Interpretive Summary: Modeling genetic differences of combined broiler chicken populations in single-step GBLUP
By Anne Zinn
The introduction of animals from a different environment or population is a common practice in livestock breeding, done to increase the genetic performance or to reduce the inbreeding in the targeted population. A study recently published in the Journal of Animal Science modeled the inclusion of a group of external birds into a local broiler chicken population, using either a fixed effect representing the origin of parents, unknown parent groups, or metafounders for the purpose of genomic evaluations. Additionally, research determined whether the predictive ability or LR method gave a more reliable estimate of the accuracy of genomic estimated breeding value.
In order to obtain accurate, unbiased information, the inclusion of animals from an external group into local genetic evaluations need to be correctly modeled; this can be done by including an extra fixed effect. For the majority of the traits in this study, the best scenario included an extra fixed effect and an extra adjustment as unknown parent groups with estimated inbreeding or metafounders. However, results demonstrated that predictions can be biased when the group to define the metafounders greatly differ in the number of animals, so it may be better to use unknown parent groups in such a situation, but further research is needed. Additionally, the unknown parent groups without estimated inbreeding did not converge.
Overall, the use of metafounders reduced the condition number of the system, resulting in faster convergence for both models, but this result cannot be generalized to other models and datasets. The estimation of accuracy using the LR model is less sensitive to the fixed effects of the model than predictive ability, which means it could be used as a proper method to estimate accuracy of genomic estimated breeding values, independently of model specification.
The full paper can be found on the Journal of Animal science webpage.