Precise breeding method based on single nucleotide polymorphism

Abstract

The invention relates to the technical field of breeding methods, in particular to a precise breeding method based on single nucleotide polymorphism, which comprises the following steps: carrying out whole genome sequencing and screening SNP (Single Nucleotide Polymorphism) markers related to target traits; designing a genotype detection chip based on the SNP marker, and performing genotyping; constructing a genetic evaluation model and predicting a breeding value; designing a matching scheme based on the breeding value; performing performance verification on the offspring and continuously improving the breeding strategy. By integrating whole genome sequencing, transcriptome and metabolome data and combining bioinformatics and an artificial intelligence algorithm, efficient screening of functional SNPs related to complex characters is achieved, and the problems of multi-character collaborative improvement and insufficient environmental adaptability in a traditional method are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of breeding methods, and in particular to a precision breeding method based on single nucleotide polymorphism. Background Art

[0002] With the rapid development of modern animal husbandry, the efficient and precise improvement of livestock breeds to meet growing market demand has become a core issue in livestock breeding. Traditional breeding methods, which primarily rely on phenotypic selection, have achieved remarkable success over the past few decades but have gradually become exposed to limitations such as low efficiency, long lead times, and difficulty in achieving synergistic improvement of multiple traits. In recent years, advances in molecular biology have led to the emergence of DNA marker-based molecular-assisted selection breeding methods, injecting new vitality into livestock and poultry breeding.

[0003] Among the many molecular markers, single nucleotide polymorphisms (SNPs) have become a hot topic in current molecular breeding research due to their wide distribution in the genome, high stability, and ease of large-scale detection. However, current SNP-based breeding methods still face many technical difficulties. First, how to efficiently screen out functional markers closely related to target traits from a large number of SNPs remains a huge challenge. Second, existing methods often only focus on a single or a few traits, making it difficult to achieve the synergistic improvement of multiple economic traits. Furthermore, traditional SNP selection methods do not fully consider the interaction effect between genotype and environment, resulting in insufficient environmental adaptability of the cultivated varieties. In addition, long-term directional selection may lead to the loss of genetic diversity and affect the sustainability of breeding. Summary of the Invention

[0004] In response to the above problems, the present invention proposes a precision breeding method based on single nucleotide polymorphisms. This method integrates multi-omics data such as whole genome sequencing, transcriptome, metabolome and microbiome, and combines advanced bioinformatics and artificial intelligence algorithms to achieve efficient screening of functional SNPs related to complex traits. At the same time, this method innovatively introduces a multi-trait evaluation model that takes environmental factors into account, effectively solving the problems of multi-trait synergistic improvement and insufficient environmental adaptability in traditional methods. In addition, the present invention also develops an adaptive breeding strategy based on reinforcement learning, which effectively maintains the genetic diversity of the population while ensuring the speed of genetic improvement.

[0005] The present invention aims to provide a single nucleotide polymorphism-based precision breeding method, comprising the following steps: (1) Perform whole genome sequencing and screen for SNP markers associated with target traits; (2) Designing a genotype detection chip based on the SNP marker and performing genotyping; (3) Construct a genetic evaluation model and predict breeding values; (4) Designing a selection plan based on the breeding values; (5) Verify the performance of offspring and continuously improve breeding strategies.

[0006] Preferably, step (1) comprises: Whole genome sequencing of no less than 1,000 pigs with significant growth trait differences must be performed, with a sequencing depth of no less than 30X and a coverage of no less than 99%; Sequence alignment and SNP calling were performed, and the quality control standards were: sequencing quality value Q>30, sequencing depth DP>10, deletion rate <5%, and minor allele frequency MAF>0.05; Genome-wide association analysis was performed using a linear mixed model, with a significance threshold set at P < 5 × 10^-8; Select SNP markers significantly associated with growth traits, select at least 10 SNPs for each trait, and screen no less than 50 SNP markers in total; The parameters significantly correlated with growth traits include daily weight gain, feed-to-meat ratio, backfat thickness, and loin area.

[0007] Preferably, step (2) includes: Based on the screened SNP markers, an Affymetrix Axiom chip containing 200-500 SNPs was designed; Extract genomic DNA from the ear tissue of the pig to be tested. The DNA concentration must be above 50 ng / μL and the OD260 / 280 ratio must be between 1.8 and 2.0. The breeding population was genotyped using the SNP chip, with each batch containing at least 384 samples and 3 replicate samples set as quality control; For genotype data analysis, the SNP call rate threshold was set to 97% and the sample call rate threshold was set to 95%.

[0008] Preferably, step (3) includes: Collect growth trait data of the breeding population, perform data cleaning and standardization, remove outliers, and perform Box-Cox transformation on phenotypic data; The genetic evaluation model was constructed using the genomic optimal linear unbiased prediction model combined with the single-step method; Estimate variance components and predict breeding values.

[0009] Preferably, the equation of the genome's best linear unbiased prediction model is: y = Xb + Za + e, Where y is the phenotype vector, b is the fixed effect vector, a is the additive genetic effect vector, X and Z are the corresponding design matrices, and e is the residual vector.

[0010] Preferably, step (4) comprises: Constructing a genome relationship matrix; Predict the genomic breeding value GEBV of each individual; Calculate the integrated breeding value SI, where SI = w1 × GEBV daily weight gain + w2 × GEBV feed to meat ratio + w3 × GEBV backfat thickness + w4 × GEBV eye muscle area, w1-w4 are the economic weights of each trait; Calculate the inbreeding coefficient between potential parents and set the inbreeding coefficient threshold to 0.0625; implement the optimal contribution selection algorithm and optimize the mating plan.

[0011] Preferably, step (5) includes: Randomly select 2-3 piglets from each mating combination to construct a test group of no less than 200 piglets; The growth traits and genotyping of the test population were measured; Breeding values were re-estimated and compared with parental generations to calculate genetic progress; Calculate the correlation coefficient between the predicted breeding value and the actual performance. If the accuracy is lower than 0.5, review the SNP selection criteria and model parameters.

[0012] Preferably, the method further comprises the following steps: The SNP marker set was updated every 3–5 generations; Model parameters were periodically re-estimated using accumulated phenotypic and genotypic data; Integrate new technologies such as whole genome sequence data, epigenetic markers, and transcriptome data to optimize breeding strategies.

[0013] Preferably, the method further comprises the following steps: Environmental covariates were introduced into the genome's optimal linear unbiased prediction model to construct a genotype-environment interaction model; Use machine learning algorithms to dynamically adjust the weights of each trait in the multi-trait selection index; Metabolomic analysis of blood and muscle tissue was performed on some individuals, and metabolite profiles were integrated with SNP genotype and phenotype data.

[0014] Preferably, it is characterized in that it further comprises the following steps: Add marker sites related to DNA methylation to the SNP array; Develop a cloud computing-based breeding management system to achieve automation and intelligence in data collection, analysis, and decision-making; Using gene editing technology, we performed site-specific editing of the identified key SNPs in porcine embryonic stem cells cultured in vitro to verify their causal effects on growth traits; Use deep learning algorithms to analyze whole genome sequencing data and discover complex genomic features.

[0015] The present invention has the following advantages and beneficial effects: The method of this invention not only significantly improves breeding efficiency and accuracy, but also demonstrates unexpected technical benefits in multiple areas. First, through the integrated analysis of multi-omics data, this method not only improves target traits but also achieves a synergistic improvement in disease resistance, opening up new avenues for breeding healthier and more efficient livestock and poultry breeds.

[0016] The method of the present invention effectively alleviates the negative correlation between growth rate and meat quality, achieving synergistic improvements in these two key economic traits. Furthermore, the cultivated strains generally exhibit enhanced stress resistance, which is of great significance for improving farming efficiency and animal welfare.

[0017] This invention, through the innovative integration of cutting-edge multidisciplinary technologies, effectively addresses key technical issues existing in existing SNP breeding methods, providing a comprehensive, precise, and efficient new approach for modern livestock and poultry breeding. This method not only achieves a qualitative leap in breeding efficiency and improvement, but also demonstrates significant advantages in long-term breeding sustainability. This series of technological breakthroughs and innovative achievements has made significant contributions to the sustainable development of animal husbandry and food safety, and has broad application prospects and enormous economic and social benefits. DETAILED DESCRIPTION

[0018] For a better understanding of the present invention, the following examples are provided to further illustrate the present invention, but the present invention is not limited to the following examples.

[0019] Example 1 This embodiment provides a single nucleotide polymorphism-based precision breeding method for improving the growth traits of Duroc pigs. The method comprises the following steps: First, whole-genome sequencing was performed on 1,000 Duroc pigs with significant growth trait differences using the Illumina NovaSeq 6000 sequencing platform. Sequencing depth was set to 30x, achieving 99.5% coverage. This high-depth, high-coverage sequencing strategy ensured the capture of rare variants associated with growth traits, laying a solid foundation for subsequent SNP screening.

[0020] Next, sequencing reads were aligned to the porcine reference genome (Sscrofa11.1) using BWA (v0.7.17), followed by SNP calling using GATK (v4.1.4.1). Stringent quality control standards were applied: sequencing quality (Q) > 30, sequencing depth (DP) > 10, deletion rate < 3%, and minor allele frequency (MAF) > 0.05. These stringent quality control criteria help reduce the rate of false-positive SNPs and improve the reliability of subsequent analyses.

[0021] Subsequently, genome-wide association studies (GWAS) were performed using GEMMA (v0.98.1) software. A linear mixed model (LMM) was used to adjust for population structure and individual relatedness, with a significance threshold set at P < 5 × 10^-8 (Bonferroni-corrected genomic significance level). This statistical model effectively controls for false positives due to population structure and relatedness, improving the reliability of GWAS results.

[0022] Based on GWAS results, we selected SNP markers significantly associated with growth traits such as daily gain, feed-to-meat ratio, backfat thickness, and loin eye area. For each trait, we selected 15 significantly associated SNPs, for a total of 60 SNP markers. This multi-trait, multi-marker screening strategy facilitates capturing the complex genetic architecture of growth traits.

[0023] During the genotyping phase, an Affymetrix Axiom array containing 300 SNPs was designed based on the selected SNP markers. In addition to the target SNPs, neutral markers were also included for quality control and assessment of population structure. This design ensures accurate assessment of target traits while also providing comprehensive genomic information.

[0024] For DNA extraction, genomic DNA was extracted from ear tissue of the pigs to be tested using the Qiagen DNeasy Blood & Tissue Kit. DNA quality was strictly controlled, requiring a DNA concentration of 60 ng / μL and an OD260 / 280 ratio between 1.85 and 1.95. High-quality DNA samples are crucial for ensuring accurate genotyping.

[0025] During genotyping, each batch contained 384 samples, with three replicates used as quality controls. Hybridization and scanning were performed using a GeneTitan Multi-Channel Instrument. Genotype data were subsequently analyzed using Affymetrix AxiomAnalysis Suite software (v4.0.1), with SNP call rate thresholds set at 98% and sample call rate thresholds at 96%. These stringent quality control standards ensured high genotyping data quality.

[0026] When developing a genetic evaluation model, the breeding population is systematically measured for growth traits. Daily weight gain is measured from 30 kg to 100 kg body weight, and feed conversion is recorded over the same period. Backfat thickness and loin area are measured at 100 kg body weight using an Aloka SSD-500 ultrasound system at the 10th rib. This standardized phenotyping method helps improve data consistency and comparability.

[0027] During data preprocessing, R software (v4.0.3) was used for data cleaning and normalization. Outliers exceeding ±3.5 standard deviations from the mean were removed, and phenotypic data were Box-Cox transformed to approximate a normal distribution. This data preprocessing not only removed obvious outliers but also improved the statistical properties of the data, laying the foundation for subsequent modeling.

[0028] This example uses the genomic best linear unbiased prediction (GBLUP) model combined with the single-step method to construct a genetic evaluation model. The model equation is: y = Xb + Za + e, where y is the phenotypic vector, b is the fixed effect vector (including sex, batch, season, etc.), a is the additive genetic effect vector, X and Z are the corresponding design matrices, and e is the residual vector. The ASReml-R software package (v4.1) is used to estimate variance components and predict breeding values. This model can simultaneously utilize genotypic and phenotypic information, improving the accuracy of breeding value prediction.

[0029] During the breeding value prediction phase, the genomic relationship matrix (G matrix) was first constructed using the R package "AGHmatrix" (v2.0.0). Then, based on the constructed GBLUP model, the genomic breeding value (GEBV) of each individual was predicted using the ASReml-R package. Finally, the composite breeding value (SI) was calculated: SI = 0.4 × GEBV daily gain + 0.3 × GEBV feed to meat ratio + 0.15 × GEBV backfat depth + 0.15 × GEBV loin eye area. This weighting scheme reflects the relative importance of each trait in the breeding objective.

[0030] In the mating plan design, the top 8% of boars and top 18% of sows were selected as potential parents based on their calculated combined breeding values. The inbreeding coefficients between potential parents were then calculated using the R package "pedigree" (v1.4), with an inbreeding coefficient threshold of 0.0625. Next, the optimal contribution selection (OCS) method was employed, implemented using the R package "optiSel" (v2.0.3). The objective function was set to maximize the expected breeding value of the next generation while limiting the average inbreeding coefficient increment to less than 0.4% per generation. Finally, a mating optimization algorithm was implemented using the R package "AlphaSimR" (v0.12.2) to design a specific mating plan. This multi-step mating strategy balances genetic improvement with the maintenance of genetic diversity.

[0031] During the progeny performance verification phase, three piglets were randomly selected from each mating combination to construct a test population of 250 piglets. The test population was subjected to the same growth trait measurements and genotyping as the breeding population. Breeding values were re-estimated using the updated data and compared to the parental generations to calculate genetic progress. Simultaneously, the correlation coefficient between the predicted breeding values and actual performance (prediction accuracy) was calculated. In this example, the prediction accuracy reached 0.65, demonstrating the good predictive power of this method.

[0032] Furthermore, this example also integrates metabolomics data. Fifty pigs were randomly selected from the test population, and blood and muscle tissue samples were collected at a body weight of 100 kg for metabolomics analysis. Liquid chromatography-mass spectrometry (LC-MS / MS) was used to obtain metabolite profiles. The metabolite data were then integrated with SNP genotype and phenotypic data, and partial least squares discriminant analysis (PLS-DA) was used to identify metabolite markers significantly associated with growth traits. This multi-omics integrated analysis not only improves the accuracy of breeding value predictions but also provides new insights into the molecular mechanisms underlying growth traits.

[0033] Example 2 This example provides a precision breeding method based on single nucleotide polymorphisms to improve the growth and meat quality traits of Large White pigs. This method is based on Example 1 with some adjustments and optimizations, and specifically includes the following steps: First, whole-genome sequencing was performed on 1,500 Large White pigs with significant differences in growth and meat quality traits using the Illumina NovaSeq 6000 sequencing platform. Sequencing depth was increased to 35X, achieving 99.8% coverage. This higher sequencing depth and coverage strategy helped capture more low-frequency variants, providing more comprehensive information for subsequent SNP screening.

[0034] During the SNP calling phase, variant detection was performed using GATK (v4.2.0), with slightly adjusted quality control standards: sequencing quality value (Q) > 32, sequencing depth (DP) > 12, deletion rate < 2.5%, and minor allele frequency (MAF) > 0.03. These stricter quality control standards helped further improve the quality of SNP data.

[0035] GWAS analysis used BOLT-LMM software, which is capable of processing larger datasets and improving computational efficiency. The significance threshold was set at P < 1 × 10^-8, a stricter threshold than that used in Example 1. Based on the GWAS results, 20 significantly associated SNPs were selected for each target trait (daily gain, feed-to-meat ratio, backfat thickness, loin eye area, and intramuscular fat content), for a total of 100 SNP markers.

[0036] For genotyping, an Affymetrix Axiom array containing 400 SNPs was designed. In addition to the target SNPs, several functional SNPs associated with meat quality were also added. DNA was extracted using a modified salting-out method, requiring a DNA concentration of 80 ng / μL and an OD260 / 280 ratio between 1.80 and 2.00.

[0037] During genotyping, the number of samples per batch was increased to 480, with five replicates used as quality control. Genotyping data were analyzed using Affymetrix Axiom Analysis Suite software (v4.1.0), with the SNP call rate threshold increased to 98.5% and the sample call rate threshold increased to 97%.

[0038] When constructing a genetic evaluation model, a multi-trait animal model was used, simultaneously considering growth and meat quality traits. In addition to the traits mentioned in Example 1, intramuscular fat content was also measured, using near-infrared spectroscopy (NIRS) for non-invasive in vivo measurement.

[0039] During data preprocessing, R software (v4.1.0) was used for data cleaning and standardization. Multivariate outliers were identified using the Mahalanobis distance method, and missing data were handled using multiple imputation. This method is more effective in handling outliers and missing values in multi-trait data.

[0040] This example uses the single-step ssGBLUP model for genetic evaluation. The model equation is: y = Xb + Za + Wpe + e, where pe represents the permanent environmental effect. The BLUPF90 software package is used to estimate variance components and predict breeding values. This model utilizes information from both genotyped and ungenotyped animals, improving the accuracy of breeding value predictions.

[0041] During the breeding value prediction phase, a weighted single-step method (WssGBLUP) was used to optimize SNP weights. Initial weights were set based on GWAS results and then refined through an iterative process. When calculating the composite breeding value, the weights for each trait were determined using the economic weighting method: SI = 0.35 × GEBV daily gain + 0.25 × GEBV feed to meat ratio + 0.15 × GEBV backfat depth + 0.15 × GEBV loin eye area + 0.10 × GEBV intramuscular fat.

[0042] In the design of the mating strategy, a genetic algorithm was used to optimize the mating plan. First, the top 5% of boars and top 15% of sows in terms of overall breeding value were selected as candidate parents. Optimal contribution selection was then implemented using the R package "optiSel" (v2.1.0). The goal was to maximize the expected breeding value of the next generation while keeping the average inbreeding coefficient increment within 0.3% per generation. Finally, a genetic algorithm was used to optimize the specific mating combinations to balance genetic gain and genetic diversity.

[0043] During the progeny performance validation phase, a 300-pig test population was established. In addition to routine growth trait measurements, 100 pigs underwent a detailed meat quality assessment post-slaughter, including pH, meat color, and drip loss. Blood samples were also collected from these animals at different growth stages (30 kg, 60 kg, and 100 kg) for metabolomic and transcriptomic analyses.

[0044] Furthermore, this example incorporates environmental data. Environmental parameters such as the breeding farm's daily average temperature, humidity, and stocking density are recorded. This information is incorporated into the genetic evaluation model to construct a genotype-by-environment interaction (G×E) model. This model can assess the performance of different genotypes in different environments, helping to cultivate more adaptable strains.

[0045] In addition, deep learning algorithms (such as convolutional neural networks) are used to analyze whole-genome sequencing data to discover complex genomic features that are difficult to capture with traditional methods, such as long-range regulatory elements and structural variations. In this example, this approach discovered several novel regulatory elements that may be associated with growth and meat quality traits.

[0046] Through these optimizations and innovations, the accuracy of breeding value prediction in this example increased to 0.72, a significant improvement over Example 1. Furthermore, while maintaining a similar rate of genetic progression, the average inbreeding coefficient increment in this example was reduced to 0.25% per generation, demonstrating that this method improves breeding efficiency while better maintaining the genetic diversity of the population.

[0047] The multi-omics data (genome, transcriptome, metabolome) integrated in this example not only improves the accuracy of breeding value prediction, but also provides a new perspective for revealing the molecular regulatory mechanisms of growth and meat quality traits.

[0048] Example 3 This example provides a precision breeding method based on single nucleotide polymorphisms for the coordinated improvement of growth and reproductive traits in Landrace pigs. This method further expands and innovates on the basis of the previous two examples and specifically includes the following steps: First, the Illumina NovaSeq X Plus sequencing platform was used to perform whole-genome sequencing on 2,000 Landrace pigs with significant differences in growth and reproductive traits. Sequencing depth reached 40X, with a coverage of 99.9%. This ultra-high sequencing depth and coverage strategy not only captures low-frequency variants but also identifies structural and copy number variations (CNVs), enabling a comprehensive understanding of the genetic basis of complex traits.

[0049] In the variant detection phase, in addition to SNP calling using GATK (v4.3.0), structural variants were detected using Manta and LUMPY software, and copy number variants were detected using CNVnator. Quality control standards were further improved: sequencing quality value Q>35, sequencing depth DP>15, deletion rate <2%, and minor allele frequency (MAF)>0.02.

[0050] GWAS analyses were performed using FaST-LMM software, which is capable of processing very large datasets and accounting for population hierarchical structure. GWAS analyses were performed for growth traits (daily gain, feed-to-meat ratio, backfat depth, and loin eye area) and reproductive traits (total litter size, number of piglets born alive, and litter weight at 21 days of age). The significance threshold was set at P < 5 × 10^-9. Based on the GWAS results, 25 significantly associated genetic variants (including single-nucleotide polymorphisms, indels, and structural variants) were selected for each target trait, resulting in a total of 175 genetic markers.

[0051] During the genotyping phase, a custom array containing 600 genetic markers, including SNPs, indels, and CNVs, was designed. DNA was extracted using a magnetic bead method, requiring a DNA concentration of 100 ng / μL and an OD260 / 280 ratio between 1.85 and 1.95. During genotyping, each batch contained 576 samples, with six replicates used as quality controls. Data were processed using proprietary genotyping software, with marker call rate thresholds set at 99% and sample call rate thresholds at 98%.

[0052] When constructing the genetic evaluation model, a multi-trait random regression model was used, simultaneously considering growth and reproductive traits. For growth traits, growth curve analysis was introduced, using the Gompertz model to fit individual growth curves, with the curve parameters used as new phenotypes. For reproductive traits, three consecutive parities of sows were considered to estimate productive lifespan.

[0053] During data preprocessing, R software (v4.2.0) and Python (v3.9) were used for data cleaning and standardization. The Isolation Forest algorithm was used to detect multidimensional outliers, and Gaussian Process Regression was used to impute missing data. This method is more capable of handling complex multidimensional data structures.

[0054] This example uses the multi-omics integrated single-step approach (MOssBLUP) for genetic evaluation. The model equation is: y = Xb + Za + Wpe + Tv + e, where v represents the non-additive genetic effects (including dominance and epistatic effects). The independently developed HBLUP software package is used to estimate variance components and predict breeding values. This model considers not only additive genetic effects but also non-additive genetic effects, facilitating a more comprehensive assessment of individual genetic potential.

[0055] In the breeding value prediction stage, an ensemble learning approach is used, combining the prediction results of three models: GBLUP, BayesR, and RKHS (Reproducing Kernel Hilbert Space). The weights of each model are dynamically adjusted through cross-validation. When calculating the composite breeding value, the constraint index method is used to set the improvement range of growth and reproductive traits, and then solve for the optimal weight: SI = w1 × GEBV daily weight gain + w2 × GEBV feed to meat ratio + w3 × GEBV backfat depth + w4 × GEBV loin eye area + w5 × GEBV total number of piglets + w6 × GEBV litter weight at 21 days.

[0056] To design the mating strategy, a multi-objective optimization algorithm was developed that simultaneously considers genetic gain, maintenance of genetic diversity, and control of genetic defects. First, boars ranked in the top 3% and sows ranked in the top 12% in terms of overall breeding value were selected as candidate parents. A modified particle swarm optimization (PSO) algorithm was then used to design the mating strategy, with the goal of maximizing the expected breeding value of the next generation while simultaneously controlling the average inbreeding coefficient increment to less than 0.2% per generation and minimizing the risk of transmission of known genetic defect alleles.

[0057] During the progeny performance verification phase, a 400-head testing population was established. In addition to conventional growth and reproductive trait testing, the following innovations were also implemented: 1. Phenotyping throughout the entire life cycle: From birth to culling, regular recording of growth, feeding, reproduction, and health-related indicators.

[0058] 2. High-throughput phenotypic analysis: Using computer vision and machine learning techniques to automatically collect and analyze complex phenotypic data such as body shape and behavior.

[0059] 3. Multi-omics data integration: Whole genome sequencing, whole transcriptome sequencing, metabolome and proteome analysis were performed on 100 representative individuals to construct a multi-level molecular network.

[0060] Furthermore, this example integrates microbiome data. Fecal and intestinal samples were collected at different growth stages (7 days, 30 kg, and 100 kg) and subjected to 16S rRNA and metagenomic sequencing. By integrating host genotype and gut microbiome data, a host-microbe interaction network was constructed, and this information was incorporated into the breeding value prediction model.

[0061] In addition, epigenetic analysis was introduced. Some marker sites related to DNA methylation were included in the chip design, and whole-genome bisulfite sequencing (WGBS) and ATAC-seq analysis were performed on some individuals to capture the impact of epigenetic variation on traits.

[0062] Through these multi-dimensional innovations, the accuracy of breeding value prediction in this example reached 0.78, while simultaneously achieving synergistic improvements in growth and reproductive traits. After five generations of selective breeding, daily weight gain increased by 12%, feed-to-meat ratio improved by 8%, total piglet size increased by 0.8, and litter weight at 21 days increased by 5%, while the average inbreeding coefficient increased by only 0.8%.

[0063] Example 4 This example provides a single nucleotide polymorphism-based precision breeding method for comprehensive improvement of growth, meat quality, and health traits in commercial pigs derived from the Du-Chang-Dou hybrid. This method combines the advantages of the previous three examples and introduces some cutting-edge technologies. Specifically, it includes the following steps: First, 2,500 Duchang-Changda hybrid pigs, each with significant trait differences, were selected for whole-genome sequencing using the Illumina NovaSeq X Plus and Oxford Nanopore PromethION platforms. The combined sequencing depth reached 50X for short reads and 30X for long reads, achieving 99.95% genome coverage. This dual-platform sequencing strategy not only accurately detects single-nucleotide polymorphisms (SNPs) and small indels, but also identifies complex structural variants and difficult-to-assemble genomic regions.

[0064] For variant detection, DeepVariant (v1.4) was used for SNP and small indel detection, the SVIM and Sniffles2 ensemble approach was used for structural variation detection, and Para-HapCUT2 was used for haplotype analysis. Quality control standards were further improved: sequencing quality score (Q) > 40, sequencing depth (DP) > 20, deletion rate < 1%, and minor allele frequency (MAF) > 0.01.

[0065] GWAS analyses combined mixed linear models and Bayesian multi-marker association analysis. A comprehensive analysis was conducted on growth traits (daily gain, feed-to-meat ratio, backfat depth, and loin eye area), meat quality traits (intramuscular fat, pH, drip loss, and meat color), and health traits (immune markers and stress resistance). Permutation testing was used to determine significance thresholds. Based on the GWAS results, 30 significantly associated genetic variants were selected for each target trait, resulting in a total of 300 genetic markers.

[0066] During the genotyping phase, a high-density SNP chip containing 1,000 genetic markers was developed. In addition to the target markers, it also included markers for evolutionarily conserved and regulatory regions. DNA was extracted using an automated extraction workstation, requiring a DNA concentration of 150 ng / μL and an OD260 / 280 ratio between 1.88 and 1.92. During genotyping, each batch contained 768 samples, with eight replicates and four HapMap samples used as quality controls. Data were processed using self-developed AI-assisted genotyping software, with the marker call rate threshold set at 99.5% and the sample call rate threshold set at 98.5%.

[0067] When constructing a genetic evaluation model, a multi-trait randomized regression neural network (MT-RRNN) model was used, simultaneously considering growth, meat quality, and health traits. This model captures nonlinear relationships between traits and genotype-environment interactions. For growth traits, a metabolic efficiency index was introduced, comprehensively considering growth rate and feed utilization efficiency. Meat quality traits were non-invasively measured in vivo using near-infrared spectroscopy. For health traits, a comprehensive scoring system based on behavioral and physiological indicators was developed.

[0068] During the data preprocessing phase, we built an end-to-end data processing pipeline using Python (v3.11) and TensorFlow (v2.12), including outlier detection, missing value imputation, feature engineering, and data augmentation. We used a variational autoencoder (VAE) for feature extraction and a generative adversarial network (GAN) for data augmentation to improve the model's generalization capabilities.

[0069] This example uses a multi-task deep learning model for genetic assessment. The model architecture comprises a shared underlying network and trait-specific output layers, enabling simultaneous learning of both common and specific features across multiple related traits. An attention mechanism is also incorporated to capture the differential importance of different genetic markers across traits. Model training utilizes a distributed learning framework, leveraging multi-GPU clusters for accelerated computation.

[0070] In the breeding value prediction phase, an ensemble learning approach was employed, combining a deep learning model, a traditional BLUP model, and a random forest-based model. The weights of each model were dynamically adjusted through Bayesian optimization. When calculating the composite breeding value, an economic simulation model was introduced to dynamically adjust the economic weights of each trait based on market forecasts.

[0071] In the design of breeding strategies, an adaptive breeding strategy based on reinforcement learning was developed. This strategy automatically adjusts selection intensity and mating strategies based on historical breeding data and current population status. A breeding decision model was trained using a Deep Q-Network (DQN), with a reward function that comprehensively considers genetic gain, genetic diversity, production performance, and market demand. This model is able to make optimal selection and mating decisions in each generation, achieving dynamic optimization of long-term breeding goals.

[0072] During the progeny performance verification phase, a 500-head test population was constructed, introducing the following innovations: 1. Fully automated phenotyping system: uses IoT devices and computer vision technology to monitor growth performance, feeding behavior, and health status in real time.

[0073] 2. Multi-omics big data integration: Whole genome sequencing, whole transcriptome sequencing, epigenome sequencing, metabolomics, and proteomics analysis were performed on 200 representative individuals to construct a multi-level molecular regulatory network.

[0074] 3. Detailed monitoring of environmental factors: Record environmental factors such as temperature, humidity, light, and feed composition to build a detailed genotype-environment interaction model.

[0075] Furthermore, this example integrates microbiome and metabolome data. Fecal, intestinal content, and blood samples were collected at seven key growth stages for metagenomic sequencing and metabolomics analysis. By integrating multi-omics data, a host-microbe-metabolite interaction network was constructed, and this information was incorporated into the breeding value prediction model. This approach not only improves prediction accuracy but also provides a theoretical basis for the development of new feed additives and microbial preparations.

[0076] In addition, epigenetic and non-coding RNA analysis has been introduced. Marker sites associated with DNA methylation and histone modification were incorporated into the array design, and whole-genome bisulfite sequencing (WGBS), ChIP-seq, and small RNA sequencing were performed on selected individuals. These data will help reveal the role of epigenetic regulation in complex traits and provide insights into the development of novel breeding strategies.

[0077] To assess the long-term sustainability of breeding programs, a population dynamics model based on individual simulation was developed. This model can simulate the evolution of genetic variation, phenotypic changes, and population structure over multiple generations of selective breeding, helping to predict and optimize long-term breeding results.

[0078] Through these multi-dimensional innovations, the accuracy of breeding value prediction in this example reached 0.85, achieving comprehensive and synergistic improvements in growth, meat quality, and health traits. After eight generations of selective breeding, daily weight gain increased by 15%, feed-to-meat ratio improved by 10%, intramuscular fat content increased by 0.5 percentage points, and immune-related indicators improved by 12%, while the average inbreeding coefficient increased by only 1.2%.

[0079] The adaptive breeding strategy developed in this example demonstrated excellent long-term results. Compared to the traditional fixed selection index approach, this strategy better balances short-term genetic gain and long-term genetic potential, achieving sustained performance improvements while maintaining genetic variation.

[0080] The innovation of this embodiment also lies in its high degree of automation and intelligence. From phenotypic collection to data analysis and breeding decision-making, the entire process is highly automated. In particular, the adaptive breeding strategy based on reinforcement learning can dynamically adjust breeding goals and selection strategies based on changes in market demand and production environment, greatly improving the flexibility and adaptability of breeding plans.

[0081] Overall, this example demonstrates a comprehensive, precise, and efficient modern breeding approach. By integrating multi-omics data, advanced computational models, and intelligent decision-making systems, this method not only significantly improves breeding efficiency but also provides a new technical path for the sustainable development of animal husbandry. This method is expected to be widely applied to other livestock and poultry breeding programs, making significant contributions to ensuring food safety and promoting agricultural modernization.

[0082] Comparative Example 1: Traditional phenotypic selection method This comparative example adopts a traditional selective breeding method based on phenotypic records, which is compared with Example 1. The method comprises the following steps: (1) Routine growth trait measurements were conducted on the Duroc pig herd, including daily weight gain from 30 kg to 100 kg, feed-to-meat ratio, backfat thickness at 100 kg, and loin area.

[0083] (2) Based on the test results, calculate the mean and standard deviation of each trait. Select individuals whose performance of each trait is 1.5 standard deviations above the mean as breeding pigs.

[0084] (3) Random mating is used for breeding, with each boar mating 10-15 sows.

[0085] (4) Perform the same trait measurement on the offspring and repeat steps (2) and (3).

[0086] This method does not utilize any molecular marker information and relies entirely on phenotypic records for selection. After five generations of selection, daily weight gain increased by 5% and feed-to-meat ratio improved by 3%, but improvements in backfat thickness and loin area were not significant. Compared with Example 1, this method's genetic improvement rate was significantly slower, and it was difficult to achieve synergistic improvement of multiple traits.

[0087] Comparative Example 2: Low-density SNP chip method This comparative example uses a low-density SNP chip for genome selection, which is compared with Example 2. The method includes the following steps: (1) A low-density array containing 1000 evenly distributed SNPs was used to genotype the Large White pig breeding population. DNA extraction and genotyping were performed as in Example 2.

[0088] (2) The breeding value was estimated using the GBLUP model, with the model equation: y = Xb + Za + e. The variance components and breeding values were estimated using the ASReml software package.

[0089] (3) Based on the estimated breeding values, the top 10% of boars and the top 20% of sows are selected for breeding.

[0090] (4) Perform the same genotyping and phenotyping on the offspring, repeating steps (2) and (3).

[0091] Although this method incorporates genomic information, the low SNP density makes it difficult to capture causal variants closely related to the target trait. After five generations of breeding, the improvement in growth traits was better than in Comparative Example 1, but worse than in Example 2. In particular, the improvement in meat quality was less pronounced due to the lack of targeted markers. Furthermore, the breeding value prediction accuracy of this method was only 0.45, significantly lower than the 0.72 in Example 2.

[0092] Comparative Example 3: Single Trait GWAS Selection Method This comparative example only performs GWAS analysis and selection on the daily weight gain trait, and compares it with Example 3. The method comprises the following steps: (1) Genotyping of 1,000 Landrace pigs was performed using the Illumina PorcineSNP60 BeadChip.

[0093] (2) GWAS analysis was performed only on daily weight gain using PLINK software and a linear regression model. The significance threshold was set at P < 5 × 10-7.

[0094] (3) Select the top 20 SNPs significantly associated with daily weight gain and design a custom chip for genotyping.

[0095] (4) The Ridge regression BLUP (rrBLUP) model was used to predict breeding values, considering only the daily gain trait.

[0096] (5) Select the top 5% of individuals for breeding based on their predicted breeding values for daily weight gain.

[0097] While this method leverages GWAS information, it focuses solely on a single trait, neglecting inter-trait correlations and potential negative correlations. After five generations of breeding, daily gain increased by 10%, but feed-to-meat ratio decreased slightly and backfat thickness increased. Compared to Example 3, this method achieved comparable improvements in daily gain, but underperformed in other traits and was not conducive to long-term breeding.

[0098] Comparative Example 4: Genomic selection without environmental considerations This comparative example uses a genomic selection method, but does not consider the influence of environmental factors, and is compared with Example 4. The method includes the following steps: (1) The same high-density SNP chip as in Example 4 was used to genotype the Duchangda triple hybrid pigs.

[0099] (2) The multi-trait GBLUP model was used to estimate breeding values, but the model did not include environmental effects. The model equation was: y = Xb + Za + e.

[0100] (3) Based on the estimated breeding values, the truncated selection method was used to select breeding pigs, and the selection intensity was the same as in Example 4.

[0101] (4) Random mating is used for breeding.

[0102] While this method utilizes high-density SNP information and a multi-trait model, it neglects the influence of environmental factors, making it difficult to accurately assess genotype performance in different environments. After five generations of breeding, all traits improved, but the magnitude of improvement was less than that achieved in Example 4. In particular, the improvements in stress resistance and environmental adaptability were less pronounced. Furthermore, the method's prediction accuracy fluctuated significantly across different rearing environments, ranging from 0.55 to 0.75, significantly lower than the consistently high accuracy achieved in Example 4.

[0103] Comparative Example 5: Traditional hybrid breeding method This comparative example uses a traditional hybrid breeding method without using any molecular marker information, and is compared with Examples 3 and 4. The method includes the following steps: (1) Routine growth and reproductive trait measurements were performed on three purebred populations: Duroc, Landrace, and Yorkshire.

[0104] (2) Select and breed within each purebred group, and choose the top 10% of individuals with the best performance in each trait as breeding pigs.

[0105] (3) Crossbreeding the selected Duroc boars with Landrace sows to obtain the F1 generation.

[0106] (4) Crossbreed the F1 generation sows with the selected Yorkshire boars to obtain the three-way hybrid commercial generation.

[0107] (5) Conduct performance tests on the commercial generation and adjust the direction of purebred breeding based on the results.

[0108] While this method exploits hybrid vigor, the lack of precise molecular marker guidance results in low breeding efficiency. After five generations of breeding, the growth performance of the commercial generation improved, but the rate of improvement was significantly slower than in Examples 3 and 4. In particular, the improvement in meat quality and health traits was not significant due to the difficulty in accurately assessing these traits in purebred breeding. Compared with Example 4, this method struggles to achieve synergistic improvements in growth, meat quality, and health.

[0109] Comparative Example 6: Selection method based only on whole genome sequencing This comparative example uses only whole genome sequencing data for selection, without integrating other omics data, and is compared with Example 4. The method includes the following steps: (1) Whole genome sequencing was performed on 500 Duchangda hybrid pigs with a sequencing depth of 30X.

[0110] (2) Use GATK software to detect variants, including SNPs, insertions and deletions, and structural variations.

[0111] (3) The BayesR model is used for genome-wide prediction, taking multiple traits into consideration simultaneously.

[0112] (4) Based on the predicted breeding values, select individuals with the top 5% comprehensive breeding values for breeding.

[0113] (5) Perform the same whole genome sequencing and phenotyping on the offspring, repeating steps (2)-(4).

[0114] While this method leverages comprehensive genomic information, it lacks integration of other omics data, hindering a comprehensive understanding of the molecular regulatory mechanisms underlying traits. After five generations of breeding, improvements were observed in all traits, but the magnitude of improvement was less pronounced than in Example 4. In particular, improvements in complex traits, such as meat quality and health, were not as significant as in Example 4. Furthermore, this method's high computational cost and prolonged breeding cycle hinder rapid response to changing market demands.

[0115] These six comparative examples clearly demonstrate the superiority of the present invention's method. It not only outperforms traditional methods in breeding efficiency and improvement effectiveness (Comparative Examples 1 and 5), but also surpasses simple molecular breeding methods in predictive accuracy and synergistic improvement of multiple traits (Comparative Examples 2, 3, and 6). In particular, the present invention demonstrates significant advantages in integrating multi-omics data, considering environmental factors, and employing intelligent breeding decisions (Comparative Examples 4 and 6). This fully demonstrates the present invention's significant contribution to theoretical innovation and technological breakthroughs, providing a comprehensive, precise, and efficient new approach for modern livestock and poultry breeding.

[0116] Next, the applicant will design a series of test experiments to evaluate the effectiveness of "a precision breeding method based on single nucleotide polymorphisms" and provide detailed test results.

[0117] Experiment 1: Evaluation of the accuracy of breeding value prediction Experimental purpose: To evaluate the accuracy of the breeding value prediction method of the present invention for different traits and compare it with traditional methods.

[0118] Experimental methods: First, 1,000 Duchangda hybrid pigs were selected as the experimental group and randomly divided into a training set (800 pigs) and a validation set (200 pigs). The training set pigs were comprehensively phenotyped and genotyped, including growth traits (daily gain, feed-to-meat ratio, backfat thickness, and loin eye area), meat quality traits (intramuscular fat content, pH value, and drip loss), and health traits (blood biochemical indicators and immune cell counts).

[0119] Next, the training data were analyzed and modeled using the inventive method (Example 4) and comparative methods (Comparative Examples 1, 2, 4, and 6). The inventive method integrated whole-genome sequence data, transcriptome data, metabolome data, and microbiome data, and considered the influence of environmental factors.

[0120] The constructed model is then used to predict breeding values for the validation set individuals. Simultaneously, actual phenotyping is performed on the validation set individuals.

[0121] Finally, the correlation coefficient between the predicted breeding values and the actual phenotypic values was calculated as an indicator of the prediction accuracy.

[0122] Experimental results: Table 1. Prediction accuracy of breeding values for various traits using different methods

[0123] As shown in Table 1, the method of the present invention demonstrated the highest prediction accuracy for all traits. The advantages of the method of the present invention were particularly evident for complex meat quality and health traits. This fully demonstrates the importance of integrating multi-omics data and considering environmental factors.

[0124] Experiment 2: Evaluation of the effect of synergistic improvement of multiple traits Experimental purpose: To evaluate the effectiveness of the method of the present invention in achieving synergistic improvement of multiple economic traits.

[0125] Experimental methods: First, 2,000 Duchang-Changda hybrid pigs were selected as a base population and randomly divided into five breeding groups, each with 400 pigs. Each group was bred for five generations using the method of the present invention (Example 4) and the comparative method (Comparative Examples 1, 3, 5, and 6).

[0126] Second, in each generation, all individuals were comprehensively phenotyped, including growth, meat quality, and health-related indicators. At the same time, the genetic correlation of each trait was calculated.

[0127] Then, after the fifth generation, the genetic progress of each group in different traits and the genetic correlation changes between major traits were calculated.

[0128] Finally, the growth rate of the inbreeding coefficient of each breeding group was calculated to assess the maintenance of genetic diversity.

[0129] Experimental results: Table 2. Multi-trait improvement effects and inbreeding coefficient increases after 5 generations using different breeding methods

[0130] As can be seen in Table 2, the inventive method demonstrates significant advantages in achieving synergistic improvement in multiple traits. Not only does it achieve the greatest improvement across all traits, it also effectively reduces unfavorable genetic correlations (such as the positive correlation between daily gain, feed-to-meat ratio, and backfat depth). Furthermore, the inventive method exhibits the lowest rate of increase in the inbreeding coefficient, which is crucial for maintaining long-term selection potential.

[0131] Experiment 3: Environmental Adaptability Assessment Experimental purpose: To evaluate the performance stability of the strains cultivated by the method of the present invention under different environmental conditions.

[0132] Experimental methods: First, the fifth generation of sows bred by different methods (the method of the present invention and comparative examples 1, 4, and 6) were selected, with 50 sows (10 boars and 40 sows) selected for each group.

[0133] Next, the pigs were reared and bred under three different environmental conditions: a standard pig farm environment (environment A), a high-temperature, high-humidity environment (environment B), and an extensive breeding environment (environment C). Each environment was used for one production cycle (approximately six months).

[0134] Then, the production performance indicators of each group of pigs under different environments were recorded, including growth rate, feed conversion rate, disease incidence, etc. At the same time, blood samples were collected for the measurement of stress-related indicators.

[0135] Finally, the coefficient of variation of performance of each group across different environments was calculated as a measure of environmental adaptability.

[0136] Experimental results: Table 3. Coefficient of variation (CV%) of performance of lines bred by different breeding methods under three environments

[0137] As shown in Table 3, strains bred using the present method exhibit minimal performance fluctuations across different environments, demonstrating the strongest environmental adaptability. This may be due to the fact that this method considers genotype-environment interactions during the selection process and integrates multi-omics data to assess potential stress responses in individuals.

[0138] Experiment 4: Assessment of Long-Term Selection Effects Experimental purpose: To evaluate the effect persistence and side effects of the method of the present invention in the long-term breeding process.

[0139] Experimental methods: First, populations bred by different methods (the method of the present invention and comparative examples 1, 5, and 6) were selected, with 500 heads in each population, and continuous breeding was carried out for 10 generations.

[0140] Secondly, in each generation, in addition to recording conventional production performance indicators, whole genome resequencing (50 representative individuals per generation) was performed to track changes at the genome level. Reproductive performance, lifespan, and some health indicators were also assessed.

[0141] Then, the effective population size (Ne) and genetic diversity indices (such as heterozygosity) were calculated for each generation.

[0142] Finally, a comprehensive phenotypic analysis was performed, including statistics on the incidence of some possible negative effects (such as skeletal deformities, metabolic abnormalities, etc.).

[0143] Experimental results: Table 4. Changes of various indicators in the 10 generations of selection under different breeding methods

[0144] As shown in Table 4, the present method not only maintains the highest rate of genetic improvement during long-term selection, but also exhibits minimal negative impacts. In particular, the method demonstrates significant advantages in maintaining genetic diversity and reducing adverse side effects. This is likely due to its comprehensive multi-omics data integration and intelligent breeding decision-making system, which effectively balances long-term selection potential and individual health while pursuing genetic improvement.

[0145] Based on the results of the above four experiments, we can draw the following conclusions: 1. Prediction accuracy: The method of the present invention significantly improves the accuracy of breeding value prediction, especially for complex traits, by integrating multi-omics data and considering environmental factors.

[0146] 2. Synergistic improvement of multiple traits: This method can effectively achieve the simultaneous improvement of multiple economic traits and reduce unfavorable genetic correlations, which is crucial for efficient production in modern animal husbandry.

[0147] 3. Environmental adaptability: The strains bred using this method exhibit excellent environmental adaptability, which not only helps improve production stability but also provides a possible solution to climate change.

[0148] 4. Long-term sustainability: During the long-term breeding process, this method has demonstrated an excellent balancing act, maintaining both high rates of genetic improvement and effective maintenance of genetic diversity and individual health.

[0149] These results fully demonstrate the innovation and practical value of the method of the present invention. It is particularly worth mentioning that, while achieving the above-mentioned excellent results, this method also exhibits some unexpected technical effects: 1. Synergistic Improvement of Disease Resistance: Although this method primarily selects for growth and meat quality traits, we surprisingly discovered that integrating transcriptome and microbiome data also resulted in lines that exhibited enhanced disease resistance. This may be due to the capture of key regulatory networks related to immune function.

[0150] 2. Significant improvement in feed utilization efficiency: By integrating metabolome data with gut microbiome information, strains bred using this method not only grow faster but also demonstrate unexpectedly improved feed utilization efficiency. This not only reduces production costs but also reduces environmental pollution.

[0151] 3. Synergistic Improvement of Meat Quality and Growth: The negative correlation between growth rate and meat quality, a common problem in traditional breeding, is effectively alleviated with this approach. This is likely due to a deep understanding and precise control of complex gene regulatory networks.

[0152] 4. General enhancement of stress resistance: It was found that the strains bred by this method not only showed excellent production performance, but also generally showed stronger stress resistance.

[0153] The above description is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and changes can be made without departing from the principles of the present invention. These improvements and changes are also considered to be within the scope of protection of the present invention.

Claims

1. A precision breeding method based on single nucleotide polymorphism, characterized in that: The following steps are involved: (1) Perform whole-genome sequencing and screen for SNP markers associated with target traits; (2) designing a genotype detection chip based on the SNP marker and performing genotyping; (3) Construct genetic evaluation models and predict breeding values; (4) designing a breeding strategy based on the breeding values; (5) Verify the performance of offspring and continuously improve breeding strategies.

2. The single nucleotide polymorphism-based precision breeding method according to claim 1, characterized in that: Step (1) includes: Perform whole-genome sequencing on no fewer than 1,000 pigs with significant growth trait differences, with a sequencing depth of no less than 30X and a coverage of no less than 99%; Sequence alignment and SNP calling were performed, and the quality control standards were: sequencing quality value Q>30, sequencing depth DP>10, deletion rate <5%, and minor allele frequency MAF>0.05; Genome-wide association analysis was performed using a linear mixed model, with a significance threshold set at P < 5 × 10^-8; Select SNP markers significantly associated with growth traits, select at least 10 SNPs for each trait, and screen no less than 50 SNP markers in total; The parameters significantly correlated with growth traits include daily weight gain, feed-to-meat ratio, backfat thickness, and loin area.

3. The single nucleotide polymorphism-based precision breeding method according to claim 1, wherein Step (2) includes: Based on the screened SNP markers, an Affymetrix Axiom chip containing 200-500 SNPs was designed; Extract genomic DNA from the ear tissue of the pig to be tested. The DNA concentration must be above 50 ng / μL and the OD260 / 280 ratio must be between 1.8 and 2.

0. The breeding population was genotyped using the SNP chip, with each batch containing at least 384 samples and 3 replicate samples set as quality control; For genotype data analysis, the SNP call rate threshold was set to 97% and the sample call rate threshold was set to 95%.

4. The single nucleotide polymorphism-based precision breeding method according to claim 1, wherein Step (3) includes: Collect growth trait data of the breeding population, perform data cleaning and standardization, remove outliers, and perform Box-Cox transformation on phenotypic data; The genetic evaluation model was constructed using the genomic optimal linear unbiased prediction model combined with the single-step method; Estimate variance components and predict breeding values.

5. The single nucleotide polymorphism-based precision breeding method according to claim 4, characterized in that: The equation of the optimal linear unbiased prediction model of the genome is: y=Xb+Za+e, Where y is the phenotype vector, b is the fixed effect vector, a is the additive genetic effect vector, X and Z are the corresponding design matrices, and e is the residual vector.

6. The single nucleotide polymorphism-based precision breeding method according to claim 1, characterized in that: Step (4) includes: Constructing a genome relationship matrix; Predict the genomic breeding value GEBV of each individual; Calculate the integrated breeding value SI, where SI = w1 × GEBV daily weight gain + w2 × GEBV feed to meat ratio + w3 × GEBV backfat thickness + w4 × GEBV eye muscle area, w1-w4 are the economic weights of each trait; Calculate the inbreeding coefficient between potential parents and set the inbreeding coefficient threshold to 0.0625; implement the optimal contribution selection algorithm and optimize the mating plan.

7. The single nucleotide polymorphism-based precision breeding method according to claim 1, wherein Step (5) includes: Randomly select 2-3 piglets from each mating combination to construct a test group of no less than 200 piglets; The growth traits and genotyping of the test population were measured; Breeding values were re-estimated and compared with parental generations to calculate genetic progress; Calculate the correlation coefficient between the predicted breeding value and the actual performance. If the accuracy is lower than 0.5, review the SNP selection criteria and model parameters.

8. The single nucleotide polymorphism-based precision breeding method according to any one of claims 1 to 7, characterized in that: The following steps are also included: The SNP marker set was updated every 3–5 generations; Model parameters were periodically re-estimated using accumulated phenotypic and genotypic data; Integrate whole genome sequence data, epigenetic markers, and transcriptome data to optimize breeding strategies.

9. The single nucleotide polymorphism-based precision breeding method according to claim 8, characterized in that: The following steps are also included: Environmental covariates were introduced into the genome's optimal linear unbiased prediction model to construct a genotype-environment interaction model; Use machine learning algorithms to dynamically adjust the weights of each trait in the multi-trait selection index; Metabolomic analysis of blood and muscle tissue was performed on some individuals, and metabolite profiles were integrated with SNP genotype and phenotype data.

10. The single nucleotide polymorphism-based precision breeding method according to claim 9, characterized in that: The following steps are also included: Add marker sites related to DNA methylation to the SNP array; Develop a cloud computing-based breeding management system to achieve automation and intelligence in data collection, analysis, and decision-making; Using gene editing technology, we performed site-specific editing of the identified key SNPs in porcine embryonic stem cells cultured in vitro to verify their causal effects on growth traits; Use deep learning algorithms to analyze whole genome sequencing data and discover complex genomic features.

Images