Login | Request Account (DAF staff only)

Improved methods of predicting genetic merit in plant breeding programs using linear mixed models

Hunt, C. H. (2021) Improved methods of predicting genetic merit in plant breeding programs using linear mixed models. PhD thesis, University of Queensland, 134 pages.

[img]
Preview
PDF
4MB

Article Link(s): https://doi.org/10.14264/06b268c

Abstract

Plant breeders are constantly faced with many challenges, particularly in testing of genotypes. These may include the variability among and within environments, seed availability and resources. Despite careful planning and management there are always uncontrolled factors that can be minimised by using appropriate statistical analysis techniques, which in some circumstances are very complex. The implementation of genomic prediction and subsequent selection has opened new avenues for in-depth exploration of better statistical methods for optimising plant breeding.
For many years Australian plant breeding trials have been analysed using techniques that include the genetic parentage and adjustments for the spatial arrangement of the genotypes in the field. This can be extended further to allow for inter-plot competition. Inter-plot competition is of particular value for trials that have two row plots. The added advantage of including pedigrees in the analysis allows for the possible detection of particular families that may be prone to competition effects.

The relationship between genotypes can be calculated using knowledge of the parent lines. This can be extended to also include the relationship calculated using marker information. We have developed a model that teases apart the parentage and the marker relationships to investigate possible increases in prediction accuracy. We observe that the difference in accuracy is largely affected by the environment (GxE) with some trials analysed optimally with only marker information and others best analysed with both pedigree and markers.

As a further improvement of the analysis, we investigate the effect of partitioning the genetic variance into additive and dominance effects while simultaneously allowing for the spatial field effects and GxE. We have found that including dominance has an effect on the accuracy of the additive effects, which in turn has an effect on selection. This study also showed that the presence of a dominance effect has a strong environmental interaction.

The final study considers the optimal combinations of testers and lines for early generation trials. The presence/magnitude of dominance and GxE has a detrimental effect on the selections of early generation hybrids that use only a single tester. We investigate this issue using trials that have two testers and compare results between the testers. Results vary between environments, in most cases the use of a single tester has limited capacity to genomically predict the performance of the lines crossed with a second tester.

Plant breeding programs require careful planning and construction of trials, with one of the most important aspects being the composition set of genotypes in each trial. This is inherently more complex for breeding programs in hybrid crops. All of the above knowledge can aid the design of a training set of genotypes that will help achieve the best genetic gain.

Item Type:Thesis (PhD)
Business groups:Crop and Food Science
Keywords:mixed models, spatial effects, pedigrees, additive effects, dominance effects, genotype by environment interaction, genomic prediction, training dataset, hybrid prediction, hybrid breeding
Subjects:Science > Botany > Genetics
Agriculture > Agriculture (General) > Agriculture and the environment
Agriculture > Agriculture (General) > Plant growing media. Potting soils
Agriculture > Agriculture (General) > Soils. Soil science
Agriculture > Agriculture (General) > Methods and systems of culture. Cropping systems
Plant culture > Seeds. Seed technology
Deposited On:28 Sep 2021 00:22
Last Modified:28 Sep 2021 00:22

Repository Staff Only: item control page

Downloads

Downloads per month over past year

View more statistics