Stiftung Tierärztliche Hochschule Hannover (TiHo)

Implementation of strategies to utilize molecular genetic information for selection in horse and dog breeding

Stock, Kathrin Friederike

Advances in molecular genetics have facilitated large scale genotyping and therewith utilization of molecular genetic information for selection in many species, including horses and dogs. Genotype-based selection strategies have been developed, but implementation conditions need to be studied in a trait- and population-specific manner. In this work a simulation approach was chosen to investigate implementation issues of marker-assisted selection (MAS), using the example of radiographic health of the limbs in the Warmblood horse, and genomic selection, using the example of canine hip dysplasia (CHD) in the German shepherd dog. Simulation parameters were chosen according to the results of previous population and molecular genetic studies to resemble specific data and pedigree structures as close as possible. For the horse, a multi-generation pedigree was simulated under the assumption of a multiple-trait scenario including one continuous trait and four genetically correlated binary traits. Quantitative trait locus (QTL) information referred to one of the binary traits. Genetic analyses were performed on the basis of phenotype or phenotype and genotype information on population samples of different size using Bayesian methods via Gibbs sampling. Implementation of the Gibbs sampler with a proper prior for the genetic relationship matrix was based on comparison of flat and proper prior analyses which revealed that reliable results can only be obtained under a proper prior. Comparison of results from analyses with and without consideration of genotype information showed that estimation of genetic parameters based on phenotypes and genotypes resulted in overestimation of the polygenic heritability of the binary trait with polygenic and QTL determinants (QTL trait), but increased accuracy of additive genetic correlation estimates. Use of information on two subsequent generations of animals was more efficient in increasing effective sample size and reducing bias of parameter estimates than mere increase of the number of informative animals from one generation. Combined use of phenotype and genotype information on parents and offspring will therefore be beneficial for identification of favorable and unfavorable genetic correlations between traits of interest, facilitating the design of successful multiple-trait selection schemes. Selection response with respect to a QTL trait can be maximized by increasing accuracy of genetic evaluation through combined use of phenotype and genotype information and selecting on the basis of polygenic breeding values and genotypes. Comparison of the applied Gibbs sampling approach with the standard methods, BLUP and REML, indicated superiority of estimating genetic parameters and predicting breeding values in mixed linear-threshold models using Gibbs sampling. Higher success in identifying genetically inferior and superior animals justifies the use of this computationally more demanding approach. Comparison of different selection schemes under realistic conditions such as small numbers of animals with trait information, low heritability of the QTL trait, and low frequency of the favorable QTL allele showed clear superiority of MAS even in a multiple-trait selection scenario with just a single QTL of small size being known for one of the traits. According to the simulation results, horse breeding should benefit from utilizing molecular genetic information by implementing MAS. For the dog, a multi-generation pedigree was simulated under realistic distributions of CHD phenotypes and biallelic markers for CHD. Results of an association study with regard to additive and dominance effects were used to derive genomic breeding values. Short-term responses to different strategies of sire selection were studied with special focus on the possible impact of random versus non-random availability of offspring phenotypes for genetic evaluation of sires. Pre-selection of phenotypes interfered with accurate prediction of breeding values. Response to selection based on the biased breeding values was significantly smaller than genomic selection. Given the equal accuracy of genomic breeding values for all individuals and its independence from amount and distribution of phenotype information on relatives, implementation of genomic selection is expected to result in a fast and considerable increase of breeding progress. Comparative study of long-term selection responses illustrated the limitations of phenotypic selection and the advantages of genomic selection. Use of genomic breeding values for parent selection resulted in large genetic gain due to efficient elimination of unfavorable marker alleles, accompanied by fast phenotypic improvement. In the simulation rigid single-trait selection against CHD on the basis of genomic breeding increased the proportion of dogs free of signs of CHD from initial 61 percent to more than 99 percent within less than ten generations. Although these figures may not be achieved in practical breeding in absence of the idealized selection conditions assumed here, they illustrate the great potential of genomic selection in speeding-up the breeding progress. Together with continued phenotypic monitoring, simulation figures allow definition of time periods after which decline of selection response may require using additional or new markers. Trait- and population-specific simulations were presented as valuable tools to quantify expected benefits from revision of existing breeding programs and introduction of utilization of molecular genetic information in horse and dog breeding.



Citation style:

Stock, Kathrin Friederike: Implementation of strategies to utilize molecular genetic information for selection in horse and dog breeding. Hannover 2008. Tierärztliche Hochschule.


Use and reproduction:
All rights reserved