Faculty of Technical Sciences

TY - ABST

T1 - Impact of genetic and environmental covariance between dam and offspring on prediction of breeding values

AU - Rome, Helene Julie Sophie

AU - Chu, Thinh Tuan

AU - Marois, Danye

AU - Huang, Chyong-Huoy

AU - Madsen, Per

AU - Jensen, Just

PY - 2021

Y1 - 2021

N2 - Improper modeling of maternal effects can cause inflation of predicted breeding values. Adding maternal genetic effect and permanent environmental maternal effect into prediction model in broilers reduced already the inflation. Nevertheless, we hypothesize that including a correlation between the additive genetic effect (a) and the maternal genetic effect (m) but also between the permanent environmental maternal (pe) and residual (e) belonging to the dam into evaluation models could reduce inflation. In our study, we estimated those correlations in broilers and investigated their impact on accuracy and inflation of breeding values. Body weight was recorded in both males and females in order to estimate the environmental covariance between dam and offspring in males, variance components and breeding values were estimated with bivariate models, where BW in males and BW females were considered as two different traits. Four models were tested: a basic model without covariance (Basic), a model including the genetic covariance (Coram), a model including the environmental covariance (Corpee) and a model including both covariance (Corampee). The correlation between a and m was found to be negative (from -0.27 to -0.37) whereas the correlation between pee and e was positive (from 0.26 to 0.32). Adding genetic and/or environmental covariance reduced the inflation of breeding values. Using simulation of similar models, we show that the models accounting for genetic and environmental correlations between dam and offspring increased prediction accuracy. However, standard cross-validation strategies lead to wrong choice of models. When accuracy was computed as the correlation between true simulated breeding value and estimated breeding value a strong gain in accuracy were observed, whereas no gain in accuracy was observed when the accuracy was computed using Legarra and Reverter regression methods due to biases in both predicted breeding values and corrected phenotypes. So adding the genetic and environmental covariance might improve the realized genetic gain while controlling inflation of breeding value in broilers.

AB - Improper modeling of maternal effects can cause inflation of predicted breeding values. Adding maternal genetic effect and permanent environmental maternal effect into prediction model in broilers reduced already the inflation. Nevertheless, we hypothesize that including a correlation between the additive genetic effect (a) and the maternal genetic effect (m) but also between the permanent environmental maternal (pe) and residual (e) belonging to the dam into evaluation models could reduce inflation. In our study, we estimated those correlations in broilers and investigated their impact on accuracy and inflation of breeding values. Body weight was recorded in both males and females in order to estimate the environmental covariance between dam and offspring in males, variance components and breeding values were estimated with bivariate models, where BW in males and BW females were considered as two different traits. Four models were tested: a basic model without covariance (Basic), a model including the genetic covariance (Coram), a model including the environmental covariance (Corpee) and a model including both covariance (Corampee). The correlation between a and m was found to be negative (from -0.27 to -0.37) whereas the correlation between pee and e was positive (from 0.26 to 0.32). Adding genetic and/or environmental covariance reduced the inflation of breeding values. Using simulation of similar models, we show that the models accounting for genetic and environmental correlations between dam and offspring increased prediction accuracy. However, standard cross-validation strategies lead to wrong choice of models. When accuracy was computed as the correlation between true simulated breeding value and estimated breeding value a strong gain in accuracy were observed, whereas no gain in accuracy was observed when the accuracy was computed using Legarra and Reverter regression methods due to biases in both predicted breeding values and corrected phenotypes. So adding the genetic and environmental covariance might improve the realized genetic gain while controlling inflation of breeding value in broilers.

M3 - Conference abstract for conference

T2 - 72nd Annual Meeting of the European Federation of Animal Science

Y2 - 29 August 2021 through 3 September 2021

ER -