The Genetic Architecture of Complex Traits and their Prediction. (360G-Wellcome-099670_B_12_Z)

The prediction of risk for disease is an area that is expanding in both the commercial and academic worlds; however, most prediction methods do not incorporate all of the available information on risk. I plan to develop computationally efficient phenotype prediction methods utilising as much of the relevant genetic information as possible, which will be tested in simulation and then applied to data from the UK Biobank. The accurate prediction of phenotypes relies on an accurate understanding of genetic architecture, including the overall effect size distribution and the relation between allele frequency and effect size. The Bayesian approach I plan to takewill incorporate knowledge about genetic architecture and learn about genetic architecture in an adaptive way. I will test whether the inclusion of environmental effects and gene-environment interaction effects improves phenotype prediction. It is known that the inclusion of genotype-phenotype data on close relatives improves phenotype prediction, and I will test whetherdata of this sort will enable more clinically useful phenotype prediction. This investigation will reveal fundamental facts about the nature of variationin disease risk while at the same time giving practical methods for risk prediction.