Genome selective breeding method based on poplar multi-generation hybrid population growth traits and application

Abstract

The invention discloses a genome selective breeding method based on poplar multi-generation hybrid population growth traits and application, and belongs to the technical field of poplar genetic breeding. Based on a multi-generation hybrid population of 762 American black poplar plants, a growth character genome selective breeding research is systematically developed, an optimal reference genome is selected, and SNP loci significantly related to the growth character of the poplar are identified through hybrid population re-sequencing and whole genome association analysis. The influence of 10 GS statistical models and different marker screening strategies on the whole genome selection prediction precision is evaluated, the optimal genome selection prediction method for the growth traits of the populus americana is established, and compared with random markers, the prediction precision is averagely improved by 114.48%. According to the method, early prediction of poplar growth traits can be realized, and fast-growing excellent parents and filial generations can be screened out, so that the forest tree breeding period is shortened, the forest tree breeding precision is improved, the test scale is reduced, and the breeding process of forest tree excellent germplasm is accelerated.

Description

Technical Field

[0001] The invention relates to a method and application of genome selection breeding based on growth traits of multi-generation hybrid population of poplars, belonging to the technical field of poplar biological breeding. Background Art

[0002] Poplars are important afforestation species in my country and are widely used in timber production, pulp and paper making, and ecological restoration. Existing new poplar varieties are difficult to meet breeding needs, and there is an urgent need to accelerate the genetic improvement process of poplars and increase genetic gain.

[0003] Traditional genetic improvement relies on long-term phenotypic evaluation and intergenerational alternation, and the breeding cycle is long. The whole genome selection (GS) breeding method is considered to be an efficient breeding method. Since it was proposed in 2001, it has played an increasingly important role in shortening the breeding cycle and improving breeding efficiency. Its advantage is that it can use whole genome markers and statistical models to evaluate individual genomic breeding values ​​(GEBV), simplify a large amount of phenotypic measurement work in the breeding process, and achieve early prediction. Growth, as an important breeding goal for poplar genetic improvement, is difficult to achieve early selection of growth traits through conventional breeding in the actual breeding process. Genomic selection can solve this problem to a certain extent. At present, there are few reports on genomic selection of growth traits in multi-generation hybrid populations of poplars.

[0004] Genomic prediction often has better prediction effects for closely related groups, especially for multi-generation hybrid groups. In addition to kinship, statistical models, marker density, etc. will affect the prediction accuracy (PA). At present, a variety of models have been developed for whole genome selection research. Different statistical models often have different prior assumptions and different assumptions about the distribution of marker effects. At present, no model can be well applied to most whole genome selection studies. In actual breeding, it is necessary to conduct a systematic evaluation in combination with breeding goals to select the most suitable model. For high-density markers, the introduction of too many invalid markers in actual analysis often causes the problem of large "p" and small "n", resulting in overfitting of the model. At the same time, too many markers will also waste a lot of computing resources.

[0005] Therefore, it is an urgent technical challenge to systematically carry out genomic selection breeding of multi-generation hybrid populations of American black poplar (Populus deltoides), establish a genetic improvement breeding system for whole-genome selection of growth traits, select fast-growing, high-yield and excellent parents and asexual lines, achieve genetic improvement of poplars, and increase genetic gain. Summary of the invention

[0006] The purpose of the present invention is to provide a genomic selection breeding method and application based on the growth traits of multi-generation hybrid populations of poplars, by determining the optimal reference genome, identifying genome-wide association study (GWAS) significant markers, selecting the optimal GS model and screening efficient prediction markers, to achieve accurate screening of fast-growing and excellent individuals, shorten the breeding cycle, and achieve genetic improvement of American black poplar.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for genomic selection breeding based on growth traits of multi-generation hybrid populations of poplars, comprising the following steps:

[0009] Step 1: Resequencing of the genome selection population and determination of the phenotype of the reference population

[0010] Multi-generation hybridization was used to construct the genomic selection reference population and breeding population. The genomic selection population was sequenced using the Illumina HiSeq6000 high-throughput sequencing platform. Paired-end PE150 (PE refers to paired-end sequencing, and 150 refers to the number of bases measured at both ends) sequencing was performed on the genomic selection population. For the reference population, tree height (H) and diameter at breast height (DBH) were measured.

[0011] Step 2: Reference genome selection and genotyping

[0012] Three high-quality genomes were selected: Populus trichocarpa, Populus deltoides 'Lux', and Populus deltoides 'Danhong' (DHY), and the population resequencing data were compared and evaluated. The most suitable reference genome was selected based on the comparison rate. Finally, Populus deltoides 'Danhong' was used as the reference genome for genotyping and quality control to obtain high-quality single nucleotide polymorphism (SNP) markers.

[0013] Step 3: Genome-wide association analysis of growth traits

[0014] Based on the genotype data and growth phenotype data of the reference population, the GWAS model was selected for association analysis to identify the quantitative trait nucleotides (QTNs) that affect the phenotype;

[0015] Step 4: Evaluation of the optimal genomic selection model for growth traits

[0016] A 5-fold cross validation was used to evaluate the performance of 10 GS models at different random marker densities and screen the optimal statistical model for growth traits;

[0017] Step 5: Screening strategy for optimal GS markers for growth traits

[0018] The screening strategy includes screening based on linkage disequilibrium (LD) levels among markers, correlation between markers and traits, and using equal amounts of random markers as controls;

[0019] Step 6: Establishment of the optimal prediction system for whole genome selection

[0020] The optimal reference genome, a subset of markers significantly associated with growth traits, the optimal whole-genome selection statistical model and the optimal marker screening strategy constitute the optimal prediction system for whole-genome selection of poplar growth traits.

[0021] Preferably, in step 2, according to the evaluation results, the core parent Danhong Poplar genome is further selected as a reference for genotyping; SNP calling and filtering are performed using GATK; the parameters of GATK are set as: QD<2.0, FS>50, MQ<20, MQRankSum>-12.5 and ReadPosRankSum>-8.0.

[0022] Preferably, in step 3, the genome-wide association analysis is performed using the IIIVmrMLM model.

[0023] Preferably, in step 4, the 10 GS models are GBLUP, rrBLUP, BayesA, BayesB, BayesC, BRR, RF, GBDT, KRR and DNNGP.

[0024] Preferably, in step 4, the different random marker densities are allSNP, 0.5SNP, 0.2SNP, 0.1SNP, 0.05SNP and 0.01SNP.

[0025] Preferably, in step 4, the 5-fold cross-validation method uses 80% of the reference population as the training population and 20% as the test population, and repeats the iteration 20 times; and uses the mean Pearson correlation coefficient between the test population breeding value and the actual observation value as an indicator for evaluating the prediction accuracy.

[0026] Preferably, in step 5, the linkage disequilibrium level (LD level) between the markers is LD≤1, LD≤0.8, LD≤0.5, LD≤0.3, LD≤0.2 and LD≤0.1; the level of correlation between the marker and the trait is p≤1, p≤0.5, p≤0.1, p≤0.01, p≤0.001 and p≤0.0001.

[0027] Preferably, in step 6, the optimal reference genome is the Populus danhongensis genome; the number of growth trait significantly associated marker subsets (tree height and breast height diameter significantly associated marker subsets) is 27; the optimal whole genome selection statistical models are RF and rrblup, respectively; and the optimal marker screening strategy is based on the GWAS association analysis p≤0.01 level.

[0028] The above-mentioned method of genomic selection breeding based on growth traits of multi-generation hybrid populations of poplars is applied in poplar breeding.

[0029] Preferably, the GEBV value of the breeding population is predicted using the optimal GS prediction system for growth traits, and excellent parents and progeny are selected based on the prediction results of tree height and breast height traits.

[0030] The beneficial effects of the present invention are as follows:

[0031] Compared with the prior art, the present invention establishes a method and application of genomic selection breeding of growth traits of multi-generation hybrid populations of American poplar (Populus deltoides), including an optimal reference genome, a subset of markers significantly associated with growth traits, an optimal whole genome selection statistical model and an optimal marker screening strategy; compared with other genomes, the average matching rate of the optimal reference genome is improved by 8%; compared with random markers, the prediction accuracy of tree height and breast diameter of the optimal whole genome prediction system is improved by 115.29% and 113.68% respectively, with an average improvement of 114.48%, and the accuracy and selection efficiency of genomic prediction of growth traits of poplar controlled by multiple genes are improved.

[0032] The present invention is further described below through specific implementation modes and drawings, but it is not intended to limit the protection scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Statistics of the comparison rates of the whole genome resequencing data under different reference genomes in Example 1 of the present invention;

[0034] Figure 2 The distribution of high-quality SNP markers on 19 chromosomes in Example 1 of the present invention;

[0035] Figure 3 The Manhattan plot of the whole gene association analysis of growth traits of the reference population in Example 1 of the present invention;

[0036] Figure 4 It is a box plot of the prediction accuracy of whole genome selection of growth traits of 10 statistical models at 6 SNP marker densities in Example 2 of the present invention;

[0037] Figure 5 This is a graph of model prediction accuracy under different marker screening strategies in Example 2 of the present invention;

[0038] Figure 6 It is a scatter plot of prediction accuracy under the optimal prediction system for growth traits in Example 2 of the present invention;

[0039] Figure 7 This is the kinship network of excellent asexual lines under the optimal prediction system for growth traits in Example 2 of the present invention. DETAILED DESCRIPTION

[0040] The present invention is further described below in conjunction with specific examples. Operations not described in detail in the following examples can be implemented by referring to molecular cloning and the instructions for use of related kits.

[0041] Unless otherwise specified, the reagents involved in the following examples are all conventional reagents available on the market, and the methods used are all commonly used methods in the art.

[0042] Example 1

[0043] A method for genomic selection breeding based on growth traits of multi-generation hybrid populations of poplars, comprising the following steps:

[0044] (1) Population resequencing and phenotyping

[0045] The Illumina HiSeq6000 high-throughput sequencing platform was used to perform double-end PE150 sequencing on the genome selection population; for the reference population, the tree height (H) and diameter at breast height (DBH) were measured, and the phenotypic statistical results are shown in Table 1 below:

[0046] Table 1 Statistical results of growth traits

[0047]

[0048] (2) Reference genome selection and genotyping

[0049] The genome of Populus trichocarpa, which is widely used as a reference genome, and the genomes of Populus deltoides'Lux' and Populusdeltoides'Danhong', which are related to the population, were selected as reference genomes to compare and evaluate the population resequencing data.

[0050] The specific implementation is as follows: First, the data quality control was performed using the default parameters of FASTP. Then, the quality-controlled data were aligned to the reference genome using the bwa-mem algorithm. Finally, the alignment rate of each sample was counted using the multiqc software and visualized using the ggbeeswarm package in the R software. The alignment results showed that the alignment rates of Populus trichocarpa and Populus I-69 were 89.04% and 91.21%, respectively, which were much lower than the 97.69% alignment rate of the Danhong Populus genome ( Figure 1 ); the core Danhong Populus genome was further selected as the reference genome, and GATK was used for variation detection. The GATK parameters were set as follows: QD<2.0, FS>50, MQ<20, MQRankSum>-12.5, and ReadPosRankSum>-8.0. VCFtools software was used to strictly filter the genotypes. The filtering parameters were: sequencing depth greater than 5×, minimum allele frequency not less than 0.05, deletion rate less than 5%, and serious deviation from Hardy-Weinberg equilibrium markers. After genotyping and quality control of 762 trees, 114 high-quality SNP markers were finally obtained, all of which were evenly distributed on 19 chromosomes ( Figure 2 ), and finally, filled using BEAGLE software for subsequent analysis;

[0051] (3) Genome-wide association analysis of growth traits

[0052] Based on the high-quality SNP markers obtained above, principal component analysis was performed using Tassel software, and the kinship matrix was calculated. Combined with the reference population phenotype, the IIIVmrMLM model was used for analysis. Genome scanning was used to obtain all marker P-values, and markers with P-values ​​< 0.01 were retained as potential QTNs for the next step of analysis. The potential QTNs obtained by genome scanning were incorporated into the multi-locus model, and the marker effect was estimated by Bayesian, and the likelihood ratio was output to test QTNs with LOD scores greater than 3.0. For significant and suggested QTNs, the threshold P value and LOD score were set to 0.05 / m and 3.0, respectively, where m is the number of potential QTNs. The Manhattan plot of the GWAS results was drawn using IIIVmrMLM software ( Figure 3), the number of QTNs with tree height and DBH exceeding the threshold were 27 and 27 respectively; significant QTNs will be added as covariates in the subsequent GS model.

[0053] Example 2: Application of the breeding method based on genomic selection of growth traits of multi-generation hybrid populations of poplars in poplar breeding

[0054] (1) Evaluation of the optimal genomic selection model for growth traits

[0055] The 5-fold cross-validation method was used to evaluate the difference in the prediction ability of the 10 mainstream GS models for two growth traits of Populus deltoides. In the 5-fold cross-validation method, 80% of the reference population was used as the training population and 20% as the test population. The process was repeated 20 times. The 10 models included 2 BLUP models (RR-BLUP and GBLUP), 4 Bayes models (BayesA, BayesB, BayesC, Bayesian LASSO and BRR), 3 machine learning models (GBDT, FR, XGBoost, SVR) and a deep learning model DNNGP.

[0056] To improve the accuracy of model evaluation, vcflib software was used to randomly generate 6 sets of high-density marker subsets; the 6 random marker densities were allSNP, 0.5SNP, 0.2SNP, 0.1SNP, 0.05SNP and 0.01SNP; the mean Pearson correlation coefficient between the test population breeding value and the actual observed value was used as an indicator to evaluate the prediction accuracy; the prediction ability of each model for growth traits under the 6 marker densities was as follows Figure 4 As shown; Based on the prediction performance of the models under 6 marker densities, it is believed that the RF model is the best for H trait: 0.266; the rrblup model is the best for DBH trait: 0.340;

[0057] (2) Screening strategy for optimal GS markers for growth traits

[0058] Based on the optimal model of growth traits, three strategies were used to generate different marker subsets (Table 2). According to the prediction results, the prediction of the optimal model of growth traits under different strategies was statistically analyzed, such as Figure 5 As shown in the figure, there is no significant difference in the prediction ability of traits between LD screening markers and random markers, and the LD level has almost no effect on the prediction ability of the model; there are significant differences in the prediction accuracy of growth traits under different P values; with the increase of correlation, the breast diameter shows a trend of first increasing and then decreasing, while with the increase of correlation, the tree height shows a trend of first increasing and then stabilizing ( Figure 5), when the P value of the marker and trait reached the 0.01 level, the model had a good prediction result; after adding the significant marker as a covariate to the model, the prediction accuracy was finally evaluated, and a scatter density map was drawn based on the GEBV value and the actual observed value, and the linear regression equation was fitted. The final prediction accuracy of the two traits H and DBH was 0.730 and 0.563 ( Figure 6 ), compared with random sampling, the improvement is 113.68% and 115.29% respectively;

[0059] Table 2 Genomic selection marker screening strategy

[0060]

[0061] (3) Prediction of breeding population genome and screening of superior clones

[0062] Based on the optimal GS prediction system for growth traits, the GEBV values ​​of 502 clones in the breeding population were predicted. Subsequently, the breeding population was sorted according to the GEBV values, and excellent clones were selected according to a 20% seed retention rate. The ranking results of growth traits were integrated to select fast-growing excellent clones to achieve genetic improvement of growth traits of Populus deltoides. Table 3 shows: 28 Populus deltoides materials with excellent growth traits. The genetic relationship of excellent clones was further integrated, and a genetic relationship network was constructed based on the information of paternal and maternal parents. The Cytoscape software was used for visualization, and the importance was indicated by the size of the point. The excellent female parents DHY and 27-4 and the excellent male parents 17-2, 16-27, and 278-3 with good growth performance were selected ( Figure 7 ).

[0063] Table 3 Results of 28 clones of Populus deltoides with excellent growth traits

[0064]

[0065] The present invention is based on a multi-generation hybrid population of 762 American black poplars, systematically conducts research on genomic selection breeding for growth traits, selects the most suitable reference genome, identifies QTNs significantly correlated with poplar growth traits through hybrid population resequencing and whole genome association analysis, evaluates the effects of 10 GS statistical models and different marker screening strategies on the prediction accuracy of whole genome selection, and establishes the most suitable genomic selection prediction method for growth traits of American black poplars. Compared with random markers, the prediction accuracy is improved by an average of 114.48%.

[0066] The present invention can achieve early prediction of poplar growth traits, screen out fast-growing and excellent parents and offspring, thereby shortening the tree breeding cycle, improving the tree breeding accuracy, reducing the test scale and accelerating the breeding process of excellent tree germplasm.

[0067] Although the purpose, concept and embodiments of the present invention have been described in detail above, a person skilled in the art will recognize that various improvements and modifications may be made to the present invention without departing from the scope of the claims, and such improvements and modifications shall still fall within the scope of protection of the present invention.

Claims

1. A method for genomic selection and breeding based on growth traits of multi-generation hybrid populations of poplars, comprising the following steps: Step 1: Resequencing of the genome selection population and determination of the phenotype of the reference population Multi-generation hybridization was used to construct the genomic selection reference population and breeding population, and the genomic selection population was sequenced using the Illumina HiSeq6000 high-throughput sequencing platform. For the reference population, tree height and DBH were measured; Step 2: Reference genome selection and genotyping Three high-quality genomes were selected: Populus trichocarpa, Populus I-69, and Populus danhongensis. The population resequencing data were compared and evaluated. The most suitable reference genome was selected based on the comparison rate. Finally, Populus danhongensis was used as the reference genome for genotyping quality control to obtain high-quality single nucleotide polymorphism markers. Step 3: Genome-wide association analysis of growth traits Based on the genotype data and growth phenotype data of the reference population, the GWAS model was selected for association analysis to identify the quantitative trait nucleotides that affect the phenotype; Step 4: Evaluation of the optimal genomic selection model for growth traits A 5-fold cross validation was used to evaluate the performance of 10 GS models at different random marker densities and screen the optimal statistical model for growth traits; Step 5: Screening strategy for optimal GS markers for growth traits The screening strategy included screening based on the level of linkage disequilibrium between markers, the correlation between markers and traits, and using an equal amount of random markers as controls.

2. The method for selective breeding based on the growth trait genomics of multi-generation hybrid population of poplar as claimed in claim 1, characterized in that: In step 1, for the reference population, the tree height and breast diameter are measured.

3. The method for selective breeding based on the growth trait genomics of multi-generation hybrid population of poplar as claimed in claim 1, characterized in that: In step 2, based on the evaluation results, the core parent Populus danhongensis was further selected for resequencing data alignment; SNP calling and filtering were performed using GATK; the parameters of GATK were set to: QD<2.0, FS>50, MQ<20, MQRankSum>-12.5, and ReadPosRankSum>-8.

0.

4. The method for selective breeding based on the growth trait genomics of multi-generation hybrid population of poplar as claimed in claim 1, characterized in that: In step 3, genome-wide association analysis was performed using the IIIVmrMLM model.

5. The method for selective breeding based on the growth trait genomics of multi-generation hybrid population of poplar as claimed in claim 1, characterized in that: In step 4, the 10 GS models are GBLUP, rrBLUP, BayesA, BayesB, BayesC, BRR, RF, GBDT, KRR and DNNGP.

6. The method for selective breeding based on the growth trait genomics of multi-generation hybrid population of poplar as claimed in claim 1, characterized in that: In step 4, the different random marker densities are all SNP, 0.5 SNP, 0.2 SNP, 0.1 SNP, 0.05 SNP and 0.01 SNP; The 5-fold cross-validation method uses 80% of the reference population as the training population and 20% as the test population, and the process is repeated 20 times; and the mean Pearson correlation coefficient between the test population breeding value and the actual observed value is used as an indicator for evaluating the prediction accuracy.

7. The method for selective breeding based on the growth trait genomics of multi-generation hybrid population of poplar as claimed in claim 1, characterized in that: In step 5, the linkage disequilibrium levels among the markers are 1, 0.8, 0.5, 0.3, 0.2 and 0.1; the correlation levels between the markers and the traits are p≤1, p≤0.5, p≤0.1, p≤0.01, p≤0.001 and p≤0.0001.

8. Application of the method for genomic selection breeding based on growth traits of multi-generation hybrid populations of poplars as claimed in any one of claims 1 to 7 in poplar breeding.

9. Application of the method for genomic selection breeding based on growth traits of multi-generation hybrid population of poplars in poplar breeding as claimed in claim 8, characterized in that: The GEBV value of the breeding population was predicted using the optimal GS prediction system for each trait, and excellent parents and progeny were selected by integrating the prediction results of multiple traits.

10. Application of the method for genomic selection breeding based on growth traits of multi-generation hybrid population of poplars as claimed in claim 9 in poplar breeding, characterized in that: The optimal prediction system for whole genome selection of poplar growth traits was composed of the most suitable reference genome, a subset of markers significantly associated with each trait, the optimal whole genome selection statistical model and the optimal marker screening strategy. The most suitable reference genome was the Danhong Populus genome; the number of tree height and breast diameter in the subset of markers significantly associated with each trait was 27 each; the optimal whole genome selection statistical models were RF and rrblup respectively; and the optimal marker screening strategy was based on the GWAS association analysis at the p≤0.01 level.

Images