Genetic evaluation with uncertain parentage: a comparison of methods

Summary

The most common method for genetic evaluation when parents are unknown is best linear unbiased prediction with genetic groups (BLUP-G). With this method unknown parents are assumed to be unrelated to any other animals in the population. This assumption is unrealistic in most situations. If a finite number of potential parents can be identified and the probabilities of being the true parent can be assigned to these, genetic evaluation can be obtained given the uncertainty of parentage without introducing genetic groups into the model. The correct numerator relationship matrix with uncertain parentage (Ā) is derived. Rules are given to efficiently compute Ā and Ā^-1. Computer simulation was used to compare BLUP-G with BLUP using Ā. The simulated population consisted of ten sires and 200 dams per breeding season. The dams were always known; the sires were unknown for 10% or 30% of the males and 30% of the females. The number of potential sires was three (BLUP-Ā₁ or ten (BLUP-Ā₂), including the true sire in both cases. Equal probabilities were assigned to each potential sire. The increase in response with BLUP-Ā₁ and BLUP-Ā₂ relative to BLUP-G ranged from 4% to 8% in the fifth breeding season. Selection with BLUP-Ā₁ or BLUP-Ā₂ resulted in higher inbreeding, 17% and 12%, respectively, than with BLUP-G. Estimates of response to selection were unbiased with BLUP-Ā₁ and BLUP-Ā₂, but not unbiased with BLUP-G. Mean square error of estimated genetic means and mean prediction error variance were higher with BLUP-G than with blup-Ā₁ or BLUP-Ā₂.