Joint Estimation of Additive and Dominance Effects of Markers Using a Genomic Model with a Residual Polygenic Effect


  • Zengting Liu vit Germany
  • Hatem Alkhoder vit Germany
  • Friederich Reinhardt vit Germany
  • Reinhard Reents vit Germany


genomic model, substitution effect, dominance effect, SNP marker


Routine genotyping of female animals in genomic selection programme opens up the opportunity to estimate non-additive genetic effects of SNP markers by using direct phenotypes of the genotyped cows. Non-additive effects of markers, e.g. dominance effects, provide useful information for genomic mating and herd management. We extended our current BLUP SNP genomic model, which includes a residual polygenic effect, to additionally estimate dominance effects of SNP markers. A new estimation algorithm was implemented to improve the rate of convergence for all estimated effects of the new genomic model. Parallel computing technique was applied to reduce the total clock time of the full estimation process. To study the convergence behaviour of the genomic dominance model, lactation yield deviations of cows and associated effective data contribution were obtained from German routine genetic evaluation for milk production and somatic cell scores based on a random regression test-day model. Genotypic and phenotypic data of a total of 17,635 cows were used to estimate dominance effects and breeding values of the SNP markers together with residual polygenic effects. Dominance variance of 5% and 10% of the total genetic variance were assumed to investigate the influence of dominance variance. Interim solutions of the model effects were compared to the final estimates from a long iteration process with 10,000 rounds. Based on the correlations between the dominance effect and breeding value estimates, it seems that the dominance effects of a large number of SNP markers can be accurately estimated and properly separated from the breeding value effects within a reasonable time frame. Further results on the convergence behaviour of the new genomic model are discussed.