Breeding method for wheat breeding

Abstract

The invention discloses a wheat breeding method, and relates to the technical field of crop intelligent breeding, and the method comprises the following steps: S1, obtaining a whole genome DNA sample of a wheat germplasm to be bred and plant tissues in different growth periods, detecting genome variation sites through a liquid phase breeding chip, and analyzing a gene space expression mode by combining a space-time transcriptome technology, a multispectral unmanned aerial vehicle and a root system in-situ sensor are used for collecting phenotype data, and a genotype and phenotype association database is established. According to the wheat breeding method provided by the invention, by integrating genomics, phenotypic omics and artificial intelligence technologies, the precision and intelligence of the breeding process are realized, the breeding efficiency and the achievement quality are remarkably improved, the dependence of traditional breeding on phenotypic selection is broken through, a genetic mechanism formed by crop traits can be comprehensively analyzed, and the method is suitable for popularization and application. And the matching relationship between the genetic loci and the environmental factors is accurately evaluated, so that a more scientific and reasonable breeding scheme is designed.

Description

A breeding method for wheat breeding Technical Field

[0001] This invention relates to the field of intelligent crop breeding technology, specifically to a breeding method for wheat. Background Technology

[0002] As one of the world's most important food crops, wheat breeding is of strategic significance for ensuring national food security and promoting sustainable agricultural development. Traditional breeding methods mainly rely on phenotypic selection and experience accumulation, resulting in long breeding cycles and low efficiency, making it difficult to meet the urgent needs of modern agriculture for high-yield, high-quality, and multi-resistant varieties. With the rapid development of genomics, phenomics, and artificial intelligence technologies, crop breeding is gradually transforming towards precision and intelligence. Modern breeding research has begun to integrate genomic variation sites, spatiotemporal expression patterns, and multidimensional phenotypic data to construct a systematic breeding technology system, providing new technical pathways for wheat variety improvement. This breeding method, which integrates multidisciplinary technologies, can more comprehensively analyze the genetic mechanisms of crop trait formation, laying the foundation for breeding superior varieties adapted to different ecological environments.

[0003] However, existing breeding technologies still face some challenges in practical applications. Association analysis of genotype and phenotypic data is often limited to single environments or specific growth stages, making it difficult to comprehensively reflect the dynamic performance of crops under complex environmental conditions. Breeding program design relies heavily on human experience, lacking a systematic assessment of the matching relationship between genetic loci and environmental factors, resulting in lower breeding efficiency. Methods for monitoring and screening phenotypic dynamics under environmental stress conditions are not precise enough, making it difficult to accurately capture changes in the expression of key traits. Furthermore, there is still room for improvement in the efficiency of breeding transformation from the laboratory to the field, hindering the rapid application and promotion of breeding results. To address these challenges, we propose a breeding method for wheat. Summary of the Invention

[0004] To address the aforementioned technical problems, a breeding method for wheat is provided. This technical solution solves the problems that the genotype-phenotype association analysis is limited by a single environment and a specific growth period, the breeding scheme relies on human experience and lacks matching assessment of genetic loci and environmental factors, the dynamic monitoring and screening of phenotypes under environmental stress is not precise, and the conversion efficiency from laboratory to field needs to be optimized.

[0005] To achieve the above objectives, the technical solution adopted by this invention is as follows: a wheat breeding method, comprising the following steps: S1, obtaining whole-genome DNA samples of the wheat germplasm to be bred and plant tissues at different growth stages, detecting genomic variation sites using liquid-phase breeding chips, analyzing gene spatial expression patterns using spatiotemporal transcriptomics technology, collecting phenotypic data using multispectral UAVs and root in-situ sensors, and constructing a genotype-phenotype association database; S2, importing multi-year environmental parameters of the target production area into this database, constructing a coupled AI framework for genome prediction and crop growth models, analyzing key genetic loci of the target trait, and forming a genetic locus-environment correlation database; S3. Combining the high-generation germplasm superior gene segment data with the matching relationship table, the genetic locus adaptability and gene set potential are analyzed through a preset AI matching algorithm to generate a parental intelligent matching scheme. After initial screening, a set of parental candidate matching schemes is constructed. S4. The hybrid offspring of the candidate schemes are placed in an intelligent breeding chamber, and multi-gradient temperature, light, water, and fertilizer and stress simulation conditions are set. Phenotypic dynamics are monitored in real time, and a set of superior offspring materials is generated through generational screening using an AI framework. S5. Combining the representative phenotypes of superior offspring across multiple generations with actual environmental data of the production area, a spatiotemporal and trait constraint function is established to screen materials, update the database and AI algorithm parameters, and generate an intelligent breeding scheme.

[0006] Preferably, step S1 includes: selecting wheat materials of different ecotypes and genetic backgrounds from the wheat germplasm resource bank, extracting whole-genome DNA samples from the young spikelet stage and grain-filling stage of each material, as well as leaf and root tissue samples from the corresponding stages; sequencing the whole-genome DNA samples using a liquid-phase breeding chip to identify SNPs, InDels, and structural variation sites in the samples, and obtaining chromosomal location information of genomic variation sites; sequencing the tissue samples using spatiotemporal transcriptomics technology to obtain gene expression data at different growth stages and different tissue parts; taking pictures of wheat at the jointing, heading, and maturity stages using a multispectral UAV, and simultaneously collecting root growth data in the soil using in-situ root sensors; performing noise reduction and normalization preprocessing on the two types of phenotypic data, and then storing them in association with genomic variation site data and gene expression data to construct a structured genotype-phenotype association database.

[0007] Preferably, S1 further includes: performing radiometric and geometric correction on the multispectral images collected by the UAV, extracting the region of a single wheat plant through an image segmentation algorithm, and calculating phenotypic parameters such as plant height, leaf area index, number of ears, and chlorophyll content; and deploying in-situ root sensors in a pre-embedded manner to collect root length, root diameter, root surface area, and root vitality data at fixed time intervals, with the collected data being uploaded to the data processing terminal in real time via a wireless transmission module.

[0008] Preferably, S2 includes: collecting meteorological data, soil data, and agricultural production records of the target production area for several consecutive years; performing outlier removal and missing value completion preprocessing on the data; importing the preprocessed environmental parameters into a coupled AI framework of genome prediction and crop growth model, wherein the genome prediction model is based on a genotype-phenotype association database and establishes a mapping relationship between genomic variation sites and target traits through machine learning algorithms, and the crop growth model simulates the growth and development process of wheat under different environmental conditions; achieving coupling of the two models through a two-way feedback mechanism; and based on the coupled AI framework, screening out key genetic loci associated with high yield, disease resistance, and stress tolerance target traits, and clarifying the expression effect of each genetic locus under different environmental conditions.

[0009] Preferably, the process of forming a matching table of genetic loci and environmental factors in S2 includes: analyzing the expression intensity changes of each key genetic locus under different temperature gradients, moisture conditions, and soil fertility to determine the environmental adaptation range of the locus; statistically analyzing the annual variation range and quarterly distribution characteristics of each environmental factor in the target production area to divide the environmental factor into grade intervals; matching the environmental adaptation range of the key genetic locus with the grade intervals of the environmental factors in the target production area to label the trait contribution value of each genetic locus under different environmental grades; sorting the genetic loci according to the chromosome positioning order and presenting the genetic locus name, chromosome position, target trait correlation, environmental adaptation range, and contribution value information under each environmental grade in tabular form.

[0010] Preferably, S3 includes: extracting superior gene segments related to the target trait from the high-generation germplasm population, determining the chromosomal location, genetic distance, and linkage relationship of each gene segment; calculating the fitness coefficient of key genetic loci among different germplasm materials based on the matching relationship table of genetic loci and environmental factors, wherein the calculation basis of the fitness coefficient includes the recombination rate of genetic loci, the direction of epistatic effect, and the consistency of environmental response; evaluating the aggregation potential of superior gene segments in hybrid offspring; and generating multiple parental combination schemes using a greedy algorithm or genetic algorithm based on the fitness coefficient and aggregation potential score, wherein each scheme specifies the selection of the father and mother and the hybridization method, and annotates the aggregation efficiency of the target genetic loci in the scheme.

[0011] Preferably, the initial screening process for the parental intelligent combination scheme in S3 is as follows: A core indicator system for initial screening is established, including: aggregation efficiency of target key genetic loci, overlap of parental flowering periods, carrier status of cytoplasmic male sterility genes, and genetic background complementarity; each combination scheme is quantitatively scored, and schemes with aggregation efficiency not less than 80%, flowering period overlap not less than 70%, absence of cytoplasmic male sterility genes, and genetic background complementarity scores not less than 85 are included in the candidate set; the feasibility of the initially selected schemes is verified, analyzing the seed setting rate, pollen viability, and key reproductive traits of the parents under actual planting conditions; the scoring results and feasibility verification conclusions are integrated to form a set of parental candidate combination schemes to be verified.

[0012] Preferably, S4 includes: setting multi-gradient parameters of temperature, light, water and fertilizer in the intelligent breeding cabin in stages according to the growth requirements of wheat at different growth stages; setting stress simulation conditions according to the main stress types of the target production area, including drought, salinity and low temperature; setting three sets of repeated experiments for each parameter gradient and stress condition, and setting a control group without stress; the intelligent breeding cabin adjusts each parameter in real time through an automatic control system.

[0013] Preferably, step S4 further includes: collecting phenotypic indicators such as plant height, stem diameter, leaf color value, ear length, and number of grains per ear using built-in sensors during the seedling, jointing, heading, and maturity stages of the hybrid offspring, with a collection frequency of once every 5 days, and increased to once every 2 days during key growth periods; transmitting the collected dynamic phenotypic data to the constructed AI framework in real time, comparing it with the predicted phenotypic values ​​of the target traits, and calculating the phenotypic conformity; screening offspring materials based on phenotypic conformity, removing materials with a phenotypic conformity of less than 60% during the seedling stage, removing materials with a phenotypic conformity of less than 75% from the jointing to heading stage, and retaining materials with a phenotypic conformity of ≥85% and meeting the stress resistance indicators at the maturity stage; and repeating the above monitoring and screening process for the selected superior offspring materials through generational advancement.

[0014] Preferably, S5 includes: collecting multi-generational representative data of superior offspring materials in the intelligent breeding cabin, including records of yield traits, quality traits, and stress resistance traits of each generation, and simultaneously collecting actual planting environment data of different plots in the target production area; constructing a spatiotemporal-trait constraint function, with the actual planting environment of the target production area as the independent variable, and the constraints being a yield deviation of ≤5% from the local main cultivated variety, a target resistance trait compliance rate of ≥95%, and quality indicators meeting relevant national standards; substituting the phenotypic data of superior offspring materials into the constraint function, and selecting materials that meet all constraints as final candidate varieties; extracting the full life cycle data of the final candidate varieties, including genotype information, representative data of each generation, and environmental response data, and updating them to the genotype and phenotype association database; and optimizing the model parameters of the AI ​​framework based on the newly added data, adjusting the genetic locus weights and environmental response coefficients.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: The wheat breeding method proposed in this invention, by integrating genomics, phenomics, and artificial intelligence technologies, achieves precision and intelligence in the breeding process, significantly improving breeding efficiency and the quality of results. It breaks through the dependence of traditional breeding on phenotypic selection, can comprehensively analyze the genetic mechanisms of crop trait formation, and accurately assess the matching relationship between genetic loci and environmental factors, thereby designing more scientific and reasonable breeding programs. With the help of the real-time monitoring and screening functions of the intelligent breeding cabin and AI framework, it can precisely capture the dynamic changes of key traits under environmental stress, ensuring the selection of superior varieties adapted to complex environmental conditions. By continuously updating the genotype and phenotypic association database and optimizing AI algorithm parameters, it also realizes the rapid iteration and upgrading of breeding results, providing strong support for the continuous improvement of wheat varieties, and has positive significance for ensuring national food security and promoting sustainable agricultural development. Attached Figure Description

[0016] Figure 1 is a flowchart of the method of the present invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Referring to Figure 1, a wheat breeding method includes the following steps: S1, obtaining whole-genome DNA samples of wheat germplasm to be bred and plant tissues at different growth stages, detecting genomic variation sites through liquid-phase breeding chips, analyzing gene spatial expression patterns using spatiotemporal transcriptomics technology, collecting phenotypic data using multispectral UAVs and root in-situ sensors, and constructing a genotype-phenotype association database; S1 includes: selecting wheat materials of different ecotypes and genetic backgrounds from the wheat germplasm resource bank, extracting whole-genome DNA samples from the young spikelet stage and grain-filling stage of each material, as well as leaf and root tissue samples from the corresponding stages.

[0019] Liquid-phase breeding chip design is based on wheat whole-genome sequencing data. Bioinformatics analysis is used to screen for polymorphic SNPs, InDel sites, and structural variation sites, and specific probes or primers are designed.

[0020] These probes or primers have high sequence specificity, can accurately identify target mutation sites, and have a coverage of more than 95% of the wheat genome, representing major functional genes and regulatory regions of wheat.

[0021] Liquid-phase breeding chips were used to sequence whole-genome DNA samples, identify SNPs, InDels, and structural variation sites in the samples, and obtain chromosomal location information of genomic variation sites.

[0022] The tissue sample processing workflow for spatiotemporal transcriptomics technology includes: rapidly freezing collected leaf and root tissue samples in liquid nitrogen to preserve RNA integrity; breaking the samples into fine particles using a tissue homogenizer; and extracting total RNA using an RNA extraction kit; converting RNA into cDNA via reverse transcription and constructing a library. For spatial localization, laser capture microdissection technology is employed, achieving precision down to the single-cell level to ensure the spatial accuracy of gene expression data. The standardization and normalization method for gene expression data uses the RPKM algorithm to eliminate the influence of sequencing depth and gene length on expression level calculations, ensuring comparability of gene expression levels between different samples.

[0023] Multispectral drones were used to take pictures of wheat during the jointing, heading, and maturity stages, and root growth data in the soil were collected simultaneously using in-situ root sensors.

[0024] Image segmentation and feature extraction employ a deep learning-based convolutional neural network algorithm model. This model, trained on a large amount of wheat image data, can accurately identify individual wheat plant regions and calculate phenotypic parameters such as plant height, leaf area index, number of ears, and chlorophyll content.

[0025] The in-situ root sensors are deployed via pre-embedding, with a density of at least one sensor per square meter. The depth of deployment is based on the distribution characteristics of wheat roots, covering a soil layer of 0-60 cm. The collected data is uploaded to the data processing terminal in real time via a wireless transmission module.

[0026] Before associating and storing multi-source heterogeneous data, data fusion and calibration techniques are employed. Radiometric and geometric corrections are performed on multispectral images acquired by UAVs to eliminate image distortion and illumination differences. Image data is synchronized temporally and spatially with root system data acquired by sensors to ensure temporal and spatial consistency between the two types of data. Genomic variation site data, gene expression level data, and phenotypic data are associating and storing data to construct a structured genotype-phenotype association database.

[0027] S1 further includes: performing radiometric and geometric correction on the multispectral images collected by the UAV, extracting the region of a single wheat plant through an image segmentation algorithm, and calculating phenotypic parameters such as plant height, leaf area index, number of ears, and chlorophyll content; and deploying in-situ root sensors in a pre-embedded manner to collect root length, root diameter, root surface area, and root vitality data at fixed time intervals, with the collected data being uploaded to the data processing terminal in real time via a wireless transmission module.

[0028] S2. Based on the database, import environmental parameters of the target production area over many years, construct a coupled AI framework for genome prediction and crop growth model, analyze key genetic loci of the target trait, and form a matching relationship table between genetic loci and environmental factors; S2 includes: collecting meteorological data, soil data and agricultural production record data of the target production area for several consecutive years, and performing outlier removal and missing value completion preprocessing on the data.

[0029] The machine learning algorithm chosen for the genome prediction model is the random forest algorithm, whose architecture consists of multiple decision trees, and improves prediction accuracy through ensemble learning.

[0030] During training, the data in the genotype-phenotype association database is divided into training, validation, and test sets. The grid search method is used to optimize hyperparameters and determine the optimal number of decision trees, maximum depth, and other parameters.

[0031] The mathematical modeling basis of crop growth models is a process-based model that considers physiological processes such as photosynthesis, respiration, water absorption and transport during wheat growth and development.

[0032] The parameterization scheme is set based on the biological characteristics of wheat and the environmental characteristics of the target production area, and is calibrated using field trial data.

[0033] The specific implementation logic of the two-way feedback mechanism is as follows: the genome prediction model transmits the predicted genotype impact data on the phenotype to the crop growth model. The crop growth model simulates the wheat growth and development process based on environmental parameters and feeds back the actual phenotype data to the genome prediction model. The data exchange frequency is once per hour, the format is structured data, and the trigger condition is the end of the model running cycle.

[0034] Meteorological data, soil data, and agricultural production records from the target production area were collected over several consecutive years. The data underwent preprocessing, including outlier removal and missing value completion. The preprocessed environmental parameters were then imported into a coupled AI framework of genome prediction and crop growth models. The genome prediction model, based on a genotype-phenotype association database, establishes a mapping relationship between genomic variation sites and target traits using machine learning algorithms. The crop growth model simulates the growth and development of wheat under different environmental conditions. A bidirectional feedback mechanism was used to couple the two models. Based on the coupled AI framework, key genetic loci associated with high yield, disease resistance, and stress tolerance were identified, and the expression effects of each genetic locus under different environmental conditions were clarified.

[0035] The biological basis for dividing the environmental adaptation range is the growth and development performance and genetic variation patterns of wheat under different environmental conditions, while the statistical basis is the degree of association between environmental factors and phenotypic traits determined by cluster analysis and principal component analysis.

[0036] The specific calculation method for quantifying trait contribution values ​​is as follows: based on the expression effects of key genetic loci under different environmental levels, a weighted average method is used to calculate their contribution to the target trait. Validation data are obtained from field trials and simulation experiments. The expression effects of each genetic locus under different environmental conditions are clarified.

[0037] The process of forming a matching table of genetic loci and environmental factors in S2 includes: analyzing the expression intensity changes of each key genetic locus under different temperature gradients, water conditions, and soil fertility to determine the environmental adaptation range of the locus; statistically analyzing the annual variation range and quarterly distribution characteristics of each environmental factor in the target production area to divide the environmental factor into grade intervals; matching the environmental adaptation range of the key genetic locus with the grade intervals of the environmental factors in the target production area to label the trait contribution value of each genetic locus under different environmental grades; sorting the genetic loci according to the chromosome positioning order and presenting the genetic locus name, chromosome position, target trait correlation, environmental adaptation range, and contribution value under each environmental grade in tabular form.

[0038] S3. Combining the superior gene segment data of high-generation germplasm with the matching relationship table, the genetic locus fitness and gene set potential are analyzed through a preset AI matching algorithm to generate intelligent parental matching schemes. After initial screening, a set of candidate parental matching schemes is constructed. S3 includes: extracting superior gene segments related to the target trait from high-generation germplasm populations, and determining the chromosomal position, genetic distance, and linkage relationship of each gene segment. The fitness coefficient is calculated as follows: Fitness coefficient = α × (1 - recombination rate) + β × epistatic effect direction coefficient + γ × environmental response consistency coefficient; where α, β, and γ are weighting coefficients, and their values ​​are determined based on genetic research and experimental data.

[0039] The recombination rate was obtained from genetic linkage maps and molecular marker-assisted selection data, and the epistatic effect was estimated using two-way ANOVA.

[0040] The aggregation potential score uses an analytic hierarchy process (AHP) to establish an evaluation model that considers factors such as the genetic effects of gene segments, linkage relationships, and environmental adaptability.

[0041] The greedy algorithm in this scenario works by selecting the combination with the highest fitness coefficient and aggregation potential score from all possible parental combinations, continuing until the target trait requirement is met or the maximum number of iterations is reached. The encoding scheme uses binary encoding, translating parental materials into binary strings. The fitness function is defined as a weighted sum of indicators such as aggregation efficiency and flowering overlap. The iteration terminates when the maximum number of iterations is reached or the fitness function value no longer significantly increases. The genetic algorithm is applied similarly, but incorporates selection, crossover, and mutation operations to increase population diversity.

[0042] Superior gene segments related to the target trait were extracted from the high-generation germplasm population, and the chromosomal location, genetic distance, and linkage relationships of each gene segment were determined. Based on the matching table of genetic loci and environmental factors, the fitness coefficients of key genetic loci among different germplasm materials were calculated. The calculation of fitness coefficients was based on the recombination rate of genetic loci, the direction of epistatic effects, and the consistency of environmental responses. The aggregation potential of superior gene segments in hybrid offspring was evaluated. Based on the fitness coefficients and aggregation potential scores, multiple parental combination schemes were generated using a greedy algorithm or a genetic algorithm. Each scheme specified the selection of paternal and maternal parents and the hybridization method, and the aggregation efficiency of the target genetic loci in the scheme was marked. The thresholds of each item in the initial screening index system were determined based on genetic principles, field trial data, and breeding target requirements. An aggregation efficiency of not less than 80% is to ensure a high aggregation probability of the target genetic locus in the hybrid offspring, and a flowering period overlap of not less than 70% is to ensure the success rate of hybridization pollination.

[0043] The initial screening process for the parental intelligent mating schemes in S3 is as follows: A core indicator system for initial screening is established, including: aggregation efficiency of target key genetic loci, overlap of parental flowering periods, carrier status of cytoplasmic male sterility genes, and genetic background complementarity. Each mating scheme is quantitatively scored, and schemes with aggregation efficiency of no less than 80%, flowering period overlap of no less than 70%, absence of cytoplasmic male sterility genes, and genetic background complementarity scores of no less than 85 points are included in the candidate set. The feasibility of the initially selected schemes is verified by analyzing the seed setting rate, pollen viability, and key reproductive traits of the parents under actual planting conditions. The scoring results and feasibility verification conclusions are integrated to form a set of candidate parental mating schemes to be verified.

[0044] S4. Place the hybrid offspring of the candidate schemes into the intelligent breeding chamber, set multi-gradient temperature, light, water, and fertilizer conditions, and simulate stress conditions, monitor phenotypic dynamics in real time, and generate a set of superior offspring materials through generational screening using an AI framework; S4 includes: setting the multi-gradient parameters of temperature, light, water, and fertilizer in the intelligent breeding chamber in stages according to the growth requirements of wheat at different growth stages. The temperature gradient is set to 10-30℃ with a control accuracy of ±0.5℃; the light intensity gradient is set to 0-2000μmol / m²·s with a control accuracy of ±50μmol / m²·s; the water gradient is set to 30%-90% of field capacity with a control accuracy of ±2%; the fertilizer gradient is set with different nitrogen, phosphorus, and potassium concentrations according to the wheat growth stage and nutrient requirements. The stress simulation conditions are set according to the main stress types of the target production area: the soil water potential range for drought stress is -0.5MPa to -1.5MPa, the salt concentration for saline-alkali stress is 0.2%-0.8%, and the temperature for low temperature stress is 0-5℃. Three replicate experiments were set up for each parameter gradient and stress condition, and a control group without stress was also set up; the intelligent breeding cabin adjusts each parameter in real time through an automatic control system.

[0045] The S4 further includes: collecting phenotypic indicators such as plant height, stem diameter, leaf color value, ear length, and number of grains per ear through built-in sensors during the seedling, jointing, heading, and maturity stages of the hybrid offspring. The built-in sensors include laser rangefinders, displacement sensors, and spectral sensors, with accuracies of ±1mm, ±0.01mm, and ±1nm, respectively.

[0046] The sensors are evenly distributed within the intelligent breeding chamber to ensure comprehensive collection of plant phenotypic data. Data collection is conducted every 5 days, increasing to every 2 days during critical growth periods.

[0047] The collected phenotypic dynamic data is transmitted in real time to the constructed AI framework via a wireless transmission protocol, and compared with the predicted phenotypic values ​​of the target trait to calculate the phenotypic conformity.

[0048] The comparative analysis process between phenotypic dynamic data and AI predictions is as follows: First, the collected phenotypic data is preprocessed, including data cleaning and outlier removal. The preprocessed data is then compared with the AI ​​predictions, and the error between the two is calculated. Based on the error, the phenotypic conformity is calculated as: Phenotypic conformity = (1 - Error / Predicted value) × 100%.

[0049] The rationale for setting screening thresholds for different growth stages is as follows: The seedling stage is a critical period for wheat growth and development, and is highly sensitive to environmental changes. Therefore, the screening threshold is lower, eliminating materials with a phenotypic conformity of less than 60% to ensure that offspring materials have a good growth foundation. The jointing to heading stage is a period in which wheat's vegetative and reproductive growth proceed simultaneously, which is crucial for the formation of target traits. The screening threshold is raised to 75% to select materials with high genetic potential. The maturity stage is a critical period for the formation of wheat yield and quality. Materials with a phenotypic conformity ≥85% and meeting the stress resistance index are retained to ensure that the final selected materials have excellent comprehensive traits.

[0050] Progeny materials were screened based on phenotypic conformity. Materials with phenotypic conformity below 60% were removed at the seedling stage, and materials with phenotypic conformity below 75% were removed from the jointing to heading stage. Materials with phenotypic conformity ≥ 85% and meeting the stress resistance index were retained at maturity. The selected superior progeny materials were then promoted through generations, and the above monitoring and screening process was repeated.

[0051] S5. By combining the representative traits of superior offspring across multiple generations with actual environmental data from the production area, a spatiotemporal and trait constraint function is established to screen materials, update the database and AI algorithm parameters, and generate an intelligent breeding program.

[0052] The S5 includes: collecting multi-generational representative data of superior offspring materials in the intelligent breeding cabin, including records of yield traits, quality traits, and stress resistance traits of each generation, and simultaneously collecting actual planting environment data of different plots in the target production area.

[0053] The mathematical model for the spatiotemporal-trait constraint function is a linear programming model, with the actual planting environment of the target production area as the independent variable, and the constraints as follows: yield deviation from the local main variety ≤ 5%, target resistance trait compliance rate ≥ 95%, and quality indicators meeting relevant national standards. The mathematical expressions for each constraint are as follows: Yield deviation constraint: (predicted yield - main variety yield) / main variety yield × 100% ≤ 5%; Resistance compliance rate constraint: number of compliant samples / total number of samples × 100% ≥ 95%; Quality indicator constraint: values ​​of each quality indicator are within the national standard range.

[0054] The weighting principle was determined based on breeding objectives and market demand, with yield, resistance, and quality weighted at 0.4, 0.3, and 0.3, respectively. Phenotypic data of superior progeny materials were substituted into the constraint function to select materials that met all constraints as final candidate varieties. The full life-cycle data of the final candidate varieties, including genotype information, representative typographic data for each generation, and environmental response data, were extracted and updated to the genotype-phenotype association database.

[0055] The AI ​​framework model parameters were optimized based on the newly added data, employing an incremental learning algorithm to adjust the genetic locus weights and environmental response coefficients. The optimization objective of the parameter adjustments was to improve the model's prediction accuracy and generalization ability, with the convergence criterion being that the model's performance metrics on the validation set no longer showed a significant improvement.

[0056] Data on the representative traits of superior progeny materials across multiple generations were collected within the intelligent breeding chamber, including records of yield, quality, and stress resistance traits for each generation. Simultaneously, actual planting environment data from different plots within the target production area were collected. A spatiotemporal-trait constraint function was constructed, using the actual planting environment of the target production area as the independent variable and constraints including a yield deviation of ≤5% from the local main cultivated variety, a target resistance trait compliance rate of ≥95%, and quality indicators meeting relevant national standards. Phenotypic data of superior progeny materials were substituted into the constraint function to select materials that met all constraints as final candidate varieties. Full life-cycle data of the final candidate varieties, including genotype information, representative trait data for each generation, and environmental response data, were extracted and updated to the genotype-phenotype association database. Based on the newly added data, the model parameters of the AI ​​framework were optimized, adjusting the genetic locus weights and environmental response coefficients.

[0057] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A breeding method for wheat, characterized in that, Includes the following steps: S1. Obtain whole-genome DNA samples and plant tissues at different growth stages from the wheat germplasm to be bred. Detect genomic variation sites using liquid-phase breeding chips, analyze gene spatial expression patterns using spatiotemporal transcriptomics technology, and collect phenotypic data using multispectral UAVs and in-situ root sensors to construct a genotype-phenotype association database. S2. Based on this database, import multi-year environmental parameters from the target production area, construct an AI framework coupling genome prediction and crop growth models, analyze key genetic loci of the target traits, and form a table of matching relationships between genetic loci and environmental factors. S3. Combine superior genes from high-generation germplasm. The data segment and matching relationship table are analyzed using a preset AI matching algorithm to determine the genetic locus suitability and gene set potential, generating intelligent parental matching schemes. After initial screening, a set of candidate parental matching schemes is established. S4: The hybrid offspring of the candidate schemes are placed in an intelligent breeding chamber, and multi-gradient and stress simulation conditions of temperature, light, water, and fertilizer are set. Phenotypic dynamics are monitored in real time, and a set of superior offspring materials is generated through generational screening using an AI framework. S5: Combining the representative phenotypes of superior offspring across multiple generations with actual environmental data of the production area, a spatiotemporal and trait constraint function is established to screen materials, update the database and AI algorithm parameters, and generate intelligent breeding schemes.

2. The wheat breeding method according to claim 1, characterized in that, S1 includes: selecting wheat materials of different ecotypes and genetic backgrounds from the wheat germplasm resource bank, extracting whole-genome DNA samples from the young spikelet stage and grain-filling stage of each material, as well as leaf and root tissue samples from the corresponding stages; sequencing the whole-genome DNA samples using liquid-phase breeding chips to identify SNPs, InDels, and structural variation sites in the samples, and obtaining chromosomal location information of genomic variation sites; sequencing tissue samples using spatiotemporal transcriptomics technology to obtain gene expression data at different growth stages and different tissue parts; using a multispectral UAV to take pictures at the wheat jointing stage, heading stage, and maturity stage, and simultaneously collecting root growth data in the soil through in-situ root sensors; performing noise reduction and normalization preprocessing on the two types of phenotypic data, and then storing them in association with genomic variation site data and gene expression data to construct a structured genotype-phenotype association database.

3. The wheat breeding method according to claim 2, characterized in that, S1 further includes: performing radiometric and geometric correction on the multispectral images collected by the UAV, extracting the region of a single wheat plant through an image segmentation algorithm, and calculating phenotypic parameters such as plant height, leaf area index, number of ears, and chlorophyll content; and deploying in-situ root sensors in a pre-embedded manner to collect root length, root diameter, root surface area, and root vitality data at fixed time intervals, with the collected data being uploaded to the data processing terminal in real time via a wireless transmission module.

4. The wheat breeding method according to claim 3, characterized in that, S2 includes: collecting meteorological data, soil data, and agricultural production records of the target production area for several consecutive years; performing outlier removal and missing value completion preprocessing on the data; importing the preprocessed environmental parameters into a coupled AI framework of genome prediction and crop growth model, wherein the genome prediction model is based on a genotype-phenotype association database and establishes a mapping relationship between genomic variation sites and target traits through machine learning algorithms, and the crop growth model simulates the growth and development process of wheat under different environmental conditions; achieving coupling of the two models through a two-way feedback mechanism; and based on the coupled AI framework, screening out key genetic loci associated with high yield, disease resistance, and stress tolerance target traits, and clarifying the expression effect of each genetic locus under different environmental conditions.

5. A wheat breeding method according to claim 4, characterized in that, The process of forming a matching table of genetic loci and environmental factors in S2 includes: analyzing the expression intensity changes of each key genetic locus under different temperature gradients, water conditions, and soil fertility to determine the environmental adaptation range of the locus; statistically analyzing the annual variation range and quarterly distribution characteristics of each environmental factor in the target production area to divide the environmental factor into grade intervals; matching the environmental adaptation range of the key genetic locus with the grade intervals of the environmental factors in the target production area to label the trait contribution value of each genetic locus under different environmental grades; sorting the genetic loci according to the chromosome positioning order and presenting the genetic locus name, chromosome position, target trait correlation, environmental adaptation range, and contribution value under each environmental grade in tabular form.

6. A wheat breeding method according to claim 5, characterized in that, S3 includes: extracting superior gene segments related to the target trait from the high-generation germplasm population, determining the chromosomal location, genetic distance, and linkage relationship of each gene segment; calculating the fitness coefficient of key genetic loci among different germplasm materials based on the matching relationship table of genetic loci and environmental factors, the calculation basis of the fitness coefficient includes the recombination rate of genetic loci, the direction of epistatic effect, and the consistency of environmental response; evaluating the aggregation potential of superior gene segments in hybrid offspring; and generating multiple parental combination schemes using a greedy algorithm or genetic algorithm based on the fitness coefficient and aggregation potential score, each scheme specifying the selection of the father and mother and the hybridization method, and labeling the aggregation efficiency of the target genetic loci in the scheme.

7. A wheat breeding method according to claim 6, characterized in that, The initial screening process for the parental intelligent mating schemes in S3 is as follows: A core indicator system for initial screening is established, including: aggregation efficiency of target key genetic loci, overlap of parental flowering periods, carrier status of cytoplasmic male sterility genes, and genetic background complementarity. Each mating scheme is quantitatively scored, and schemes with aggregation efficiency of no less than 80%, flowering period overlap of no less than 70%, absence of cytoplasmic male sterility genes, and genetic background complementarity scores of no less than 85 points are included in the candidate set. The feasibility of the initially selected schemes is verified by analyzing the seed setting rate, pollen viability, and key reproductive traits of the parents under actual planting conditions. The scoring results and feasibility verification conclusions are integrated to form a set of candidate parental mating schemes to be verified.

8. A wheat breeding method according to claim 7, characterized in that, The S4 includes: setting multi-gradient parameters of temperature, light, water and fertilizer in the intelligent breeding cabin in stages according to the growth requirements of wheat at different growth stages; setting stress simulation conditions according to the main stress types of the target production area, including drought, salinity and low temperature; setting three sets of repeated experiments for each parameter gradient and stress condition, and setting a control group without stress; the intelligent breeding cabin adjusts each parameter in real time through an automatic control system.

9. A wheat breeding method according to claim 8, characterized in that, S4 further includes: collecting phenotypic indicators such as plant height, stem diameter, leaf color value, ear length, and number of grains per ear using built-in sensors during the seedling, jointing, heading, and maturity stages of the hybrid offspring, with a collection frequency of once every 5 days, and increased to once every 2 days during key growth periods; transmitting the collected dynamic phenotypic data in real time to the constructed AI framework, comparing it with the predicted phenotypic values ​​of the target traits, and calculating the phenotypic conformity; screening offspring materials based on phenotypic conformity: removing materials with a phenotypic conformity of less than 60% during the seedling stage, removing materials with a phenotypic conformity of less than 75% from the jointing to heading stage, and retaining materials with a phenotypic conformity of ≥85% and meeting the stress resistance indicators at the maturity stage; and repeating the above monitoring and screening process for the selected superior offspring materials through generational advancement.

10. A wheat breeding method according to claim 9, characterized in that, S5 includes: collecting multi-generational representative data of superior offspring materials in the intelligent breeding cabin, including records of yield traits, quality traits, and stress resistance traits of each generation, and simultaneously collecting actual planting environment data of different plots in the target production area; constructing a spatiotemporal-trait constraint function, with the actual planting environment of the target production area as the independent variable, and the constraints being a yield deviation of ≤5% from the local main cultivated variety, a target resistance trait compliance rate of ≥95%, and quality indicators meeting relevant national standards; substituting the phenotypic data of superior offspring materials into the constraint function, and selecting materials that meet all constraints as final candidate varieties; extracting the full life cycle data of the final candidate varieties, including genotype information, representative data of each generation, and environmental response data, and updating them to the genotype and phenotype association database; and optimizing the model parameters of the AI ​​framework based on the newly added data, adjusting the genetic locus weights and environmental response coefficients.

Images