Marker and haplotype-based association mapping and genomic prediction to unravel the complex genetic architecture of grain yield and yield stability in spring bread wheat
Grain yield is the most important economic trait in wheat breeding. The detailed understanding of the genetic architecture of grain yield is crucial and the determining factor to optimize genomics-assisted selection strategies in wheat. First, we performed a marker and haplotype-based genome-wide association study (GWAS) for grain yield (GY) and yield stability coefficient (Pi) on 4,302 advanced breeding lines from five CIMMYT international bread wheat trails grown in multiple (optimally irrigated and stress) environments. All lines were genotyped using genotyping-by-sequencing. A haplotype map was built based on linkage disequilibrium between markers. Twenty-nine markers and 16 haplotypes were associated with GY and Pi across two and three germplasm trials with allelic effects ranging from 2 to 11% across environments. Secondly, we performed genomic prediction, testing eight different prediction models incorporating single markers (base model), haplotypes, epistatic interactions, and significant markers/haplotypes identified in GWAS. Initial results show that by including haplotypes and epistatic interactions among haplotypes (main effect and genome-wide), prediction accuracies range between 0.33-0.49 for GY, a 3 to 22.5% improvement over the base model. Despite the identification of significant marker/haplotype trait associations across traits and environments in GWAS, accounting for these markers in genomic prediction does not improve the prediction models. Our results suggest that the haplotype-based approach can increase prediction ability, but that the knowledge of the genetic architecture of grain yield might not have significant consequence on genomic-assisted selection.