New fast Bayes method estimating genomic estimated breeding value

Abstract

The invention discloses a new fast Bayes method estimating a genomic estimated breeding value. The method comprises the steps of performing productivity phenotype test and genome sequencing on all individuals, and acquiring SNP sites of a genome; screening out qualified SNP sites, and supplementing missing genotypes; dividing all individuals into a reference group and an estimation group randomly;calculating an expected effect value of each SNP site in a Bayes model of the reference group according to a phenotype value of the reference group and an iterative conditional expectation algorithm;accumulating expected effect values of all SNP sites so as to obtain a genomic estimated breeding value (GEBV) of the estimation group; and calculating a related coefficient of the GEBV and the phenotype value or a true breeding value so as to acquire estimation accuracy. The method has the characteristics of being fast in calculation speed and high in stability, and maintaining the estimation accuracy of the breeding value.

Description

technical field [0001] The invention relates to the field of genome breeding, in particular to a fast Bayesian method for estimating genome breeding values. Background technique [0002] The traditional breeding value estimation method is called best linear unbiased prediction (Best Linear Unbiased Prediction, BLUP), which mainly estimates individual breeding value based on individual phenotype and pedigree records. However, the traditional breeding method cannot estimate the Mendelian sampling error term contained in the breeding value very well, because the traditional breeding method cannot track the specific information of alleles in the process of generation transmission. With the development of high-throughput sequencing technology, it has become possible to obtain high-density SNP genotype data in various species, and modern breeding is gradually moving towards the field of genome breeding. In 2001, Meuwissen et al. proposed to use genome-wide markers to estimate the...

AI Technical Summary

Problems solved by technology

In 2009, Meuwissen et al. invented a fast Bayesian method that only proposed a priori distribution for the labeling effect. They assumed that the distribution of the labeling effect was a Laplace distribution (or called a double exponential distribution), but this method cannot Ensure that the SNP effect value can converge under any circumstances, especially when the marker density is very large. The reason is that there are many exponential terms in the solution formula, which may cause the problem of numerical out-of-bounds

Images