Exploration des approches pangénomiques en amélioration variétale chez l'orge à six rangs dans l'Est du Canada
|Abstract:||The emergence of high-throughput genotyping and the development of statistical methods linking genotype to phenotype have led to pangenomic approaches, performed on a genome-wide scale, exploitable in plant breeding. First, these approaches were used to examine the association between genotype and phenotype in genome-wide association studies (GWAS) in order to identify quantitative trait loci (QTLs) useful in marker-assisted selection (MAS). More recently, these approaches have been explored in genomic prediction which aims, on the one hand, to identify the most promising crosses (genomic mating), and on the other hand, to identify the most promising lines within a set of progeny (genomic selection). In both cases, these predictions are based on a statistical model linking genotype to phenotype in a training population. These genome-wide approaches offer great potential but are still emerging and many questions remain unanswered in barley. Our study focuses on some of these questions and is divided into four areas of research. Genome-wide approaches require a large number of single nucleotide polymorphism (SNP) markers. Thus, in the first part of this project, we optimized the protocol of genotyping by sequencing (GBS). This part details the entire process, from the preparation of GBS libraries until the production of a high-quality SNP catalog. As an illustration, we generated a catalog of 30,000 SNPs with a broad chromosome distribution and high genotype accuracy. In the second part, using phenotypic and genotypic data from a breeding population, we compared the effectiveness of three GWAS approaches (Single-SNP, Multi-SNP and Haplotype-based) to detect QTLs for important agronomic traits. The Multi-SNP and Haplotype-based approaches identified more QTLs than the Single-SNP approach. The overlap between the approaches was limited, as each approach uncovered a different subset of previously validated QTLs. In the third part we studied the impact of three factors on the accuracy of genomic selection: (1) the performance of different statistical models (including epistasis or not), (2) the number of SNP markers included in the model as well as (3) their localization (genic/non-genic regions). The model that incorporates both the additive and epistatic effects of SNPs showed the best performance even though the differences between the models were modest. With as few as 2K SNP, the accuracy of genomic selection remained comparable to that based on the entire catalog (35K), while a significant decrease in accuracy was observed at 500 SNPs. In most cases, the use of SNPs located in genic regions, even coding regions, did not provide a significant improvement.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||14 December 2019|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.