Detection and use of single gene effects in large animal populations

Unbiased estimation of single gene effects can only be achieved by estimating them simultaneously with other environmental and polygenic effects in mixed inheritance models. As in large animal populations the vast majority of animals are however not genotyped, missing genotypes have to be estimated. Currently used methods as iterative peeling or MCMC are unpractical for large datasets. Recently an alternative method to estimate missing gene content, dened as the number of copies of a particular allele was developed. Unknown gene content is approximated from known genotypes based on the additive relationships between animals. In this study the proposed method was tested for the detection of candidate gene effects for bovine transmembrane GHR on rst lactation milk, fat and protein test-day yields in Holsteins. The GHR gene was estimated to show moderate to small gene substitution effects of 295 g/day for milk, -8.14 g/day for fat yield and -1.83 g/day for protein yield for a phenylalanine replacement by a tyrosine (frequency 23.3%). Only 961 mostly recent sires out of 2,755,041 animals were genotyped. The accuracy of the procedure was then estimated by doing 15 simulations using gene dropping and adjustment of the observed 12,858,741 records using the estimated parameters. The new method to estimate missing gene content resulted to be functional and accurate as relative bias in the estimation of allele frequency was very low (0.2%) as were the biases for moderate allele substitution effects (milk: 3.7%; fat yield: 3.3%). Biases were larger for traits with smaller substitution effects (protein yield: 55.3%). The new method has the potential to allow even in very large animal population with few genotyped animals reliable estimation and use of moderate to large single gene effects.