Rapid and accurate genome prediction method and equipment based on heritability model

Abstract

The invention belongs to the field of animal and plant breeding prediction, and discloses a rapid and accurate genome prediction method based on a heritability model, which comprises the following steps: calculating the frequency and linkage imbalance score of a marked secondary allele; refining the SNP heritability model to obtain an optimal SNP heritability model; selecting an optimal layering number; acquiring layer genotype data formed by genotype data of all individuals of the to-be-predicted data set in corresponding layers, calculating a diagonal weight matrix of each layer of SNP estimated heritability assignment to the corresponding layer based on the optimal SNP heritability model, and calculating a genetic relationship matrix between individuals of the layer genotype data corresponding to each layer; and fitting the multi-random hybrid model to obtain the genome estimation breeding value of each phenotype of all individuals in the data set to be tested. The method is based on a linear model framework, the calculation complexity is low, the key steps can be subjected to parallel processing, and the calculation speed is high; according to the method, the character specific SNP heritability model is constructed, and the model representation of a complex character genetic structure is optimized, so that the prediction performance is improved.

Description

Technical Field

[0001] The present invention belongs to the field of animal and plant breeding prediction, and in particular relates to a fast and accurate genome prediction method and device based on a heritability model. Background Art

[0002] With the widespread application of high-density single nucleotide polymorphism (SNP) genotyping technology covering the entire genome, whole genome prediction technology has gradually become the mainstream means of evaluating and predicting individual genetic value in the field of animal and plant breeding. Currently, the models used in whole-genome selection are mainly divided into two categories: direct and indirect methods. The representative direct method is the GBLUP method. Under the framework of a mixed linear model, it constructs an individual relationship matrix based on genotype information, estimates variance components using the maximum likelihood method, and then solves the model to obtain individual breeding values. Because the efficiency of model solution is only affected by the number of individuals, GBLUP has low computational complexity and high operational efficiency. However, its model assumption of marker equal contribution is relatively simple, resulting in poor prediction accuracy. The representative indirect method is the Bayesian method. It first constructs a model in a training population, and based on Bayesian prior information theory, uses a hidden Markov iterative process to calculate marker effect values. Then, in a validation population, marker effects are accumulated according to individual genotypes to obtain individual breeding values. The indirect method uses different prior assumptions to optimize the model representation of the genetic structure of the trait. Compared with GBLUP, it has higher prediction accuracy. However, its assumption introduces many unknown parameters, and the iterative process of parameter solution cannot be parallelized, resulting in poor computational efficiency. Summary of the Invention

[0003] The technical problem to be solved by the present invention is that in genomic prediction, the traditional GBLUP method has low prediction accuracy due to the limitation of the marker equal contribution model assumption, and the Bayesian method has low computational efficiency due to the complex parameter solution process. A fast and accurate genomic prediction method based on the heritability model is proposed, which optimizes the model representation of the genetic structure of complex traits based on the trait-specific heritability model, called MAPS (Marker Adjusted Prioritization and Stratification).

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A rapid and accurate genome prediction method based on a heritability model comprises the following steps:

[0006] Step S1, obtain the phenotype file, genotype file, and marker map file, construct a training set, the training set includes the genotype data and phenotype data of each individual, and calculate the minor allele frequency and linkage disequilibrium score of the marker;

[0007] Step S2, constructing a SNP heritability model, and training the SNP heritability model to obtain the optimal SNP heritability model;

[0008] Step S3, based on the optimal SNP heritability model, the estimated heritability of all single nucleotide polymorphisms (SNPs) is calculated and all single nucleotide polymorphisms (SNPs) are divided into K layers, a likelihood ratio test is performed under the K layers to obtain the corresponding significance P value, the K value is traversed within the set value range, and the K value corresponding to the minimum significance P value is selected as the optimal number of layers;

[0009] Step S4, based on the optimal number of stratifications determined in step S3 and the SNPs corresponding to each stratum, obtain the layer genotype data consisting of the genotype data of all individuals in the dataset to be predicted in the corresponding stratum; based on the optimal SNP heritability model determined in step S2, calculate the estimated heritability of the SNPs in each stratum and assign it to the diagonal weight matrix of the corresponding stratum; calculate the kinship matrix between individuals in the layer genotype data corresponding to each stratum; fit a multi-random mixed model to obtain the genomic estimated breeding value of each phenotype of all individuals in the dataset to be tested.

[0010] As described above, in the genotype file in step S1, 0, 1, and 2 are used to represent the three alleles aa, Aa, and AA, respectively. The map file contains the SNP name, the chromosome number where the SNP is located, and the physical position of the SNP on the chromosome.

[0011] As described above, the minor allele frequency in step S1 is calculated based on the following steps:

[0012] Call R software to save the genotype file as a genotype matrix. In the genotype matrix, rows represent individuals, columns represent markers, and elements are genotypes. The elements of each row of the genotype matrix are the genotype data of the corresponding individual. The mean of each column of the genotype matrix is calculated, and half of the column mean is used as the minor allele frequency (MAF) of the marker. If the minor allele frequency (MAF) exceeds 0.5, the corresponding minor allele frequency (MAF) is replaced by 1-MAF.

[0013] The linkage disequilibrium score in step 1 is calculated based on the following formula:

[0014]

[0015] Among them, ξ j is the jth single nucleotide polymorphism SNP j The corresponding linkage disequilibrium score, SNP j The columns of the genotype matrix corresponding to the SNP f The squared Pearson correlation coefficient between the columns of the corresponding genotype matrix, SNP fis the jth single nucleotide polymorphism SNP j Other single nucleotide polymorphisms within 1Mb upstream and downstream of the map file.

[0016] As mentioned above, the SNP heritability model in step S2 is based on the following formula:

[0017]

[0018] in, is the jth single nucleotide polymorphism SNP j The estimated heritability of is the jth single nucleotide polymorphism SNP j The marginal effect estimate of j is the jth single nucleotide polymorphism SNP j The linkage disequilibrium score, p j is the jth single nucleotide polymorphism SNP j The minor allele frequency of , α is a model hyperparameter.

[0019] As described above, the marginal effect estimates in step S2 are obtained based on the following steps:

[0020] The kinship matrix G among all individuals was calculated using the genotype file, and a single random mixed linear model was used to perform genome-wide association analysis on the training set to obtain the single random mixed model log-likelihood value l0 and the marginal effect estimate of the single nucleotide polymorphism (SNP).

[0021] As described above, training the SNP heritability model in step S2 includes:

[0022] Set the model hyperparameter α range, set the model hyperparameter α to the initial value α0,

[0023] Calculate the estimated heritability of all single nucleotide polymorphisms (SNPs), assign the estimated heritability of all single nucleotide polymorphisms (SNPs) to the diagonal weight matrix W, use the genotype file to calculate the new kinship matrix M between all individuals, fit a single random mixed linear model to obtain the log-likelihood value l1, construct the chi-square statistic 2(l1-l0), and perform a likelihood ratio test to obtain the significant P value.

[0024] The model hyperparameter α was updated according to the Brent method. The significance P value calculation was repeated each time until the significance P value reached the minimum. The corresponding model hyperparameter α was the optimal model hyperparameter. The optimal SNP heritability model was constructed based on the optimal model hyperparameter.

[0025] The optimal number of layers in step S3 is determined based on the following steps:

[0026] The K-means clustering algorithm is used to cluster the estimated heritability of all single nucleotide polymorphisms (SNPs) as features, and all single nucleotide polymorphisms (SNPs) are divided into K layers.

[0027] Assign the estimated heritability corresponding to each layer to the diagonal weight matrix of the corresponding layer, and calculate the new kinship matrix between individuals of the layer genotype data corresponding to each layer.

[0028] Using the new kinship matrix, the multi-random mixture model is fitted to obtain the log-likelihood value l K , construct the chi-square statistic 2(l K -l0), perform the likelihood ratio test to obtain the corresponding significance P value,

[0029] K is traversed within the value range, and the significance P value is calculated each time to obtain the significance P value under all values of K. The K with the smallest significance P value is selected as the optimal number of stratifications.

[0030] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned prediction method when executing the computer program.

[0031] A computer-readable storage medium stores a computer program, which implements the steps of the above-mentioned prediction method when executed by a processor.

[0032] A computer program product includes a computer program, which implements the steps of the above prediction method when executed by a processor.

[0033] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0034] 1. The present invention is based on a linear model framework, with low computational complexity, and key steps can be processed in parallel, resulting in fast computational speed. 2. The present invention optimizes the model representation of the genetic structure of complex traits by constructing a trait-specific SNP heritability model, thereby improving prediction performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a comparison chart of the prediction accuracy of the genome best linear unbiased prediction GBLUP and the Bayesian method BayesR and the method described in Example 1 of the present invention (MAPS). DETAILED DESCRIPTION

[0036] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts are within the scope of protection of the present invention.

[0037] Example 1:

[0038] A rapid and accurate genome prediction method based on a heritability model comprises the following steps:

[0039] Step S1, obtain phenotype file, genotype file, and marked map file, marked as single nucleotide polymorphism SNP. Obtain the phenotypic data corresponding to each individual based on the phenotype file, obtain the genotype data of each individual based on the genotype file, and confirm that the individuals with both phenotypic data and genotype data are the total individuals. Construct a training set and a validation set, randomly divide the total individuals into 5 groups, select the phenotypic data and genotype data of 4 groups of individuals as the training set, and the genotype data of the remaining 1 group of individuals as the validation set. The data set to be predicted and the validation set only include the genotype data of the individual, and the phenotypic data of the individual in the prediction data set are unknown. According to the genotype matrix and map file generated by the genotype file, calculate the minor allele frequency (MAF) and linkage disequilibrium score (LDsc) of the marker;

[0040] Step S1-1: Call R software to read the phenotype file, genotype file, and map file. The genotype file must be in numeric format, using 0, 1, and 2 to represent the three alleles aa, Aa, and AA, respectively. Check and adjust the order of individuals in the phenotype and genotype files to ensure consistency. The map file contains three columns of information: the SNP name, the chromosome number where the SNP is located, and the physical location of the SNP on the chromosome.

[0041] Step S1-2, calls R software and the genotype file read by described step S1-1 is saved as genotype matrix with matrix format, and makes row representative individual, column representative mark, element is genotype, the element of every row of genotype matrix is the genotype data of corresponding individual, genotype matrix each column is averaged, using half of the column mean as the MAF (minor allele frequency) of mark, if MAF exceeds 0.5, then corresponding MAF is replaced by 1-MAF, ensure that the scope of MAF is between 0 and 0.5, phenotype file includes the phenotypic data of each individual, builds training set, and training set includes individual phenotypic data and corresponding genotypic data; Build verification set, only include individual genotypic data in verification set;

[0042] Step S1-3, call R software to save the map file read in step S1-1 in data.frame format, calculate the LDsc (linkage disequilibrium score) corresponding to each SNP in combination with the genotype matrix, and use the jth single nucleotide polymorphism SNP as the j For example, the jth single nucleotide polymorphism SNP j The corresponding linkage disequilibrium score ξ j Calculated as: j-th single nucleotide polymorphism SNP j The columns of the genotype matrix corresponding to the SNP j Other SNPs within 1Mb upstream and downstream of the map file f The corrected sum of squares of the Pearson correlation coefficients between the columns of the corresponding genotype matrix is expressed as formula (1), where SNP j The columns of the genotype matrix corresponding to the SNP f The square of the Pearson correlation coefficient between the columns of the corresponding genotype matrix, N is the number of individuals,

[0043]

[0044] In this example, Duroc pig sample data is used as an example to illustrate the practical application of the genome prediction method of the present invention, and the classic GBLUP direct method and BayesR indirect method are used as benchmarks for comparison.

[0045] In step S1, Duroc pig genotype data (Table 1 shows the genotype data table for Duroc pigs), phenotypic data (Table 2 shows the phenotypic data table for Duroc pigs), and map files (Table 3 shows the map files for Duroc pigs) were obtained from a public dataset (http: / / gigadb.org / dataset / 100894). Individuals with both phenotypic and genotypic data were used as the total individuals. Training and validation sets were constructed. The seven phenotypes in Table 2 are back fat thickness (BF) at 100 kg, loin depth (LMD) at 100 kg, lean meat percentage (LMP) at 100 kg, total number of teats (TTN), number of teats on the left side (LTN), number of teats on the right side (RTN), and time per day (TPD) for Duroc pigs.

[0046] Table 1

[0047] ID SNP1 SNP2 SNP3 … SNP258662 0001 0 0 1 … 0 0002 0 0 2 … 2 0003 2 1 0 … 1 … … … … … … 2770 2 1 0 … 2

[0048] Table 2

[0049] ID BF LMD LMP TTN LTN RTN TPD 0001 12.16 47.5 52.44 11 6 5 84.9648 0002 9.63 44.2 54.61 15 8 7 63.8484 0003 11.69 46.2 53.95 12 6 6 86.4532 … … … … … … … … 2770 9.34 46.4 57.98 12 6 6 57.6142

[0050] Table 3

[0051] SNP CHR POS SNP1 1 29039 SNP2 1 41138 SNP3 1 49386 … … … SNP258662 18 55971779

[0052] Step S2, constructing a SNP heritability model, and training the SNP heritability model to obtain the optimal SNP heritability model;

[0053] Step S2-1, based on the phenotypic data and genotypic data of the training set read in step S1, the kinship matrix G between all individuals is calculated using the genotype file based on the VanRaden method, and a genome-wide association study (GWAS) is performed on the training set using a single random mixed linear model (MLM) to obtain the single random mixed model log-likelihood value l0 and the marginal effect estimates β1, β2, ..., β m , m is the total number of SNPs;

[0054] Step S2-2, combining the SNP marginal effect estimate calculated in step S2-1 and the MAF and LDsc calculated in step S1, constructs a SNP heritability model, which is expressed as formula (2), with SNP j For example, is the jth single nucleotide polymorphism SNP j The estimated heritability of is the jth single nucleotide polymorphism SNP j The marginal effect estimate of j is the jth single nucleotide polymorphism SNP j The linkage disequilibrium score LDsc, p j is the jth single nucleotide polymorphism SNP j The minor allele frequency MAF, α is the model hyperparameter;

[0055]

[0056] Step S2-3, set the range of α to [-3, 1], give an initial value α0,

[0057] The Brent method is used to optimize the hyperparameter α in the SNP heritability model, including:

[0058] Step S2-4, calculate the estimated heritability of all SNPs based on formula (2) According to the characteristics of the VanRaden algorithm, the estimated heritability of all single nucleotide polymorphisms (SNPs) is assigned to a diagonal weight matrix W. The new kinship matrix M between all individuals is calculated using the genotype file. A single random mixed linear model is fitted to obtain the log-likelihood value l1. Based on the log-likelihood value l0 of the single random mixed model obtained in step S2-1, a chi-square statistic 2(l1-l0) is constructed. A likelihood ratio test is performed to obtain the significance P value P0.

[0059] Step S2-5, update the model hyperparameter α according to the Brent method, repeat the above calculation in step S2-4 each time until the significance P value reaches a minimum, take the model hyperparameter α at this time as the optimal model hyperparameter, and construct the optimal SNP heritability model based on the optimal model hyperparameter;

[0060] Step S3, based on the calculation results of the heritability model established in S2, the K-means clustering algorithm is used in combination with the likelihood ratio test to optimize marker clustering and achieve ranking and stratification of SNPs;

[0061] Step S3-1, based on the estimated heritability of all single nucleotide polymorphisms (SNPs) obtained based on the optimal SNP heritability model in step S2, the estimated heritability of all single nucleotide polymorphisms (SNPs) is clustered as a feature using the K-means clustering algorithm, and all single nucleotide polymorphisms (SNPs) are divided into K layers, where K is 20 in this embodiment, and the genotype data of all individuals in the corresponding layer constitute the layer genotype data;

[0062] Step S3-2: For each layer of SNPs, the estimated heritability corresponding to the 1st, 2nd, ..., Kth layers is assigned to the diagonal weight matrix W1, W2, ..., W of the corresponding layer according to the VanRaden algorithm. K , respectively calculate the new kinship matrix M1, M2, ..., M between individuals of the genotype data corresponding to the 1st, 2nd, ..., Kth layers K ;

[0063] Step S3-3, using the step S3-2 to obtain a new kinship matrix M1, M2, ..., M K , fitting multiple random mixture models to obtain the log-likelihood value l K , based on the l0 obtained in step S2-1, construct the chi-square statistic 2(l K -l0), perform the likelihood ratio test to obtain the corresponding significance P value P K ;

[0064] Step S3-4, K is traversed in the range of 1 to b, and steps S3-1 to S3-3 are repeated in parallel using multiple threads to obtain the P values P1, P2, ..., P under all values of K. b, select K with the smallest P value as the optimal number of layers;

[0065] Step S4, based on the optimal number of stratifications and the SNPs corresponding to each stratum determined in step S3, obtain the layer genotype data consisting of the genotype data of all individuals in the data set to be predicted in the corresponding stratum; based on the optimal SNP heritability model determined in step S2, calculate the estimated heritability of the SNPs in each stratum and assign it to the diagonal weight matrix of the corresponding stratum; calculate the kinship matrix between individuals in the layer genotype data corresponding to each stratum; fit a multi-random mixed model to obtain the genomic estimated breeding value (GEBV) of each phenotype of all individuals in the data set to be tested; the log-likelihood value and the genomic estimated breeding value are both output results of the multi-random mixed model.

[0066] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods.

[0067] To verify the effect of the present invention, Duroc pig individuals with both phenotypic data and genotypic data were used as the total individuals, and the total individuals were randomly divided into 5 groups. The genotypic data of the individuals in the first group were selected as the validation set, and the phenotypic data and genotypic data of the remaining 4 groups of individuals were used as training sets to obtain the training set and validation set corresponding to one training validation. Similarly, the second group validation set, the third group validation set, the fourth group validation set, the fifth group validation set and the corresponding second group training set, the third group training set, the fourth group training set, and the fifth group training set were selected in turn; the above steps were repeated five times in total to obtain the training sets and validation sets corresponding to 25 training validations.

[0068] For each training and validation run, steps S2 and S3 were performed on the training set. Based on the optimal number of strata and the SNPs corresponding to each stratum determined in step S3, the genotype data of the corresponding validation set in the corresponding stratum were used to construct the stratum genotype data. Based on the optimal SNP heritability model determined in step S2, the estimated heritability of each SNP in each stratum was assigned to the diagonal weight matrix of the corresponding stratum. The kinship matrix between the individuals in the stratum genotype data of each stratum was calculated. A multi-random admixture model was fitted to obtain the genomic estimated breeding value (GEBV) for each phenotype of all individuals in the validation set. The log-likelihood and genomic estimated breeding value were both outputs of the multi-random admixture model. The Pearson correlation coefficient between the genomic estimated breeding value and the actual phenotypic data of each phenotype of all individuals in the validation set was calculated as the prediction accuracy. The prediction accuracy for the same phenotype for all individuals in the validation set from the 25 training and validation runs was averaged.

[0069] The prediction accuracy of the GBLUP method and the BayesR method was tested on the individuals in the validation set of 25 training validations. For the GBLUP method and the BayesR method, the prediction accuracy of the same phenotype of all individuals was averaged. Figure 1 As shown in the figure, compared with the prediction accuracy of the best linear unbiased prediction (GBLUP) and the Bayesian method BayesR, the method described in the present invention (MAPS) achieved good prediction results in the seven traits of Duroc pigs; among them, the prediction accuracy of BF, LMP, LTN, RTN, TTN and TPD traits of Duroc pigs was significantly higher than that of the GBLUP method and the BayesR method, which shows that the method described in this article (MAPS) has good performance in the field of genomic prediction.

[0070] Example 2:

[0071] In this embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0072] Example 3:

[0073] In this embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0074] Example 4:

[0075] In this embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A fast and accurate genome prediction method based on a heritability model, characterized in that: The following steps are involved: Step S1, obtain the phenotype file, genotype file, and marker map file, construct a training set, the training set includes the genotype data and phenotype data of each individual, and calculate the minor allele frequency and linkage disequilibrium score of the marker; Step S2, constructing a SNP heritability model, and training the SNP heritability model to obtain the optimal SNP heritability model; Step S3, based on the optimal SNP heritability model, the estimated heritability of all single nucleotide polymorphisms (SNPs) is calculated and all single nucleotide polymorphisms (SNPs) are divided into K layers, a likelihood ratio test is performed under the K layers to obtain the corresponding significance P value, the K value is traversed within the set value range, and the K value corresponding to the minimum significance P value is selected as the optimal number of layers; Step S4, based on the optimal number of strata determined in step S3 and the SNPs corresponding to each stratum, obtaining the stratum genotype data consisting of the genotype data of all individuals in the corresponding stratum of the dataset to be predicted, calculating the diagonal weight matrix of the estimated heritability of the SNPs in each stratum and assigning it to the corresponding stratum based on the optimal SNP heritability model determined in step S2, and calculating the kinship matrix between individuals in the stratum genotype data corresponding to each stratum; Fit multiple random mixture models to obtain genomic estimated breeding values for each phenotype of all individuals in the test data set.

2. A rapid and accurate genome prediction method based on a heritability model according to claim 1, characterized in that: In the genotype file in step S1, 0, 1, and 2 are used to represent the three alleles aa, Aa, and AA, respectively. The map file contains the SNP name, the chromosome number where the SNP is located, and the physical position of the SNP on the chromosome.

3. A rapid and accurate genome prediction method based on a heritability model according to claim 1, characterized in that: The minor allele frequency in step S1 is calculated based on the following steps: Call R software to save the genotype file as a genotype matrix. In the genotype matrix, rows represent individuals, columns represent markers, and elements are genotypes. The elements of each row of the genotype matrix are the genotype data of the corresponding individual. The mean of each column of the genotype matrix is calculated, and half of the column mean is used as the minor allele frequency (MAF) of the marker. If the minor allele frequency (MAF) exceeds 0.5, the corresponding minor allele frequency (MAF) is replaced by 1-MAF. The linkage disequilibrium score in step 1 is calculated based on the following formula: Among them, ξ j is the jth single nucleotide polymorphism SNP j The corresponding linkage disequilibrium score, SNP j The columns of the genotype matrix corresponding to the SNP f The squared Pearson correlation coefficient between the columns of the corresponding genotype matrix, SNP f is the jth single nucleotide polymorphism SNP j Other single nucleotide polymorphisms within 1Mb upstream and downstream of the map file.

4. The rapid and accurate genome prediction method based on the heritability model according to claim 1, characterized in that: The SNP heritability model in step S2 is based on the following formula: in, is the jth single nucleotide polymorphism SNP j The estimated heritability of is the jth single nucleotide polymorphism SNP j The marginal effect estimate of j is the jth single nucleotide polymorphism SNP j The linkage disequilibrium score, p j is the jth single nucleotide polymorphism SNP j The minor allele frequency of , α is a model hyperparameter.

5. The rapid and accurate genome prediction method based on the heritability model according to claim 1, characterized in that: The marginal effect estimate in step S2 is obtained based on the following steps: The kinship matrix G among all individuals was calculated using the genotype file, and a single random mixed linear model was used to perform genome-wide association analysis on the training set to obtain the single random mixed model log-likelihood value l0 and the marginal effect estimate of the single nucleotide polymorphism (SNP).

6. A rapid and accurate genome prediction method based on a heritability model according to claim 5, characterized in that: The training of the SNP heritability model in step S2 includes: Set the model hyperparameter α range, set the model hyperparameter α to the initial value α0, Calculate the estimated heritability of all single nucleotide polymorphisms (SNPs), assign the estimated heritability of all single nucleotide polymorphisms (SNPs) to the diagonal weight matrix W, use the genotype file to calculate the new kinship matrix M between all individuals, fit a single random mixed linear model to obtain the log-likelihood value l1, construct the chi-square statistic 2(l1-l0), and perform a likelihood ratio test to obtain the significant P value. The model hyperparameter α was updated according to the Brent method. The significance P value calculation was repeated each time until the significance P value reached the minimum. The corresponding model hyperparameter α was the optimal model hyperparameter. The optimal SNP heritability model was constructed based on the optimal model hyperparameter.

7. A rapid and accurate genome prediction method based on a heritability model according to claim 5, characterized in that: The optimal number of layers in step S3 is determined based on the following steps: The K-means clustering algorithm is used to cluster the estimated heritability of all single nucleotide polymorphisms (SNPs) as features, and all single nucleotide polymorphisms (SNPs) are divided into K layers. Assign the estimated heritability corresponding to each layer to the diagonal weight matrix of the corresponding layer, and calculate the new kinship matrix between individuals of the layer genotype data corresponding to each layer. Using the new kinship matrix, the multi-random mixture model is fitted to obtain the log-likelihood value l K , construct the chi-square statistic 2(l K -l0), perform the likelihood ratio test to obtain the corresponding significance P value, K is traversed within the value range, and the significance P value is calculated each time to obtain the significance P value under all values of K. The K with the smallest significance P value is selected as the optimal number of stratifications.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the prediction method according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the prediction method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the prediction method according to any one of claims 1 to 7 are implemented.

Images