Author: Dylan Larkin
Publisher:
Published: 2020
Total Pages: 506
ISBN-13:
DOWNLOAD EBOOK →
Genomic selection (GS) is an important tool for increasing genetic gain for economically important traits in breeding programs. Genomic selection uses molecular markers across the entire genome in order to predict the performance of breeding lines for a trait of interest prior to phenotyping. A training population (TP) of elite germplasm, representative of the University of Arkansas wheat breeding program, was developed in order to predict important agronomic and Fusarium head blight (FHB) resistance traits within the University of Arkansas wheat breeding program through cross-validation and forward prediction. A genome-wide association study (GWAS) was performed on the TP to identify novel FHB resistance loci for deoxynivalenol (DON) accumulation, Fusarium damaged kernels (FDK), incidence (INC), and severity (SEV). Significantly loci were used as fixed effects in a GS model (GS+GWAS) and compared to a naïve GS (NGS) model, where the NGS models had significantly higher prediction accuracies (PA) than the GS+GWAS models for all four FHB traits. The GWAS identified novel loci for all four FHB traits, most notably on chromosomes 3BL and 4BL. Multivariate GS (MVGS) models using correlated traits as covariates were also compared to NGS models and the MVGS models significantly outperformed the NGS models for all four traits. The same TP was also evaluated for five agronomic traits, including grain yield (GY), heading date (HD), maturity date (MD), plant height (PH), and test weight (TW), where MVGS models were compared to NGS models. Again, MVGS models significantly outperformed NGS models for all five agronomic traits, especially when there were strong genetic correlations between predicted traits and covariates. Additionally, MVGS models were tested using GY data for genotypes only grown in some environments to predict said genotypes in missing environments. This method significantly improved PA for GY between 6% and 21% for four of six tested environments. The abovementioned TP was then used for forward prediction to predict GY for untested F4:6 breeding lines and DON, FDK, and SEV for F4:7 breeding lines. The MVGS models were comparable to phenotypic selection and had higher selection accuracies for two of three breeding cycles for GY, both cycles for DON, and at least one cycle for FDK and SEV. The MVGS model also had higher PAs for all four traits compared with the NGS models. These results show that GS, and MVGS, can be successfully implemented in a wheat breeding program over multiple breeding cycles and can be effective alongside phenotypic selection for economically important traits. The MVGS models are particularly effective when predicted traits share strong genetic correlations with covariate traits, and covariate traits have a higher heritability than the predicted traits.