Efficiency of an approximate animal model for maternal and direct genetic effects of litter size in swine

Abstract

A simulated population of 120 sows and 24 boars in service per year was used to compare genetic evaluations for first-parity litter size and selection response when approximate and complete animal models were used. The complete animal model included maternal and direct effects. The approximate model did not account for covariances between maternal breeding values, or between maternal and direct breeding values, but included a random litter effect of birth of sow. After 10 yr of selection, the overall response (maternal plus direct) was 2.94, 1.61, and .77 pigs using the approximate model and 3.11, 1.65, and .67 pigs using the complete model with equal weighting of maternal and direct genetic effects in the aggregated breeding value for genetic correlations between maternal and direct effects of 0, -.5, and -.9, respectively. Higher reduction was obtained in direct genetic response that was 5, 8, and 19% lower using the approximate model than the complete model for genetic correlations of 0, -.5, and -.9, respectively. Use of the approximate model over 10 yr of selection resulted in a bias of estimation of direct genetic trend of .24, .54, and .48 pigs and in bias in estimated overall response (maternal plus direct) of -.49, .67, and 1.28 pigs for genetic correlations between maternal and direct effects of 0, -.5, and -.9, respectively. Bias in environmental trend estimated from year-season effects was of the same magnitude as the bias in estimated overall response but with an opposite sign. With the approximate model, maternal genetic response was due to a correlation between maternal genetic effects and direct EBV, which for an embedded trait such as maternal genetic effects occurred even when there was no correlation between their true genetic effects. Findings suggested that over the short term the approximate model can be used without substantial loss of overall response but the use of the complete model was recommended because of unbiased estimates in genetic and environmental trend and higher direct genetic response.