A method for predicting peanut calcium environment-specific whole genome selection

Abstract

The application discloses a kind of peanut calcium environment specificity whole genome selection prediction method, it is related to plant genetic breeding and bioinformatics technical field, the method includes the following steps: S1, obtaining whole genome SNP site;S2, obtaining associated marker;S3, obtaining three types of functional target marker subset;S4, obtaining multi-scenario prediction accuracy;S5, output prediction and sorting result;The application can significantly improve the prediction accuracy of low-calcium tolerance trait under low-calcium stress environment, the application proposes "based on whole genome association analysis+multi-scenario interactive cross-environment prediction" strategy, by dividing associated marker into constitutive, low-calcium specific and calcium-specific three types, the environmental specificity layered utilization of marker information is realized, and experiment shows that, in low-calcium stress environment, the average accuracy rate reaches 0.79-0.93 using Indu site for prediction, significantly better than the method of traditional whole genome marker direct modeling.

Description

Technical Field

[0001] This invention relates to the fields of plant genetics and breeding and bioinformatics, specifically to a method for predicting peanut calcium environment-specific whole-genome selection. Background Technology

[0002] Peanut yield, quality, and stress tolerance are important traits, mostly typical quantitative traits, regulated by complex genetic networks. Due to the low genomic variation rate of peanuts, traditional molecular marker-assisted selection has limited availability of tightly linked markers, making it difficult to meet the selection accuracy requirements for complex traits in practical breeding. Genomic selection (GS), by integrating genome-wide marker information and estimating the genomic estimated breeding value (GEBV) of an individual, can achieve early prediction and selection of complex traits and has been widely used in major crops such as maize and wheat. However, the research and application of whole-genome selection technology in peanuts is still in its early stages. Existing patents related to whole-genome selection in peanuts, such as "A method for selecting hybrid combinations of high-yield, large-fruited peanuts based on whole-genome selection" (CN115443907B), mainly focus on screening hybrid combinations and parental selection for yield traits, without addressing the construction of specific prediction models under environmental stress conditions. For the specific production scenario of acidified and calcium-deficient ecological zones in the south, there is still a lack of dedicated prediction models and marker screening strategies for peanut stress tolerance under low calcium stress. Traditional whole-genome selection methods do not consider the differential contribution of environmentally specific associated loci to the prediction accuracy under different environments, and directly incorporate all whole-genome markers into the model for prediction, resulting in low accuracy of selection for complex traits under low calcium stress, which is difficult to meet the actual breeding needs. Therefore, there is an urgent need for a peanut genome-wide selection prediction breeding method that can fully utilize environmental-specific associated marker information in low-calcium stress environments, so as to improve the early prediction accuracy of low-calcium tolerance traits and shorten the variety breeding cycle. Summary of the Invention

[0003] This invention provides a method for predicting peanut calcium environment-specific whole-genome selection; To achieve the above objectives, the present invention provides the following technical solution: A method for predicting peanut calcium environment-specific genome-wide selection includes the following steps: S1. Obtaining whole-genome SNP loci: Using peanut germplasm resources as training populations, field phenotypic identification and whole-genome resequencing were carried out at multiple locations under different calcium environments over many years. Whole-genome SNP loci were obtained through quality control. S2. Obtain association markers: Based on the whole genome SNP sites, whole genome association analysis was performed under calcium-treated and non-calcium-treated environments to extract the association marker information of the top N P-values ​​of each trait, and then the association markers were obtained. S3. Obtain three types of functional target marker subsets: Based on the intersection and difference algorithm, perform data layering on the associated markers to obtain three types of functional target marker subsets; S4. Obtain multi-scenario prediction accuracy: Using the three types of functional target marker subsets as prediction variables, construct prediction models respectively, and carry out multi-scenario interactive cross-environment prediction in both calcium-applied and non-calcium-applied environments to obtain multi-scenario prediction accuracy. S5. Output prediction and ranking results: Evaluate the accuracy of the multi-scenario prediction through cross-validation, obtain the optimal prediction model for different traits, predict and rank the genomic estimated breeding values ​​of the test breeding materials based on the optimal prediction model, and output the prediction and ranking results.

[0004] As a preferred embodiment of the present invention, in step S1, the average sequencing depth of the whole genome resequencing is 20×, and the whole genome SNP sites are quality controlled by the GATK best practice process and population-level filtering to retain sites with a minimum allele frequency of less than 5% and a deletion rate of less than 20%, thereby obtaining 1,150,004 whole genome SNP sites.

[0005] In a preferred embodiment of the present invention, in step S2, the value of N is 1000. Based on the significance of the genome-wide association analysis results, the P values ​​are sorted from smallest to largest, and then the association marker information of the top 1000 is extracted to finally obtain the association marker.

[0006] In a preferred embodiment of the present invention, in step S3, the three types of functional targeting marker subsets include a purely constitutive marker subset common to both environments, a specific marker subset for the low-calcium environment, and a specific marker subset for the calcium-treated environment. Let A be the set of associated markers in the calcium-treated environment and B be the set of associated markers in the non-calcium-treated environment. Then: The subset of purely constitutive markers shared by the two environments is: Cons = A ∩ B; The specific marker subset for the low-calcium environment is: Indu=B\A; The specific marker subset of the calcium application environment is: Spec=A\B.

[0007] As a preferred embodiment of the present invention, in step S4, the multi-scenario interaction cross-environment prediction includes at least the following scenarios: First scenario: Based on a subset of specific markers in the low-calcium environment, predict the phenotype in the low-calcium environment; Second scenario: Based on the subset of purely constitutive markers shared by the two environments, predict the phenotype under low calcium environment; Third scenario: Based on the specific marker subset of the low-calcium environment, predict the phenotype under calcium application environment; Fourth scenario: Based on a subset of specific markers from the calcium application environment, predict the phenotype under the calcium application environment; Fifth scenario: Based on the subset of purely constitutive markers shared by the two environments, predict the phenotype under calcium application conditions; Scenario 6: Based on a subset of specific markers from the calcium application environment, predict the phenotype under low calcium conditions.

[0008] As a preferred embodiment of the present invention, in step S4, the prediction model includes at least one of the following: Bayesian A prediction model, Bayesian B prediction model, Bayesian CPi prediction model, genome optimal linear unbiased prediction model, ridge regression optimal linear unbiased prediction model, random forest prediction model, and support vector machine prediction model.

[0009] As a preferred embodiment of the present invention, in step S4, the traits predicted by the multi-scenario interactive cross-environment include at least one of peanut-related traits such as pod fullness, pod plumpness, kernel yield, weight of 100 pods, pod thickness, and pod width.

[0010] In a preferred embodiment of the present invention, in step S5, the cross-validation is 5-fold cross-validation, and its specific steps are as follows: A. Randomly divide the training group into 5 non-overlapping parts, take the first part as the "validation set", and merge the remaining 4 parts as the "training set" to construct the prediction validation model. B. Based on the constructed prediction and verification model, predict the phenotype of the first part, record the correlation coefficient between the predicted value and the actual value, and then obtain the prediction and verification accuracy. C. Take the second set as the "validation set", and merge the remaining 4 sets as the "training set" to construct the prediction validation model. Repeat steps A and B until all 5 sets of data have been used as the "validation set" once. D. Take the average of the prediction verification accuracies obtained in the above 5 rounds to obtain the average prediction verification accuracy.

[0011] As a preferred embodiment of the present invention, in step S5: For environment-specific major-effect associated sites with predominantly additive effects, Bayesian A prediction model, Bayesian B prediction model, Bayesian CPi prediction model, genome-optimal linear unbiased prediction model, or ridge regression-optimal linear unbiased prediction model are used for prediction. For environment-specific major-effect associated sites where non-additive effects are predominant, random forest or support vector machine prediction models are used for prediction.

[0012] This invention provides a method for predicting peanut calcium environment-specific genome-wide selection. It has the following beneficial effects: 1. This invention can significantly improve the prediction accuracy of low-calcium tolerance traits under low-calcium stress. This invention proposes a strategy of "genome-wide association analysis + multi-scenario interactive cross-environment prediction". By classifying association markers into three categories, namely constitutive, low-calcium-specific, and calcium-application-specific, it realizes the environment-specific hierarchical utilization of marker information. Experiments show that the average accuracy of prediction using Indu sites under low-calcium stress reaches 0.79-0.93, which is significantly better than the traditional method of direct modeling using genome-wide markers. This provides an effective technical solution to the problem of insufficient accuracy in selecting low-calcium tolerance traits in peanuts in acidified and calcium-deficient ecological zones in the south. 2. This invention significantly shortens the breeding cycle and reduces breeding costs. Using the method of this invention, the environmental specific breeding value of an individual can be predicted with high accuracy at the seedling stage, without waiting for the plant to mature and conduct field phenotypic identification in multiple environments over many years. Compared with traditional phenotypic selection, it can significantly shorten the breeding cycle and reduce the redundant costs of field screening in multiple environments. 3. This invention provides a precise technical path for peanut breeding. This invention initially establishes a breeding decision-making process of "whole genome selection prediction + molecular marker-assisted verification" for acidified and calcium-deficient ecological zones in the south. It can directly identify the key genetic regions that regulate peanut response to soil calcium deficiency stress, and provide a solid data foundation and candidate site support for the discovery of low calcium tolerance molecular markers and the creation of superior parents. 4. This invention has good prospects for promotion and application. It can be applied to variety improvement work in the entire southern flower production area, and provides important theoretical guidance and technical support for promoting the precision breeding of crops in complex ecological zones in my country. Attached Figure Description

[0013] Figure 1 This is a flowchart of a peanut calcium environment-specific whole-genome selection prediction method according to the present invention; Figure 2 This is a schematic diagram comparing the prediction accuracy of peanut pod fullness percentage under different calcium environments according to the present invention. Figure 3 This is a schematic diagram comparing the prediction accuracy of peanut pod fullness under different calcium environments according to the present invention. Figure 4 This is a schematic diagram comparing the prediction accuracy of peanut kernel yield under different calcium environments according to the present invention. Figure 5 This is a schematic diagram comparing the prediction accuracy of peanut 100-shell weight under different calcium environments according to the present invention. Figure 6 This is a schematic diagram comparing the prediction accuracy of peanut pod thickness under different calcium environments according to the present invention. Figure 7 This is a schematic diagram comparing the prediction accuracy of peanut pod width under different calcium environments according to the present invention. Figure 8 This is a schematic diagram comparing the prediction accuracy of peanut whole genome based on different association marker subsets under different calcium treatments according to the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0015] Example 1: Please see Figures 1-7 This invention provides a method for predicting peanut calcium environment-specific whole-genome selection, comprising the following steps: S1. Obtaining whole-genome SNP loci: Using peanut germplasm resources as training populations, field phenotypic identification and whole-genome resequencing were conducted at multiple locations under different calcium environments over several years. Whole-genome SNP loci were obtained through quality control. Specifically, 213 core peanut germplasm resources were selected as training populations. Whole-genome resequencing was performed on these training populations with an average sequencing depth of 20×. Transcriptome sequencing was also performed simultaneously. The raw sequencing data was quality filtered to remove low-quality reads and adapter contamination sequences. High-quality reads were aligned to the peanut reference genome for SNP detection and genotype determination. Quality control was achieved using the GATK best practice workflow and population-level filtering. Loci with a minimum allele frequency of less than 5% and a deletion rate of less than 20% were retained. A total of 1,150,004 high-quality SNP loci were obtained for subsequent genome-wide association analysis and genome-wide selection modeling. S2. Obtaining Association Markers: Based on whole-genome SNP loci, genome-wide association analysis was performed under calcium-treated and non-calcium-treated conditions to extract the association marker information of the top N P-values ​​for each trait, and then the association markers were obtained. Specifically, 213 training population materials were planted under two treatment conditions, calcium-treated (Ca) and non-calcium-treated (noCa), and field phenotypic identification was carried out. The investigated traits included 6 core quantitative traits related to peanut yield: pod fullness weight percentage (FFWR), pod plumpness (PDP), kernel yield (RK), 100-pod weight (HPW), pod thickness (PT), and pod width (PW). Genome-wide association analysis (GWAS) was performed on more than 1,150,004 high-quality whole-genome SNP loci under two different treatments. Based on the significance of the GWAS results, the P-values ​​were sorted from smallest to largest. For each trait, the top 1,000 association markers with the highest P-values ​​were extracted as the set of association markers for that trait in that environment. S3. Obtaining three types of functional target marker subsets: Based on the intersection and difference algorithm, the associated markers are layered to obtain three types of functional target marker subsets. Specifically, using the intersection and difference algorithm, the extracted marker information is layered. The associated marker set under the calcium-enhanced environment is A, and the associated marker set under the non-calcified environment is B. The data layers are as follows: (1) The subset of purely constitutive markers shared by the two environments is: Cons=A∩B. These markers are associated with the target trait under different calcium nutrition environments, representing the basic genetic regulatory sites of the trait; (2) The subset of specific markers for low calcium environment is: Indu=B\A. These markers are only activated and expressed under low calcium stress conditions and are key genetic loci in response to calcium deficiency stress. (3) The specific marker subset of calcium application environment is: Spec=A\B. These markers play a role under normal calcium supply conditions, but may be inhibited or lose their effect under calcium deficiency conditions. Using the above hierarchical strategy, genome-wide markers are divided into three subsets of functional target markers with different environmental response characteristics; S4. Obtaining Multi-Scenario Prediction Accuracy: Using the three functional target marker subsets as predictor variables, prediction models are constructed for each subset. Multi-scenario interactive cross-environment predictions are conducted under both calcium-treated and non-calcium-treated environments to obtain multi-scenario prediction accuracy. Specifically, the three marker subsets (Cons, Indu, Spec) are used as predictor variables, and prediction models are constructed for each subset. The constructed prediction models include seven classic models: Bayesian A prediction model, Bayesian B prediction model, Bayesian CPi prediction model, genomic optimal linear unbiased prediction model, ridge regression optimal linear unbiased prediction model, random forest prediction model, and support vector machine prediction model. It should be noted that all seven classic models must be constructed, and multi-scenario interactive cross-environment predictions are conducted under each model. The multi-scenario interactive cross-environment predictions must include at least the following scenarios: Scenario 1: Predicting phenotypes under low-calcium environments based on a subset of specific markers. Second scenario: Based on a subset of purely constitutive markers shared by both environments, predict phenotypes under low-calcium conditions; Scenario 3: Predicting phenotypes under calcium application conditions based on a subset of specific markers in low-calcium environments; Fourth scenario: Predicting phenotypes under calcium application environment based on a subset of specific markers from the calcium application environment; Fifth scenario: Predict phenotypes under calcium application conditions based on a subset of purely constitutive markers shared by both environments; Scenario 6: Predicting phenotypes under low-calcium conditions based on a subset of specific markers from calcium application environments; S5. Output Prediction and Ranking Results: The prediction accuracy across multiple scenarios is evaluated through cross-validation to obtain the optimal prediction model for different traits. Based on the optimal prediction model, the estimated genomic breeding values ​​of the test breeding materials are predicted and ranked, and the prediction and ranking results are output. Specifically, 5-fold cross-validation is used to evaluate the prediction accuracy: the training population is randomly divided into 5 equal parts. Four parts are used as the training set to train the model, and the remaining part is used as the validation set to evaluate the prediction accuracy. This process is repeated 5 times, and the average value is taken. The steps include: A. Randomly divide the training group into 5 non-overlapping parts, take the first part as the "validation set", and merge the remaining 4 parts as the "training set" to build a prediction validation model. B. Based on the constructed prediction and verification model, predict the phenotype of the first part, record the correlation coefficient between the predicted value and the actual value, and then obtain the prediction and verification accuracy. C. Take the second set as the "validation set", and merge the remaining 4 sets as the "training set" to build a prediction validation model. Repeat steps A and B until all 5 sets of data have been used as the "validation set" once. D. Take the average value of the prediction verification accuracy obtained in the above 5 rounds to obtain the average prediction verification accuracy. For environment-specific major-effect associated sites with predominantly additive effects, Bayesian A prediction model, Bayesian B prediction model, Bayesian CPi prediction model, genome-optimal linear unbiased prediction model, or ridge regression-optimal linear unbiased prediction model are used for prediction. For environment-specific major-effect associated sites with predominantly non-additive effects, random forest or support vector machine prediction models are used for prediction. The evaluation results show that: (1) Under low calcium stress, the average accuracy of predicting phenotypes using specific loci (Indu_noCa) was the best among all traits, with prediction values ​​ranging from 0.79 to 0.93. Under the same conditions, its prediction accuracy was significantly higher than that of predicting using common constitutive loci (Cons_noCa). (2) Under normal calcium application conditions, the predictive performance of the Spec_Ca combination is generally better than that of the constitutive marker (Cons_Ca). (3) For environmentally specific major-effect associated sites with additive effects, the prediction performance of Bayesian A prediction model, Bayesian B prediction model, Bayesian CPi prediction model, genome best linear unbiased prediction model or ridge regression best linear unbiased prediction model is generally robust; while for traits with a large proportion of non-additive effects, machine learning models (random forest prediction model or support vector machine prediction model) show certain advantages in some scenarios. Based on the above evaluation results, the optimal combination of marker subset and prediction model is determined for different traits and different environmental conditions. The established optimal prediction model was applied to the early screening of breeding materials in the low-calcium ecological environment of southern China. DNA was extracted from the materials during the seedling stage for genotyping. The genotyping data was input into the optimal prediction model to calculate the estimated genomic breeding value (GEBV) for each individual. The breeding materials were comprehensively ranked according to GEBV, and materials with higher GEBV rankings were selected for subsequent breeding procedures (such as hybridization, backcrossing, variety comparison trials, etc.). Using the above methods, the environmental-specific breeding value of individuals can be predicted with high precision at the seedling stage, enabling early screening of breeding materials and targeted selection of superior traits, thereby significantly shortening the breeding cycle and reducing breeding costs.

[0016] Example 2: The difference between this embodiment and Embodiment 1 is as follows: the size of the training population can be adjusted according to actual breeding needs, and is not limited to 213 copies; the depth of whole-genome resequencing can be selected between 15× and 30× according to actual conditions; the number of associated markers extracted in the GWAS analysis is not limited to the top 1000 P-values, and can be adjusted between 500 and 5000 according to the genetic architecture and marker density of the trait; the choice of prediction model is not limited to the 7 types listed in Embodiment 1, and other suitable models (such as deep learning models, ensemble learning models, etc.) can be introduced according to actual data and trait characteristics; the number of folds in cross-validation is not limited to 5 folds, and 3 folds, 5 folds, or 10 folds can be selected for cross-validation according to the size of the training population. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting peanut calcium environment-specific whole-genome selection, characterized in that, Includes the following steps: S1. Obtaining whole-genome SNP loci: Using peanut germplasm resources as training populations, field phenotypic identification and whole-genome resequencing were carried out at multiple locations under different calcium environments over many years. Whole-genome SNP loci were obtained through quality control. S2. Obtain association markers: Based on the whole genome SNP sites, whole genome association analysis was performed under calcium-treated and non-calcium-treated environments to extract the association marker information of the top N P-values ​​of each trait, and then the association markers were obtained. S3. Obtain three types of functional target marker subsets: Based on the intersection and difference algorithm, perform data layering on the associated markers to obtain three types of functional target marker subsets; S4. Obtain multi-scenario prediction accuracy: Using the three types of functional target marker subsets as prediction variables, construct prediction models respectively, and carry out multi-scenario interactive cross-environment prediction in both calcium-applied and non-calcium-applied environments to obtain multi-scenario prediction accuracy. S5. Output prediction and ranking results: Evaluate the accuracy of the multi-scenario prediction through cross-validation, obtain the optimal prediction model for different traits, predict and rank the genomic estimated breeding values ​​of the test breeding materials based on the optimal prediction model, and output the prediction and ranking results.

2. The peanut calcium environment-specific whole-genome selection prediction method according to claim 1, characterized in that, In step S1, the average sequencing depth of the whole genome resequencing is 20×. The whole genome SNP sites are quality controlled using the GATK best practice process and population-level filtering to retain sites with a minimum allele frequency of less than 5% and a deletion rate of less than 20%, thereby obtaining 1,150,004 whole genome SNP sites.

3. The peanut calcium environment-specific whole-genome selection prediction method according to claim 1, characterized in that, In step S2, the value of N is 1000. Based on the significance of the genome-wide association analysis results, the P values ​​are sorted from smallest to largest, and then the association marker information of the top 1000 is extracted to finally obtain the association marker.

4. The peanut calcium environment-specific whole-genome selection prediction method according to claim 1, characterized in that, In step S3, the three types of functional targeting marker subsets include a purely constitutive marker subset common to both environments, a specific marker subset for the low-calcium environment, and a specific marker subset for the calcium-treated environment. Let A be the set of associated markers in the calcium-treated environment and B be the set of associated markers in the non-calcium-treated environment. Then: The subset of purely constitutive markers shared by the two environments is: Cons = A ∩ B; The specific marker subset for the low-calcium environment is: Indu=B\A; The specific marker subset of the calcium application environment is: Spec=A\B.

5. The peanut calcium environment-specific whole-genome selection prediction method according to claim 4, characterized in that, In step S4, the multi-scenario interaction cross-environment prediction includes at least the following scenarios: First scenario: Based on a subset of specific markers in the low-calcium environment, predict the phenotype in the low-calcium environment; Second scenario: Based on the subset of purely constitutive markers shared by the two environments, predict the phenotype under low calcium environment; Third scenario: Based on the specific marker subset of the low-calcium environment, predict the phenotype under calcium application environment; Fourth scenario: Based on a subset of specific markers from the calcium application environment, predict the phenotype under the calcium application environment; Fifth scenario: Based on the subset of purely constitutive markers shared by the two environments, predict the phenotype under calcium application conditions; Scenario 6: Based on a subset of specific markers from the calcium application environment, predict the phenotype under low calcium conditions.

6. The peanut calcium environment-specific whole-genome selection prediction method according to claim 1, characterized in that, In step S4, the prediction model includes at least one of the following: Bayesian A prediction model, Bayesian B prediction model, Bayesian CPi prediction model, genome optimal linear unbiased prediction model, ridge regression optimal linear unbiased prediction model, random forest prediction model, and support vector machine prediction model.

7. The peanut calcium environment-specific whole-genome selection prediction method according to claim 1, characterized in that, In step S4, the traits predicted by the multi-scenario interactive cross-environment include at least one of peanut-related traits such as pod fullness, pod plumpness, kernel yield, 100-pod weight, pod thickness, and pod width.

8. The peanut calcium environment-specific whole-genome selection prediction method according to claim 1, characterized in that, In step S5, the cross-validation is 5-fold cross-validation, and its specific steps are as follows: A. Randomly divide the training group into 5 non-overlapping parts, take the first part as the "validation set", and merge the remaining 4 parts as the "training set" to construct the prediction validation model; B. Based on the constructed prediction and verification model, predict the phenotype of the first part, record the correlation coefficient between the predicted value and the actual value, and then obtain the prediction and verification accuracy. C. Take the second set as the "validation set", and merge the remaining 4 sets as the "training set" to construct the prediction validation model. Repeat steps A and B until all 5 sets of data have been used as the "validation set" once. D. Take the average of the prediction verification accuracies obtained in the above 5 rounds to obtain the average prediction verification accuracy.

9. The peanut calcium environment-specific whole-genome selection prediction method according to claim 1, characterized in that, In step S5: For environment-specific major-effect associated sites with predominantly additive effects, Bayesian A prediction model, Bayesian B prediction model, Bayesian CPi prediction model, genome-optimal linear unbiased prediction model, or ridge regression-optimal linear unbiased prediction model are used for prediction. For environment-specific major-effect associated sites where non-additive effects are predominant, random forest or support vector machine prediction models are used for prediction.

Images