Leveraging Historical Data for Estimating Genetic Gain and Implementing Genomic Selection in a Student Led Barley Breeding Program

Authors

Sydney Graham, University of Nebraska-LincolnFollow

First Advisor

Reka Howard

Date of this Version

5-2025

Document Type

Thesis

Citation

A thesis presented to the faculty of the Graduate College at the University of Nebraska in partial fulfillment of requirements for the degree of Master of Science

Major: Statistics

Under the supervision of Professor Reka Howard

Lincoln, Nebraska, May 2025

Comments

Copyrght 2025, Sydney Graham. Used by permission

Abstract

In Nebraska, winter feed barley presents an emerging market for producers and an opportunity to diversify cropping systems. The University of Nebraska Barley Breeding Program aims to develop high-yielding, winter-hardy varieties. A unique aspect of this program is that doctoral students serve as barley breeders and are responsible for crossing, data collection, and advancement decisions. While this provides hands-on experience for the students, the impact of student leadership has not been examined.

This study used a historical data set to evaluate the realized genetic gain of the breeding program, and as a training population for genomic selection. The dataset consisted of 302 genotypes from the advanced yield trial evaluated from 2002 to 2022 in three Nebraska locations. The rate of realized genetic gain for yield was estimated by regressing estimated genotypic means on the year they entered the trial. Additionally, SNP data for 189 genotypes was generated with the USDA-SoyWheOatBar-3K array, and daily weather covariates were collected. The 2024 observation nursery was genotyped and used as a testing set for genomic selection. Genomic selection models included four variations of GBLUP (G+E, G+E+GxE, G+W, G+W+GxW), four Bayesian models (Bayes A, Bayes B, Bayes C, Bayes LASSO), and two machine learning approaches (random forest, support vector machine). The models were applied for both winter survival and grain yield.

The realized genetic gain for yield was 62.4 kilograms per hectare per year. For grain yield, the GBLUP and Bayesian genomic selection models performed well, and the subset of the training population used was more important than which model was selected. The highest predictive accuracy was achieved using the Bayes A model trained on Lincoln data only (r = 0.420). For prediction of winter survival, the highest prediction accuracy was for the RF model using Lincoln only training data (r = 0.263). While improvements can be made for winter survival, implementing genomic selection for yield can aid future student barley breeders and provide continuity between transitions.

Advisor: Reka Howard