A system and method for selecting parents of rapeseed hybrid combination

Abstract

This invention relates to the field of intelligent crop breeding technology, and discloses a rapeseed hybrid parent selection system and method. The method includes collecting and analyzing genotype data among rapeseed varieties, removing false-positive variations introduced by sequencing errors, constructing a molecular marker matrix and extracting features to obtain a first feature vector; classifying rapeseed varieties based on the first feature vector to obtain their genetic diversity levels; constructing initial parent combinations based on the genetic diversity levels, and predicting the adaptability of the hybrid offspring of the initial parent combinations in multiple environments by training an environmental adaptability prediction model, and calculating their environmental adaptability phenotypic values; finally, optimizing the initial parent combinations based on the adaptability scores to obtain optimized parent combinations. This invention, through a data-driven approach, improves the targeting and effectiveness of parent combination selection, significantly enhancing breeding efficiency and the environmental adaptability of hybrid offspring.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent crop breeding, and more specifically, to a system and method for selecting parents of a rapeseed hybrid combination. Background Art

[0002] The existing parent selection of rapeseed hybrid combinations mainly relies on the experience of breeders and lacks scientific quantitative methods. With the development of molecular biology and big data analysis technology, it is possible to use genomic information to assist in the parent selection of rapeseed.

[0003] In recent years, some studies have attempted to apply whole genome selection (GS) prediction to crop breeding. A Chinese patent with publication number CN116564407A discloses a method for efficiently predicting the flowering period of chrysanthemum based on whole genome selection. This method obtains significantly associated SNP sites through genome-wide association analysis (GWAS) and uses a support vector machine (SVM) model to predict the flowering period of chrysanthemum with an accuracy of 0.90~0.95. This provides a new idea for early prediction of crop phenotypes.

[0004] Patent publication number EP3326093A1 discloses an improved computer-implemented method for predicting the true agronomic value of plants. The method obtains genotype and phenotypic data of a population line and inputs it into a mixed model combining fixed effects and random effects to predict the agronomic value and breeding value of the plant. This lays the foundation for quantitative evaluation of plant phenotypes.

[0005] In addition, a Chinese patent application with publication number CN117594129A discloses an optimal prediction system for poplar growth traits based on whole genome selection, as well as its construction method and application. This method uses GWAS to identify SNP sites significantly associated with poplar growth traits, and compares the effects of different machine learning models and the number of SNP markers on GS prediction accuracy to establish an optimal GS prediction system. This system can quickly and accurately calculate the breeding value of poplar growth traits, with a prediction accuracy of 0.77, which accelerates the selection and breeding process of excellent forest germplasm.

[0006] However, existing methods mainly focus on the prediction of single traits and do not fully consider the impact of environmental factors. Combining genotype with environment to predict the performance of multiple traits of parental hybrid offspring in different environments is of great significance for guiding the optimal selection of hybrid parents. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides a system and method for selecting parents of a rapeseed hybrid combination.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for selecting parents of a rapeseed hybrid combination, comprising:

[0010] Collect genotype data that characterize the genetic differences between rapeseed varieties; eliminate false positive variations introduced by sequencing errors in the genotype data and construct a molecular marker matrix; perform feature extraction on the molecular marker matrix to obtain the first eigenvector; classify rapeseed varieties based on the first eigenvector to obtain the genetic diversity level of rapeseed varieties;

[0011] Initial parent combinations were constructed based on the genetic diversity levels of rapeseed varieties, and environmental adaptability prediction models were constructed and trained;

[0012] The genotype data of each rapeseed variety in the initial parent combination are obtained and combined to generate an initial genotype data group; the initial planting environment data of the hybrid offspring of the initial parent combination are obtained, and based on the initial genotype data group and the initial planting environment data of the hybrid offspring of the initial parent combination, an environmental adaptability prediction model is used to predict the adaptability performance of the hybrid offspring of the initial parent combination in multiple environments, and the environmental adaptability phenotypic values ​​of the hybrid offspring are obtained, and the adaptability scores of the hybrid offspring of each initial parent combination in different environments are obtained according to the environmental adaptability phenotypic values; the initial parent combination is optimized based on the adaptability score to obtain an optimized parent combination.

[0013] Furthermore, the method of extracting features from the molecular marker matrix to obtain a first feature vector includes:

[0014] The allele frequencies in the molecular marker matrix are coded with 0 and 1 to form a two-dimensional matrix with varieties as rows and marker sites as columns, which is used as the original molecular marker matrix;

[0015] The original molecular marker matrix was preprocessed to eliminate dimension differences and generate a standardized molecular marker matrix. The features that explained the largest proportion of genetic variation were extracted from the standardized molecular marker matrix as the first eigenvector.

[0016] Furthermore, the method for classifying rapeseed varieties based on the first eigenvector to obtain the genetic diversity level of the rapeseed varieties includes:

[0017] Evaluate the clustering quality under different cluster numbers and select the optimal cluster number K*;

[0018] Taking the first eigenvector as input, the K-means algorithm was used to divide the rapeseed varieties into K* genetic diversity levels as the K-means clustering result;

[0019] The Euclidean distance between rapeseed varieties is calculated based on the first eigenvector as the genetic distance between rapeseed varieties, a cluster tree is constructed, and rapeseed varieties are divided into different genetic diversity levels according to the structure of the cluster tree as the hierarchical clustering result;

[0020] Compare the K-means clustering results and the hierarchical clustering results, and select the best clustering result as the final genetic diversity level of rapeseed varieties;

[0021] The method for constructing an initial parent combination based on the genetic diversity level of rapeseed varieties comprises:

[0022] Setting genetic diversity thresholds for different levels of genetic diversity;

[0023] Set breeding target thresholds based on the genetic diversity level and genetic diversity threshold of the breeding target;

[0024] All parent combinations are traversed, and the parent combination in which the average genetic distance between all parents is greater than the breeding target threshold is used as the initial parent combination.

[0025] Furthermore, the method for constructing and training an environmental adaptability prediction model includes:

[0026] The genotype data of each rapeseed variety in the historical parent combination is obtained and combined to generate a historical genotype data group; the historical planting environment data and historical phenotypic data of the hybrid offspring of the historical parent combination are obtained;

[0027] A neural network is used to construct an environmental adaptability prediction model, which includes an input layer, several hidden layers and an output layer. The input layer receives historical genotype data sets and historical planting environment data, and the output layer outputs the predicted environmental adaptability phenotypic values ​​of the hybrid offspring of the historical initial parent combinations, with the historical phenotypic data serving as true labels. The number of neurons in the output layer corresponds to the number of environmental adaptability phenotypes to be predicted.

[0028] Furthermore, the method for obtaining the adaptability scores of the hybrid offspring of each initial parent combination in different environments according to the environmental adaptability phenotype value includes:

[0029]

[0030] in, represents the fitness score of the hybrid offspring of the i-th initial parent combination in the j-th environment, represents the value of the kth adaptive phenotype of the hybrid offspring of the i-th initial parent combination in the j-th environment, represents the mean of the kth adaptive phenotype of the hybrid offspring of all initial parent combinations in the jth environment, represents the standard deviation of the kth adaptive phenotype of the hybrid offspring of all initial parent combinations in the jth environment, Indicates The hybrid offspring of the initial parent combination is Under the environment The rank of the adaptive phenotype, represents the hybrid offspring of all initial parent combinations The mean of the fitness phenotype ranks, represents the weight of the kth adaptive phenotype, Indicates The importance coefficient of each adaptive phenotype, Indicates The importance coefficient of the fitness phenotype rank, Indicates The sensitivity parameter of the adaptive phenotype, Indicates The sensitivity parameter of the adaptive phenotype rank, n represents the number of adaptive phenotypes, Indicates The maximum value of the adaptive phenotype among the hybrid offspring of all initial parent combinations, Indicates The maximum rank of a fitness phenotype among the hybrid offspring of all initial parental combinations.

[0031] Furthermore, the method of optimizing the initial parent combination based on the adaptability score to obtain the optimized parent combination includes:

[0032] Obtain the adaptability scores of the hybrid offspring of the initial parent combinations under different target planting environments, and form an adaptability score matrix with the initial parent combinations as rows and the target planting environments as columns;

[0033] Based on the adaptability score matrix, the mean and variance of the adaptability scores of the hybrid offspring of each initial parent combination under all target planting environments were calculated, and the mean of the adaptability score under different planting environments was screened to be greater than the mean adaptability threshold. And the variance is less than the adaptive variance threshold The parent combination is used as a cross-domain optimal parent combination for multi-target planting environments;

[0034] For the target planting environment j, select the target planting environment j whose adaptability score is greater than the adaptability threshold. parent combination to form a directional preferred parent combination for the target planting environment j.

[0035] Furthermore, the adaptive threshold The calculation methods include:

[0036] ;

[0037] in, represents the adaptability score of the hybrid offspring of the first initial parent combination in the jth target planting environment, represents the adaptability score of the hybrid offspring of the second initial parent combination in the jth target planting environment, represents the adaptability score of the hybrid offspring of the mth initial parent combination in the jth target planting environment, is the median function, is the absolute median deviation function, is the regulating factor, is the adaptability threshold of the jth target planting environment.

[0038] A rapeseed hybrid combination parent selection system, which is used to implement the above-mentioned rapeseed hybrid combination parent selection method, the system comprising:

[0039] Data collection module: used to collect genotype data that characterizes the genetic differences between rapeseed varieties; used to obtain historical planting environment data and historical phenotypic data of the hybrid offspring of historical parent combinations; collect the genotype data of each rapeseed variety in the initial parent combination and combine them to generate an initial genotype data set; collect the initial planting environment data of the hybrid offspring of the initial parent combination;

[0040] Data preprocessing module: used to eliminate false positive variants introduced by sequencing errors in genotype data, screen markers with significant polymorphism, and construct a molecular marker matrix;

[0041] Feature extraction module: used to extract features from the molecular marker matrix and obtain the first feature vector;

[0042] Cluster analysis module: used to classify rapeseed varieties based on the first eigenvector using a clustering algorithm to obtain the genetic diversity level of the rapeseed varieties;

[0043] Parent combination construction module: used to construct initial parent combinations based on the genetic diversity levels of rapeseed varieties;

[0044] Phenotypic value prediction module: used to construct and train an environmental adaptability prediction model, and obtain the environmental adaptability phenotypic value of the hybrid offspring of the initial parent combination according to the initial genotype data group of the initial parent combination, the initial planting environment data and the environmental adaptability prediction model; and obtain the adaptability score of the hybrid offspring of each initial parent combination in different environments according to the environmental adaptability phenotypic value;

[0045] Parent combination optimization module: optimize the initial parent combination based on the fitness score of the hybrid offspring of the initial parent combination to obtain the optimized parent combination.

[0046] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the above-mentioned method for selecting parents of a rapeseed hybrid combination is implemented.

[0047] A computer-readable storage medium stores a computer program, and when the computer program is executed, the method for selecting parents of a rapeseed hybrid combination is implemented.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] By making full use of molecular marker information and phenotypic information, the genetic diversity level of rapeseed varieties is evaluated based on the molecular marker data of rapeseed varieties, and parent combinations are constructed on this basis to predict the environmental adaptability of hybrid offspring, and finally select hybrid combinations with high adaptability to the environment and excellent traits. The entire method process is mainly data-driven, which greatly improves the pertinence and effectiveness of parent combination.

[0050] When constructing the initial parent combination, the concept of genetic diversity threshold is introduced. According to the requirements of breeding goals, a reasonable breeding target threshold is set to strengthen the purpose and direction of parent combination. At the same time, the genetic distance between parents in the parent combination is considered, and parents with large genetic distance are preferred, which is conducive to the formation of hybrid vigor.

[0051] By constructing an environmental adaptability prediction model to predict the environmental adaptability of hybrid offspring, the interaction effect between genotype and environment is fully considered. The prediction model is constructed using a neural network and historical breeding data is input for training, which greatly improves the accuracy and reliability of the prediction and provides a more accurate estimate of the trait performance of hybrid offspring.

[0052] In the process of optimizing parent combinations, directional optimization and cross-domain optimization strategies were proposed for specific target planting environments and multi-target planting environments, respectively. Directional optimization ensures the adaptability of hybrid offspring in the target environment, while cross-domain optimization takes into account the comprehensive performance in different environments, improves the robustness of hybrid combinations to environmental changes, and meets diverse breeding needs.

[0053] The system is highly automated, with multiple functional modules working together, efficiently integrating multiple links such as data collection, preprocessing, feature extraction, cluster analysis, parent combination construction, phenotype prediction, and parent combination optimization. Each module can be flexibly split and combined to adapt to different breeding processes, and has strong practicality and promotion value.

[0054] The method for selecting parents of rapeseed hybrid combinations proposed in this invention can be widely applied to the selection of hybrid parents of other crops, providing a new idea for intelligent design breeding. By integrating phenotypic, genotypic, and environmental data, using artificial intelligence and big data analysis technology, the entire breeding process can be optimized, the selection of excellent varieties can be accelerated, and strong support can be provided for the development of modern agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0056] Figure 1 This is a flow chart of a method for selecting parents of a rapeseed hybrid combination in the present invention;

[0057] Figure 2 It is a flow chart of a method for extracting features from a molecular marker matrix to obtain a first feature vector in a method for selecting parents of a rapeseed hybrid combination of the present invention;

[0058] Figure 3 It is a flow chart of a method for optimizing an initial parent combination to obtain an optimized parent combination in a method for selecting parents of a rapeseed hybrid combination of the present invention;

[0059] Figure 4 The present invention is a functional module diagram of a rapeseed hybrid combination parent selection system. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] Example 1

[0062] See also Figure 1 As shown, this embodiment provides a method for selecting parents of a rapeseed hybrid combination, comprising:

[0063] Step S1000, collecting genotype data that characterizes genetic differences between rapeseed varieties; eliminating false positive variations introduced by sequencing errors in the genotype data, and constructing a molecular marker matrix; performing feature extraction on the molecular marker matrix to obtain a first eigenvector; and classifying the rapeseed varieties based on the first eigenvector to obtain the genetic diversity level of the rapeseed varieties;

[0064] Further, step S1000 includes:

[0065] Step S1100, collecting genotype data describing genetic differences between rapeseed varieties; performing systematic phenotypic identification on rapeseed varieties under multiple representative environmental conditions to obtain phenotypic data of rapeseed varieties; obtaining planting environment data of rapeseed varieties, including meteorological data, soil physical and chemical property data, and management measures data;

[0066] Specifically, whole genome sequencing, resequencing, gene chips and other technologies are used to obtain genotype data such as SNP (single nucleotide polymorphism) and InDel of rapeseed varieties to characterize genetic differences between varieties. Under multiple representative environmental conditions (such as cold, drought, saline-alkali and other stress environments), rapeseed varieties are systematically phenotypic identified, focusing on important agronomic traits such as yield, resistance, and quality. Data on the planting environment of rapeseed varieties are collected simultaneously, including meteorological data (temperature, humidity, light, etc.), soil physical and chemical properties data (fertility, pH value, etc.), and management measures data (sowing time, density, fertilization, etc.). In the process of genotype data collection, attention should be paid to the representativeness of samples and sequencing quality. Samples should cover the main groups and special germplasm resources of rapeseed, and the sequencing depth should meet the needs of variant detection. In the process of phenotypic data collection, strict field test plans and phenotypic identification standards should be formulated, and experiments should be repeated at multiple points for many years to ensure the reliability of data and the accuracy of adaptability evaluation. The collection of planting environment data should be comprehensive and systematic to provide support for accurately characterizing the interaction between genotype and environment. High-quality multi-dimensional data is the basis of intelligent parent matching, so strict quality control must be carried out during the data collection phase to provide reliable data support for subsequent analysis.

[0067] Step S1200, remove false positive variants introduced by sequencing errors in genotype data, screen markers with significant polymorphisms, and construct a molecular marker matrix; perform outlier detection and correction on phenotypic data, and fill in missing data;

[0068] Specifically, the original genotype and phenotype data often contain noise and missing values, and direct use will affect the accuracy of subsequent analysis. During the genome sequencing process, some erroneous sequencing results may be generated due to technical or experimental operations. These errors may lead to the misidentification of DNA sequences. False positive variation refers to the variation detected in the sequencing data that is actually caused by sequencing errors and is not a variation that actually exists in the genome of the organism. For example, a certain position is mistakenly considered to have a SNP, but in fact the position is consistent in all individuals and has no variation. Polymorphism refers to the presence of two or more alleles at a certain locus, and these alleles have a high frequency in the population. A marker with significant polymorphism means that the marker has a large variation between different individuals. A variety of quality control and preprocessing methods are used to improve data quality. When filtering SNP markers, markers with large information content and strong discrimination can be screened based on indicators such as mismatch rate, missing rate, and minor allele frequency (MAF). Phenotypic outlier correction can use statistical methods such as the 3σ principle and box plots to automatically identify outliers, and then breeding experts can identify and adjust them in combination with agronomic knowledge. When filling missing data, it is necessary to select an appropriate algorithm based on the missing mechanism and missing rate of the data. Through a series of preprocessing steps, the integrity and availability of the data can be significantly improved, laying a data foundation for mining genetic laws.

[0069] Step S1300, extracting features from the molecular marker matrix to obtain a first feature vector; and classifying rapeseed varieties based on the first feature vector;

[0070] Further, step S1300 includes:

[0071] Step S1310, extracting features from the molecular marker matrix to obtain a first feature vector;

[0072] Furthermore, if Figure 2 As shown, step S1310 includes:

[0073] Step S1311, using 0 and 1 to encode the allele frequencies in the molecular marker matrix to form a two-dimensional matrix with varieties as rows and marker sites as columns, as the original molecular marker matrix;

[0074] Step S1312, preprocessing the original molecular marker matrix to eliminate dimension differences, generating a standardized molecular marker matrix, and extracting the feature that explains the largest proportion of genetic variation from the standardized molecular marker matrix as the first feature vector;

[0075] Specifically, allele frequency refers to the frequency of a particular allele at a locus in a population. Each locus usually has two or more alleles (for example, A and a), and the allele frequency describes the proportion of these alleles in a population. In genetic research, it is often necessary to encode alleles for data analysis. The encoding of 0 and 1 is suitable for dimorphic alleles, that is, there are only two alleles at a locus. In genomic analysis, many SNP markers are dimorphic, that is, there are only two possible nucleotide changes at a specific position. Encoding with 0 and 1 can simplify data storage and processing. 0 encoding indicates that the individual has one allele at the locus (usually the reference allele, or the allele with a value of 0 in binary representation). 1 encoding indicates that the individual has another allele at the locus (usually the alternative allele, or the allele with a value of 1 in binary representation). First, according to the results of molecular marker detection, the allele frequency is marked with 0 and 1 encoding, and the original data matrix with varieties as rows and marker sites as columns is constructed. Since the dimensional differences of different marker loci may be large, the original matrix needs to be preprocessed by centering and standardization to eliminate the dimensional effect. Based on the standardized matrix, principal component analysis is used to extract several eigenvectors. Among them, the eigenvector corresponding to the first principal component explains the largest proportion of genetic variation and can be used as a comprehensive representation of the genetic composition of the variety.

[0076] Step S1320, based on the first eigenvector, using a clustering algorithm to classify the rapeseed varieties to obtain the genetic diversity level of the rapeseed varieties;

[0077] Further, step S1320 includes:

[0078] Step S1321, evaluating the clustering quality under different clustering numbers, and selecting the optimal clustering number K*;

[0079] The step S1321 includes:

[0080] S13211: Set the number of clusters The value range is ,in and Represents the number of clusters The minimum and maximum values ​​of

[0081] S13212: For each Value, use clustering algorithm to cluster the first eigenvector and get the corresponding clustering result , repeat the clustering process N* times, and select the best clustering result as the current Clustering results under value ;

[0082] S13213: Based on clustering results , calculate the current Clustering quality evaluation index under value ; Clustering quality evaluation index It consists of two parts: intra-class distance and inter-class distance;

[0083] The current Clustering quality evaluation index under value The calculation methods include:

[0084]

[0085] in, Indicates the current Value The optimal clustering , Indicates The center point of the optimal cluster, Indicates The number of samples for the optimal clustering, Representation sample To the optimal cluster center The distance represents the total number of samples, K is the number of clusters, Indicates Cluster Centers and Cluster Centers The distance between.

[0086] This formula comprehensively considers the intra-class compactness (numerator) and inter-class separation (denominator). The numerator measures the average distance from the samples in each optimal cluster to the optimal cluster center, and the denominator measures the average distance between different optimal cluster centers. The smaller the value, the smaller the distance within the class, the larger the distance between classes, and the better the clustering effect. At the same time, in order to punish too large a number of clusters, A penalty term is introduced in .

[0087] S13214: Different Clustering quality evaluation index under value Compare and choose Smallest The value is taken as the optimal number of clusters K*.

[0088] Specifically, the distance metric in the above steps is You can choose common distance metrics such as Euclidean distance and Manhattan distance. Cluster center The mean vector of samples in each cluster can be used, the total number of samples M is the number of rapeseed varieties, K min and K max The setting of needs to take into account both the precision of clustering and the computational efficiency. Generally, the appropriate value range can be selected according to the number of rapeseed varieties and the level of genetic diversity. The setting of the number of repeated clustering N* is mainly to reduce the random influence of the clustering algorithm and improve the stability of the clustering results. It can usually be set between 10 and 50.

[0089] Introducing clustering quality evaluation index Q K The main purpose of Q is to quantitatively evaluate the clustering effect under different clustering numbers and provide an objective basis for the selection of the optimal clustering number K*. K The intra-class distance in the index reflects the compactness of the cluster, that is, the samples in the same cluster should be similar enough; the inter-class distance reflects the separation of the cluster, that is, there should be a clear distinction between different clusters. Using the ratio of the two as a measure of cluster quality can better balance the compactness and separation of clusters. K Taking the inverse of the denominator in , we can convert it into the average distance between clusters, which is more in order of magnitude with the average distance within the cluster in the numerator. Considering that the increase in the number of clusters K will increase the risk of overfitting, we K Introducing a penalty term , so that a larger penalty is given when the number of clusters is large. The introduction of is mainly to balance the impact of data volume on penalty intensity, and finally choose Q K The smallest K value is taken as the optimal clustering number K*, which can minimize the model complexity while ensuring the clustering effect.

[0090] For example, assuming that the number of rapeseed varieties , examine the number of clusters exist For each value in the range , use K-means algorithm to perform 10 clusterings, get 10 clustering results, select The smallest as current The optimal clustering result under the value Then based on Calculate Different Under value Value, for example, when hour, 0.15; when hour, ;when hour, ...; By comparison, we can find that hour, The minimum value was 0.12, so K*=4 was finally selected as the optimal cluster number to divide rapeseed varieties into four genetic diversity levels.

[0091] In summary, the above steps introduce the clustering quality evaluation index based on the optimal clustering result , which can more accurately evaluate the clustering effects at different clustering granularities and adaptively determine the optimal clustering number K*, thereby obtaining more reliable results for the classification of genetic diversity of rapeseed varieties; this method improves the objectivity and theoretical support of cluster number selection, and lays a good foundation for subsequent hybrid parent selection.

[0092] Step S1322, using the first eigenvector as input, using the K-means algorithm to divide the rapeseed varieties into K* genetic diversity levels as the K-means clustering result;

[0093] Step S1323, calculating the Euclidean distance between the rape varieties based on the first eigenvector as the genetic distance between the rape varieties, constructing a cluster tree, and dividing the rape varieties into different genetic diversity levels according to the structure of the cluster tree as a hierarchical clustering result;

[0094] Step S1324, comparing the K-means clustering result and the hierarchical clustering result, and selecting the best clustering result as the final rapeseed variety genetic diversity level;

[0095] Specifically, cluster analysis is the process of dividing varieties into different groups based on their genetic similarities. The first eigenvector comprehensively reflects the genetic composition of the variety and is an ideal input for cluster analysis. Based on the determined optimal cluster number K*, the K-means algorithm is used to cluster the first eigenvector, and the varieties are divided into K* genetic diversity levels by minimizing the sum of squared deviations within the family. In contrast, the idea of ​​hierarchical clustering can also be adopted. The Euclidean distance between varieties is calculated as its genetic distance based on the first eigenvector. A bottom-up method is used to merge the two nearest clusters until all varieties are in one cluster, a clustering tree is constructed, and the varieties are classified according to the hierarchical structure of the tree. For different clustering results, indicators such as silhouette coefficient and Davies-Bouldin index can be used to objectively evaluate the clustering quality, and the results obtained by the method with better clustering performance are selected. For example, rapeseed varieties are divided into low genetic diversity, medium genetic diversity and high genetic diversity. The combined use of K-means clustering and hierarchical clustering can reveal the genetic diversity structure of rapeseed germplasm resources from different angles and provide a basis for the reasonable division of genetic diversity levels.

[0096] Step S2000, constructing an initial parent combination based on the genetic diversity level, constructing and training an environmental adaptability prediction model; using the environmental adaptability prediction model, predicting the adaptability performance of the hybrid offspring of the initial parent combination in multiple environments, obtaining the environmental adaptability phenotypic values ​​of the hybrid offspring, and obtaining the adaptability scores of the hybrid offspring of each initial parent combination in different environments according to the environmental adaptability phenotypic values; optimizing the initial parent combination based on the adaptability score to obtain an optimized parent combination.

[0097] Furthermore, step S2000 includes:

[0098] Step S2100, constructing an initial parent combination based on the genetic diversity level of rapeseed varieties;

[0099] Furthermore, step S2100 includes:

[0100] Step S2110, setting genetic diversity thresholds for different genetic diversity levels;

[0101] Step S2120, setting a breeding target threshold according to the genetic diversity level and genetic diversity threshold of the breeding target;

[0102] Step S2130, traversing all parent combinations, and taking the parent combination in which the average genetic distance between all parents is greater than the breeding target threshold as the initial parent combination;

[0103] Specifically, genetic diversity is an important basis for hybrid breeding, and the genetic diversity level of the parental combination directly affects the genetic improvement potential of the hybrid offspring. Therefore, when constructing the initial parental combination, it is necessary to set a breeding target threshold as a basic constraint to ensure that the parental combination has sufficient genetic differences and lay the foundation for the expression of hybrid vigor. Parental combinations in which the average genetic distance between all parents in the parental combination is less than or equal to the breeding target threshold cannot be used as the initial parental combination, and the next parental combination should be judged whether it can be used as the initial parental combination.

[0104] First, genetic diversity thresholds of different genetic diversity levels are set by those skilled in the art based on experience; exemplary, genetic diversity thresholds of different genetic diversity levels are as follows:

[0105] Low genetic diversity: 0.1-0.2;

[0106] Medium genetic diversity: 0.2-0.4;

[0107] High genetic diversity: 0.4-0.5;

[0108] If the breeding goal is to keep genetic diversity within a medium range, a specific value, such as 0.3, can be selected as the breeding goal threshold. The parents in the parent combination are rapeseed varieties used for hybridization. The parent combination is a combination of two or more parents selected for hybridization. Rapeseed varieties can be used as parents to produce new varieties through hybrid breeding. The Euclidean distance between all parents in the parent combination is used as the genetic distance. Since the space of possible parent combinations is very large, exhaustive search is computationally infeasible. Therefore, heuristic search algorithms such as genetic algorithms and simulated annealing algorithms are needed to efficiently explore the optimal combination under the constraints of the breeding goal threshold. Through systematic construction and optimization, high-quality initial parent combinations that take into account both genetic diversity and breeding goals can be obtained, laying the foundation for subsequent analysis.

[0109] Step S2200, obtaining and combining the genotype data of each rapeseed variety in the historical parent combination to generate a historical genotype data group; obtaining the historical planting environment data and historical phenotypic data of the hybrid offspring of the historical parent combination; using the historical genotype data group, the historical planting environment data and historical phenotypic data of the hybrid offspring of the historical parent combination as sample data sets, and constructing and training an environmental adaptability prediction model;

[0110] Further, step S2200 includes:

[0111] Step S2210, using a neural network to construct an environmental adaptability prediction model, including an input layer, a plurality of hidden layers and an output layer, wherein the input layer receives a historical genotype data set and historical planting environment data, and the output layer outputs predicted environmental adaptability phenotype values ​​of hybrid offspring of historical initial parent combinations, with the historical phenotype data serving as true labels; the number of neurons in the output layer corresponds to the number of environmental adaptability phenotypes to be predicted;

[0112] Step S2220, dividing the sample data set into a training set, a validation set, and a test set according to a certain ratio, wherein the training set is used for model training, the validation set is used for model hyperparameter optimization, and the test set is used for evaluating model performance;

[0113] Step S2230, training the neural network model using the training set data, calculating the predicted output through forward propagation, updating the parameters through back propagation, and continuously iterating and optimizing until the model converges or reaches a preset number of iterations;

[0114] Step S2240, using the validation set data to optimize the hyperparameters of the trained model, and finding the optimal hyperparameter combination through grid search;

[0115] Step S2250, using the test set data to evaluate the performance of the optimized model, calculate the evaluation index, and comprehensively judge the prediction ability and generalization ability of the model;

[0116] Specifically, when constructing a neural network model, the number of neurons in the output layer corresponds to the number of environmental adaptability phenotypes to be predicted. The number of adaptive phenotypes refers to the number of phenotypic traits used to describe the environmental adaptability of the initial parent combination. In different studies, different numbers of adaptability-related phenotypes may be selected. The number of adaptive phenotypes determines the output dimension of the adaptive prediction model. The setting of the hidden layer needs to take into account the expressive power and computational efficiency of the model, and generally adopts a multi-layer fully connected structure. During the model training process, the input data is transformed and nonlinearly activated through each layer through forward propagation to obtain the predicted adaptive phenotypic value. The adaptive phenotypic value refers to the phenotypic data related to adaptability exhibited by the initial parent combination under specific environmental conditions, such as the measured values ​​of traits such as yield, resistance, and quality. It is a direct measure of the environmental adaptability of the initial parent combination. The error between the predicted output and the true value is used as the loss function, and the back propagation algorithm is used to adjust the parameters of each layer along the direction of gradient descent to achieve parameter learning and model optimization. Commonly used loss functions include mean square error, cross entropy, etc., and optimization algorithms include stochastic gradient descent, Adam, etc. By setting appropriate batch size, learning rate and other hyperparameters, the rate and stability of model learning can be controlled. In the hyperparameter optimization stage, the performance of the model under different hyperparameter combinations is systematically evaluated through cross-validation and other methods, and the hyperparameter combination with the strongest generalization ability is selected to improve the robustness and applicability of the model. The trained and optimized model is unbiasedly evaluated using the test set data to objectively measure its predictive ability in practical applications. The mean square error reflects the degree of deviation between the predicted value and the true value, and the determination coefficient reflects the degree to which the predicted value explains the variation of the true value. These two indicators jointly describe the goodness of fit and predictive ability of the model. Through the careful construction and optimization of the environmental adaptability prediction model, the complex nonlinear relationship between phenotype and environmental factors can be fully explored, the expression ability and generalization ability of the model can be improved, and a strong support can be provided for the accurate prediction of the environmental adaptability of the initial parent combination. The optimized prediction model can effectively guide the screening of parent combinations adapted to specific environments, accelerate the adaptive breeding process, improve the success rate of rapeseed hybrid combinations, and reduce breeding costs. At the same time, the constructed environmental adaptability prediction model can be embedded in the breeding decision-making system to assist breeders in the design and optimization of key links such as parent selection and new variety breeding, thereby improving breeding efficiency and accuracy. In addition, the model training and hyperparameter optimization methods used in the prediction model can also be extended to the research of other crops and traits, providing new ideas for accelerating the discovery of excellent crop genotypes and the cultivation of new varieties.

[0117] Step S2300, obtaining the genotype data of each rapeseed variety in the initial parent combination and combining them to generate an initial genotype data group; obtaining the initial planting environment data of the hybrid offspring of the initial parent combination, and using the environmental adaptability prediction model to predict the adaptability performance of the hybrid offspring of the initial parent combination in multiple environments based on the initial genotype data group and the initial planting environment data of the hybrid offspring of the initial parent combination, and obtaining the environmental adaptability phenotypic value of the hybrid offspring, and obtaining the adaptability score of the hybrid offspring of each initial parent combination in different environments based on the environmental adaptability phenotypic value;

[0118] Methods for obtaining the adaptability scores of hybrid offspring of each initial parent combination in different environments according to the environmental adaptability phenotypic values ​​include:

[0119]

[0120] in:

[0121] : represents the fitness score of the hybrid offspring of the i-th initial parent combination in the j-th environment, ranging from 0 to 100;

[0122] : represents the value of the kth adaptive phenotype of the hybrid offspring of the i-th initial parent combination in the j-th environment, which is obtained from the environmental adaptability phenotype value output by the environmental adaptability prediction model;

[0123] : The mean of the kth adaptive phenotype of the hybrid offspring of all initial parent combinations in the jth environment;

[0124] : The standard deviation of the kth adaptive phenotype of the hybrid offspring of all initial parent combinations in the jth environment;

[0125] : No. The hybrid offspring of the initial parent combination is Under the environment The rank of each adaptive phenotype is calculated from the environmental adaptability phenotype value in the order of the phenotype value from small to large.

[0126] : The hybrid offspring of all initial parent combinations The mean of the adaptive phenotype ranks;

[0127] : represents the weight of the kth adaptive phenotype, reflecting the relative importance of the phenotype to the adaptability. It is set by technicians in this field based on experience, and the sum of all weights is 1;

[0128] : No. The importance coefficient of an adaptive phenotype controls the influence of the adaptive phenotype value on the adaptive score, and the value is between 0 and 1, which is set by those skilled in the art based on experience;

[0129] : No. The importance coefficient of the adaptive phenotype rank controls the influence of the phenotype rank on the adaptive score. The value is between 0 and 1 and is set by technicians in this field based on experience. and The sum is 1;

[0130] : No. The sensitivity parameter of an adaptive phenotype controls the rate at which the adaptability of the phenotype decreases when it deviates from the mean. It is a positive number. The larger the value, the more sensitive it is to deviations and the faster the adaptability decreases. It is set by experts in the field based on experience.

[0131] : No. The sensitivity parameter of the adaptive phenotype rank controls the rate at which the adaptability decreases when the rank deviates from the mean. It is a positive number. The larger the value, the more sensitive it is to the deviation and the faster the adaptability decreases. It is set by technicians in this field based on experience.

[0132] n: number of adaptive phenotypes;

[0133] : No. The maximum value of a fitness phenotype among the hybrid offspring of all initial parent combinations;

[0134] : No. The maximum rank of a fitness phenotype among the hybrid offspring of all initial parent combinations;

[0135] when When, with The increase of will gradually increase, and the positive contribution of the adaptive phenotype value to the adaptive score will become greater and greater; When, with The increase of will gradually decrease, and the negative impact of the adaptive phenotype value on the adaptive score will become smaller and smaller; The increase of will gradually increase, and the rank of the fitness phenotype will have an increasingly greater impact on the fitness score.

[0136] The adaptive phenotype rank provides a robust measure of adaptability. By calculating the rank order of the hybrid offspring of each parental combination on a specific adaptive phenotype, the impact of extreme values ​​or outliers on the scoring results can be reduced. Even if some parental combinations show abnormally high or low predicted values ​​on individual phenotypes, their rank order can still objectively reflect their relative performance. The rank achieves comparability between different adaptive phenotypes. Due to the large differences in the dimensions and numerical ranges of different adaptive phenotypes, direct comparison of their predicted values ​​may be biased towards phenotypes with larger values. The rank order provides a unified dimensionless indicator, so that the contributions of different phenotypes to adaptability can be weighed on the same scale. Introducing rank-related terms in the formula can balance and correct the adaptive score calculated solely based on the phenotypic value. On the one hand, the greater the deviation of the phenotypic value from the mean (i.e., the more prominent the performance in the standardized phenotypic value item) The higher the rank order (i.e., the more prominent the performance in the rank-related items), the better the adaptability. The weighted average of the two items comprehensively considers the absolute performance and the relative performance, and can more comprehensively evaluate the adaptability of the parent combination. The introduction of rank enhances the interpretability of the adaptability score. Breeding experts can intuitively judge the relative advantages and disadvantages of the parent combination among the same type of materials based on the rank order, which is convenient for selection and decision-making. Compared with pure numerical scores, rank provides more intuitive reference information. By setting the phenotypic importance coefficient α and the rank importance coefficient β in the formula, the influence weight of the two types of information on the final score can be flexibly controlled, which provides adjustment space for adapting to different breeding goals and preferences. Breeding experts can adjust the values ​​of α and β based on experience, highlight the role of phenotypic values ​​or rank order, and obtain evaluation results that are more in line with breeding needs.

[0137] The formula comprehensively considers the information of the two dimensions of adaptive phenotype value and rank, balances their effects on the environmental adaptability of hybrid offspring, more comprehensively evaluates the environmental adaptability of the hybrid offspring of parent combinations, more accurately evaluates the advantages and disadvantages of parent combinations, and provides a reference for screening out excellent hybrid combinations; the introduction of weight parameters can be used to perform differentiated weighting according to the importance of different adaptive phenotypes to improve the rationality of the score; the normalization method is used to convert adaptive phenotypes of different dimensions into dimensionless indicators for easy comparison and synthesis; the setting of sensitivity parameters can adjust the sensitivity of the score to the deviation of the adaptive phenotype from the optimal value, thereby improving the discrimination of the score; the formula is flexible in form, and parameters can be adjusted and modified according to actual needs. It has strong scalability and can be appropriately adjusted according to different environments and breeding goals. The introduction of ranks improves the robustness, comparability, comprehensiveness and interpretability of fitness scores, enabling them to more effectively guide the optimization of parental combinations. At the same time, the weighted average of rank-related items and phenotypic value items not only improves the accuracy of scores, but also provides space for flexible response to different breeding needs. These advantages help to improve the efficiency of parent selection and accelerate the process of hybrid combination selection.

[0138] The said The hybrid offspring of the initial parent combination is Under the target planting environment The rank of the adaptive phenotype The calculation methods include:

[0139] (1) Collect phenotypic values: Obtain the hybrid offspring of all initial parent combinations in the Under the target planting environment The value of the adaptive phenotype ;

[0140] (2) Sorting: sort all Arrange in order from smallest to largest;

[0141] (3) Assigning rank values: Assign rank values ​​in the order of sorting, with the smallest Corresponding to rank 1, the second smallest The corresponding rank value is 2, and so on. The specific steps are as follows:

[0142] Find the hybrid offspring of all initial parent combinations in Under the target planting environment The set of fitness phenotype values ;

[0143] Sort the values ​​in the set from small to large, denoted as ;

[0144] Each The position of , for example, the smallest Corresponding 1, the second smallest Corresponding is 2, and so on.

[0145] For example, suppose there are four initial parent combinations In a target planting environment The fitness phenotype values ​​are as follows:

[0146]

[0147] Arranged in order from smallest to largest:

[0148]

[0149] Therefore, the rank value for:

[0150] ;

[0151] ;

[0152] ;

[0153] ;

[0154] This method ensures that the rank values ​​reflect the order of the adaptive phenotypes from small to large, making it easier to comprehensively consider the impact of the adaptive phenotypes on the score in the formula.

[0155] Step S2400, optimizing the initial parent combination based on the fitness scores of the hybrid offspring of the initial parent combination to obtain an optimized parent combination;

[0156] Furthermore, if Figure 3 As shown, step S2400 includes:

[0157] Step S2410, obtaining the adaptability scores of the hybrid offspring of the initial parent combinations under different target planting environments, and forming an adaptability score matrix with the initial parent combinations as rows and the target planting environments as columns;

[0158] Step S2420, based on the adaptability score matrix, calculate the mean and variance of the adaptability score of each initial parent combination hybrid offspring under all target planting environments, and select the adaptability score means under different planting environments greater than the adaptability mean threshold. And the variance is less than the adaptive variance threshold The parent combination is used as a cross-domain optimal parent combination for multi-target planting environments;

[0159] Step S2430: for the target planting environment j, select the target planting environment j whose adaptability score is greater than the adaptability threshold. parent combinations to form a directional preferred parent combination for the target planting environment j;

[0160] Adaptive Threshold The calculation methods include:

[0161] ;

[0162] in:

[0163] : represents the adaptability score of the hybrid offspring of the first initial parent combination in the jth target planting environment;

[0164] : represents the adaptability score of the hybrid offspring of the second initial parent combination in the jth target planting environment;

[0165] : represents the adaptability score of the hybrid offspring of the mth initial parent combination in the jth target planting environment;

[0166] : median function;

[0167] : Absolute median deviation function, used to calculate the median of the absolute deviations between each value in the data set and the median;

[0168] : regulatory factor;

[0169] : The adaptability threshold under the jth target planting environment;

[0170] m: total number of initial parent combinations;

[0171] Modulating Factor Used to adjust the effect of the MAD function on the adaptive threshold. A larger Z value will increase the width of the threshold, so that more parent combinations meet the adaptive threshold. In this case, the screening criteria are looser and more combinations can be included. A smaller Z value will reduce the width of the threshold, so that fewer parent combinations meet the adaptive threshold. In this case, the screening criteria are stricter and only the best combinations are included. Choose an appropriate Z value based on experience and historical data. For example, a common Z value selection range is between 1 and 3. Through dynamic thresholding The settings can flexibly adjust the screening criteria according to actual conditions to improve the accuracy and adaptability of screening.

[0172] Specifically, the adaptability score is evaluated based on the performance of the hybrid offspring of the initial parent combination in each target planting environment. The target planting environment refers to the expected planting conditions, such as different climate zones, soil types, management measures, etc. By constructing an environmental adaptability prediction model, the performance of the hybrid offspring of each initial parent combination in different target planting environments can be predicted and given an adaptability score. These scores can be comprehensively evaluated through some indicators such as yield, disease resistance, quality, etc. The final adaptability score matrix, the rows represent the initial parent combinations, the columns represent different target planting environments, and each element in the matrix represents the adaptability score of the hybrid offspring of the corresponding parent combination in a specific environment. The mean of the adaptability score represents the overall performance level of the hybrid offspring of the parent combination in all target planting environments, while the variance reflects the stability of its performance. Set an adaptability mean threshold and an adaptability variance threshold, and screen out the hybrid offspring of parent combinations whose mean adaptability score in all target planting environments exceeds the mean threshold and whose variance is lower than the variance threshold. This ensures that the hybrid offspring of the selected parent combination not only performs well in most environments, but also performs stably and is not easily affected by environmental changes. Step S2430 sets an adaptability threshold and selects parent combination hybrid offspring with adaptability scores higher than the threshold, so as to ensure that the selected parent combination hybrid offspring has the best performance in a specific planting environment and is suitable for the planting requirements in the specific environment.

[0173] Example 2

[0174] This embodiment provides a system for selecting parents of a rapeseed hybrid combination based on the embodiment 1. Figure 4 As shown, including:

[0175] Data collection module: used to collect genotype data that characterizes the genetic differences between rapeseed varieties; used to obtain historical planting environment data and historical phenotypic data of the hybrid offspring of historical parent combinations; collect the genotype data of each rapeseed variety in the initial parent combination and combine them to generate an initial genotype data set; collect the initial planting environment data of the hybrid offspring of the initial parent combination;

[0176] Data preprocessing module: used to eliminate false positive variants introduced by sequencing errors in genotype data, screen markers with significant polymorphism, and construct a molecular marker matrix;

[0177] Feature extraction module: used to extract features from the molecular marker matrix and obtain the first feature vector;

[0178] Cluster analysis module: used to classify rapeseed varieties based on the first eigenvector using a clustering algorithm to obtain the genetic diversity level of the rapeseed varieties;

[0179] Parent combination construction module: used to construct initial parent combinations based on the genetic diversity levels of rapeseed varieties;

[0180] Phenotypic value prediction module: used to construct and train an environmental adaptability prediction model, and obtain the environmental adaptability phenotypic value of the hybrid offspring of the initial parent combination according to the initial genotype data group of the initial parent combination, the initial planting environment data and the environmental adaptability prediction model; and obtain the adaptability score of the hybrid offspring of each initial parent combination in different environments according to the environmental adaptability phenotypic value;

[0181] Parent combination optimization module: optimize the initial parent combination based on the fitness score of the hybrid offspring of the initial parent combination to obtain the optimized parent combination.

[0182] Example 3

[0183] This embodiment discloses an electronic device, including a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit. When the central processing unit executes the computer program, the method for selecting parents of a rapeseed hybrid combination provided above is implemented.

[0184] Since the electronic device introduced in this embodiment is an electronic device used to implement a method for selecting parents of a rape hybrid combination in the embodiment of this application, based on the method for selecting parents of a rape hybrid combination introduced in the embodiment of this application, a person skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application will not be described in detail here. As long as a person skilled in the art implements the electronic device used in the method for selecting parents of a rape hybrid combination in the embodiment of this application, it belongs to the scope of protection of this application.

[0185] Example 4

[0186] This embodiment discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the above-mentioned method for selecting parents of a rapeseed hybrid combination is implemented.

[0187] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters, weights and thresholds in the formula are set by technicians in this field according to actual conditions.

[0188] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0189] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0190] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0191] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0192] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0193] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0194] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

[0195] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for selecting parents of a rapeseed hybrid combination, characterized in that: include: Collect genotype data to characterize genetic differences among rapeseed varieties; Eliminate false positive variants introduced by sequencing errors in genotype data and construct a molecular marker matrix; Extracting features from the molecular marker matrix to obtain a first eigenvector; classifying rapeseed varieties based on the first eigenvector to obtain genetic diversity levels of the rapeseed varieties; Initial parent combinations were constructed based on the genetic diversity levels of rapeseed varieties, and environmental adaptability prediction models were constructed and trained; The genotype data of each rapeseed variety in the initial parent combination is obtained and combined to generate an initial genotype data group; the initial planting environment data of the hybrid offspring of the initial parent combination is obtained, and the adaptability performance of the hybrid offspring of the initial parent combination in multiple environments is predicted using an environmental adaptability prediction model based on the initial genotype data group and the initial planting environment data of the hybrid offspring of the initial parent combination, and the environmental adaptability phenotypic value of the hybrid offspring is obtained, and the adaptability score of the hybrid offspring of each initial parent combination in different environments is obtained according to the environmental adaptability phenotypic value; the initial parent combination is optimized based on the adaptability score to obtain an optimized parent combination; The method of extracting features from the molecular marker matrix to obtain a first feature vector comprises: The allele frequencies in the molecular marker matrix are coded with 0 and 1 to form a two-dimensional matrix with varieties as rows and marker sites as columns, which is used as the original molecular marker matrix; Preprocess the original molecular marker matrix to eliminate dimension differences, generate a standardized molecular marker matrix, and extract the feature that explains the largest proportion of genetic variation from the standardized molecular marker matrix as the first eigenvector; The method for classifying rapeseed varieties based on the first characteristic vector to obtain the genetic diversity level of the rapeseed varieties includes: Evaluate the clustering quality under different cluster numbers and select the optimal cluster number K*; Taking the first eigenvector as input, the K-means algorithm was used to divide the rapeseed varieties into K* genetic diversity levels as the K-means clustering result; The Euclidean distance between rapeseed varieties is calculated based on the first eigenvector as the genetic distance between rapeseed varieties, a cluster tree is constructed, and rapeseed varieties are divided into different genetic diversity levels according to the structure of the cluster tree as the hierarchical clustering result; Compare the K-means clustering results and the hierarchical clustering results, and select the best clustering result as the final genetic diversity level of rapeseed varieties; The method for constructing an initial parent combination based on the genetic diversity level of rapeseed varieties comprises: Setting genetic diversity thresholds for different levels of genetic diversity; Set breeding target thresholds based on the genetic diversity level and genetic diversity threshold of the breeding target; Traverse all parent combinations, and the parent combination whose average genetic distance between all parents is greater than the breeding target threshold is used as the initial parent combination; The method for constructing and training an environmental adaptability prediction model comprises: The genotype data of each rapeseed variety in the historical parent combination is obtained and combined to generate a historical genotype data group; the historical planting environment data and historical phenotypic data of the hybrid offspring of the historical parent combination are obtained; A neural network is used to construct an environmental adaptability prediction model, which includes an input layer, N hidden layers and an output layer, where N is an integer greater than 0. The input layer receives a historical genotype data set and historical planting environment data, and the output layer outputs the predicted environmental adaptability phenotypic values ​​of the hybrid offspring of the historical initial parent combination, with the historical phenotypic data serving as the true label. The number of neurons in the output layer corresponds to the number of environmental adaptability phenotypes to be predicted.

2. The method for selecting parents of a rapeseed hybrid combination according to claim 1, characterized in that: The method for obtaining the adaptability scores of the hybrid offspring of each initial parent combination under different environments according to the environmental adaptability phenotype value comprises: ； in, represents the fitness score of the hybrid offspring of the i-th initial parent combination in the j-th environment, represents the kth adaptive phenotype value of the hybrid offspring of the i-th initial parent combination in the j-th environment, represents the mean of the kth adaptive phenotype value of the hybrid offspring of all initial parent combinations in the jth environment, represents the standard deviation of the kth adaptive phenotype value of the hybrid offspring of all initial parent combinations in the jth environment, Indicates The hybrid offspring of the initial parent combination is Under the environment The rank of the fitness phenotype values, represents the hybrid offspring of all initial parent combinations The mean of the ranks of the fitness phenotype values, represents the weight of the kth adaptive phenotype, Indicates The importance coefficient of each adaptive phenotype, Indicates The importance coefficient of the rank of the fitness phenotype values, Indicates The sensitivity parameter of the adaptive phenotype, Indicates The sensitivity parameter of the rank of the adaptive phenotype values, n represents the number of adaptive phenotypes, Indicates The maximum adaptive phenotypic value among all the hybrid offspring of the initial parent combinations. Indicates The maximum rank of the fitness phenotype value among the hybrid offspring of all initial parent combinations.

3. The method for selecting parents of a rapeseed hybrid combination according to claim 1, characterized in that: The method of optimizing the initial parent combination based on the adaptability score to obtain the optimized parent combination includes: Obtain the adaptability scores of the hybrid offspring of the initial parent combinations under different target planting environments, and form an adaptability score matrix with the initial parent combinations as rows and the target planting environments as columns; Based on the adaptability score matrix, the mean and variance of the adaptability scores of the hybrid offspring of each initial parent combination under all target planting environments are calculated, and the mean of the adaptability score under different planting environments is selected to be greater than the mean adaptability threshold. And the variance is less than the adaptive variance threshold The parent combination is used as a cross-domain optimal parent combination for multi-target planting environments; For the target planting environment j, select the target planting environment j whose adaptability score is greater than the adaptability threshold. parent combination to form a directional preferred parent combination for the target planting environment j.

4. The method for selecting parents of a rapeseed hybrid combination according to claim 3, characterized in that: The adaptive threshold The calculation methods include: ； in, represents the adaptability score of the hybrid offspring of the first initial parent combination in the jth target planting environment, represents the adaptability score of the hybrid offspring of the second initial parent combination in the jth target planting environment, represents the adaptability score of the hybrid offspring of the mth initial parent combination in the jth target planting environment, is the median function, is the absolute median deviation function, is the regulating factor, is the adaptability threshold of the jth target planting environment.

5. A system for selecting parents of a rapeseed hybrid combination, which is used to implement a method for selecting parents of a rapeseed hybrid combination according to any one of claims 1 to 4, characterized in that: The system comprises: Data collection module: used to collect genotype data that characterizes the genetic differences between rapeseed varieties; used to obtain historical planting environment data and historical phenotypic data of the hybrid offspring of historical parent combinations; collect the genotype data of each rapeseed variety in the initial parent combination and combine them to generate an initial genotype data set; collect the initial planting environment data of the hybrid offspring of the initial parent combination; Data preprocessing module: used to eliminate false positive variants introduced by sequencing errors in genotype data, screen markers with significant polymorphism, and construct a molecular marker matrix; Feature extraction module: used to extract features from the molecular marker matrix and obtain the first feature vector; Cluster analysis module: used to classify rapeseed varieties based on the first eigenvector using a clustering algorithm to obtain the genetic diversity level of the rapeseed varieties; Parent combination construction module: used to construct initial parent combinations based on the genetic diversity levels of rapeseed varieties; Phenotypic value prediction module: used to construct and train an environmental adaptability prediction model, and obtain the environmental adaptability phenotypic value of the hybrid offspring of the initial parent combination according to the initial genotype data group of the initial parent combination, the initial planting environment data and the environmental adaptability prediction model; and obtain the adaptability score of the hybrid offspring of each initial parent combination in different environments according to the environmental adaptability phenotypic value; Parent combination optimization module: optimize the initial parent combination based on the fitness score of the hybrid offspring of the initial parent combination to obtain the optimized parent combination.

6. An electronic device comprising a memory, a central processing unit, and a computer program stored in the memory and executable on the central processing unit, characterized in that: When the central processing unit executes the computer program, the method for selecting parents of a rapeseed hybrid combination according to any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method for selecting parents of a rapeseed hybrid combination according to any one of claims 1 to 4 is implemented.

Images