Haplotype-based genomic selection efficiency over generations

• Ano de defesa: 2019
• Autor(a) principal: Pereira, Helcio Duarte
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Plantas - Melhoramento genético; Genômica; Modelos matemáticos; Genética molecular; Marcadores genéticos; Genética Quantitativa
• Link de acesso: https://locus.ufv.br//handle/123456789/28705

It is expected that genomic selection based on haplotypes makes better use of linkage disequilibrium among markers and QTLs than models based on single markers, and hence, happened a lower decrease in prediction accuracies over time. The objective of this work is to evaluate the efficiency of genomic selection based on haplotypes in a context of advanced generations of prediction and high marker density. We simulated phenotypes and genotypes of 5,000 unrelated individuals (founders) that gave rise to another 5,000 individuals in the next five generations, without selection. A total number of 60,000 SNPs with density of 0.01 cM was adopted in a genome of 10 chromosomes. An additional scenario of lower marker density (6,000 SNPs) was also evaluated from a sample of that whole set. The haplotype length of 10 SNPs was adopted and the haplotypes calling was made within each chromosome after SNP phasing. The trait investigated in predictions, under a GBLUP model, had broad sense heritability of 0.30. The use of lower marker density harmed the prediction accuracies and coincidence of the best individuals predicted. On average the accuracies decreased in 24% for the model based on SNPs and in 65% for the model based on haplotypes. For coincidence those reductions were 18 and 35% for SNPs and haplotypes, respectively. Considering the best scenario, namely, high marker density, the predictions based on single SNP ranged from 0.3891 to 0.4864 while the predictions based on haplotypes ranged from 0.2515 to 0.3886, for the first and the last prediction generations, respectively. For coincidence of the best predicted individuals those values ranged from 0.2466 to 0.3859 for SNP model and from 0.1959 to 0.3349 for haplotype model. In general the accuracies based on haplotypes were 30% lower than based on SNPs. For coincidence this difference was 21% on average. There is reduction, both in prediction accuracy and coincidence, over the generations with the use of SNPs or haplotypes. Our investigation does not support the use of haplotypes for prediction over generations.