Efforts to understand the genetic aetiology of complex traits have gained a lot of momentum in the last decade. Advancement of next generation sequencing and the ever-decreasing price of genotyping platforms have allowed us to carry out a vast number of genome wide association studies (GWAS). Until relatively recently, GWAS was guided by the common disease – common genetic variation paradigm. However, recent findings and developments have made us look at the bigger picture, including rare genetic variation. In addition, methodological developments are guiding the translation of GWAS findings. For instance, diverse statistical methods can be applied on genetically informative data to estimate the genetic correlation between complex diseases. The latter can have important medical implications, as genetically correlated diseases might be responsive to the same treatments. Also, approaches such as Mendelian randomization (MR) can help investigations of causal factors in disease when is unfeasible to carry out randomized control trials.
My focus is on the application of statistical methods in complex trait genetics – I cover a range of phenotypes, ranging from eye disease to cancer. I begin with a general introduction describing the methods used along this thesis as well as the important concepts. I make particular emphasis of methodological approaches and challenges during association studies of rare variants, as well as approaches used for the estimation of genetic correlation and for MR. In the latter part of the introduction, I present an overview of the traits examined and the gaps this thesis covers.
The second chapter displays a mapping study of exonic variants with central corneal thickness (CCT), an endophenotype of keratoconus. This work led to the identification of a missense mutation in WNT10A, associated to a 2-fold increase in risk of keratoconus.
The following three chapters interrogate epidemiological aspects of refractive error (RE) and myopia through genetic approaches. Many studies have observed a strong correlation between myopia, time outdoors, and education level. One hypothesis to explain these associations is that less time outdoors and more education translates into more time performing near-work activities, which may promote eye elongation and the development of myopia. Another hypothesis is that light induces dopamine release, suppressing the eye elongation. Further, some studies suggested that vitamin D might play a role in the development of the condition. To investigate some of these hypotheses I carried out gene mapping and MR approaches. Chapter 3 introduces a GWAS study of conjunctival ultraviolet autofluorescence (CUVAF). CUVAF has excellent potential as a biomarker of sun exposure compared to survey data. Understanding the aetiology behind this biomarker is potentially helpful when assessing the hypotheses of sun exposure and myopia. In chapters 4 and 5, I present two MR studies assessing the causal relationship between RE (level of myopia), vitamin D and education levels. Using a sample of 37,382 individuals of European ancestry and 8,376 from Asian ancestry and SNPs in the DHCR7 and CYP2R1 genes as instrumental variables (IVs), we ruled out a causal association of vitamin D on RE. Chapter 5 describes the MR study of education and myopic RE. In this, using data from three different cohorts and an education level polygenic risk score derived from alleles effects from GWAS summary data, we estimated that approximately every 2 years of additional education result in an increase of myopia.
Following to chapter 6, I performed polygenic assessments of age-related macular degeneration (AMD) and primary open angle glaucoma (POAG). . Using genome-wide array data on Australian cases and controls, we estimated the array heritability of both diseases, and the variance explained by the genome-wide associated loci. Further, we assess whether there is some genetic overlap between AMD and POAG, beyond the signal seen at ABCA1, which at genome-wide significance level is associated to both. In addition, we investigated whether the difference in prevalence of POAG in males and females, can be due (at least in part) to genetics. Our analyses suggest that risk to POAG is conferred by many genetic variants of small effects and that the genetic overlap between POAG and AMD is not restricted to ABCA1. Moreover, we found evidence of genetic differences between genders in POAG.
In the last results’ chapter I show the work investigating the genetic architecture of epithelial ovarian carcinomas (EOC) and its subtypes. I look into the array heritability of each subtype and their pairwise genetic correlations. Moreover, I examine their genetic overlap with risk factors including obesity, smoking behaviour, diabetes, age at menarche and height. Overall, this work shows that EOC and its subtypes do not have a large array heritability and that the genetic architecture of the subtypes is homogenous. Finally, I show evidence of a genetic overlap of EOC with obesity and diabetes.
In the last chapter, I discuss and propose future directions for the field of statistical genetics, particularly in the areas of mapping, correlation and causation.