Quantitative traits are typically influenced by a myriad of genes with their additive, dominance and epistatic effects, whereat their number and kind of interplay are hardly known. Although additive genetic effects are of major interest in breeding applications, the knowledge about non-additive genetic effects is important for the understanding of phenomena like heterosis or inbreeding depression
... [Show full abstract] as well as for an appraisal of broad sense heritability. When analyses of genome-wide marker information try to make allowance for non-additive effects, the separation of non-zero effects from unimportant ones is an issue e.g. for the improvement of genetic value prediction and deeper insights into the relative importance of additive and non-additive effects for genetic variation. A previously published spike and slab variable selection approach was adapted to this problem by specifying a complexity parameter for each kind of genetic effect included in a statistical model. These parameters represent the proportion of non-zero effects within each source of genetic variation. With aid of the complexity parameters, an empirical selection procedure was appended to identify the significant non-zero genetic effects a posteriori. The suitability of this approach was verified via simulations with either a small number or a multitude of QTL with additive and dominance effect. The accuracy of genetic value prediction was always at a high level in both scenarios. Despite that most of the genetic variation was due to additive effects, the contribution of dominance effects could be assessed. The spike and slab approach had, however, difficulties not only to determine the amount of dominance variation but also the number of dominance effects; its proportion was underestimated to a larger extent than the proportion of additive effects.