Genomic Selection Accuracy Parametric and Nonparametric Statistical Methods with Additive and Dominance Genetic Architectures

In most genomic prediction studies only additive effects will be used in models for estimating genomic breeding values (GEBV). However, dominance genetic effects are an important source of variation for complex traits, considering them into account may improve the accuracy of GEBV. In the present study, performed applying simulated data, the effect of different heritability values (0.1, 0.3 and 0.5) and different values for the proportion of dominance variance to phenotypic variance (0, 0.05 and 0.15) on genomic selection accuracy in parametric (LASSO, A and B Bayes) and non-parametric (RKHS) statistical methods were studied. Correlations between the true and genomic breeding values, as a measure for the accuracy of genomic predictions under different scenarios were calculated using R software. The results of the present study revealed that, under all statistical methods as heritability values increased, the accuracy of genomic predictability increased. Also, as the value of dominance variance to phenotype variance increased genomic accuracy slant was slow under parametric methods but in the non-parametric method accuracy continued to increase. Under non parametric method the average mean square error was more reduced as the ratio of dominance variance to phenotype variance increased. Therefore, it may be concluded that under non- parametric method as the ratio of dominance variance to phenotype variance increased the accuracy of genomic predictions would be more increased than that of under parametric methods.