The role of the theoretical covariance between SNPs in the design of experiments in genomic evaluations
Final Report Abstract
The aim of genome‐based association studies is to identify and localise genomic markers that are associated with trait expression. The estimated size of a marker effect give rise for further investigation of the biological functionality of the corresponding DNA region. But only if a marker effect is large enough, it can be detected by some statistical testing procedure. Hence, the precision of parameter estimates is highly important in order to increase the number of truly detected DNA variants and to reduce the number of false‐positive detections. To increase the chance (i.e. power) for discovering a trait‐associated variant with genome‐based association studies, a sufficiently large sample is required. An optimal sample size can be estimated under given framework conditions. An experimenter can specify most of the parameters required for an experimental design but the distribution of marker genotypes is actually available after the experiment had taken place. Hence, to design a future experiment, we suggest approximating this information from genetic theory. The approximation depends on the underlying breeding populations (half‐ or full‐sib families) because non‐random mating influences the extent of dependence between markers. Then, knowing the genetic information of selected parents, the number of progeny can be derived to guarantee a certain power of association analysis later on. We could show that, if the dependence between markers is taken into account, the experimental design leads to a more efficient use of animal resources than a classical method used so far. Not only sample size influences the precision of parameter estimates but also the ability of a statistical approach to cope with the dependence among markers which occurs due to the proximity of markers on the genome. We investigated two regression approaches that were able to take information on the proximity of markers into account. Especially, when markers were grouped according to the strength of association among each other, genetic effects captured by markers were estimated more precisely. With this project, we underpinned the importance of employing the dependence among markers, either by grouping them accordingly or by including a kind of similarity matrix in a statistical approach for genome‐based association studies. Furthermore, we contributed statistical and computational tools to design future experiments for fine‐mapping of trait‐associated variants in breeding populations.
Publications
- A modified penalised regression approach to precision‐related questions in genomic evaluations, 69th Annual Meeting of the EAAP in Dubrovnik, Croatia, August 27‐30, 2018
Doschoris, M. & Wittenburg, D.
- Investigating the effects of cluster analysis on grouped ridge regression in genome‐wide association studies, DAGStat Conference in Munich, Germany, March 18 ‐ 22, 2019
Doschoris, M
(See online at https://dx.doi.org/10.13140/RG.2.2.21407.61609) - Planning experimental designs for genomic evaluations: sample size and statistical power calculation, 70th Annual Meeting of the EAAP in Ghent, Belgium, August 26‐29, 2019
Wittenburg, D. & Doschoris, M.
- (2020) Design of experiments for fine‐ mapping quantitative trait loci in livestock populations, BMC Genet., 2020, 21, 66
Wittenburg, D., Bonk, S., Doschoris, M. & Reyer, H.
(See online at https://doi.org/10.1186/s12863-020-00871-1) - Design of Experiments for Fine‐Mapping Quantitative Trait Loci in Livestock Populations, PAGXXVIII Conference in San Diego, US, January 11‐15, 2020
Wittenburg, D. & Doschoris, M.
(See online at https://dx.doi.org/10.1186/s12863-020-00871-1)
