Estimation of candidate gene effects in dairy cattle populations.

Standard programs for animal models can be modified to include the effect of a candidate gene, even if only a fraction of the population is genotyped. The elements of the incidence matrix for this effect is 0 or 1 for genotyped individuals and for the probabilities of the candidate gene genotypes for individuals that were not genotyped. The effects of a diallelic candidate gene that were estimated by this method were compared on simulated populations with three alternative estimation methods: analysis of genetic evaluations, yield deviations, and daughter yield deviations, all of which were derived from a standard animal model. The bases of comparison were mean squared error and bias. Four types of experimental designs were considered: genotyping sires only, genotyping cows only, genotyping half of the cows but no sires, and genotyping half of the sires and half of the cows. The estimates that were derived from the three alternative methods all underestimated the simulated effects. The genetic evaluations were more biased than were the yield deviations and the daughter yield deviations. The proposed method was significantly biased only for the design in which half of the daughters, but no sires, were genotyped. Bias and mean squared errors were always lowest by the proposed method.