|
|
Models that reflect the multifaceted contributions of genomic loci have a potential to facilitate unprecedented quantification of the genetic architecture underlying various traits and increase genomic selection (GS) prediction accuracies. To evaluate the performance of such models, simulation studies are essential. Therefore, a R/CRAN package called simplePHENOTYPES developed by the Lipka Lab is first discussed. This package uses real marker data to simulate pleiotropic quantitative trait nucleotides (QTNs) that behave in either an additive, dominance, or epistatic manner. We first demonstrate how simulating traits from simplePHENOTYPES can be used to evaluate the ability of a multi-trait genome-wide association study (GWAS) model to distinguish between linkage and pleiotropy. Similarly, we show how simulated traits can be used to quantify how well an stepwise model selection GWAS procedure can identify two-way epistatic QTN and correctly classify them as epistatic.
We end the presentation with two demonstrations on how simulation studies can be used to make important practical recommendations for GS breeding programs. We first show a study that illustrates how accounting for large-effect marker-trait associations in GS models does not guarantee an increase in prediction accuracy. Finally, we present a recent simulation study that evaluates the impact of training set composition from two different species (Miscanthus sinensis and Miscanthus sacchariflorus) on the GS prediction accuracy in an interspecific Miscanthus × giganteus F2 population.
|
|
