Document Type

Thesis - Open Access

Award Date

2026

Degree Name

Master of Science (MS)

Department / School

Agronomy, Horticulture, and Plant Science

First Advisor

Sunish Sehgal

Abstract

Wheat (Triticum aestivum L.) is a major cereal crop worldwide, but recent gains in grain yield (GY) have not kept pace with projected demand. GY is a highly quantitative trait with low heritability that requires multi-year, multi-location phenotyping, limiting selection efficiency. Targeting higher heritability yield component traits, such as kernel size and weight, and incorporating them into the genomic selection (GS) pipeline has the potential to accelerate genetic gain for GY. The objectives of this study were to (i) understand the relationship between GY and the kernel-size traits, including thousand kernel weight (TKW), kernel length (KL), kernel width (KW), and kernel area (KA), within the South Dakota State University (SDSU) winter wheat breeding program, and (ii) evaluate the genomic predictive ability (PA) for GY, grain protein content (GPC), test weight (TW), and the four kernel-size traits (TKW, KL, KW, and KA). A total of 1,391 hard winter wheat (HHW) genotypes from advanced trials (ADV: AYT + EYT) and preliminary yield trials (PYT) were evaluated at Aurora (Brookings) and Dakota Lakes (Pierre), South Dakota, during the 2023 and 2024 growing seasons. Genotyping-by-sequencing (GBS) of all lines resulted in 15,683 high-quality SNPs. Two GS models, ridge regression best linear unbiased prediction (rrBLUP) and reproducing kernel Hilbert spaces (RKHS), were compared using five-fold cross-validation (CV) within the ADV population, and RKHS was subsequently used for forward prediction of PYT lines under two training scenarios: (1) ADV as the training set, and (2) ADV augmented with a representative core subset of PYT lines (PYT-Core). All kernel-size traits showed positive and significant correlations with GY (r = 0.27–0.46, p < 0.001), and their broad-sense heritability values were generally higher than those observed for GY. In cross-validation, KW achieved the highest PA (up to 0.54), followed by KL (0.33–0.48), while TKW was the most environmentally variable kernel trait (PA range: 0.05–0.39). The rrBLUP and RKHS models performed comparably; however, RKHS was better across most environments. Forward prediction using ADV alone yielded moderate PA for kernel traits (TKW: 0.21–0.34; KL: 0.24–0.39; KW: 0.14–0.55) and lower PA for GY (0.15–0.35). Incorporating the PYT-Core into the training set improved PA across most trait-environment combinations, with the largest enhancement observed for KL (up to 0.54), TKW (up to 0.46), and KA (up to 0.46) in 2024, alongside improvements for GY, GPC, and TW. These results suggest that, along with GY, kernel size traits like KL and KW could also be included in the GS pipeline at the SDSU winter wheat breeding program. Including a small representative subset of PYT lines in the training set generally improved forward prediction accuracy and may help increase genetic gain in applied breeding programs.

Publisher

South Dakota State University

Download

Share

COinS
 

Rights Statement

In Copyright