Studying The Genetic Architecture Of Complex Traits In A Population Isolate PDF Download

My thesis project is concerned with tapping the potential of population isolates for the dissection of complex trait architecture. Specifically, isolates can aid the identification of variants that are usually rare in other populations. This thesis principally contains in depth investigations into genetic imputation and heritability analysis in isolates. We approached both of these studies from two main angles; first from a methodological standpoint where we created extensive simulation datasets in order to investigate how the specificities of an isolate should determine strategies for analyses. Secondly, we demonstrated such concepts through analysis of genetic data in the known isolate of Cilento. Imputation is a crucial step to performing association analyses in an isolate and represents a cost-efficient method for gaining dense genetic data for the population. The effectiveness of imputation is of course dependent on its accuracy. Hence, we investigated the wide range of possible strategies to gain maximal imputation accuracy in an isolate. We showed that software using algorithms which specifically evoke known characteristics of isolates were, unexpectedly, not as successful as those designed for general populations. We also demonstrated a very small study specific imputation reference panel performing very strongly in an isolate; particularly for rare variants. For many complex traits, there exist discordances between estimates of heritabilities from studies in closely related individuals and from studies on unrelated individuals. In particular, we noted that most researchers consider dominant (non-additive) genetic effects as unlikely to play a significant role despite contrasting results from previous studies on isolates. Our second analysis revealed possible mechanisms to explain such disparate published heritability estimates between isolated populations and general populations. This allowed us to make interesting deductions from our own heritability analyses of the Cilento dataset, including an indication of a non-null dominance component involved in the distribution of low-density lipoprotein level measurements (LDL). This led us to perform genome-wide association analyses of additive and non-additive components for LDL in Cilento and we were able to identify genes that had been previously linked to the trait in other studies. In the contexts of both of our studies, we observed the importance of retaining genotype uncertainty (genotype dosage following imputation or genotype likelihoods from sequencing data). As a prospective of this thesis, we have proposed ways to incorporate this uncertainty into certain methods used in this project. Our findings for imputation strategies and heritability analysis will be highly valuable for the continued study of the isolate of Cilento but will also be instructive to researchers working on other isolated populations and also applicable to the study of complex diseases in general.