Method for sorting seeds

The method employs NIR spectroscopy and multivariate analysis to develop discrimination models for selecting seeds with desired traits, addressing inefficiencies in current seed sorting by predicting internal seed quality, thereby reducing 'seed loss' and improving yield.

WO2024248014A1PCT designated stage expired Publication Date: 2024-12-05KYUSHU UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2024/019621
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-30
Filing Date
2024-05-29
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Current seed sorting methods are inefficient in distinguishing high-quality seeds from low-quality ones, leading to significant 'seed loss' due to reliance on external characteristics and inability to fine-tune sorting conditions, especially when biochemical changes within seeds affect germination and yield without visible changes.

Method used

A method using near-infrared (NIR) spectroscopy and multivariate discriminant analysis to construct a dataset from spectral data of seeds, correlating it with quality tests to derive a discrimination model for selecting seeds with desired traits like germination ability, stress tolerance, and disease resistance, allowing for non-destructive and precise selection.

Benefits of technology

Enables the efficient and precise selection of high-quality seeds by predicting seeds with desired traits, improving seed yield and reducing 'seed loss' through accurate discrimination models that account for internal biochemical changes, enhancing seed quality control beyond traditional sorting methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2024019621_05122024_PF_FP_ABST
    Figure JP2024019621_05122024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention provides a method for sorting plant seeds having a desired trait, the method including the following first to seventh steps: (First step) A step for extracting some seeds from a seed group. (Second step) A step for constructing a data set with respect to the extracted seeds, the step for constructing a data set including the following (a) to (c): (a) beaming light onto the extracted seeds and acquiring spectral data and / or spectral data derived therefrom; (b) subjecting the seeds for which spectral data has been acquired to quality inspection with respect to the desired trait, and determining the advance quality rate of the extracted seeds, the advance quality rate being defined by the following formula: [Advance quality rate (%)] = [number of seeds having desired trait from among seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation] × 100; (c) associating the spectral data for each seed and the result of the quality inspection. (Third step) A step for applying multivariate discriminant analysis to the data set constructed in the second step and thereby deriving candidates for a good / poor discrimination model for calculating a discrimination score for the desired trait in each seed. (Fourth step) A step for selecting a discrimination model from the candidates for a good / poor discrimination model derived in the third step, the step including selecting as a good / poor discrimination model a candidate that satisfies the criterion that the conformity rate exceeds the advance quality rate under a standard sorting condition, where the recovery rate and the conformity rate are defined by the following formulae, and a sorting condition that the advance quality rate and the recovery rate have the same value is defined as the "standard sorting condition": [Recovery rate (%)] = [Number of seeds selected on the basis of discrimination score from among seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation] × 100 [Conformity rate (%)] = [Number of seeds with desired trait from among seeds subjected to quality evaluation and selected on the basis of discrimination score] / [number of seeds selected on the basis of discrimination score from among seeds subjected to quality evaluation] × 100 (Fifth step) A step for beaming light onto the remaining seeds not extracted from the seed group in the first step and / or other seeds obtained under essentially the same conditions as those for said seeds and acquiring spectral data, and applying the discrimination model selected in the fourth step to the spectral data to determine the discrimination score for each seed. (Sixth step) A step for determining a threshold for the discrimination score on the basis of the discrimination scores determined in the fifth step. (Seventh step) A step for comparing the discrimination scores of the respective seeds determined in the fifth step with the threshold for the discrimination score determined in the sixth step, and thereby recovering seeds predicted to have the desired trait.
Need to check novelty before this filing date? Find Prior Art

Description

How to Select Seeds

[0001] The present invention relates to a method for selecting seeds, and more particularly to a method for selecting plant seeds having desired traits by analyzing spectral data of the seeds using multivariate statistical analysis.

[0002] The world's population continues to grow, projected to reach 8 billion by 2022, 8.5 billion by 2030, and 9.7 billion by 2050. Therefore, the stable production and supply of food is becoming increasingly important for the survival of human society. One effective way to address this issue is to develop high-yielding varieties and spread them worldwide. However, no matter what variety is used for crop production, the superior characteristics of that variety cannot be fully realized without high-quality seedlings. In ancient Japan, there is a saying, "naehansaku," which means "half of growth depends on the seedlings." Going further, high-quality seedlings cannot be produced without high-quality seeds. Therefore, how to produce and prepare high-quality seeds, how to preserve their quality without deterioration, and how to evaluate their current quality are urgent technological challenges that humanity must address together.

[0003] The importance of seed quality control is strongly recognized by public institutions both in Japan and abroad. The International Seed Testing Association (ISTA) provides guidelines on quality testing methods for almost all seeds of major crop species. The Japanese government also enforces the Seed and Seedlings Act, which sets quality standards for seeds for agricultural production distributed on the market. One reason for such official monitoring is that seed quality (especially germination and genetic purity) is usually very difficult to determine by appearance or other simple criteria. The accidental spread of low-quality seeds could lead not only to regional food shortages but also to poverty among farmers.

[0004] Official quality control certainly helps prevent seed-related crop failures, but there are always two sides to everything. For example, if official quality control requires seeds to meet a 90% quality rate, a seed lot with a quality rate just 1% lower than 90% (i.e., an 89% quality rate) will lose its commercial value. However, in reality, fierce quality competition among seed suppliers has rendered even such legal standards irrelevant or uncompetitive. As a result, a large amount of seeds that have no commercial value continue to be produced despite their potential to contribute to crop production, a situation that can be called ruthless "seed loss."

[0005] Even for major crop species, the quality rate of unsorted seeds after harvest typically falls below 90%. The occurrence of defective seeds is partly due to incomplete seed maturation in the mother plant, but damage, deterioration, or contamination during the post-harvest seed preparation process can also be a contributing factor. Only after multiple stages of sorting can a high quality rate within legal standards be achieved. The large-scale equipment used for such sorting includes spiral gravity separators, trommel rotary separators, vibration separators, magnetic separators, weight separators, and color separators. However, if the final quality rate after sorting falls below the target, the seeds finally prepared through sowing, cultivation, harvesting, and sorting simply end up as a pile of garbage. Currently, there is no way to improve the quality rate even by the "last 1 percent" from this stage.

[0006] The seed sorting devices described above are based on the premise that seed quality is reflected in characteristics detectable from the outer surface or exterior of the seed. Characteristics detectable from the outer surface or exterior of the seed include shape, surface texture, adhesiveness, specific gravity, and color. The above-mentioned devices are suitable for mass sorting and can prepare most harvested seeds to a high-quality state. In contrast, these devices have the disadvantage of being difficult to fine-tune their operating conditions. Furthermore, when focusing on the biochemical characteristics of seeds, changes in the hidden embryo and endosperm can have a much greater impact on seed function than changes in the exposed seed coat. For example, oxidation of endosperm lipids and infestation of the embryo by certain pests do not immediately cause changes in the appearance or weight of the seed, but they significantly impair germination ability. It is presumed that "seed loss" is caused by such defective seeds.

[0007] Near-infrared (NIR) spectroscopy is an effective method for nondestructively evaluating differences and changes in the chemical composition of materials, primarily composed of organic compounds. In other words, NIR spectroscopy has a strong affinity with food, pharmaceuticals, and agricultural products, and its application to their quality control is expanding. Imaging spectroscopy, which has been developed primarily in the field of remote sensing, is also attracting attention as a technology for quality control of these products, which are characterized by heterogeneity. Near-infrared imaging spectroscopy, which combines these new technologies, is highly expected to provide a key to eliminating "seed loss."

[0008] In fact, there have been reports of the application of NIR imaging spectroscopy to assess seed quality in terms of survival, growth potential, genetic purity, and the presence or absence of insect or fungal damage (Non-Patent Documents 1 to 4). However, the fact that seed loss continues to occur clearly indicates that this technology has not yet reached a practical level or has not yet been fully implemented in society.

[0009] Rahman A. et al., Assessment of seed quality using non-destructive measurement techniques: A Review., Seed Science Research, Volume 26, Issue 4, December 2016, pp.285-305.Feng L. et al., Hyperspectral imaging for seed quality and safety inspection: A Review, Plant Methods, 15, Article number: 91 (2019)Venkatesan S. et al., Role of near-infrared spectroscopy in seed quality evaluation: A review. Agric Rev. 2020;41.Long W. et al., A Review of Artificial Intelligence Methods for Seed Quality Inspection based on Spectral Imaging and Analysis., J. Phys.: Conf. Ser., 1769, 012013.

[0010] The methods reported in the above-mentioned prior art documents are all useful because they enable non-destructive seed selection under certain conditions. However, since these prior art methods are limited to application to specific plant seeds and specific traits, there remains a strong need for the development of new, more versatile seed selection methods that can be applied to a wider range of varieties and traits.

[0011] As a result of intensive research into the above-mentioned problems, the present inventors have found that when a data set for each plant seed constructed using an optical analysis method or the like is applied to multivariate discriminant analysis to generate a large number of candidate seed discriminant models, and a discriminant model with high predictive accuracy is selected from the candidate models, it is possible to select a model with high predictive accuracy by introducing the concept of "standard selection conditions," and have completed the present invention by conducting further research based on this finding. That is, the present invention is as follows.

[0012] [1] A method for selecting plant seeds having desired traits, the method comprising the following steps 1 to 7: (step 1) extracting some seeds from a seed population; (step 2) constructing a dataset for the extracted seeds, the step of constructing the dataset comprising the following steps (a) to (c): (a) irradiating the extracted seeds with light to obtain spectral data and / or derivative spectral data thereof; (b) subjecting the seeds for which spectral data has been obtained to a quality inspection for the desired trait, and determining a preliminary rate of acceptable seeds for the extracted seeds, wherein the preliminary rate of acceptable seeds is defined by the following formula: [Preliminary rate of acceptable seeds (%)] = [Number of seeds having the desired trait among the seeds subjected to quality evaluation] / [Number of seeds subjected to quality evaluation] × 100; (c) correlating the spectral data of each seed with the results of the quality inspection; (step 3) applying multivariate discriminant analysis to the dataset constructed in step 2 to derive candidate quality discrimination models for calculating a discrimination score for the desired trait in each seed; (Step 4) defining the recovery rate and the conformance rate by the following formulas: [Recovery rate (%)] = [number of seeds selected based on the discrimination score among seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation] × 100 [Conformance rate (%)] = [number of seeds having desired traits among seeds subjected to quality evaluation and selected based on the discrimination score] / [number of seeds selected based on the discrimination score among seeds subjected to quality evaluation] × 100, and defining the sorting conditions under which the preliminary conformance rate and the recovery rate have the same value as each other as "standard sorting conditions", a step of selecting a discrimination model from the candidate conformance models derived in Step 3, the step including selecting, under the standard sorting conditions, a candidate that satisfies the criterion that the conformance rate exceeds the preliminary conformance rate as a conformance model; (Step 5) irradiating the remaining seeds not taken out of the seed population in Step 1 and / or other seeds obtained under substantially the same conditions as the remaining seeds to obtain spectral data, and applying the discrimination model selected in Step 4 to the spectral data to determine the discrimination score of each seed; (Sixth step) determining a threshold value for the discriminant score based on the discriminant score determined in the fifth step;(Step 7) A step of recovering seeds predicted to have desired traits by comparing the discrimination scores of each seed determined in Step 5 with the discrimination score threshold determined in Step 6. [2] The selection method according to [1], which is a method for selecting plant seeds having two or more desired traits. [3] The selection method according to [1] or [2], wherein the trait is at least one trait selected from the group consisting of germination ability, germination vigor, genotype, stress tolerance, dormancy, disease resistance, insect resistance, QTL characteristics, eating quality, heading date, and morphological characteristics. [4] The selection method according to any one of [1] to [3], characterized in that the prediction of two or more desired traits is performed by any one of the following methods (i) to (iii): (i) a method in which seeds are individually scored using a discrimination model for each trait, focusing on the quality of each trait, and seeds with all scores above a threshold are predicted to be good seeds; (ii) a method in which seeds with all desired traits being good are selected using a single discrimination model, regardless of the quality of each trait, by classifying seeds with good traits as good if all desired traits are good and classifying seeds otherwise as bad, and seeds are scored using the single discrimination model and predicted to be good seeds with scores above a threshold; or (iii) a method in which a discrimination model for each trait is selected using the quality of each trait, and then scores obtained using the discrimination model for each trait are integrated to obtain an integrated score, and the quality of each seed is predicted based on the ranking of the integrated score. [5] The selection method according to any one of [1] to [4], further comprising a step of evaluating the accuracy of the discrimination model. [6] The selection method according to any one of [1] to [5], wherein the plant is a horticultural crop.

[0013] According to the present invention, plant seeds having desired traits can be selected non-destructively, highly efficiently, and highly accurately from a population of plant seeds having desired traits and those not having the desired traits.

[0014] This figure shows the appearance of vegetable seeds divided into quality categories. The numbers below the boxes represent the scores of each seed calculated using the discrimination model in Table 2. This figure shows the near-infrared reflectance spectra of vegetable seeds divided into quality categories. A to I indicate the average reflectance spectra for each category, and J indicates the reflectance spectrum of an individual seed. This is a flow diagram showing the procedure for deriving a quality discrimination model. This is a schematic diagram showing the procedure for deriving a quality discrimination model. This figure shows the relationship between the distribution range of discrimination scores and discrimination accuracy. A indicates a hypothetical situation using a high-precision discrimination model, and B indicates a low-precision model. The preliminary pass rate is set to 80%. Black and gray triangles indicate the preliminary pass rate and the standard threshold (see Table 3, footnote 2), respectively. Abbreviations: eCDF (empirical cumulative distribution function). (a) Density distribution of discrimination scores for all seeds and eCDF of the proportion of seeds exceeding the horizontal axis score. (b) Density distribution of discrimination scores for good and bad seeds. (c) Distribution range of discrimination scores for good and bad seeds (boxplot and scatter plot). (d) Relationship between the set value of the lower score threshold and discrimination accuracy. (e) Relationship between seed recovery rate and discrimination accuracy. This figure shows the relationship between the selected wavelength and variable importance in the quality discrimination model. Abbreviations: SPRC (Standardized Partial Regression Coefficient), standard partial regression coefficient; VIP (Variable Importance in Projection), variable importance in projection. This figure shows the relationship between the lower score threshold and accuracy index in discriminating quality of cauliflower seeds. A. Internal validity verification, B. External validity verification. When germination traits and hybridization traits are sequentially predicted using models m-Ca47g and m-Ca47x, respectively, this figure shows the relationship between the set value of the lower threshold for the discrimination score of both models and the accuracy index in discriminating quality seeds (F1 hybrids with normal germination ability). Gray triangles indicate the standard threshold (see Table 3, footnote 2). This figure shows the internal validity verification of the quality discrimination model. A. Pumpkin variety 5, B. Pea variety 19, C. D. Lettuce variety 803, D. Green onion variety 22, E. Tomato variety 94, G. Green onion variety 51, H. Tomato variety 221, F, I, and K. Cauliflower varieties 47. F indicates the results and accuracy of discrimination based on germination traits only, I indicates the results and accuracy of discrimination based on hybrid traits only, and J indicates the results and accuracy of discrimination based on whether the seeds are good seeds (F1 hybrids with normal germination ability) or not.K shows the results and accuracy of quality discrimination using the integrated score, reflecting predictions of both traits based on the two discriminant models used in F and I. The symbols, abbreviations, and graphs (a) to (e) are the same as those in Figure 5. Figure 1 shows the external validation of the quality discrimination model. A. Pumpkin cultivar 5, B. Pea cultivar 19, C. Lettuce cultivar 803, D. Green onion cultivar 22, E. Tomato cultivar 94, F to I. Cauliflower cultivar 47. A to F correspond to Figures 8A to F, and G to I correspond to Figures 8I to K. The symbols, abbreviations, and graphs (a) to (e) are the same as those in Figure 5. Figure 2 shows the PR / rPR curves for quality discrimination of cauliflower seeds. A. PR curve, B. rPR curve. Figure 1 shows the relationship between the recall rate for discriminating each trait and the precision rate (A) and relative precision rate (B) for discriminating good seeds (F1 hybrids with normal germination ability) when sequentially discriminating germination traits and hybrid traits using models m-Ca47g and m-Ca47x, respectively. Precision rate and relative precision rate are expressed as a curve, and overall discrimination accuracy is expressed as the volume under the curve. These are also referred to as "curves" and "AUC (area under the curve)" in accordance with the expressions in Figures 11 and 12. Abbreviations: (r)PR ((relative) Precision-Recall), relationship between (relative) precision rate and recall rate; AUC (Area Under Curve). Figure 1 shows the PR curve for each combination of the good / bad discrimination model and dataset. Abbreviations: PR (Precision-Recall), relationship between precision rate and recall. Figure 1 shows the rPR curve for each combination of the good / bad discrimination model and dataset. Relative precision rate is defined by the following equation: [Relative precision] = (precision - initial precision) / (1 - initial precision). Note that in this specification, "precision" is synonymous with "post-test quality rate," and "initial precision" is synonymous with "pre-test quality rate." The relationship between recall and relative precision is referred to as the "rPR curve," and the area under the curve is referred to as "rPR-AUC." Figures showing image displays of quality discrimination scores. A. Pumpkin variety 5, B. Pea variety 19. Figures showing germination and early growth characteristics of seeds classified by quality discrimination scores. A and B. Pea variety 19, C. Lettuce variety 803, D. Green onion variety 22, E. Cauliflower variety 47. Image B is an image in which only pixels corresponding to the green leaves of A are extracted.The rankings shown in the upper rows of A to D and E are predicted by the discriminant model for germination traits, and the rankings shown in the lower row of E are predicted by the discriminant model for mating traits.

[0015] The present invention will be specifically described below.

[0016] The present invention provides a method for selecting plant seeds having desired traits (hereinafter, may be referred to as "the method of the present invention"), comprising the following 1st to 7th steps: (1st step) a step of extracting some seeds from a seed population; (2nd step) a step of constructing a dataset for the extracted seeds, the step of constructing the dataset comprising the following (a) to (c): (a) irradiating the extracted seeds with light to obtain spectral data and / or derivative spectral data thereof; (b) subjecting the seeds for which the spectral data has been obtained to a quality inspection for the desired trait, and determining a preliminary pass rate for the extracted seeds, wherein the preliminary pass rate is defined by the following formula: [Preliminary pass rate (%)] = [number of seeds having the desired trait among the seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation] × 100; (c) correlating the spectral data of each seed with the results of the quality inspection; (Step 3) deriving candidate good / bad discrimination models for calculating a discrimination score for a desired trait in each seed by applying multivariate discriminant analysis to the data set constructed in Step 2; (Step 4) selecting a discrimination model from the candidate good / bad discrimination models derived in Step 3, which includes defining recovery rate and precision rate by the following formulas: [Recovery rate (%)] = [number of seeds selected based on the discrimination score among seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation] x 100 [Precision rate (%)] = [number of seeds having the desired trait among seeds subjected to quality evaluation and selected based on the discrimination score] / [number of seeds subjected to quality evaluation selected based on the discrimination score] x 100, and when the selection conditions under which the preliminary good / bad product rate and the recovery rate are equal are defined as "standard selection conditions," selecting a candidate that satisfies the criterion that the precision rate exceeds the preliminary good / bad product rate under the standard selection conditions as a good / bad product discrimination model; (Step 5) A step of irradiating light onto the remaining seeds not removed from the seed population in Step 1 and / or other seeds obtained under substantially the same conditions as those of the seeds to obtain spectral data, and applying the discrimination model selected in Step 4 to the spectral data to determine a discrimination score for each seed; (Step 6) A step of determining a threshold value for the discrimination score based on the discrimination score determined in Step 5;(Step 7) A step of recovering seeds predicted to have desired traits by comparing the discrimination score of each seed determined in Step 5 with the discrimination score threshold determined in Step 6.

[0017] In the method of the present invention, the term "plant seeds" refers to seeds of any plant and is not particularly limited. The plant is not particularly limited as long as it is a seed plant, and may be, for example, either angiosperms or gymnosperms. Furthermore, when the plant is angiosperms, the plant may be either dicotyledonous or monocotyledonous. Furthermore, when the plant is dicotyledonous, the plant may be either sympetalous or dioppetalous. In one aspect, the method of the present invention can be used to select seeds of plants with high added value. Examples of such plants with high added value include, but are not limited to, horticultural crops and plants that can be used as building timber. Horticultural crops include vegetables, fruit trees, and ornamental plants. Examples of vegetables include, but are not limited to, pumpkin, pea, lettuce, green onion, tomato, cauliflower, bitter melon, okra, onion, myoga (Japanese ginger), soybean, butterbur, asparagus, chive, broad bean, celery, carrot, mizuna (potherb mustard), komatsuna (Japanese mustard spinach), shungiku (garland chrysanthemum), radish, broccoli, spinach, Chinese cabbage, arugula, lotus root, turnip, avocado, cucumber, paprika, garlic, corn, zucchini, parsley, cilantro, eggplant, and bell pepper. Examples of fruit trees include, but are not limited to, plum, fig, akebia, acerola, avocado, olive, orange, persimmon, quince, guava, cranberry, walnut, grapefruit, cherry, and pomegranate. Examples of flowering plants include, but are not limited to, morning glory, celosia, cosmos, zinnia, columbine, globe amaranth, petunia, periwinkle, snapdragon, sunflower, impatiens, portulaca, portulaca, balloon vine, marigold, gypsophila, snapdragon, calendula, sweet pea, stock, dianthus, daisy, nigella, nemesia, nemophila, poppy, verbena, pansy, viola, corn poppy, cornflower, lupine, and forget-me-not. Examples of plants that can be used as building lumber include, but are not limited to, cedar, cypress, Japanese cypress, chestnut, zelkova, cherry, beech, walnut, falcata, and Japanese red pine.

[0018] In the method of the present invention, the term "desired trait" encompasses both traits possessed by the seed itself and traits that are expressed in a plant germinated from the seed when the seed is grown. Examples of the "traits" to be selected in the method of the present invention include, but are not limited to, germinability, germination vigor, genotype, stress tolerance, dormancy, disease resistance, insect resistance, QTL characteristics, eating quality, heading date, and morphological characteristics such as leaf size.

[0019] (First Step) The first step of the present invention is characterized by extracting a portion of seeds from a seed population. The number of seeds constituting the seed population may be two or more, with no particular upper limit. For example, the seed population may typically be a seed population consisting of 2 to 10,000,000 seeds, preferably 1,000 to 10,000,000 seeds, and more preferably 10,000 to 10,000,000 seeds, but is not limited to these. Alternatively, the seed population may be expressed in terms of weight. For example, the seed population may be a seed population weighing 1 g to 10,000 kg, preferably 1 kg to 10,000 kg, and more preferably 10 kg to 10,000 kg, but is not limited to these.

[0020] Furthermore, the "portion" of the seed population may vary depending on the "traits" to be selected, but is usually about 50 to 50,000 seeds, preferably 100 to 30,000 seeds, and more preferably 200 to 3,000 seeds, but is not limited to these.

[0021] (Second Step) The second step of the present invention is to construct a data set for the extracted seeds. The construction of the data set includes at least the following steps (a) to (c):

[0022] (a) irradiating the extracted seeds with light to obtain spectral data and / or its derived spectral data; (b) subjecting the seeds for which spectral data has been obtained to a quality inspection for desired traits and determining a preliminary yield rate of the extracted seeds; the preliminary yield rate in (b) is defined by the following formula: [Preliminary yield rate (%)] = [number of seeds having the desired traits among the seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation] × 100; and (c) correlating the spectral data of each seed with the results of the quality inspection.

[0023] In this specification, the "preliminary non-defective rate" may be referred to as the "initial conformance rate."

[0024] In (a) of constructing the dataset, the method of irradiating the seeds with light to obtain spectral data (reflection, absorption, and / or transmission spectral data) and / or its derived spectral data can be a method commonly known in the art of optical analysis. The light irradiated onto the seeds is not particularly limited as long as it can obtain spectral data. Examples of light irradiated onto the seeds include, but are not limited to, microwaves, terahertz waves, infrared light, visible light, ultraviolet light, X-rays, and gamma rays. In a preferred embodiment of the present invention, the light may be visible light or infrared light. Spectral data can be obtained using, for example, the methods and conditions used in the examples of this application, as well as the methods taught in Japanese Patent No. 6782408, but are not limited thereto.

[0025] Furthermore, methods known per se in the technical field of optical analysis may be used to generate derived spectral data from the obtained spectral data. For example, the derived spectral data can be generated by performing reciprocal (1 / R) transformation, logarithmic transformation, standard normalization, smoothing and smoothed differentiation using a Savizky-Golay (SG) filter, or any combination thereof on the obtained spectral data (R). Note that either one of the spectral data and the derived spectral data may be used, or two or more of them may be used. In a preferred embodiment, the spectral data used may be one of the spectral data and the derived spectral data generated therefrom that can lead to a prediction model with the highest accuracy.

[0026] Furthermore, any known device may be used to acquire spectral data. For example, the spectral characteristics of biological tissue can be measured by point measurement using a fiber optic spectrometer, or by using a hyperspectral camera, a type of remote sensor, to measure (measure) the characteristics together with coordinate information (images) (area measurement). For measurement, an appropriate device with high detection sensitivity in the target wavelength range can be selected. In the method of the present invention, devices equipped with photodetectors such as, but not limited to, CCD, CMOS, CQD, InGaAs, HgCdTe (MCT), and Type II superlattice (T2SL) can be used. Furthermore, when acquiring spectral data, a reflectance-corrected image can be generated and a seed recognition model can be applied, if necessary (see ST1-3 in Figure 3).

[0027] The quality test in (b) of constructing the dataset is also not particularly limited as long as it allows the preliminary pass rate for the desired trait to be determined. For example, if the desired trait is "germination ability," this can be easily confirmed by subjecting the seeds from which the spectral data was acquired to a germination test in accordance with, for example, the International Seed Testing Standards established by the International Standards Authority of the United Nations (ISTA) or a similar germination test. When the number of seeds subjected to the germination test is A (number of seeds), and the number of seeds that actually germinated is B (number of seeds), the preliminary pass rate (%) (i.e., germination rate) can be determined as 100B / A (%), based on the formula: [number of germinated seeds (B) among the seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation (A)] × 100.

[0028] To explain the "correspondence between the spectral data of each seed and the quality test results" in (c) of constructing the dataset specifically using the case where the trait is "germination ability," the spectral data (reflection, absorption, or transmission spectral data or derivative spectral data) of each seed obtained by irradiating light on seeds A, B, and C are designated SA, SB, and SC, respectively. Furthermore, when seeds A, B, and C are subjected to a quality test, the test results show that seed A germinates normally, seed B germinates abnormally, and seed C fails to germinate. In (c), the spectral data is intended to be correlated with the quality results, such that SA indicates normal germination, SB indicates abnormal germination, and SC indicates failure to germinate. Note that dummy variables may be used to facilitate the determination of quality. For example, in the case of "germination ability," a variable such as 1 (good) can be assigned to normal germination and -1 (bad) can be assigned to abnormal germination and failure to germinate, but this is not limiting.

[0029] (Third Step) The third step of the present invention is a step of deriving candidates for a good / bad discrimination model for calculating a discriminant score for a desired trait in each seed by applying multivariate discriminant analysis to the data set constructed in the second step.

[0030] The multivariate discriminant analysis in this step may be performed using a method known per se, and any method may be used as long as the desired effects of the present invention can be obtained. In the method of the present invention, the multivariate statistical analysis technique may be, but is not limited to, partial least squares (PLS) or principal component analysis (PCA). The multivariate discriminant analysis may preferably be partial least squares-discriminant analysis (PLS-DA).

[0031] As an example, a case where multivariate discriminant analysis is performed using PLS-DA will be outlined.

[0032] PLS-DA models an equation such as the following equation 1 using set explanatory variables and target variables.

[0033] (Formula 1) y=a+b 1 x 1 +b 2 x 2 +b 3 x 3 +b 4 x 4 +…

[0034] Here, x 1 , x 2 , x 3 and x 4 is the explanatory variable, y is the objective variable, a is the intercept (constant), and b 1 , b 2 , b 3 and b 4 is the partial regression coefficient (constant).

[0035] Using the spectral data acquired in the second step as explanatory variables and the results obtained by the quality inspection as objective variables, candidate quality discrimination models for calculating discriminant scores for desired traits in each seed are derived. The method for deriving candidate quality discrimination models may also be a method known per se, but the method used in the examples of the present invention will be described below as an example.

[0036] The derived spectral data Xs is generated from the reflection, absorption, or transmission spectral data (R) obtained by irradiating light onto the seeds, including the data itself (X1), using the above-mentioned conversion method and any combination thereof. In the discriminant analysis, all of the derived spectra are attempted to be used as explanatory variables, but if the predictive performance of the derived discriminant model is equivalent, it is sufficient to adopt the simplest derived pattern possible.

[0037] (Procedure for Derivation of Candidate Good / Fail Discrimination Model) Final preparation of the dataset is performed as follows: A reflection, absorption, or transmission spectrum is acquired, and its derived spectrum is generated to prepare explanatory variable data. Standardization is performed on either or both of the dependent variable and explanatory variable, as necessary. A discriminant model is derived using multivariate discriminant analysis. For example, a discriminant model can be derived using linear sparse modeling as follows: For all combinations of dependent variable and explanatory variable data, initial values ​​for the weights of each explanatory variable are first determined using PLS regression or Ridge regression. Next, a solution path is calculated using Adaptive LASSO regression, representing the relationship between the regularization coefficient λ, the partial regression coefficient, and the number of explanatory variables for which the partial regression coefficient does not become zero. If the number of explanatory variables that minimize the sum of squared residuals in the regression is p, candidate combinations of explanatory variables suitable for reducing the number of explanatory variables to p or less can be determined from the solution path. The final discriminant model is derived using PLS-DA. A discriminant model is derived using 2 to p explanatory variables determined from the solution path in adaptive LASSO regression. The derived discriminant model is set as a candidate for the good / bad discriminant model. Note that at this stage, many candidate discriminant models are presented. The discriminant model that gives a higher relative score to good seeds is selected in the next step 4.

[0038] (Fourth Step) The fourth step of the present invention is a step of selecting a discrimination model from the candidate pass / fail discrimination models derived in the third step.

[0039] In the fourth step, the "recovery rate," "suitability rate," and "standard selection conditions" are defined as follows:

[0040] [Recovery rate (%)] = [Number of seeds selected based on the discriminant score among the seeds subjected to quality evaluation] / [Number of seeds subjected to quality evaluation] × 100

[0041] [Fairness rate (%)] = [Number of seeds having desired traits among the seeds subjected to quality evaluation and selected based on the discriminant score] / [Number of seeds selected based on the discriminant score among the seeds subjected to quality evaluation] × 100

[0042] [Standard sorting conditions]: Sorting conditions that make the [Preliminary non-defective rate] and [Recovery rate] the same value

[0043] When the recovery rate, precision rate, and standard selection conditions are defined as described above, candidates that satisfy the criterion that the precision rate exceeds the preliminary yield rate are selected as quality discrimination models, thereby making it possible to select a discrimination model capable of discriminating quality for a desired trait from the many candidate models obtained in the third step. Note that the greater the precision rate exceeds the preliminary yield rate, the better the discrimination model is considered to be. Furthermore, in order to enable easy adjustment of the precision rate as needed, it is preferable to select a discrimination model that has a high rank correlation between increases and decreases in the recovery rate and increases and decreases in the precision rate (i.e., a relationship in which a decrease in the recovery rate reliably increases the precision rate, and an increase in the recovery rate reliably decreases the precision rate).

[0044] In this specification, the "matching rate" may be referred to as the "post-test non-defective rate."

[0045] (Step 5) Step 5 of the present invention is a step of irradiating light onto the remaining seeds not removed from the seed population in Step 1 and / or other seeds obtained under substantially the same conditions as the seeds to obtain spectral data, and applying the discrimination model selected in Step 4 to the spectral data to determine the discrimination score for each seed. In other words, Step 5 is a step of obtaining spectral data for non-training seeds from the seed population and applying the discrimination model obtained in Step 4 to this to determine the discrimination score for each seed. The method and conditions for obtaining the spectral data may be the same as those used for the training seeds, or may be partially different as long as the desired effect is obtained.

[0046] (Step 6) Step 6 of the present invention is a step of determining a threshold value for the discrimination score based on the discrimination score determined in Step 5. If the preliminary non-defective rate in the seed population is "k %," the score that is in the top "k %" can be set as the threshold value for standard selection (standard threshold). Note that, if a discrimination model having a high rank correlation between increases and decreases in the recovery rate and increases and decreases in the subsequent non-defective rate is selected in Step 4 above, the precision rate can be adjusted by appropriately increasing or decreasing the threshold value based on the standard threshold value.

[0047] (Step 7) Step 7 of the present invention is a step of recovering seeds predicted to have the desired traits by comparing the discrimination score of each seed determined in Step 5 with the discrimination score threshold determined in Step 6. The recovery of seeds may be achieved by selecting and removing seeds predicted to have the desired traits, or by selecting and removing seeds predicted not to have the desired traits and recovering the seeds not removed as seeds having the desired traits.

[0048] In one embodiment of the present invention, the method of the present invention may be a method for selecting plant seeds having two or more desired traits.

[0049] For example, examples of two or more traits include, but are not limited to, a combination of two or more of the traits listed above (i.e., germination ability, germination vigor, genotype, stress tolerance, dormancy, disease resistance, insect resistance, QTL characteristics, eating quality, heading date, and morphological characteristics such as leaf size).

[0050] When the method of the present invention is a method for selecting two or more desired traits, the prediction of the two or more desired traits is carried out by any one of the following methods (i) to (iii):

[0051] (i) A method in which seeds are scored individually using a discriminant model for each trait, focusing on the quality of each trait, and seeds with all scores above a threshold are predicted to be good seeds ("sequential prediction using multiple models").

[0052] (ii) A method in which a single discriminant model is selected by ignoring the quality of each trait, determining that seeds are good if all of the desired traits are good, and determining that seeds are bad otherwise, and then scoring seeds using this single discriminant model, and predicting that seeds with scores above a threshold are good seeds ("direct prediction using a single model").

[0053] (iii) A method in which a discrimination model for each trait is selected by focusing on the quality of each trait, the scores obtained using the discrimination model for each trait are integrated to form an integrated score, and the quality of each seed is predicted based on the ranking of the integrated score ("direct prediction using an integrated model").

[0054] Using either method, selection of seeds having two or more desired traits can be achieved.

[0055] In one aspect of the present invention, the method of the present invention may further include a step of evaluating the accuracy of the discriminant model. The accuracy evaluation is an evaluation of the "generalized performance" of the selected discriminant model. Specifically, the accuracy evaluation of the good / bad discrimination model can be performed by the following method.

[0056] The discriminant model obtained by carrying out the above-mentioned steps is subjected to internal validity verification when applied to the internal data used in its derivation (i.e., the seed population from which the dataset was constructed), and external validity verification when applied to external data obtained independently of the derivation of the model.

[0057] To evaluate the overall accuracy of a pass / fail discrimination model that is independent of the set value of the lower score threshold, a receiver operating characteristic (ROC) curve or a precision-recall (PR) curve can be used. In particular, in quality inspections that consider the adoption of defective products (false positives) as more serious than the rejection of non-defective products (false negatives), a PR curve, which is sensitive to false positives, is often used.

[0058] In both ROC and PR curves, the closer the AUC (Area Under Curve) is to the maximum value of 1, the better the discriminant model is. However, while the minimum value for ROC-AUC is always 0.5, the minimum value for PR-AUC is the initial precision rate, i.e., the preliminary non-defective rate, which varies depending on the data set to which it is applied.

[0059] Although PR-AUC is suitable for verifying which of several discriminant models exhibits superior performance for a specific data set, it is not suitable for verifying whether a single discriminant model exhibits equivalent performance for different data sets, because it is affected by the prior non-defective rate.

[0060] To compensate for the above-mentioned shortcomings in the PR curve, we define the relative precision expressed by the following formula:

[0061] [Relative precision] = (precision - initial precision) / (1 - initial precision)

[0062] The relationship between relative precision and recall is called the rPR (relative precision-recall) curve, and the area under the curve is called the rPR-AUC. The rPR-AUC is independent of the dataset to which it is applied and always takes a value in the closed interval [0, 1].

[0063] When n models are used in pass / fail discrimination, the relationship between precision or relative precision and recall forms a hypersurface occupying an (n+1)-dimensional space. For convenience, the relationships between precision and relative precision and recall are referred to as PR curves and rPR curves, and the ratios below the hypersurfaces they form are referred to as PR-AUC and rPR-AUC, regardless of the number of discriminant models used.

[0064] For all derived discrimination models, PR and rPR curves are created when each is applied to a data set derived from the target crop species. Furthermore, PR-AUC and rPR-AUC are calculated from each curve. The accuracy of the discrimination model is evaluated based on the shape of the created curve and the AUC value.

[0065] The present invention will be explained in more detail in the following examples, but the present invention is not limited to these examples in any way.

[0066] [Distinguishing vegetable seeds with quality that meets shipping requirements] (Materials) Vegetable seeds (pumpkin, pea, lettuce, green onion, tomato, cauliflower) harvested between 2015 and 2022 provided by Tokita Seed Co., Ltd. were used as test materials (Table 1).

[0067]

[0068] Seeds were irradiated with near-infrared light by applying a DC voltage to two 24V 250W aluminum mirror halogen lamps (JTR24V250W10H / 5-AL, GX5.3 base, manufactured by Kahoku Lighting Solutions Co., Ltd.). Near-infrared hyperspectral images were captured using a line-scan hyperspectral camera (CV-N801HS, manufactured by Sumitomo Electric Industries, Ltd.). The near-infrared lens was a 30 mm focal length, image-side telecentric lens (manufactured by Sumitomo Electric Industries, Ltd.). The working distance during imaging was 28 cm, the spatial resolution was 90 ppi, and the wavelength sampling interval was 6 nm. The wavelength sensitivity range of the camera used was 980 to 2,350 nm. The reflectance at each wavelength was calibrated to the equivalent of diffuse reflectance based on the images of a standard reflector with a reflectance of 99% and a dark current (reflectance of 0%), which constitute a contrast target (SRT-MS-050, Labsphere, USA). The reflectance spectrum of the seeds was calculated by averaging the reflectance spectra recorded for each pixel in the area occupied by each seed.

[0069] The appearance of the seeds was photographed using an 8k color line scan camera (EV71C4CCL8005-BA0 manufactured by e2v) under illumination with a white LED light source (IDBA-HMS150WHV-S manufactured by Leimac). The lens was an object-side telecentric lens (FT04-150CL manufactured by Optoart) with an optical magnification of 0.4x, and the spatial resolution was 2,032 ppi. To adjust the white balance of the camera, a standard reflector with a reflectance of 25% (when the seed surface was dark) or 50% (when the seed surface was light) that constitutes the contrast target (SRT-MS-050) was used.

[0070] (Method) I. Dataset Construction (1) Seeds were uniformly irradiated with light from a halogen lamp, and near-infrared hyperspectral images containing spectral information in the near-infrared wavelength range were captured. For some seeds, high-resolution external images were also captured using an 8k line camera (Figure 1).

[0071] (2) From the near-infrared hyperspectral image, the near-infrared reflectance spectrum of the area occupied by each seed was calculated (Figure 2).

[0072] (3) The seeds were subjected to a germination test in accordance with the international seed testing regulations established by the International Seed Testing Association (ISTA), and the germination characteristics of each seed were classified into three groups: normal germination, abnormal germination, and non-germination.

[0073] (4) Cauliflower variety 47 (hereafter, variety number omitted) is used as an F1 hybrid, but due to incomplete self-incompatibility, it may produce self-fertilized seeds. The hybrid traits of each seed, whether F1 or self-fertilized, were determined by isozyme analysis using seedlings.

[0074] (5) Dummy variables were assigned to each good and bad seed, and these were used as the objective variables in the discriminant analysis described below. When discriminating between germination traits, the value of the dummy variable was set to 1 (good) for normally germinated seeds and -1 (bad) for others, and so on. When discriminating between cauliflower hybrid traits, F1 hybrids were set to 1 and others were set to -1. Furthermore, when discriminating between germination and hybrid traits and whether they all met the requirements for good seeds, F1 hybrids with normal germination ability were set to 1 and others were set to -1.

[0075] II. Derivation of a model for discriminating between good and bad vegetable seeds (6) A discriminant analysis was performed to predict whether a seed has the desired traits (normal germination for all crop species, and being an F1 hybrid for cauliflower) based on the near-infrared reflectance spectrum of the seed.

[0076] (7) The procedure of the discriminant analysis conformed to the method described in Japanese Patent Application Laid-Open No. 2023-125301. In the present invention, smoothed differentiation of the near-infrared reflectance spectrum and the column combination of spectra with different differential orders were omitted (FIG. 3).

[0077] (8) The near-infrared reflectance spectrum (R), which serves as an explanatory variable, was subjected to reciprocal (1 / R) and logarithmic (pseudo-absorbance, -logR) transformations, as well as standard normalization (SNV, Standard Normal Variate; effective wavelength range: 980-2,200 nm) and smoothing using a Savizky-Golay (SG) filter (five-point, third-order polynomial approximation) in arbitrary combinations to generate 11 derived spectra. A total of 12 spectra, including the original spectrum, were used as explanatory variables (Figure 4).

[0078] (9) The seed quality discrimination model was derived using a multivariate linear sparse modeling technique that combines adaptive LASSO (Least Absolute Shrinkage and Selection Operator) and PLS-DA (Partial Least Squares-Discriminant Analysis). At this stage, many candidate models are presented, but it is necessary to select the one that gives a higher relative score to good seeds. The model selection method is specifically described below.

[0079] III. Selection of a Good / Bad Discrimination Model (10) In order to devise a method for selecting a desirable discrimination model from the many candidates presented above, the following thought experiment was carried out.

[0080] 1) Assume two discriminant models: a high-precision model that assigns clearly separated scores to good and bad seeds, and a low-precision model that assigns widely overlapping scores. Here, the pre-qualifying rate is set to 80%, and the former assigns scores to good and bad seeds that follow a normal distribution with means of 1 and -1 and standard deviations of 0.25, while the latter assigns scores to good and bad seeds that follow a normal distribution with means of 0.05 and -0.05 and standard deviations of 0.25 (Figure 5).

[0081] 2) A seed sorting operation is performed in which seeds are collected in descending order of score. When a high-precision discriminant model is used, the precision rate (post-qualification rate) remains near 100% until the collection rate reaches 80%, which matches the preliminary quality rate (Fig. 5A(e)). On the other hand, when a low-precision discriminant model is used, the precision rate quickly converges to the preliminary quality rate (Fig. 5B(e)). In either case, the precision rate and recall rate (quality seed collection rate) become equal once the collection rate matches the preliminary quality rate.

[0082] 3) When selecting good seeds using a discriminant model, seeds with scores above a predetermined lower threshold are sequentially collected after all seeds have been measured, rather than collecting the seeds with the highest scores. Since the relationship between the score and its ranking can be easily calculated, the relationship between the recovery rate and each index related to discrimination accuracy, including precision and recall (Figure 5(e)), can be redrawn as the relationship with the discrimination score (Figure 5(d)).

[0083] 4) Here, the sorting conditions that make the preliminary non-defective rate and the recovery rate equal are defined as "standard sorting conditions," and the lower limit threshold of the discrimination score of the seeds to be recovered at that time is defined as "standard threshold."

[0084] From the above, it can be seen that a requirement for an excellent discriminant model is that when selection is performed using standard selection conditions, the precision rate (i.e., the ex-post quality rate) and its equivalent, the recall rate, are significantly higher than the pre-pre-quality rate.

[0085] (11) In addition, in selecting good seeds, it is desirable to be able to adjust the matching rate as needed. That is, it is more desirable to establish a relationship in which the matching rate can be increased with high accuracy as the recovery rate is lowered (the lower threshold score is increased).

[0086] (12) The quality discrimination models for each crop seed shown in Tables 2-1, 2-2, and 2-3 below were selected based on the requirement that they satisfy the above two points. In addition, in each model in the same table, the relationship between the wavelength bands adopted as necessary explanatory variables and the standard partial regression coefficients and variable importance in projection (VIP) is shown in Figure 6.

[0087]

[0088]

[0089]

[0090] IV. Combining Multiple Discriminant Scores Embodiments in which multiple traits are assessed are described hereafter.

[0091] (13) When good cauliflower seeds are defined as those that (1) have normal germination ability and (2) are F1 hybrids, the following two methods can be used to identify good cauliflower seeds.

[0092] Method 1) The quality of germination and mating traits are scored individually using two models (m-Ca47g and m-Ca47x), and seeds with the highest scores are predicted to be good seeds.Method 2) The quality of individual traits is ignored, and only good seeds are predicted (m-Ca47).

[0093] (14) In method 1), to obtain the desired discrimination accuracy, it is necessary to set a lower score threshold for the two models. When n discrimination models are used, the relationship between the set value of the lower score threshold and each index related to discrimination accuracy forms the shape of an (n+1)-dimensional hypersurface. When n = 2 as in this example, each index forms a curved surface, but it can also be drawn in a two-dimensional form as a heat map (Figure 7). However, this method is significantly less easy to operate than method 2).

[0094] (15) In order to integrate the scores of multiple discriminant models and set only one lower threshold, we devised the following mathematical process.

[0095] 1) Let M1 to Mk be models for discriminating between good and bad traits 1 to k, and let y1(X) to yk(X) be the discriminant scores obtained by substituting explanatory variable X into these models.

[0096] 2) Tens of a number n (several hundred) of seeds making up a specific lot are measured, and scores y1(x) to yk(x) are calculated from the corresponding explanatory variables x (x1 to xn). If the pre-qualification rates for each trait are assumed to be p1 to pk, the highest n × p (p1 to pk) scores of y (y1 to yk) are defined as yp (yp1 to ypk). The standard deviation of y (y1 to yk) is defined as s (s1 to sk).

[0097] 3) Score y (y1 to yk) is converted to score z (z1 to zk) using the following formula: z = (y - yp) / s

[0098] 4) The minimum value of the score z (z1 to zk) is set as the integrated score in the discrimination using M1 to Mk.

[0099] 5) The quality of each seed is determined based on the ranking of the integrated score. Standard conditions and standard thresholds can also be determined for the determination using the integrated score.

[0100] This process was applied to m-Ca47g and m-Ca47x to produce model m-Ca47gx (Tables 3-1 and 3-2).

[0101]

[0102]

[0103] V. Accuracy evaluation of good / bad discrimination models (16) For all selected discrimination models, the distribution range of the discrimination scores for all seeds or good / bad seeds, and the relationship between the set value of the lower score threshold or the seed recovery rate and each index related to discrimination accuracy are shown in Figures 8 and 9. The configuration of the figures is the same as Figure 5.

[0104] (17) Figures 8 and 9 show the results of internal validation when each discriminant model is applied to the internal data used in its derivation, and external validation when it is applied to external data obtained independently of the model derivation. Figure 9 shows the evaluation results of the model's generalization performance.

[0105] (18) To evaluate the overall accuracy of a pass / fail discrimination model that does not depend on the set value of the lower score threshold, a receiver operating characteristic (ROC) curve or a precision-recall (PR) curve is often used. In particular, in quality inspections that consider the adoption of defective products (false positives) more serious than the rejection of non-defective products (false negatives), a PR curve, which is sensitive to false positives, is often used.

[0106] (19) In both ROC curves and PR curves, a discriminant model is considered to be superior when the AUC (Area Under Curve) is closer to 1, which is the maximum value. However, while the minimum value of ROC-AUC is always 0.5, the minimum value of PR-AUC is the initial precision rate, i.e., the preliminary non-defective rate, which varies depending on the data set to which it is applied.

[0107] (20) PR-AUC is suitable for verifying which of several discriminant models performs better on a specific dataset, but it is not suitable for verifying whether a single discriminant model performs equally well on different datasets because it is affected by the prior yield rate.

[0108] (21) To compensate for the above-mentioned shortcomings in the PR curve, we define the relative precision expressed by the following equation:

[0109] [Relative precision] = (precision - initial precision) / (1 - initial precision)

[0110] (22) The relationship between relative precision and recall is called the rPR (relative precision-recall) curve, and the area under the curve is called the rPR-AUC. The rPR-AUC is independent of the dataset to which it is applied and always takes a value in the closed interval [0, 1].

[0111] (23) When n models are used in pass / fail discrimination, the relationship between precision and relative precision and recall also forms the shape of an (n+1)-dimensional hypersurface. For convenience, the relationship between precision and relative precision and recall is referred to as the PR curve and rPR curve, and the ratios below the hypersurface formed by these curves are referred to as the PR-AUC and rPR-AUC, regardless of the number of discriminant models used.

[0112] (24) For all derived discriminant models, the PR curves (Figures 10A and 11) and rPR curves (Figures 10B and 12) are shown when applied to the datasets derived from the target crop species. The PR-AUC and rPR-AUC calculated from each curve are also listed in Tables 3-1 and 3-2.

[0113] VI. Selection of Good Seeds Based on a Discriminant Model (25) We developed software that calculates the scores of the discriminant models shown in Table 2 for each pixel or area occupied by a seed using near-infrared hyperspectral images of the seeds, converts them into pseudocolors or brightness values, and visualizes them (Figure 13).

[0114] (26) Using the software, seeds with high and low discriminant scores for germination or mating traits were collected. For each category, seed germination and early growth were evaluated (Figure 14). For cauliflower, isozyme types were analyzed and the subsequent yield rate (suitability rate) was calculated (Table 4).

[0115]

[0116] (Results and Discussion) (1) For all crop species, it was impossible to subjectively determine the quality of the seeds after the germination test from their appearance before germination and the outline of their near-infrared reflectance spectra (Figures 1 and 2).

[0117] (2) When the derived discriminant model was applied to the target variety, it was confirmed that the distribution of scores was, without exception, in the order from high to low, from good seeds to poor seeds (Figure 8 and Figures 9(b) and (c)).

[0118] (3) In the model predicting germination traits, the distribution range of scores shown by abnormally germinated seeds was located between the scores shown by normally germinated and non-germinated seeds (Figures 8A-H and Figures 9A-F(b) and (c)), and it was inferred that the discrimination score generally reflected the growth potential of the seeds.

[0119] (4) The distribution of the discriminant scores overlapped between good and bad seeds, and misclassification was unavoidable for both crop species. Furthermore, compatibility of the discriminant model was confirmed between tomato varieties 94 and 221, but the discriminant accuracy was suggested to be inferior compared to when applied to the same variety (Table 3 and Figures 12H and 12I). For leeks, it was determined that sharing a discriminant model between varieties 22 and 51 was not practical (Table 3 and Figures 12F and 12G).

[0120] (5) The distribution range of the discriminant scores was not the same when the same variety and lot of seeds were used, even when the test was conducted on different occasions (Fig. 8, Fig. 9(a) to (c) ; for tomato variety 94, different lots were used in Fig. 8 and Fig. 9 ). The derived discriminant model could be repeatedly applied to at least the same variety, but it was desirable to optimize the lower threshold score for selecting good seeds for each occasion.

[0121] (6) When a discriminant model is applied to near-infrared spectra obtained from randomly selected seeds and a score is calculated, the quality of each individual seed is unknown. Furthermore, the density distribution of the scores shows only a featureless, single peak (Figure 9C, E-I(a); A, B, and D are not randomly selected), making it difficult to estimate a lower threshold score for the desired sorting from this shape. However, it is possible to determine a standard threshold for recovering seeds with the same rate as the estimated pre-selection rate, and it is reasonable to define this as the standard sorting condition (standard condition). Therefore, it is desirable to understand the pre-selection rate of the seeds to be used as accurately as possible.

[0122] (7) If you want to increase the post-test good product rate (matching rate) compared to the standard conditions, you can increase the lower threshold of the discrimination score, and if you want to increase the recall rate (recovery rate of good seeds), you can lower the lower threshold of the score.

[0123] (8) In the above example, it is not easy to estimate how much the change in the lower threshold of the discriminant score will affect the precision and recall. Subtracting a standard threshold from the score and then dividing by the standard deviation of the score has the advantage of making the variance of the scores uniform, just like standardizing (normalizing) variables. This is also effective when selecting good seeds using a single model.

[0124] (9) By using the score conversion described above, the thresholds to be adjusted can be integrated into one even when selecting good seeds using multiple discriminant models.

[0125] (10) In cauliflower, when predicting F1 hybrids with normal germination ability, which are considered to be good seeds, higher accuracy was observed when a combination of models predicting germination and mating traits individually was used rather than when a single model directly discriminated between good and bad seeds (Table 3 and Figures 12L and 12M).

[0126] (11) When a model predicting germination traits was applied and the germination and early growth of seeds with high and low discriminant scores were compared, seeds with high scores not only had a high germination rate but also tended to show vigorous growth after germination (Figure 14). The same tendency was observed for lettuce, for which the performance of the derived quality discrimination model was somewhat inferior. From this, it can be inferred that the discriminant scores for germination traits reflect the growth potential of the seeds.

[0127] (12) By confirming that the precision rate (synonymous with the posterior precision rate and identical to the recall rate) under standard selection conditions was sufficiently higher than the pre-prediction rate, we were able to efficiently narrow down a large number of candidate discriminant models to a preferred model (Table 3 and Figure 3). Furthermore, by defining the relative precision rate and calculating the area under the rPR curve (rPR-AUC), which represents the relationship between the precision rate and the recall rate, we were able to objectively compare the discrimination accuracy regardless of the combination of model and dataset (Table 3 and Figure 12).

[0128] (13) With the technology of the present invention, it is possible to determine whether the seeds are good or bad within approximately one day of receiving the results of the seed quality evaluation test, and if possible, to obtain a specific discrimination model. Therefore, even if the occurrence of defective seeds is accidental and not due to the genetic characteristics of the variety, seed selection work can be started soon after the problem is discovered in order to restore the rate of good seeds to a level that meets the shipping standard.

[0129] (14) To make this possible, it is necessary to acquire near-infrared hyperspectral images of seeds in advance that can be linked to the results of quality evaluation tests.

[0130] (15) As mentioned above, deriving a discrimination model does not require a great deal of time or effort. It is possible to continue updating a model applied to a specific crop variety to improve its accuracy, with the aim of reusing it on other occasions. However, the reasons for the occurrence of defective seeds often vary from occasion to occasion. It is important not only to enrich the library of discrimination models compatible with various crop varieties, but also to derive the optimal quality discrimination model for each occasion when a problem occurs, and to establish a seed quality control system that can respond promptly to resolve the problem.

[0131] (16) Near-infrared hyperspectral images can be acquired at once, and the data obtained can be used to distinguish multiple traits, such as seed germination characteristics and genotype. The images contain information about grain size and are also useful for estimating seed weight. This invention provides a low-cost, space-saving means of advanced seed sorting without the need to install multiple sorting machines with different functions.

[0132] According to the present invention, plant seeds having desired traits can be selected non-destructively, highly efficiently, and highly accurately from a population of plant seeds having and not having the desired traits, and therefore the present invention is extremely useful, for example, in the seed and seedling industry.

[0133] This application is based on patent application No. 2023-089036 filed in Japan (filing date: May 30, 2023), the contents of which are incorporated in their entirety herein.

Claims

1. A method for selecting plant seeds having desired traits, comprising the following steps 1 to 7: (step 1) extracting a portion of seeds from a seed population; (step 2) constructing a dataset for the extracted seeds, the step of constructing the dataset comprising the following steps (a) to (c): (a) irradiating the extracted seeds with light to obtain spectral data and / or derived spectral data thereof; (b) subjecting the seeds for which the spectral data has been obtained to a quality inspection for the desired trait, and determining a preliminary pass rate for the extracted seeds; wherein the preliminary pass rate is defined by the following formula: [Preliminary pass rate (%)] = [Number of seeds having the desired trait among the seeds subjected to quality evaluation] / [Number of seeds subjected to quality evaluation] x 100; (c) matching the spectral data of each seed with the results of the quality inspection; (step 3) applying multivariate discriminant analysis to the dataset constructed in step 2 to derive candidates for a pass / fail discrimination model for calculating a discrimination score for the desired trait in each seed; (fourth step) defining the recovery rate and the conformance rate by the following formulas: [recovery rate (%)] = [number of seeds selected based on the discrimination score among the seeds subjected to quality evaluation] / [number of seeds subjected to quality evaluation] × 100 [conformance rate (%)] = [number of seeds having desired traits among the seeds subjected to quality evaluation and selected based on the discrimination score] / [number of seeds selected based on the discrimination score among the seeds subjected to quality evaluation] × 100, and defining the selection conditions under which the preliminary conformance rate and the recovery rate have the same value as each other as "standard selection conditions", a step of selecting a discrimination model from the candidates of the conformance model derived in the third step, the step including selecting a candidate that satisfies the criterion that the conformance rate exceeds the preliminary conformance rate under the standard selection conditions as a conformance model; (fifth step) irradiating the remaining seeds not taken out of the seed population in the first step and / or other seeds obtained under substantially the same conditions as the remaining seeds to obtain spectral data, and applying the discrimination model selected in the fourth step to the spectral data to determine the discrimination score of each seed; (Sixth step) determining a threshold value of the discriminant score based on the discriminant score determined in the fifth step;(Step 7) A step of recovering seeds predicted to have desired traits by comparing the discrimination scores of each seed determined in step 5 with the discrimination score threshold determined in step 6.

2. The method according to claim 1, which is a method for selecting plant seeds having two or more desired traits.

3. The selection method according to claim 1 or 2, wherein the trait is at least one trait selected from the group consisting of germination ability, germination vigor, genotype, stress resistance, dormancy, disease resistance, insect resistance, QTL characteristics, eating quality, heading time, and morphological characteristics.

4. A selection method according to claim 1 or 2, characterized in that prediction of two or more desired traits is carried out by any of the following methods (i) to (iii): (i) a method in which seeds are scored individually using a discrimination model for each trait, focusing on the quality of each trait, and seeds with all scores above a threshold are predicted to be good seeds; (ii) a method in which seeds are selected using a single discrimination model by ignoring the quality of each trait, in which seeds are judged to be good if all desired traits are good and poor otherwise, seeds are scored using the single discrimination model, and seeds with scores above a threshold are predicted to be good seeds; (iii) a method in which seeds are selected using a discrimination model for each trait, focusing on the quality of each trait, and then the scores obtained using the discrimination model for each trait are integrated to obtain an integrated score, and the quality of each seed is predicted based on the ranking of the integrated score.

5. The selection method according to claim 1 or 2, further comprising a step of evaluating the accuracy of the discrimination model.

6. The method of claim 1 or 2, wherein the plant is a horticultural crop.

Citation Information

Patent Citations

  • Method for selecting cypress tree superior seed using hyperspectral image technology

    KR102112088B1

  • Method for nondestructively examining component of vegetable or the like by near-infrared spectroscopy and its device

    WO2005111583A1

  • Tree seed selecting method using near infrared light

    WO2016084452A1

  • Methods for analyzing plant material, for determining plant material components and for detecting plant diseases in plant material

    WO2022175309A1