Environment characterization and genomic prediction for end-use quality traits in soft white winter wheat

Authors: AOUN, MERIEM - Washington State University; CARTER, ARRON - Washington State University; Thompson, Yvonne; Ward, Brian; Morris, Craig

Submitted to: The Plant Genome
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/8/2021
Publication Date: 12/6/2021
Citation: Aoun, M., Carter, A., Thompson, Y.A., Ward, B.P., Morris, C.F. 2021. Environment characterization and genomic prediction for end-use quality traits in soft white winter wheat. The Plant Genome. 14(3). Article e20128. https://doi.org/10.1002/tpg2.20128.
DOI: https://doi.org/10.1002/tpg2.20128

Interpretive Summary: Soft white wheat produced in the Northwestern region of the United States has some of the highest quality in the world. To meet the requirements of grain markets, millers, and bakers, maintaining or improving end-use quality is a major component of wheat breeding programs. Kernel hardness (kernel texture), flour water absorption, protein (gluten) strength, and milling quality differentiate between hard wheat and soft wheat. Compared to soft wheat flours, hard wheat flours have larger particle size and higher damaged starch, stronger gluten strength, and higher levels of non-starch polysaccharides, and that result in increasedhigher water absorption capacity. Thus, hard wheat is generally used for making bread, whereas soft wheat is used for making cookies, cakes, pastries, crackers, and Asian-style noodles. In this study, we tested the use of FA analytic model and environment clustering based on genomic prediction to understand the pattern of G×E interaction for end-use quality traits. To test these approaches, we used historical multi-environment data on 14 end-used quality traits from the Washington State University (WSU) soft white winter wheat breeding program. Using these datasets, the objectives of this study were 1) understand and compare the pattern of G×E interaction in 14 soft white wheat end-use quality traits, 2) identify possible environment clustering or outlier environments, 3) test the utility of genomic selection for end-use quality traits, and 4) test the effects of excluding outlier environments on genomic prediction accuracies for end-use quality traits.

Technical Abstract: Maintaining or improving end-use quality of wheat traits is crucial to meeting the requirements of grain markets, millers, and bakers. However, end-use quality phenotyping is laborious and expensive thus, not testing may not occur until later generations in ed in the wheat breeding programs until late generations. Understanding the pattern of genotype-by-environment (G×E) interaction is important to optimize breeding for end-use quality traits. Similarly, genomic selection can be implemented to enable more effective breeding for end-use quality. We used a multi-environment unbalanced dataset comprised of 672 soft white winter wheat breeding lines and cultivars adapted to the Pacific Northwest region of the United States which were evaluated for 14 end-use quality traits. Genetic correlations between environments based on factor analytic (FA) models showed low-to-moderate G×E interaction for most traits but high G×E interaction especially for grain protein (WPROT) and flour protein (FPROT). Environment clustering analysis based on genomic prediction approach showed no significant environment clustering for most traits, whereas clustering based on location was observed for kernel size (SKSIZE), break flour yield, flour yield (FYELD), milling score (MSCOR), flour swelling volume (FSV), and cookie diameter (CODI). Genomic prediction accuracies were high for most traits thereby justifying the use of genomic selection for soft white wheat end-use quality. Excluding outlier environments based on genetic correlations between environments was more effective in increasing the genomic prediction accuracies compared to that using environment clustering analysis. For SKSIZE, kernel weight, MSCOR, flour ash, and FSV, excluding outlier environments resulted in increased genomic prediction accuracies by 1-11%. However, for WPROT, FYELD, FPROT, and CODI, excluding outlier environments was not effective in increasing genomic prediction accuracies. This study showed the effectiveness of FA model in studying G×E interaction in multi-environment, multi-trait, and unbalanced data. Excluding outlier environments increased accuracies for traits with low-to-moderate G×E interaction but not for traits with high G×E interaction.