Over the last decade, a substantial amount of work in genetics has been done with the goal of understanding how genetic variation affects human traits and diseases, primarily via genome-wide association studies (GWAS). Until recently, these studies have focused on associations with single nucleotide variants (SNVs), largely because they have traditionally been easier to genotype. However, the genome contains diverse classes of non-SNV variation such as short tandem repeats (STRs) and structural variants (SVs) that have been shown in some cases to affect human traits. The increasing availability of deep whole genome sequencing (WGS) data, has now enabled algorithms to robustly detect high resolution structural variants and STRs, and the potential for deeper understanding of these variants. Here I present two studies that focus on characterizing the extent and functional impact of SVs and STRs in the human genome. First, I present a study in which I built a comprehensive high quality map of SVs and STRs using over 700 deeply sequenced genomes. I also describe a novel method of filtering variants using reproducibility of genotypes within genetically duplicate sample pairs, and use this information to make insights into the quality of diverse classes of variants called using different methods. I then utilize this high quality map of genetic variation to assess the impact of different classes of variation on gene expression, and show that the functional properties of unique classes of genetic variation is associated with their likelihood to affect genes and linkage to complex traits in humans.

A central question in genetics is how different classes of DNA variants affect RNA splicing and expression. While there has been substantial progress in associating single nucleotide polymorphisms and small indels with these phenotypes, only recently has affordable high throughput sequencing provided the opportunity to assess the impact of somatic, rare, and copy number variants (CNVs) on RNA splicing and expression. In this thesis, I use high throughput sequencing to investigate the effect of somatic variants in SF3B1 on RNA splicing and characterize the genetic regulation of gene expression in induced pluripotent stem cells (iPSCs). In the first part, I examine the effect of recurrent somatic mutations in the splicing factor SF3B1 on RNA splicing in three different cancer types and find that SF3B1 mutants use hundreds of cryptic 3’ splice sites that are rarely used in samples without SF3B1 mutations. Sequence properties of these cryptic 3’ splice sites suggest altered sterics may allow usage of cryptic 3’ splice sites in SF3B1 mutants. I also identify several candidate genes with out-of-frame cryptic splice sites that are used in a majority of transcripts in the mutants and may contribute to oncogenesis. In the second part, I examine the genetic regulation of gene expression in a collection of 215 human iPSCs using transcriptome and whole genome sequencing. I identify expression quantitative trait loci (eQTLs) for nearly six thousand genes including markers of pluripotency such as POU5F1, LCK, IDO1, and CXCL5. A comparison to GTEx eQTLs reveals that iPSCs are well powered statistically for finding eQTLs and have a unique regulatory landscape. I identify biallelic and multiallelic CNVs eQTLs and find that a substantial proportion of CNV eQTLs appear to affect intergenic regulatory regions. I also find that rare promoter variants weakly disrupt gene expression while rare CNVs that overlap genes tend to disrupt gene expression with relatively high effect sizes. Overall, this thesis helps define the roles of somatic, rare, and copy number variants in the regulation of gene expression and splicing and provide key insights into SF3B1-mutated cancers and iPSCs as a model system for molecular association analyses.

In order to prospectively identify individuals at risk for disease, it is important to identify markers that can be reliably measured and to understand the relation of these markers to the disease. For venous thromboembolism (VTE), large-scale genetic studies have had limited success identifying genetic variants and proteins that contribute to disease risk. An alternative method is to focus on the intermediate steps between genetics and disease, such as protein levels. In this dissertation, we set out to identify genetic variants that contribute to blood protein levels and then identify blood proteins that can be used as biomarkers for venous thromboembolism. For the whole dissertation, we utilize the Tromsø Study, a single-center, prospective, study of the inhabitants of Tromsø, Norway. These individuals allow us to identify genetic variants and proteins that are associated with VTE before any symptoms of the disease start, which is key when trying to pre-emptively identify and treat individuals that are at risk for developing VTE. In chapter two, we measure cardiovascular-relevant serum proteins using enzyme-linked immunosorbent assays. We then identify common and rare genetic variation that is associated with the levels of these proteins. In chapter three, we measure the plasma proteome using tandem-mass-tagged mass spectrometry. We then identify common and rare genetic variation that is associated with the levels of these proteins. We further investigate the underlying mechanisms of how genetic variation regulates levels of plasma proteins. In chapter four, we utilize the same plasma proteome that was measured in chapter three in order to identify proteins that are associated with risk of venous thromboembolism. Together, this work advances our understanding of how genetic variants ultimately result in diseases, via their effects on intermediate protein levels.

Genome-wide association studies (GWAS) have suggested that the underlying genetic basis of complex traits and disease is driven by large numbers of non-coding variants with modest effects that likely act by modifying gene regulation. Towards understanding the regulatory impact of genetic variation, expression quantitative trait loci (eQTL) analyses have been performed across dozens of human tissues to link the influence of genetic variants on gene expression levels. While these eQTL studies have provided important biological insights, they are still limited by not considering the contexts in which these variants function, including the stage of development and cell type. Specifically, others have shown increased disease risk in adulthood has links to fetal origins, suggesting that characterizing gene expression in fetal-like cells could identify genetic variants that are associated with adult traits, but function primarily or solely during development. Additionally, as eQTL studies are typically performed across bulk tissues, unaccounted for cellular heterogeneity present in bulk gene expression measurements can affect genotype-gene expression associations. Thus, it is important to identify regulatory variants that alter gene expression in both primitive and adult cell types and to characterize cellular heterogeneity across tissues to comprehensively understand the genetic basis of complex traits and disease.

Here, I present two studies, which utilize gene expression data from fetal-like and adult tissues to characterize cellular heterogeneity at distinct stages of human development. I have examined the cellular heterogeneity in fetal-like induced pluripotent stem cell (iPSC)-derived cardiovascular progenitor cells (CVPCs) using single cell (sc)RNA-seq data to identify cell populations that emerge as a result of the cardiac differentiation. Further, I deconvoluted 180 iPSC-CVPCs and identified factors innate to iPSCs that impacted cardiac fate. Next, I showed that mouse scRNA-seq can be used as an alternative to human scRNA-seq for the deconvolution of adult GTEx bulk tissues and considering cell composition eQTL studies powered the discovery of novel eQTLs, some of which were cell-type-associated and colocalized with GWAS disease loci.

A substantial fraction of SNPs associated with human traits and diseases through genome-wide association studies (GWAS) are likely regulatory variants as they tend to be located within enhancers and associated with differential gene expression. Thus, as a key step in implementing personalized medicine, it is important to identify regulatory variants in the human genome, and characterize their underlying molecular mechanisms. However, identifying and elucidating the functions of regulatory variants is currently challenging as these variants show similar associations with many other neutral variants due to linkage disequilibrium, can be quite far from the gene(s) they regulate, and often have cell type-specific effects. In order to overcome these challenges and interrogate the function of these regulatory variants, it could be possible to examine and integrate epigenetic information in a cell type dependent manner. Here, I present three studies which focus on the functionality of the epigenome – specifically chromatin looping and co-accessibility – in the context of gene regulation, genetics, and disease. I present a tool for computationally working with chromatin loop data, and utilize this tool to show that genetic variation is not associated with large changes in chromatin looping, but rather small modulation in contact propensity which are associated with large changes in gene expression. I then examine chromatin co-accessibility, and show that genetic variants may be able to mediate long range effects on genes and accessible sites hundreds of megabases away – across entire chromosomes – which are associated with cell type relevant genes and diseases.

Adverse events during fetal pancreas development can result in insulin resistance, impaired glucose metabolism, and loss of beta cell function, leading to an increased risk of developing diabetes in adulthood. While current quantitative trait loci (QTL) datasets have been instrumental in characterizing genetic variants associated with diabetes, they only reflect molecular associations present in mature adult tissues. Furthermore, only a fraction of diabetes-associated loci colocalize with QTLs identified in adult whole pancreas and islet tissues. Given the important role of fetal development in adult diabetes predisposition, interrogating the molecular effects of genetic variation during this crucial period could provide valuable mechanistic insights into the etiology of obesity and diabetes. First, we conducted an eQTL analysis on 107 RNA-seq samples from iPSC-derived pancreatic progenitor cells (PPC) to map genetic loci associated with gene expression and isoform usage changes during early pancreas development. Colocalization with eQTLs from adult pancreatic tissues identified genetic variants that were either specifically active during early pancreas development, specifically active in the adult pancreatic stage, or shared across both stages but had stage-unique regulatory functions. Colocalization with genome-wide association studies (GWAS) loci revealed developmental-unique eQTLs with potential roles in glucose homeostasis or diabetes, including those associated with TPD52, CDC37L1-DT, MEG3, and CDH3. Second, we conducted chromatin accessibility QTL (caQTL) analysis using matched PPC ATAC-seq samples. We found that caQTL variants were enriched in distal regulatory regions, including CTCF-binding sites and PPC-specific super enhancer regions, and were enriched for motifs of transcription factors expressed in pancreatic progenitors. Colocalization of eQTLs, caQTLs, and GWAS signals identified putative regulatory mechanisms for TPD52 expression and its impact on fasting glucose levels, as well as KIT and its impact on body mass index. Together, this body of work provides a unique and powerful resource for interrogating the molecular effects of genetic variation during early pancreas development and their potential impact on adult complex pancreatic traits and disease.