Document Type

Thesis - Open Access

Award Date

2017

Degree Name

Master of Science (MS)

Department / School

Agronomy, Horticulture, and Plant Science

First Advisor

Melanie Caffe-Treml

Keywords

error variance, genomic selection, heritability, milling and nutritional quality, prediction accuracy

Abstract

Oats can lower cholesterol, reduce risks of type-2 diabetes, and help prevent heart diseases when consumed daily. Therefore, it is important to evaluate and select breeding lines with desirable milling and nutritional quality traits. Genomic selection, which uses genotyping data to predict the breeding value of an individual, is a promising method to increase genetic gain by selecting for quality traits earlier in the line development process. In this study, we collected phenotypic data for three nutritional traits (protein, β- glucan, and fat content) and five milling quality traits (percent plumps, percent thins, percent plump groat, groat percent and thousand kernel weight) on grain samples from 465 different oat genotypes grown at four locations in South Dakota in either 2015 or 2016. To take account of field variation, we investigated four linear mixed models with the R package, minque, subject to the inclusion or exclusion of row and/or column effects. Overall, inclusion of row and column effects reduced the error variance, and accordingly, increased heritability and improve relative efficiency for most of the traits. Thus, the full model with inclusion of row and column effects was applied to predict genotypic effects for genomic selection analysis. All breeding lines were genotyped with genotyping by sequencing (GBS) technique. Genomic selection models were evaluated using five currently available genomic selection methods (RRBLUP, GAUSS, PLSR, Elastic Net, and Random Forest) along with model averaging (AVE). Prediction accuracy ranged from 0.19 to 0.74 and 0.30 to 0.70 among traits and locations in the year 2015 and 2016, respectively. Fat content and percent plump were the two traits with the highest prediction accuracy. ß-glucan content, on the other hand, had the lowest prediction accuracy in both years. Overall the prediction accuracy was moderate to high for most of the traits in this study. Our results suggested that genomic selection could offer a valuable strategy to improve genetic gain for major milling and nutritional quality traits in oats.

Library of Congress Subject Headings

Oats -- Genetics.
Oats -- Nutrition.
Oats -- Milling.
Oats -- Selection.

Description

Includes bibliographical references

Format

application/pdf

Number of Pages

89

Publisher

South Dakota State University

Share

COinS
 

Rights Statement

In Copyright