Estimation and exploitation of linkage disequilibrium in pigs.

摘要

The United States Pork Industry is an important source of income in rural America, and its continued profitability and success can be facilitated through genetic improvement for a variety of production and health traits. Prediction of genomic breeding values (GEBV) based on high density genotypes has the potential to increase genetic progress. The overall objective of this dissertation was to describe the structure of linkage disequilibrium (LD) across the pig genome, assess the potential of genotype imputation from low to high density genotypes, and estimate accuracy of genomic prediction in pure-bred pig populations using either observed or imputed high density genotypes.;The first study focused on the estimation of LD and pairwise persistence of phase across the genome of four US pig populations. Observed LD was high between adjacent SNP (0.36-0.46) and persisted at high levels as pairwise distance between SNP increased to 1 Mb (0.20-0.25). Persistence of phase is a measure of prediction reliability of markers in one population by those in another and ranged between 0.87 and 0.92 for pairwise SNP distance <10 kb. We concluded that high estimates of LD between adjacent SNP in this study are promising for the implementation of genomic selection, especially in conjunction with genotype imputation to increase cost efficiency. However, persistence of phase appears to be too low to indicate that the use of combined training panels would be advantageous for accuracy of genomic prediction at the current marker density.;The second study focused on the accuracy of genotype imputation and variables affecting imputation accuracy. Using a commercially available 10K tagSNP panel and a small reference panel of 128 haplotypes average accuracy of imputation was 0.95. Increasing the size of the haplotype reference panel led to an overall increase in imputation accuracy (IA=0.97 with 512 haplotypes), but was especially useful in increasing imputation accuracy of SNP with MAF below 0.1 and for SNP located in the chromosomal extremes. In addition, our results show that randomly sampling individuals to genotype for the construction of a reference haplotype panel is more cost efficient than specifically sampling older animals or trios with no observed loss in imputation accuracy. From these results, we expected that losses in accuracy of genomic prediction using imputed genotypes would be minimal.;In the third study we assessed the loss of prediction accuracy of GEBV obtained for Yorkshire pigs using imputed instead of observed genotypes. Accuracy of genomic evaluation using observed genotypes was high for three traits (0.65-0.68). Using genotypes imputed with high accuracy (R2=0.95) for genomic evaluation did not significantly decrease accuracy of prediction. The decrease in accuracy of genomic evaluation was significant when imputation accuracy dropped to R2=0.88. Genomic evaluation based on imputed genotypes in selection candidates is a cost efficient alternative for implementation of genomic selection in pigs. Furthermore, genotyping animals at lower cost and low density, followed by imputation, can result in increased accuracy by allowing more animals into the training panel.;In conclusion, we showed that accurate prediction of GEBV in a US Yorkshire population is possible, and cost efficiency can be increased through the use of genotype imputation in selection candidates. Furthermore, our results of LD for three other US pig populations indicate that similar or high accuracy of prediction can be expected within each of these populations. In addition, we briefly discuss how our results can be extended to prediction of breed composition, and GEBV prediction and GWAS using whole genome sequence.