Beef cattle

Accounting for Dominance Improves Estimated Breeding Value Accuracy for Age at First Calving in Girolando Cattle

---

Abstract

Dominance effects are often ignored in genetic evaluations due to analytical complexity, their non-directly transmissible nature, and their typically small contribution relative to additive genetic variance. However, traits with relatively low additive variance, such as reproductive traits, particularly in crossbred populations, may present significant dominance effects that should be accounted for to improve Estimated Breeding Value (EBV) accuracy. The objective of this study was to evaluate gains in EBV accuracy for Age of First Calving (AFC) in Girolando cattle by fitting dominance in the model. The dataset was provided by the Girolando Breed Association in Brazil and comprised 11,687 phenotyped and genotyped animals with the Zoetis custom SNP chip ZBN (55K). After quality control, 38,825 markers remained for further analysis. Two genomic best linear unbiased prediction (GBLUP) models were fitted using the BLUPF90 family programs, both including fixed effects of breed composition (Gyr and Holstein), year, and season of calving, and the random effects of contemporary group (CG), formed by herd, year, and season of birth, and the additive genetic effect. In the first model, pedigree-based heterozygosity was fitted as a covariate to account for crossbreeding effects, without modeling dominance. In the second model, dominance deviation (DD) and the inbreeding coefficient were included instead of heterozygosity, allowing for the assessment of non-additive genetic effects. A fivefold cross-validation (80:20 ratio) was applied to evaluate changes in EBV accuracy and phenotype prediction. The proportions of phenotypic variance due to additive and dominance effects were 0.07 and 0.05, respectively, and EBVs from both models were highly correlated (0.94), yet individuals' EBVs changed slightly when DD was included. In fivefold cross-validation, including DD increased EBV accuracy by 12.84% relative to the additive model, whereas gains in phenotype prediction were minor, with only a 0.06% increase in the correlation between observed and predicted phenotypes and a slight RMSE reduction (-0.11%). Fitting DD effectively partitioned additive and dominance effects, removing most dominance variance from the residual, with only minor changes in additive variance that did not affect the heritability estimates, while improving the correlation between EBVs and adjusted phenotypes, as EBVs no longer captured dominance variation. The modest gains in phenotype prediction reflect the small proportion of phenotypic variance explained by both additive and dominance components, combined with the strong influence of CG effects. These results demonstrate that, even for traits with limited dominance variance, accounting for non-additive effects improves the correlation between EBVs and adjusted phenotypes in cross-validation, while phenotypic improvements remain limited due to environmental and fixed-effect contributions. This approach can enhance selection response in crossbred populations for traits exhibiting some level of dominance effects.

Keywords: 2026

How to Cite:

Santos, M., Silva, M., Guimarães, S. & Rosa, G., (2026) “Accounting for Dominance Improves Estimated Breeding Value Accuracy for Age at First Calving in Girolando Cattle”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286071. doi: https://doi.org/10.31274/wcgalp.23846

Rights: 1