عنوان مقاله [English]
Background and objective: The availability of thousands of polymorphic single nucleotides (SNPs) spread across the genomes made it possible to use genome-wide marker information to predict total breeding value in the implementation of genomic selection. The interaction between several gene loci contributes to the phenotypic changes associated with the expression of complex polygenic traits. The present study aimed to compare the performance of different models with additive and non-additive genetic effects, especially epistasis effects in predicting variance components and genomic breeding values as well as QTN in a dairy cow population based on Bayes Lasso model.
Materials and methods: A population of dairy cattle was simulated based on a choice of four pathways to maximize the rate of genetic progress over ten generations with an effective size of 100 individuals in the base population. The genomic structure was assumed consisting of 3 chromosomes with a length of 100 cM and On each chromosome were located 1000 markers 2 allelic markers with a frequency of 0.5. 50 QTL double alleles with equal frequency were randomly assigned to each chromosome. For QTL allelic effects, a gamma distribution with a parameter slope of 0.4 was used in QMSim software and sampling was performed. Phenotypic and genotypic data of the last 10 generations were used for analysis. Marker effects, variance components and genetic parameters were estimated using the Bayes Lasso method in the form of 6 statistical models, that model 1 includes only the marker additive effects, model 2 includes the polygenic effects of animal, model 3 includes the combined effects of additive marker effects and polygenic effects of animal, model 4 includes the dominance effect of markers and polygenic effects of animal, model 5 includes the dominance and additive effects of markers in addition to the polygenic effects of animal, and finally, model 6 includes the additive, dominance and epistatic effects of markers in addition to the polygenic effects of animal. To control the type I error in estimating marker effects, Bonferroni test was used at 1% probability level.
Results: The results of this study showed that the additive model or the polygenic model alone could not show the missing or hidden variance, in such a way that adding non-additive genetic effects including dominance and epistatic effects reduced the residual variance effects. Also in the complete model including additive and non-additive effects, the total genetic variance increased. With the addition of non-additive effects in the statistical model the Hertability in the broad sense was increased and the Hertabilitity in the narrow sense of trait decreased. The accuracy of breeding value estimation in model 1 was the lowest and in models 3 to 6 was the same. Model 5 with the lowest false positive QTN and the highest true positive proportion was the appropriate model for genetic evaluation of the studied traits. By including polygenic and non-additive effects in the model, the number of false positive errors was reduced.
Conclusion: The results of this study showed that the components of additive and dominance variance markers had a small share in the genetic diversity of the studied trait, but the polygenic effects of animal and epistatic had the largest share in the exploring of genetic diversity of trait. The results also showed that the major contribution of gene loci in the exploring of traits is related to additive genes with an additive effect that have a normal distribution. Ignoring non-incremental effects causes inflammation in the additive variance of the trait.