P1-10: Scale effects on genomic modelling and prediction

PhD student: Swetlana Berger
Thesis Committee: Prof. Henner Simianer, Prof. Heike Bickeböller, Prof. Thomas Kneib
Graduation Date: 12/2014

While genome analysis in humans mostly concentrates on single or few interacting genes and only starts to consider polygenic models, animal breeders successfully use the infinitesimal model, aggregated over all loci. Between these extremes many intermediate scales are possible, and prediction of genetic or phenotypic values can be based on all these models or mixtures of those. Using genome partitioning approaches we expect a trade-off between genomic resolution and predictive performance. The impact of non-additive genetic phenomena will increase with refinement. Different scales are expected to be optimal for different traits, aims and species. With whole genome SNP-chip and sequencing data the validity of different genetic models and the impact of the chosen scale on the quality of prediction and the validity of the infinitesimal model can be tested.
In the present project the genome is not treated as a continuum, but as a sequence of conserved segments which can be traced back to an original founder population. The segment structure is the result of a stochastic (Poisson) process, and features like mean and variance of segment length are functions of demographic parameters like population size, number of generations, and mating regime. Understanding the structure will help to design more efficient procedures for genomic prediction. The project primarily aims at developing a thorough understanding of the underlying mechanisms and resulting patterns. The developed models will be validated with high throughput genomic data from different species. Finally, practical consequences for the design of genomic selection schemes will be derived.

back to overview of all projects