Investigation of the statistical distribution of the effects of single nucleotide polymorphic markers on the bias caused by pre-selection of animals for genotyping in genomic selection

Abstract

Introduction: The bias in the selection of superior animals and accurate prediction of
hereditary values of animal offspring selected in future generations is one of the important
topics of breeding; therefore, the study of oblique trends in successive generations can be
effective in the method of unbiased selective animals in the future. The statistical
distribution of the effects of single nucleotide markers can be different in different traits and
occupy a range from the normal distribution to the gamma distribution; therefore, the study
of the amount of bias caused by pre-selection for different traits can be different according
to their genetic structure.
The purpose of this study was to investigate the effect of statistical distribution of the effects
of SNPs on the biased process of estimating breeding values resulting from pre-selection of
animals by genomic selection method. The statistical distributions studied included two
distributions of normal and gamma, the biased trend of each of which was studied during
consecutive generations. The bias criterion includes the regression of actual correction
values on the estimated correction values.
Materials & Methods: Initial Simulation and historical population in the form of two
selection scenarios in three different traits with different heritability in 10 consecutive
generations by two selection scenarios of 10% and 50% and the number of three QTL
species were simulated accurately to estimate genomic breeding values using QMSim
software. The need for calculations was analyzed using R software.
Results: The regression of TBVs on their GEBVs in the first generation whose genotyping
was random was about one and unbiased; But by making choices based on superior breeding
values from the second generation onwards it created a bias. As the number of consecutive
selection generations increases so does the amount of bias but the rate of change in this bias
decreased dramatically after the second generation and remained almost constant in the
fourth generation which could be due to a reduction in genetic and phenotypic variance as
a result of continuous selection known as the Bolmer effect.
Conclusion: The results showed that in both statistical distributions, the amount and
intensity of bias and its trend are almost the same and the difference in the statistical
distribution of effects will not cause a difference in the amount and trend of bias. Moreover,
due to the stabilization of the amount of bias in the 4th generation onwards, it is possible to
correct the amount of bias from the 4th generation onwards by using a scale corrector.