Kim, Bongsong: Genomic Prediction and Association Mapping Using Publicly Available Data of German Variety Trials in Spring Barley. - Bonn, 2014. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5n-37049
@phdthesis{handle:20.500.11811/5855,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5n-37049,
author = {{Bongsong Kim}},
title = {Genomic Prediction and Association Mapping Using Publicly Available Data of German Variety Trials in Spring Barley},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2014,
month = sep,

note = {In recent decades, the implementation of best linear unbiased prediction (BLUP) has been extended beyond its initial purpose for the breeding value (BV) estimation to conduct the association mapping and genomic selection. In this study, the prospect of using BLUP was investigated for the BV estimation, AM and GS in self-pollinating crop with a German barley cultivar collection that is publicly available. Chapter 1 introduces issues of this study and provides a review of the relevant literatures. Chapters 2 and 3 address the application of BLUP with an assembled data set of German spring barley cultivars in unbalanced trials. One issue regarding this work was the absence of a method for computing a numerator relationship matrix (NRM) for selfing crop species. Therefore, the method of constructing the NRM was developed in this study, which is introduced in Chapter 2. Chapter 3 reports the application of the underlying NRM to BLUP for grain yield, scald severity and net blotch severity. Heritabilites resulted in 0.719 for grain yield, 0.491 for scald severity and 0.581 for net blotch severity, which suggests that the given phenotypic data were measured in sufficient level. Spearman’s rank correlation between BLUP estimates and mean phenotypes (MPs) were shown to be 0.854 for grain yield, 0.893 for scald severity and 0.940 for net blotch severity, which indicates that the selection depending on the BLUP may respond better than that depending on the phenotypic observation using MPs. Chapter 4 describes the measurement of the marker-trait association for the aforementioned traits in German spring-sown barley cultivars and 1181 diversity array technology (DArT) markers. Two models were fitted: (1) the BLUP that embeds a marker-based kinship matrix and a discriminant analysis of principle component matrix (KD model) and (2) the BLUP that embeds a marker-based kinship matrix and a subpopulation matrix resolved using STRUCTURE software (KS model). For the stringent evaluation of marker-trait association, the significance level of p < 0.001 in the Wald test and cross-validation were applied. In total, six marker-trait associations were detected (one for grain yield, four for scald severity and one for net blotch severity). Chapter 5 presents the genomic selection performed using ridge regression BLUP (RR-BLUP) with the same plant materials as used in Chapter 4. The increasing sizes of the training set and marker set were positively correlated with prediction accuracy. As a novel approach, marker sets that were selected based on the strength of marker-trait linkages were examined. To form the sets of markers, p-values obtained from the mapping study were referenced, and ten sets of markers were prepared by applying p-value thresholds of 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 and 1.0. The resulting prediction accuracies ranged from 0.3226 to 0.7323 for grain yield, from 0.3534 to 0.5396 for scald severity and from 0.4340 to 0.8326 for net blotch severity. A marker set formed with a decreasing p-value appeared to provide the higher prediction accuracy for all traits by overcoming the weakness of the small size of marker set, showing that the use of p-values is promising in RR-BLUP.},
url = {https://hdl.handle.net/20.500.11811/5855}
}

The following license files are associated with this item:

InCopyright