Linkage disequilibrium and genomic selection in pigs

• Ano de defesa: 2015
• Autor(a) principal: Veroneze, Renata
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Suíno - Genética; Genética quantitativa; Marcadores genéticos; Genética e Melhoramento dos Animais Domésticos
• Link de acesso: http://www.locus.ufv.br/handle/123456789/7597

Genomic selection and genomic wide association studies (GWAS) are widely used methods that aim to exploit the linkage disequilibrium (LD) between markers and quantitative trait loci (QTL). Securing a sufficiently large set of genotypes and phenotypes can be a limiting factor when implementing genomic selection that may be overcome by combining data from multiple populations or using crossbred information. The overall objective of this thesis was to characterize LD patterns in different pig populations and to evaluate whether the differences in LD determine the accuracy of genomic predictions when using different reference sets (within-, across- and multi- population) and methodologies. In this thesis I used data from pure lines and crossbred pig populations genotyped with PorcineSNP60 BeadChip. Loess regression provided a better fit to the real LD data, and more accurate LD predictions could be made, compared to nonlinear regression. It was also shown that Loess regression can be used to statistically compare the LD decay of different populations. The persistence of LD phase between crosses and the parental pig lines was found to be high, from which it was hypothesized that similar marker-QTL associations would be found in a cross and in their purebred parent populations and therefore accuracies of genomic prediction across these populations should be high. Between the pure lines the persistence of phase was low, thus higher density panels should be used to have the same marker-QTL associations across these lines. Accuracies obtained from across- and multi-population genomic prediction and from using crossbred data did however not follow the expectations based on LD. Having the same LD phase may therefore not be as important for genomic prediction accuracy as previously thought but rather the interplay between LD, genetic architecture and allele frequencies also plays a major role. Differences in allele frequencies between lines and information from GWAS on the genetic architecture of traits for the different lines were taken into account in analyses developed in the later chapters. The use of weights, based on GWAS results, was expected to lead the GBLUP model towards the real genetic architecture of the traits. This strategy was shown to have some benefit for the genomic predictions with single- and multi-population data sets. Weights obtained from GWAS in different data sets (within and combining populations) did not always lead to increased accuracies of prediction, depending on which lines the weights are applied to. Using weights from GWAS in a combined population was the best approach, resulting in higher accuracy of GBLUP predictions within single- as well as in multi-population analysis. Understanding and evaluating how the accuracy of within-, across- and multi-population genomic prediction is affected by differences in LD, in genetic architecture and in allele frequencies is key to optimize the accuracy of genomic prediction in pig breeding.