope of the appended claims.

Claims

CLAIMS:

1. A method for distinguishing between genetically modified (GMO) grain and non-GMO grain, the method comprising:

5 subjecting a grain sample to near infrared spectroscopy to produce spectral data; and determining whether the grain sample contains GMO grain or non- GMO grain based on the spectral data.

2. The method of claim 1, wherein the determining step comprises o applying the spectral data to a calibration equation that defines a spectroscopic difference between a non-GMO grain and a GMO grain to calculate a value that indicates whether the sample contains GMO grain or non-GMO grain.

3. The method of claim 2, further comprising providing an indication of the calculated value to a user. 5

4. The method of claim 2, wherein the spectroscopic data includes spectral data obtained from the grain sample at a plurality of different wavelengths, the method further comprising applying the spectral data as variables to the calibration equation to calculate the value.

5. The method of claim 4, wherein the calibration equation is derived by 0 regression between spectroscopic data for non-GMO grain and GMO grain.

6. The method of claim 4, wherein the calibration equation is derived by multiple-linear regression.

7. The method of claim 4, wherein the calibration equation is derived by partial least squares. 5

8. The method of claim 4, wherein the calibration equation is derived by neural network analysis.

9. The method of claim 1, wherein the spectral data is near-infrared reflectance data.

10. The method of claim 1, further comprising: applying the spectral data to a plurality of different calibration equations to calculate a plurality of values; and determining whether the grain sample contains GMO grain based on the calculated values.

11. The method of claim 10, wherein each of the calibration equations corresponds to a calibration between a non-GMO grain and one of a plurality of different GMO grains.

12. The method of claim 11, further comprising formulating each of the calibration equations by: subjecting a non-GMO grain to near infrared spectroscopy to generate spectroscopic data for the non-GMO grain; subjecting a plurality of different GMO grains to near infrared spectroscopy to generate spectroscopic data for each of the GMO grains; and analyzing differences between the spectroscopic data for the non-GMO grain relative to the spectroscopic data for each of the GMO grains to produce the calibration equations.

13. The method of claim 1 1, further comprising: storing a catalog of different calibration equations in a storage device; adding new calibration equations for newly marketed GMO grains to the catalog in the storage device; and retrieving selected calibration equations from the storage device for use in determining whether the grain sample contains GMO grain.

14. A system for distinguishing between genetically modified (GMO) and non-GMO grain, the system comprising: a near infrared spectroscopy device capable of producing spectral data from a grain sample; and a processor that determines whether the grain sample contains GMO or non- GMO grain based on spectral data received from the device.

15. The system of claim 14, wherein the processor applies the spectral data to a calibration equation that defines a spectroscopic difference between a non-GMO grain and a GMO grain to calculate a value, and determines whether the grain sample contains GMO grain based on the calculated value.

16. The system of claim 14, further comprising an indicator device connected to and capable of receiving output from the processor that provides a user with an indication of the calculated value.

17. The system of claim 15, wherein the spectral data produced by the near infrared spectroscopic device includes spectral data obtained from the grain sample at a plurality of different wavelengths, the processor applying the spectral data as variables to the calibration equation to calculate the value.

18. The system of claim 17, wherein the calibration equation is derived by regression between spectroscopic data for non-GMO grain and GMO grain.

19. The system of claim 18, wherein the calibration equation is derived by multiple-linear regression.

20. The system of claim 18, wherein the calibration equation is derived by partial least squares.

21. The system of claim 18, wherein the calibration equation is derived by neural network analysis.

22. The system of claim 14, wherein the processor is programmed to apply the spectral data to a plurality of different calibration equations to calculate a plurality of values, and determine whether the sample contains GMO grain based on the calculated values.

23. The system of claim 22, wherein each calibration equation corresponds to a calibration between non-GMO grain and one of a plurality of different GMO grains.

24. The system of claim 22, further comprising a storage device connected to the processor that stores the different calibration equations, the processor retrieving the calibration equations from the storage device for the application of the spectral data to the calibration equations.

25. A method for distinguishing between genetically modified (GMO) and non-GMO grain, the method comprising: measuring the infrared spectral signature of a grain sample; and determining whether the grain sample contains GMO grain based on the measurement.

26. A method for distinguishing between genetically modified (GMO) and non-GMO grain, the method comprising: analyzing a grain sample without application of wet chemical techniques, and determining whether the grain sample contains GMO grain based on the analysis.