CN105512510B - A method for assessing heritability from genomic data - Google Patents

Abstract

The invention discloses a kind of methods assessed by genomic data genetic force, for a certain quantitative character, the estimation of the marker effect of full-length genome is carried out using GBLUP algorithm by using the reference group individual of different number, and then the breeding value of estimation group is obtained, and calculate accuracy of estimation；The fitting of curve linearization(-sation) is carried out by genome accuracy of estimation and reference group's size, the inverse of the intercept of the regression equation fitted is the estimated value of genetic force；The present invention assesses the genetic force of quantitative character by the data of genome, the achievement studied may be directly applied in animals and plants quantitative character breeding, algorithm of the invention does not carry out pedigree record to individual but genes of individuals group is sequenced, the genetic force of character is predicted by full-length genome label, genetic force estimated result is mainly used in the breeding work in future, in addition, sequencing can capture Mendelian sampling error, relative record pedigree data can obtain more accurate pedigree information.

Description

A method for assessing heritability from genomic data

technical field

The invention relates to the field of genetic engineering, in particular to a method for evaluating heritability through genomic data.

Background technique

The current heritability assessment methods mainly use the genetic relationship between individuals, and use various statistical methods, such as variance analysis, correlation analysis, etc., to infer. This method requires a complete pedigree record. The workload of recording is very large or even difficult to achieve, such as aquatic animals; in addition, the traditional method of heritability assessment is to treat the genomic information as a "black box", which cannot capture the specific information of genes transmitted from parent to offspring. That is, it is impossible to prepare to capture Mendelian sampling errors, resulting in large estimation errors; in order to solve the problems of heavy pedigree recording workload and inability to accurately capture Mendelian sampling errors in traditional heritability estimation methods, it is necessary to improve the existing technology.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide an algorithm for evaluating heritability through genomic data, which overcomes the problems of large errors and cumbersome pedigree records in traditional heritability estimation, so as to solve the problems raised in the above background art.

The invention does not record the pedigree of the individual, but directly sequences the genomes of all individuals, and estimates the estimation accuracy of the genomic breeding value and then the heritability of the trait by combining the individual performance records and the genomic marker information.

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

An algorithm for evaluating heritability through genomic data. For a quantitative trait, by using different numbers of reference group individuals to estimate the marker effect of the whole genome, the breeding value of the estimated group is obtained, and the estimation accuracy is calculated. The above process is actually the specific process of genome selection. In this invention, GBLUP is used as the algorithm for calculating marker effects. The GBLUP algorithm was invented by Meuwissen et al. in 2001. Its prior distribution considers that the effect variances of all marker sites in the genome are equal. , the marker effect can be calculated by the following formula:

in, is the overall average; is the effect vector of all marker sites; the genome estimated breeding value is obtained by adding up the effects of all marker sites, wherein the genome estimated breeding value is expressed by GEBV, that is, GEBV=∑X _i g _i ; GEBV estimation accuracy is calculated by calculating GEBV The correlation coefficient with the true breeding value, where the true breeding value is represented by TBV, i.e. r _{(GEBV, TBV)} ; meanwhile, Daetwyler et al. in 2008 deduced that in the case of GBLUP algorithm estimating the breeding value, r _{(GEBV, TBV) )} another formula is:

Among them, N _p is the number of individuals in the reference group; h ² is the heritability of the trait under study; M is the number of effective genome segments that determine the trait; however, in actual production, the specific value of TBV cannot be known, so use The phenotype value replaces TBV, where the phenotype value is represented by Y, and the relationship between GEBV and Y is deduced as:

In formula (3), different values of r _{(GEBV, Y)} can be obtained by adjusting the size of N _p , and the curve equation is fitted. Linear equation:

This equation is equivalent to the linear regression model y=a+bx, where y is the reciprocal of the square of r _{(GEBV, Y)} , x is the reciprocal of N _p , and the intercept a of the equation is the reciprocal of heritability. By finding the equation The inverse of the intercept of , yields an estimate of heritability.

As a further scheme of the present invention: all individual genomes are sequenced to obtain SNP information, the SNP sites of all individuals correspond, and the missing data are filled by an imputation method.

As a further scheme of the present invention: in order to prevent the single estimation error from being large, the method of multiple hybridization verification is adopted to repeatedly randomly select the reference population and the estimated population from the population to obtain estimation results close to the true value.

As a further scheme of the present invention: using different reference group numbers combined with GBLUP algorithm to calculate the effect value of each marker in the genome, to obtain the breeding value of the estimated group, and obtain the estimate by performing correlation analysis on the breeding value and phenotype value of the estimated group Accuracy

Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention evaluates the heritability of quantitative traits through genomic data, the research results can be directly applied to the breeding of quantitative traits of animals and plants, and the algorithm of the present invention can be used in On the basis of not establishing a pedigree, the heritability of traits can be predicted through whole genome markers, which solves the problem that pedigree recording is cumbersome or even difficult to achieve, and because sequencing can capture Mendelian sampling errors, the algorithm of the present invention can be compared to recording pedigree data. Get more accurate pedigree information.

Description of drawings

FIG. 1 is an algorithm flow chart of the present invention.

Figure 2 is a trend diagram of the GEBV accuracy of the two traits of body weight and body length in the present invention changing with the size of the reference population.

3 is a trend diagram of the GEBV accuracy and reference population size of the two traits of body weight and body length in the present invention after conversion according to formula 4.

Among them, the value of the abscissa is the reciprocal value of the number of individuals in the reference group; the value of the ordinate is the reciprocal of the square of the GEBV accuracy; R ² is the coefficient of determination of the regression equation.

Detailed ways

The technical solution of the present patent will be described in further detail below in conjunction with specific embodiments.

Please refer to Figures 1-3, an algorithm for assessing heritability through genomic data. For a certain quantitative trait, by using different numbers of reference group individuals to estimate the marker effect of the whole genome, the breeding of the estimated group is obtained. value, and calculate the estimated accuracy; the curve linear fitting is performed through the genome estimated accuracy and the reference population size, and the reciprocal of the intercept of the fitted regression equation is the estimated value of heritability; it is characterized in that: the genome selected The specific process uses GBLUP as the algorithm for calculating the marker effect. The effect variance of all marker sites in the genome is equal, and the marker effect is calculated by the following formula:

in, is the overall average; is the effect vector of all marker sites; the genome estimated breeding value is obtained by adding up the effects of all marker sites, wherein the genome estimated breeding value is expressed by GEBV, that is, GEBV=∑X _i g _i ; GEBV estimation accuracy is calculated by calculating GEBV The correlation coefficient with the real breeding value, where the real breeding value is represented by TBV, that is, r _{(GEBV, TBV) is} obtained; in the case of the GBLUP algorithm estimating the breeding value, another calculation formula of r _{(GEBV, TBV)} is:

Among them, N _p is the number of individuals in the reference group; h ² is the heritability of the trait under study; M is the number of effective genome fragments that determine the trait; in actual production, the specific value of TBV cannot be known, so the table The type value replaces TBV, where the phenotype value is represented by Y, and the relationship between GEBV and Y is deduced as:

In formula (3), different values of r _{(GEBV, Y)} can be obtained by adjusting the size of N _p , and the curve equation is fitted. Linear equation:

This equation is equivalent to the linear regression model y=a+bx, where y is the reciprocal of the square of r _{(GEBV, Y)} , x is the reciprocal of N _p , and the intercept a of the equation is the reciprocal of heritability. By finding the equation The inverse of the intercept of , yields an estimate of heritability.

Sequence the genomes of all individuals to obtain SNP information. The SNP loci of all individuals correspond to each other, and the missing data are filled by the imputation method; in order to prevent the single estimation error from being large, the method of multiple hybridization verification is adopted, and randomization from the population is repeated repeatedly. Extract the reference population and estimated population to obtain estimation results close to the true value; use different numbers of reference populations combined with the GBLUP algorithm to calculate the effect values of each marker in the genome to obtain the breeding value of the estimated population. The phenotypic value is analyzed to obtain the estimated accuracy, which solves the problem of cumbersome or even difficult to complete the pedigree recording, and accurately captures the Mendelian sampling error in the transmission process of alleles.

Example 1

1. The test objects are 500 large yellow croakers, using artificial spawning technology, all large yellow croakers were born on the same day, that is, all of the same age; when the test time is two years of large yellow croakers, the measured traits are the weight and body length of all large yellow croakers .

2. Use GBS (genotyping-by-sequencing) sequencing technology to sequence the genomes of all individuals to be studied, and screen qualified SNP sites. The parameters are controlled as follows: MAF>0.05, Hardy-Weinberg equilibrium test P-value> 0.001, marker loci with a single locus deletion rate lower than 20% were left; a total of 29,748 qualified SNP markers were finally screened, and the missing loci were filled by the interpolation program of Beagle 3.3.

3. Among all 500 individuals, 20%, that is, 100 individuals, are randomly selected as the estimated group, and the rest are divided into four levels according to the number of individuals 100, 200, 300, and 400. Observe that the number of individuals in the reference group at four different levels corresponds to The changing trend of estimation accuracy; using the GBLUP algorithm to estimate all marker effects under each level, obtain the breeding value GEBV of each individual of the estimated population, and obtain the estimated accuracy by calculating the correlation coefficient between the GEBV of the estimated population and the phenotypic value. , namely r _{(GEBV, Y)} .

In order to reduce the influence of excessive single sampling error, step 3 is repeated 20 times. Since the individuals of each estimation group and reference group are randomly sampled, the results of each repetition will be slightly different, but 20 times The average of the results will be closer to the real results, and the results of 20 averages are shown in Figure 2.

4. Take the reciprocal of the reference cluster size (ie, N _p ) for each class, and take the inverse of the square of the average of the 20 results of the estimated accuracy (ie, r _{(GEBV, Y)} ) for each class, and either The relationship between is shown in Figure 3, and the final regression equation is fitted according to formula (4), as shown in the following table:

According to the results in the above table, the estimated heritability of body weight was 0.227 and body length was 0.196.

The preferred embodiments of the present patent have been described in detail above, but the present patent is not limited to the above-mentioned embodiments. Within the scope of knowledge possessed by those of ordinary skill in the art, the present invention can also be made without departing from the purpose of the present patent. Various changes.

Claims (4)

1. A method for estimating heritability through genome data comprises the steps of estimating the marker effect of a whole genome for a certain quantitative character by using different numbers of reference population individuals to further obtain the breeding value of an estimated population, and calculating the estimation accuracy; carrying out curve linear fitting through genome estimation accuracy and reference population size, wherein the reciprocal of the intercept of the fitted regression equation is an estimation value of the heritability; the method is characterized in that: the specific process of genome selection adopts GBLUP as an algorithm for calculating the marker effect, the effect variances of all marker loci of the genome are equal, and the marker effect is calculated by the following formula:

wherein,is the overall average;effect vectors for all marker sites; the estimated genomic breeding value is obtained by adding the effects of all the marker sites, wherein the estimated genomic breeding value is represented by GEBV, i.e., GEBV ═ Σ X_ig_i(ii) a The accuracy of the GEBV estimation is determined by calculating the correlation coefficient between the GEBV and the true breeding value, i.e. r_(GEBV，TBV)Obtaining; in the case of the GBLUP algorithm for estimating the breeding value, where the true breeding value is represented by TBV, r_(GEBV，TBV)Another calculation formula of (a) is:

wherein N is_pIs the number of individuals of the reference population; h is²Heritability for the trait under study; m is the number of effective genomic fragments determining the trait; in actual production, the specific value of TBV is unknown, so TBV is replaced by a tabular value, denoted Y, from which the relationship between GEBV and Y is derived as:

in the formula (3), by adjusting N_pCan obtain different r_(GEBV，Y)Fitting the curve equation, wherein the fitting mode adopts curve linearization, and the formula (3) is arranged to obtain a linear equation:

this equation corresponds to a linear regression model y ═ a + bx, where y is r_(GEBV，Y)X is N_pThe intercept a of the equation is the inverse of the heritability, and the estimate of the heritability is obtained by calculating the inverse of the intercept of the equation.

2. The method of claim 1, wherein the genome of all individuals is sequenced to obtain SNP information, SNP sites of all individuals correspond, and missing data is filled up by interpolation.

3. The method for estimating heritability through genome data according to claim 1, wherein multiple hybridization verifications are used to randomly extract the reference population and the estimated population from the population repeatedly to obtain an estimated result close to the true value, in order to prevent a single estimation error from being large.

4. The method of claim 1, wherein the effect values of each marker in the genome are calculated using different reference population numbers in combination with the GBLUP algorithm to obtain the breeding values of the estimated population, and the accuracy of the estimation is obtained by correlating the breeding values and phenotypic values of the estimated population.