Using genomic information to increase accuracy of predicting breeding values is the new standard in animal breeding and genetics. Although genotyping livestock animals is becoming cheaper, it is still not affordable for some populations. One of the biggest challenges before the implementation of genomic selection is to decide which and how many animals should be genotyped. Given that the answer depends on several population structure parameters, simulation emerges as a cost-effective alternative to test different scenarios that will provide the greatest genetic gain given the genotyping structure and artificial selection scheme in horse populations. This proposal is an important part of a project under development in Brazil (Process 2018/26465-3). The overall objectives are to evaluate how population structure and selection design can impact the genomic predictions, as well as the discovery of quantitative trait loci. Stochastic simulation will be used to create a complex population structure for two artificial selection designs based on: 1) estimated breeding value, and 2) phenotype. After the simulations are done, animals will be selected for genotyping based on seven criteria: 1) high and low estimated breeding value; 2) high and low phenotype; 3) high estimated breeding value; 4) high phenotype; 5) randomly; 6) largest number of progeny; and 7) graph theory to identify families. Four traits will be simulated to represent the selection goal of Campolina horse breeders: gait and morphology. Predictive ability will be assessed by a validation method called forward prediction, in which genotyped animals in the most recent generation will be used as validation and the remaining generations as training. The inflation/deflation will be assessed through a simple linear regression of the true breeding value (simulated) on the genomic estimated breeding value. Simulation will be done using QMSim, a well-established software for simulating livestock data. The genomic analysis (i.e. variance components estimation, breeding value estimation, predictive ability, and genome-wide association) will be done using the BLUPF90 family of programs. Population structure will be evaluated by principal component analysis and discriminant principal component analysis of pedigree and genomic relationship matrices. In summary, this project will contribute to the development of genotyping strategies applied not only to horses but to other species, aiming a better use of the public and private funding, and obtaining the most accurate genomic predictions and discovering quantitative trait loci in a more precise way. This very project has not only academic impact but will also help breeders in the difficult task of selecting animals for genotyping when resources are limited.
News published in Agência FAPESP Newsletter about the scholarship: