Search

Infectious diseases have been a major cause of mortality and morbidity throughout human history. The dramatic advances in technology, especially in genome sequencing, have opened a new avenue to study infectious diseases and changed the landscape of public health and medical science. Compared to non-communicable human diseases where human genetics is the sole focus of genetic studies, infectious diseases add another layer of complexity due to the host-pathogen dynamic. Since infection can be seen as the result of the complex interplay of host and pathogen and the environment, a comprehensive understanding requires the incorporation of all these factors. In this thesis, I explore the different applications of genetic data in studying infectious diseases through the lens of host-pathogen dynamics and their clinical or public health implication. I first used virus whole genome data to rebuild the epidemiological history of Hepatitis C virus subtype 3a (HCV-3a). I showed how World War II had likely led to the spread of HCV-3a from South Asia and becoming a global epidemic. Notably, I also incorporated host genetic data into the analysis and showed how it helped us identify and address the confounding factor of human migration when conducting an evolutionary analysis on HCV-3a. I then integrated previously published methods to investigate protein residue covariation to investigate the signals co-evolution within the Hepatitis B virus (HBV). I built a co-evolution profile for HBV and link them to previous studies to demonstrate how drug resistance is one of the major driving forces of protein residue co-evolution in HBV. Lastly, I explored the potential of integration of national-scale databases that collect pathogen (Second Generation Surveillance System) and host data (UK Biobank) respectively and conducted multiple genome-wide association studies (GWASs) and lay out how such integration can help improve our understanding of human genetic risk factors for infectious diseases. To conclude, I discuss the importance of host-pathogen data integration as well as the need for developing statistical models that can account for both the data volume as well as complexity resulted from such integration.

Background: Blood pressure lowering is an established strategy for preventing microvascular and macrovascular complications of diabetes, but its role in the prevention of diabetes itself is unclear. On the other hand, controversy exists as to whether the threshold of blood pressure for initiation of antihypertensive therapy should differ between people with and without type 2 diabetes. I aimed to integrate individual participant data from major randomised controlled trials and genetic data to fill these knowledge gaps. Objectives: This thesis sought to examine three main objectives: to investigate the effect of pharmacological blood pressure-lowering on the risk of new-onset type 2 diabetes; to investigate the separate effects of blood pressure-lowering drug classes on the risk of new-onset type 2 diabetes; to investigate the effect of pharmacological blood pressure-lowering treatment for the prevention of major cardiovascular disease in persons with and without type 2 diabetes. Methods: For the first and second objectives, I conducted a one-stage individual participant-level data meta-analysis of major randomised controlled trials using data from the Blood Pressure Lowering Treatment Trialists' Collaboration (BPLTTC). Analyses were complemented with Mendelian randomisation studies using naturally randomised genetic variants associated with systolic blood pressure and genetic variants in the gene that encodes the therapeutic targets of each drug class. For the third objective, I used one-stage individual participant-level data meta-analysis using the BPLTTC dataset. I expressed the treatment effect per 5 mmHg reduction in systolic blood pressure on the risk of developing a major cardiovascular event as the primary outcome, defined as the first occurrence of fatal or non-fatal stroke or cerebrovascular disease, fatal or non-fatal ischaemic heart disease, or heart failure causing death or requiring hospitalisation. Cox proportional hazard models, stratified by trial, were used to estimate hazard ratios (HRs) separately by type 2 diabetes status at baseline, with further stratification by baseline categories of systolic blood pressure (in 10 mmHg increments from < 120 mmHg to ≥170 mmHg). Results: For the first and second objectives, blood pressure-lowering treatment was found to reduce the risk of diabetes by 11% (hazard ratio per 5 mmHg lower systolic blood pressure 0.89 [95% confidence interval [CI] 0.84 to 0.95]). Similarly, in the Mendelian randomisation study, each 5 mmHg genetically influenced lower systolic blood pressure was associated with an 11% lower risk of diabetes (odds ratio 0.89 [95% CI 0.86 to 0.93]). Evidence from genetic data and trials was also consistent in that angiotensin-converting enzyme inhibitors and angiotensin receptor blockers reduced the risk of diabetes, and beta-blockers increased this risk. There was no effect for calcium channel blockers and findings for thiazide diuretics were inconsistent. For the third objective, over 4.2 years median follow-up (IQR 3.0 to 5.0), a 5 mmHg reduction in systolic blood pressure decreased the risk of major cardiovascular events in both groups, but with a weaker relative treatment effect in participants with type 2 diabetes (HR 0.94 [95% CI 0.91 to 0.98]) compared with those without type 2 diabetes (0.89 [0.87 to 0.92]; p for interaction=0.001). However, absolute risk reductions did not differ substantially between people with and without type 2 diabetes (absolute risk reduction -1.54 [95% CI -2.04 to -1.04] in people with diabetes and -1.61 [-1.86 to -1.36] in people without diabetes, p for interaction =1). We found no reliable evidence for heterogeneity of treatment effects by baseline systolic blood pressure in either group. Conclusions: Blood pressure lowering is an effective strategy for the prevention of new-onset type 2 diabetes. Established pharmacological interventions, however, have qualitatively and quantitively different effects on diabetes, likely due to their differing off-target effects, with angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers having the most favourable outcomes. Additionally, although the relative beneficial effects of blood pressure reduction on major cardiovascular events were weaker in participants with type 2 diabetes than in those without, absolute effects were similar. The difference in relative risk reduction was not related to the baseline blood pressure or allocation to different drug classes. Therefore, the adoption of differential blood pressure thresholds, intensities of blood pressure lowering, or drug classes used in people with and without type 2 diabetes is not warranted.

Background: The proliferation of large-scale studies with genetic data, coupled with technological advances in genetics has served as a catalyst for the development of causal inference approaches leveraging genetic data. One such approach, Mendelian randomization (MR), uses genetic variants as instrumental variables to estimate causal relationships while mitigating bias due to residual confounding. MR methods can be broadly be categorised based on the type of data utilised, specifically approaches using either individual level data or summary data from genome-wide association studies (GWAS). Objectives: To evaluate and contribute to the existing MR methods literature, introducing novel statistical frameworks for use with individual and summary level data. In this work radial multivariable Mendelian randomization (RMVMR) is presented as an approach for visualising summary MR analyses, while MR using Gene-by-Environment interactions (MR-GxE) allows for the validity of individual instruments to be evaluated. Analyses of the causal determinants of stroke are subsequently conducted within an evidence triangulation framework, highlighting the utility of employing individual and summary level MR methods in combination. Results: An appraisal of existing MR methods is conducted in chapter 1, after which RMVMR is presented as a novel approach for visualising key aspects of multivariable MR analyses in chapter 2. Interaction MR is described in chapter 3, providing an overview of how causal effects can be estimated using a single instrument in the presence of measured or unmeasured gene-by-covariate interactions. Simulation studies are performed for each MR method, highlighting the relative strengths and limitations of each approach. After conducting a preliminary review of stroke in chapter 4, applied analyses are performed in chapter 5, identifying a range of modifiable risk factors with respect to stroke. Conclusion: The statistical methods presented in this work have the potential to overcome common limitations of existing MR methods, and highlight their application to stroke as a key area of global health research.

Biodiversity monitoring underpins decision making in conservation. From small scale management, through to international policy decisions, biodiversity data are used to advise and influence those in positions of power. We live in an age of increasing data accumulation. In particular, an increase in citizen science schemes has led to a wealth of biodiversity information. Despite this increase in data collection, there are still large gaps in our knowledge, with questions that remain unanswered and subjects that continue to be neglected. This thesis focuses on biodiversity monitoring, investigating how data from current long-term monitoring schemes can be applied to novel questions, how new forms of biodiversity monitoring are desperately required and how different forms of biodiversity monitoring data can be combined to explain features of a species biology. Firstly, long-term population monitoring data are used to overcome an impasse in functional ecology. The ability to predict ecosystem service stability has so far been an out of reach goal. However, analysing correlations between species population dynamics offers an achievable method of determining whether specific functions and services are at risk of declining due to changes in species abundances across a community. Secondly, the genetic diversity of the meadow brown butterfly, Maniola jurtina, is investigated at both the spatial and temporal scale. The genetic diversity of the species is found to be stable across the study area and over time. This represents an important contribution to the field of genetic diversity monitoring. Despite being acknowledged as an increasingly important measure of biodiversity, genetic diversity monitoring schemes are extremely rare outside of socioeconomic species. This study represents one of the first examples of the monitoring of a wild species that has no direct economic value. Next, the genetic diversity of M. jurtina is investigated at the continental scale, building upon the work in the previous chapter. Across the continent there appears to be distinct population structuring, with individuals in the UK belonging to a different genetic cluster to those on the mainland. Finally, the ability to combine monitoring data with genetic and experimental data is demonstrated with an investigation into the phenology of M. jurtina. Analysis of long-term monitoring data determined that M. jurtina display a protracted flight period on chalk sites. Genetic data are used to determine whether any genetic structuring of populations is associated with these differences, whilst experimental data are used to determine the effect of drought on phenology. Overall, this thesis brings together three separate areas to demonstrate the wide range of studies that monitoring data can be applied to. Furthermore, the importance of genetic diversity monitoring is highlighted, along with a demonstration into the relative ease at which it can be accomplished. In the final chapter, the limitations of the work are discussed along with the wide range of future applications.

Despite a recent decline in prevalence, malaria still incurs a substantial burden on global health. Further progress towards elimination will require developing control strategies that are guided by detailed epidemiological surveillance. Yet, many existing surveillance approaches - including estimating parasite prevalence with community-level surveys, or tracking resistance with therapeutic efficacy studies -- are time consuming and labour intensive. There is an interest in harnessing the increasing availability of P. falciparum genetic data for surveillance, but several challenges remain outstanding. First, the relationships between genetic and epidemiological processes are poorly understood, hindering epidemiological inference from genetic data. Second, the ubiquity of mixed infections complicates analysis of malaria genetic data. Third, how the collection of genetic data can be efficiently integrated into malaria control is still being resolved. In this thesis, we first elucidate relationships between genetic diversity statistics and epidemiological parameters affecting parasite prevalence, by developing an agent-based simulation that includes a model of parasite DNA. We find many genetic diversity statistics are correlated with parasite prevalence, but that the strength and timelines of these relationships varies. In particular, statistics related to mixed infections are strongly correlated with, and rapidly responsive to, changes in parasite prevalence. Second, we leverage identity-by-descent patterns present within mixed infections to make inferences about the microepidemiology of malaria. We separate mixed infections comprised of two strains into those generated by a single mosquito bite (co-infection) and those generated by two mosquito bites (super-infection). Applying our approach to the Pf3k data resource, we find evidence for serial co-infection across several hosts, and show that the regional rate of serial co-infection is related to parasite prevalence. Finally, we develop an approach to genetic data collection that can be deployed in malaria-endemic countries. Using nanopore sequencing and working with local scientists, we infer the drug resistance status of ten samples from Zambia, in-country and in two days. In addition, we present evidence suggesting the approach could be used to identify mixed infections and infer their complexity. We end by discussing what statistical modelling approaches are most likely to permit parasite prevalence inference in the future, and outline some of the challenges associated with sequencing in the malaria-endemic world.

Recent years have offered a wealth of genetic association data, with a concurrent explosion in the availability of methods for exploring causal effects through randomly allocated genetic variants that serve as proxies for traits of interest. This thesis investigates the state of this field within the remit of cardiovascular disease. Following an introduction into cardiovascular disease and Mendelian randomization (MR), the research focuses on dietary, social and pharmacological exposures as demonstrative examples for highlighting the breadth of techniques that can be harnessed towards understanding underlying mechanisms and therapeutic opportunities. Both two-sample and one-sample MR analyses are performed, using genetic summary data from large-scale consortia and the UK Biobank. Sample sizes for individual analyses typically exceed tens of thousands of participants. A diverse array of MR methods are employed, appropriate to the setting and objective of each analysis. Considering systemic iron status as a diet-related trait, genetic instruments are identified with consequent MR analyses supporting a protective effect on risk of cardiovascular outcomes related to atherosclerosis but a detrimental effect on outcomes related to thrombosis arising from stasis of blood. Phenome-wide association study further highlights effects of systemic iron status outside the remit of cardiovascular disease. In the investigation of social factors, MR mediation analysis techniques are applied to identify the pathways by which education affects cardiovascular disease risk, with multivariable MR further used to disentangle the direct effects of education and intelligence respectively. In the investigation of pharmacological exposures, genetic instruments for antihypertensive drugs are identified and validated by comparing against corresponding estimates from clinical trials. Phenome-wide association study is used to identify possible side-effects and repurposing opportunities, with a potential detrimental effect of calcium channel blockers identified on risk of diverticulosis. The final section provides an overview of the current state of applied MR, as well as future perspectives.

Outbreaks of infectious disease pose an enormous global burden on human well-being. Effectively controlling such outbreaks requires timely and evidence-based interventions, which are increasingly being informed by statistical modelling efforts. A specific field of statistical modelling attempts to reconstruct the history of transmission events between individual cases in order to provide high-resolution epidemiological insights directly relevant to infection control. Recent years have seen considerable methodological progress in this regard, with an emphasis on using whole genome sequencing (WGS) data. However, current methods largely disregard other types of outbreak data and are therefore poorly suited to outbreaks where WGS data is unavailable or weakly informative. This thesis aims to improve the utility and widen the scope of outbreak reconstruction tools by incorporating additional types of outbreak data. I begin with an overview of this field of research (Chapter 1). I then conduct a simulation study to evaluate the realistic informativeness of WGS data for different pathogens (Chapter 2). Next, I develop methods and software for integrating commonly available epidemiological data, namely contact data (Chapter 3) and hospital ward data (Chapter 4), into the inference of transmission chains. Finally, I consider outbreak reconstruction in a more operational context. I develop a simple algorithm and software tool for visualising timed transmission networks (Chapter 5) and then apply the methods developed in this thesis to synthesized data resembling the ongoing Ebola outbreak in the Democratic Republic of Congo (Chapter 6). In the last chapter, I discuss the methodological components of this thesis and end on a more general reflection of how outbreak reconstruction tools can be realistically used in outbreak settings to reduce the burden of disease.

The prevalence of type 2 diabetes (T2D) has increased steadily over the last thirty years and has now reached epidemic proportions. The secondary complications associated with T2D have significant health and economic impacts worldwide and it is now regarded as the seventh leading cause of mortality. Therefore, understanding the underlying causes of T2D is high on government agendas. The condition is a multifactorial disorder with a complex aetiology. This means that T2D emerges from the convergence between genetics, the environment and diet, and lifestyle choices. The genetic determinants remain largely elusive, with only a handful of identified candidate genes. Genome-wide association studies (GWAS) have enhanced our understanding of genetic-based determinants in common complex human diseases. To date, 120 single nucleotide polymorphisms (SNPs) for T2D have been identified using GWAS. Standard statistical tests for single and multi-locus analysis, such as logistic regression, have demonstrated little effect in understanding the genetic architecture of complex human diseases. Logistic regression can capture linear interactions between SNPs and traits however it neglects the non-linear epistatic interactions that are often present within genetic data. Complex human diseases are caused by the contributions made by many interacting genetic variants. However, detecting epistatic interactions and understanding the underlying pathogenesis architecture of complex human disorders remains a significant challenge. This thesis presents a novel framework based on deep learning to reduce the high-dimensional space in GWAS and learn non-linear epistatic interactions in T2D genetic data for binary classification tasks. This framework includes traditional GWAS quality control, association analysis, deep learning stacked autoencoders, and a multilayer perceptron for classification. Quality control procedures are conducted to exclude genetic variants and individuals that do not meet a pre-specified criterion. Logistic association analysis under an additive genetic model adjusted for genomic control inflation factor is also conducted. SNPs generated with a p-value threshold of 10−2 are considered, resulting in 6609 SNPs (features), to remove statistically improbable SNPs and help minimise the computational requirements needed to process all SNPs. The 6609 SNPs are used for epistatic analysis through progressively smaller hidden layer units. Latent representations are extracted using stacked autoencoders to initialise a multilayer feedforward network for binary classification. The classifier is fine-tuned to discriminate between cases and controls using T2D genetic data. The performance of a deep learning stacked autoencoder model is evaluated and benchmarked against a multilayer perceptron and a random forest learning algorithm. The findings show that the best results were obtained using 2500 compressed hidden units (AUC=94.25%). However, the classification accuracy when using 300 compressed neurons remains reasonable with (AUC=80.78%). The results are promising. Using deep learning stacked autoencoders, it is possible to reduce high-dimensional features in T2D GWAS data and learn non-linear epistatic interactions between SNPs while enhancing overall model performance for binary classification purposes.

The major dental diseases are caries and periodontitis. These two common conditions are important public health problems and have complex, multifactorial aetiology. Study designs which use genetic data can provide evidence about the molecular and biological basis of disease and also help prioritize modifiable risk factors which have causal relevance for disease. To date these approaches have had limited success in dental epidemiology, possibly due to the lack of large studies with genetic data and dental phenotypes. Through a theoretical review and a pair of applied illustrations, I argue that questionnairederived and index-linked dental data provide a valuable resource for the application of modern epidemiological methods. Using these data for genetic association discovery, the main findings are novel risk loci for caries in both adult and paediatric populations, and evidence suggesting that the genetic risk factors for caries and periodontitis are partially overlapping with a range of other health traits. These newly discovered risk loci can act as proxies for variation in dental disease experience using current methods for causal effect estimation. I test the reciprocal hypotheses that dental diseases have downstream effects on cardio-metabolic traits, and that metabolic traits influence risk of dental diseases, finding some evidence supporting existing beliefs that dental diseases may have undesirable downstream effects on health. Together, the results suggest that studies which use genetic data will have an important role in the future of dental epidemiology and can help improve understanding of both the molecular and broader health and social aetiology of caries and periodontitis. Unlocking the full potential of these methods will require the community to support still-larger studies and adopt modern working practices in dental epidemiology.

Effective population size (Ne) is an important statistic in conservation science and in the broader topics of evolutionary genetics. Ne is often used to quantify the rate of evolutionary events such as losses in genetic diversity. Estimating and interpreting such quantity can however be challenging. Chapter 2 focuses on the change in allele frequency between two or more time points due to genetic drift. A new likelihood-based estimator N̂_B for contemporary Ne estimation is proposed by adopting a hidden Markov algorithm and continuous approximations. N̂_B is found to be several-fold faster than the existing methods without sacrificing accuracy. It also relaxes the upper bound of Ne to several million and which is currently limited to about 50000 due to computing limitations. Chapter 3 extends N̂_B to handle multialleleic loci through using Dirichlet-multinomial distributions. An R package is also provided and available for download. Chapter 4 explores the signatures of linkage disequilibrium (LD) between a pair of loci induced by genetic drift as a function of recombination rate and historical population sizes. E[r²] can be expressed as the weighted sum of the probability of coalescent at different time points of which information about Ne is contained. This relationship is verified by computer simulation and then applied to historical Ne estimation as illustrated in an example of Anopheles coluzzii population. A new likelihood-based routine Constrained ML is suggested in chapter 5 to estimate haplotype frequencies and r² from genotypes under Hardy-Weinberg Equilibrium. It is shown to be identical to existing EM algorithm under normal conditions but far less sensitive to initial conditions. A new “unbiased” sample size correction is also proposed to estimate r². To summarise, this work pushes the Ne estimation to its current boundary and more importantly provides suitable tools to analyse the ever-growing datasets.

This thesis explored the utility of genome wide association studies (GWAS) and meta-analysis to identify variants associated with serum pollutants levels and metabolic phenotypes. The secondary aims of these thesis were to annotate these variants with regulatory information from databases and to investigate the role of copy number variation in influencing gene expression and phenotype. In the second and third chapters, we identified single nucleotide polymorphisms (SNPs) associated with pollutant and metabolic phenotypes. An example of a novel association was the relationship between SNPs in the ABCG2 gene and octachlorodibenzo-p-dioxin (OCDD). In the third chapter, we identified associations between metabolic phenotypes and SNPs. An example of an identified association was that between serum apolipoprotein B levels and rs7412, which is consistent with other GWAS. To fine map these loci and assign functional annotation, I created Bayesian credible sets and checked for overlap between SNPs in these credible sets and regulatory marks in the Encyclopaedia of DNA Elements (ENCODE) database. Annotating these credible set SNPs with functional information revealed various histone modifications and transcription factors that overlapped. This study was also successful in identifying copy number variants (CNVs) from the PIVUS cohort. There were moderately strong associations between CNVs and some of the phenotypes studied. These associations did not appear to be mediated by the SNP within the CNVs, as the latter had higher P values of associations. In addition, I identified several clinically relevant CNV-expression quantitative trait loci (CNV-eQTL) associations a separate cohort of healthy individuals. Some of these associations were cell specific and/or context specific. Many of these CNVs contained SNPs which are lead SNPs in GWAS studies on a variety of different phenotypes. In conclusion, this thesis was successful in identifying SNPs and CNVs associated with phenotypes, as well as annotating some of these variants with regulatory information. Further work is needed to clarify the mechanisms of regulation.

Work in this thesis represents advances in addressing two key challenges in epidemiological and ecological modelling: the lack of an effective and easily deployable modelassessment tool and a statistically sound joint inferential framework for epidemic and evolutionary processes. Firstly, we present a novel statistical framework that combines classical and Bayesian reasoning in testing for mis-specifications of a spatiotemporal model by investigating the consistency of so-called latent residuals with a known sampling distribution using a classical hypothesis test. Second, we devise a statistically sound Bayesian framework which facilitates the integration of epidemiological and genetic data; specifically, we demonstrate how the transmitted sequences can be effectively imputed so that the transmission dynamics of the joint epidemic and evolution process can be accurately recovered and also any unsampled infected hosts can be naturally accommodated in the analysis. The new methodology we propose are assessed using simulation studies and they are applied to two real-world epidemic datasets which respectively describe the spread of an invasive plant and foot-andmouth disease in the UK, which shows that they may greatly enhance our ability to understand the transmission dynamics of disease and therefore lead to more efficient disease management.

Autism Spectrum Disorder (ASD) has a strong, yet heterogeneous, genetic component. Among the various methods that are being developed to help reveal the underlying molecular aetiology of the disease, one approach that is gaining popularity is the combination of gene expression and clinical genetic data, often using the SFARI Gene database, which comprises lists of curated genes considered to have causative roles in ASD when mutated in patients. We built a gene co-expression network to study the relationship between ASD-specific transcriptomic data and SFARI genes and analysed it at different levels of granularity, first as individual genes, then as clusters of genes, and finally as the complete co-expression network. No significant evidence was found of association between SFARI genes and differential gene expression patterns when comparing ASD samples to a control group, nor statistical enrichment of SFARI genes in gene co-expression network clusters that have a strong correlation with ASD diagnosis. However, classification models that incorporate topological information from the whole ASD-specific gene co-expression network can predict novel SFARI candidate genes that share features of existing SFARI genes and have support for roles in ASD in the literature. We demonstrate that only co-expression network analyses that integrate information from the whole network are able to reveal signatures linked to ASD diagnosis. These analyses successfully identify novel candidate genes associated with ASD where individual gene or cluster analyses fail. We also find a statistically significant association between the level of expression of SFARI genes and their SFARI Gene Score which confounds downstream analysis of ASD gene expression data. We present a novel approach to correct for this that is generalisable to other situations where analysis is affected by continuous sources of bias.

A growing population of cells accumulates genetic mutations. We study stochastic models of this process. Cells divide and die as a branching process, and a cell's genetic information is a sequence of nucleotides which mutates randomly at division. Motivated by biologically realistic parameters, we consider that few cells grow to many cells and mutation rates are small, proving approximations in this limit. In particular we are interested in mutation frequencies and their dependency structure along the genetic sequence; the relevance of the evolutionary tree and selection are discussed. Amongst other results, we recover a power-law distribution for mutation frequencies which is consistent with previously published cancer genetic data.

Major Depressive Disorder (MDD) is a disabling, common psychiatric disorder and the leading cause of global disability. A complex combination of genetic and environmental factors gives rise to MDD, although the exact aetiology has not been identified. Genome-wide association studies (GWAS) have established that MDD has a moderate heritability of approximately 37%. MDD has in the past also been associated with abnormalities of white matter microstructure, which represents the brain’s connectivity network. This network is also moderately heritable, providing rationale to investigate its relationship to MDD genetic risk. Over recent years, there has been considerable progress in establishing genetic contributions to MDD. These advances can be harnessed, in combination with neuroimaging and epigenomics, to understand the neurobiology of the disorder. This has only recently become possible at sufficient scale with the availability of large publicly available datasets including genomic, epigenomic, and neuroimaging data. In the current thesis, I therefore aimed to leverage genetic, epigenetic, and neuroimaging data in two large datasets, UK Biobank (N range: 6,400 – 14,800) and Generation Scotland: Scottish Family Health Study (N = 625). Specifically, I aimed to uncover links between white matter microstructure, as measured by fractional anisotropy and mean diffusivity, and (i) differential gene expression as indexed by expression quantitative trait loci (eQTLs) scores in chapter 2; here, decreased white matter integrity was found to be associated with 6 scores regulating genes previously reported to be implicated in neurological and neuropsychiatric disorders, while 2 scores regulating neurodevelopment-linked genes were associated with increased white matter integrity; (ii) MDD genetic risk stratified by the NETRIN1 Signalling Pathway, previously implicated in MDD, indexed by polygenic risk scores (PRS) in chapter 3; results indicated novel associations between the pathway-focussed PRS and decreased white matter integrity in thalamic radiations, as well as several association fibres, including superior and inferior longitudinal fasciculus; (iii) a novel wholegenome epigenetic risk score for MDD, which uncovered an association with MDD, but no significant associations with changes in white matter microstructure (chapter 4). The overall aim of the thesis was to use advanced genomic techniques to stratify genetic function and risk and explore epigenetic risk for MDD in order to identify novel links to structural brain connectivity. Overall, the three studies provide a strong rationale for integrating neuroimaging, genomic and epigenomic data. Specifically, findings in chapter 2 indicate the importance of DCAKD, SLC35A4, SEC14L4, SRA1, PLEKHM1, UBE3C, NMT1, and CPNE1, not previously found by conventional GWAS approaches. This suggests that integrating neuroimaging and genetic expression data may uncover novel associations that inform disease- or trait-specific genetic links to brain connectivity. Chapter 3 results provide a rationale for investigating the NETRIN1 Signalling Pathway and emphasise the role of thalamic connections in MDD within this biological pathway, indicating that novel associations with brain connectivity may be uncovered at a more focused level when stratifying MDD risk by biology. Finally, results from chapter 4 indicate that epigenetics play an important role in MDD risk, although further analysis including larger-scale epigenetic and neuroimaging data should be carried out to uncover the role of epigenetics in relation to brain phenotypes.

A reference interval represents the normative range for measurements from a healthy population. It can be interpreted as a prediction intervalfor a new individual from the overall population. The reference interval based on one study might not be applicable to a broader population. Meta-analysis can provide a more generalizable reference interval based on the combined population by synthesizing results from multiple studies. However, existing random effects methods may give imprecise estimates of the between-study variation with only a few studies. In addition, the normal distribution of underlying study-specific means, and equal within-study variance assumption in these methods may be inappropriate in some settings. In the first paper, we develop a mixture distribution method using the fixed effects model. It combines studies by assuming the overall population is a mixture of sub-populations comprised of individual studies. This mixture distribution method does not explicitly estimate the between-study heterogeneity, which is difficult for a random effects model with few studies. In the second paper, We propose a Bayesian nonparametric (NP) model with more flexible assumptions to extend random effects meta-analysis for estimating reference intervals. The simulation studies show the performance of the mixture distribution and NP approaches when the assumptions of normally distributed study mean and equal within-study variances do not hold. Both methods are applied to real datasets and provide more reasonable estimates for reference intervals compared with existing methods. The third paper focuses on developing a new Mendelian randomization (MR) approach, which leverages genetic data to estimate the causal effect of an exposure factor on an outcome from observational studies. We utilize genetic correlations to summarize information on a large set of genetic variants associated with the exposure factor. Our proposed two-stage random effects approach (TS-RE) can accommodate many weak and pleiotropic effects. Our approach quantifies the variation explained by all included instrumental variables instead of estimating the individual effects and thus could accommodate weak IVs. This is useful for performing MR estimation in small studies where the selection of valid IVs is unreliable and thus has a large influence on the MR estimation. Through simulation and real data analysis, we demonstrate that our approach provides a robust alternative to the existing methods.

Collection of large amounts of genetic data and advancements in computational genetics over the recent years provide us with tools to explore epigenetic mechanisms that lead to aging and lifespan. In the context of continuous DNA methylation data, with a novel cross-species DNA methylation microarray targeting conserved CpG sites across mammalian species, we are able to leverage readily available statistical models to extensively study important life history traits such as lifespan, gestation time, and time to sexual maturity across various species. DNA methylation data are often high dimensional and require regularized regression frameworks to construct practical prediction models. Based on an unprecedented mammalian DNA methylation data set, we have developed methylation-based epigenetic life history traits predictors using regularized linear regressions. The estimators can accurately predict maximum lifespan using cytosine methylation patterns collected from over 13,000 samples derived from 348 mammalian species. To extend our future inferential analyses into diverse data sources such as RNA-seq data, we have proposed an L0-regularized Poisson graphical model for exploring gene-to-gene relations. The superior theoretical properties that the L0 sparse graphical model enjoys will more effectively assist the future work of clustering and grouping large numbers of DNA methylation sites and genes. Both the applied research and methodological work will aid in the aging research goals of integrating various layers of multiomics data.

Conservation and management of wild populations requires management professionals to devise methods for collecting data that are efficient and accurate. For example, understanding population sizes and stability is important to predicting extinction risk but the lack of comparisons of costs and effectiveness of different methods limits efficient assessment of wild populations. Similarly, monitoring and tracking the spread of wildlife diseases is important for conservation and management, however disease samples and other data have to be collected efficiently and in ways that maximize our understanding of the disease-wildlife systems. In my thesis, we addressed these research gaps in three studies. First, we conducted a meta-analysis of peer-reviewed studies that compared two or more monitoring. This provided an insight into the quantitative differences that are expected among methods. Next, we developed a forward-time, agent-based model to compare different approaches for collecting wildlife data, including data needed to quantify population size or density or understand disease presence. We found that the stationary approaches have more detections or sampling events compared to the mobile approaches, but both approaches have equal number of unique detections (i.e. unique individuals detected or sampled). Our model also suggested that some sampling approaches may be better suited for very large populations compared to other sampling schemes. Finally, we reviewed our current understanding of avian malaria with particular focus on recent advances in understanding this disease and its effects on wildlife and future efforts to control further spread. The results of this thesis will allow conservation practitioners and managers to identify the most effective monitoring techniques for their species of interest, and consideration of more efficient sampling approaches for wildlife diseases.

In recent years, electronic health records (EHR) have been combined with genetic data to uncover disease biology and accelerate generation of hypotheses for drug development and treatment strategies. The goal of this dissertation is to develop novel statistical models that can address the challenges of analyzing ‘imperfect’ EHR data and to propose privacy-preserving methods that enable sensitive individual-level data sharing across EHR studies and other large genetic studies. In Chapter II, we propose a statistical method to address misclassified clinical outcomes, a common challenge in EHR data. One essential step of EHR-based genome-wide association studies is constructing a cohort of cases and controls for a specific disease from billing codes and other clinical or administrative data. Nearly always, a perfect strategy for deriving disease phenotypes from billing codes is not available, resulting in some incorrect case/control labels. Here, we propose a method to estimate the misclassification of case/control status by examining genotype information of dozens of disease associated loci. Through simulation and application to the Michigan Genomics Initiative data, we demonstrate that the method enables the evaluation of new EHR-based phenotype definition schemes and provides accurate estimates of disease association measures when phenotypes are misclassified. In Chapter III and IV, we focus on identifying overlapping samples between studies, a common challenge when aggregating information across datasets. We particularly focus on identifying duplicate or related samples when sharing the underlying individual level genetic data is restricted. We propose methods that do not require disclosure of individual identities but that can still identify genetic relatives across datasets. In Chapter III, we show that by grouping genotypes into segments and calculating summary statistics within each segment, we are able to obscure and encode individual-level genetic information. Relatives can be inferred with the coded genotypes using a likelihood model. Simulation and application to the Trans-Omics for Precision Medicine (TOPMed) program data demonstrate the utility and security of the method. In Chapter IV, we extend the method further, with a strategy that guarantees stronger encryption and is expected to work across heterogeneous populations. This secure protocol can infer genetic relatives among people of diverse ethnic backgrounds. The method works by combining a cryptographic technique, homomorphic encryption, with the robust relationship inference method previously described by Manichaikul et al (2010). Through simulations, we show that our method's performance is identical to that of implementations that use the original unencrypted genotypes. Our protocol scales well in computing time and is protected from several possible attacks. The secure protocol was again applied to TOPMed dataset. Securely identifying related samples will facilitate combination of results across datasets when there are restrictions to sharing the underlying individual level data. In conclusion, the methods developed here well enhance use of EHR data and genome data to improve accuracy of case/control status as well as decrease inclusion of relatives across studies when desired.

Over the last two decades, technological improvements have led to a tremendous reduction in the cost and speed of DNA sequencing. This has opened the door to many new applications, including the quantification of transcriptomes at the resolution of single cells (scRNA-Seq), and the discovery of genetic features associated with phenotypes of interest, also known as genome-wide association study (GWAS). scRNA-Seq has emerged in just 10 years as a major tool to investigate biological diversity. The ability to assess gene expression for individual cells enables the study of a range of biological processes at new levels of resolution. This can have many interesting applications, including the identification of novel cell types, defined by their unique transcriptomic signatures. Drawing from the clustering literature, many methods have been developed to group cells together based on their gene expression characteristics. However, all those algorithms require the tuning of hyper-parameters based on typically ad hoc recommendations. Moreover, the direct validation of the discovered cell types is generally difficult, if not impossible. The grouping of cells is also not unique for a given biological system, and there often exists a hierarchy of cell types, with ever-finer levels of resolutions. Because of all this, the discovery of reliable, replicable cell types remains a major challenge. In Chapter 2, we will delve deeper into this issue and introduce a new method called Dune that tackles the resolution-replicability trade-off in clustering. scRNA-Seq data also enable the tracking of continuous developmental changes, without the need for arbitrary discretization that stemmed purely from the data collection protocol. This allows to investigate processes such as the cell cycle, the differentiation of stem cells into different cell types, or the cellular response to a drug over time. In Chapter 3, we will investigate how to characterize patterns of gene expression along such developmental trajectories, to identify dynamic genes and drivers of differentiation, using the tradeSeq method. In Chapter 4, we will provide a general workflow called condiments for analyzing such dynamic systems in the presence of multiple conditions, such as treatment/control. GWAS represent another field that has gained major attention following the emergence of cheaper high-throughput sequencing technologies. In human populations, the problem has been extensively studied, mainly in the context of diseases such as diabetes. However, GWAS can also be applied to bacterial genomes, especially in the context of antibiotic resistance. Some concepts from the human GWAS literature are applicable in bacteria. However, characteristics of bacterial genome mean that other concepts, such as that of a reference genome, are inappropriate and irrelevant. New methods need to be developed for this specific problem. In Chapter 5, we present a new subgraph enumeration method named CALDERA that leverages the structure of the data to provides more robust analyses and facilitate the interpretation of bacterial GWAS data.

With the advancement of next-generation sequencing technology, numerous disease/ phenotypic associations with the human microbiome and human genome are uncovered and revealed. In this dissertation, we take advantage of this and explore these association patterns using machine learning methods. We first design a deep learning method MDeep for microbiome-based prediction by considering both the taxon abundance and phylogenetic tree. MDeep models the taxonomic rank by the convolutional layers and captures the phylogenetic correlation on each taxonomic rank via the convolutional operation. Our simulations and real data analysis demonstrate that MDeep outperforms competing methods in both regression and binary classifications. In order to explore the diseases/ phenotypic associations with the human genome, we propose two machine learning frameworks. The first framework, WEVar, is a supervised learning framework by integrating the pre-computed scores from representative existing scoring methods, which will benefit from each individual method by automatically learning the relative contribution of each method and produce an ensemble score for the final prediction. Using simulation and real data studies, we show both context-free WEVar and context-dependent WEVar outperform the individual scoring methods on the state-of-the-art benchmark datasets. Furthermore, we find WEVar can prioritize experimentally validated non-coding variants in an LD block. The second framework, DeepMFIVar, is a deep multimodal learning framework for the functional interpretation of genetic variants. DeepMFIVar learns a predictive model linking DNA sequence context and clinical information to quantitative epigenetic signals. The mutation effect of the 210 million genetic variants are generated by the difference of the predicted epigenetic signal for the reference and for alternative alleles. The application to DNA methylation and histone modification demonstrates that DeepMFIVar can accurately predict locus-specific epigenetic signals using DNA sequence and clinical information, and it is also capable of prioritizing variants for downstream experiments.

Sickle cell anemia is caused by a single mutation in the β-hemoglobin gene, HBB. The disease originated in Africa and affects millions of people worldwide. Sickle hemoglobin tetramers polymerize upon deoxygenation and lead to hemolysis and vaso-occlusion. Patients with high fetal hemoglobin (HbF) can have milder disease. The only FDA-approved drug is hydroxyurea that increases HbF. HbF modulates the disease by preventing the polymerization of sickle hemoglobin and reduces the pain episodes, anemia, and organ damage associated with the disease. There are five common haplotypes associated with the HbS gene and that are very loosely associated with disease severity and HbF. Understanding the genetic bases of HbF regulation is a key factor to identify potential drug targets to induce HbF for therapeutic purposes. To fully understand the mechanism behind HbF regulation, developing a fast and accurate computational method for sickle cell haplotype classification is useful for examining the variability of HbF among sickle cell patients. Moreover, investigating the cis and trans-acting regulators of HbF gene expression to pinpoint the mechanism through which they regulate HbF is essential to develop a successful treatment. The availability of high-throughput genetic data provides an excellent opportunity to study HbF regulation in sickle cell patients and normal people comprehensively. The work reported in this thesis describes a fast and accurate method for sickle cell HBB haplotype classification. I also examine the differential effect of cis and trans-acting HbF hemoglobin regulators on -globin gene expression using the GTEx database and identify BCL2L1 as a new potential trans-regulator of HbF.

The advent of time- and cost-effective technologies for genotyping and sequencing human DNA has massively increased both the type and amount of genetic data available for study. In order to best utilize this data, new methods must be developed to better assess how human history affects genetics and how genetics affects human phenotypes such as height, eye color and disease risk.This work presents five new methods that build upon each other to address this challenge. The first method leverages geographic information contained in rare genetic variation to infer the genetic ancestry of individuals at each location in the genome. It increases ancestry inference accuracy when applied to cohorts of continentally admixed individuals. This method also allows inference of local ancestry when studying cohorts containing subcontinentally admixed individuals.The second method applies the idea of highly structured geographic information in rare variation to create a better variant filtering approach for finding the causal variation in monogenic disorders. By finding better estimates of allele frequencies both within and across populations, it reduces the number of variants that must be considered as potentially disease causing. This results in decreased time and cost expenditures in necessary follow-up analyses.Due to multiple testing issues, compound heterozygous architectures and haplotype affects are difficult to detect as contributing to complex diseases or gene regulation. The next two methods present ways to detect these complex features. Compared to standard marginal association approaches, these two methods show that compound heterozygous architectures and haplotype effect models often better capture the genetic contributions to traits. The results demonstrate the need for future fine-mapping approaches that seek complex causal architectures.The final method in this work searches for causal relationships in gene expression networks. These networks are formed by genes with highly correlated expression levels. However, the correlation may be due to unobserved confounding variables. By utilizing genetic variants as instrumental variables, this method finds causal gene-on-gene effects. Knowing the direction and magnitude of gene-on-gene effects is vital to better understanding regulatory networks in disease pathways and for the identification of drug targets.

While technological innovation has dramatically increased the amount and variety of genomic data available to geneticists, no assay is perfect and both human error and technical artifacts can lead to erroneous data. A proper analysis pipeline must both detect errors, and, if possible, correct them. One common source of errors in genetic data is sample-to-sample contamination. This dissertation will identify methods to address contamination in the most common types of genetic studies. Chapter 2 focuses on methods for detecting and quantifying contamination in both array-based and next-generation sequencing (NGS) genotype data. For the array-based data, we use the observed intensities from the genotyping instruments to quantify contamination with two distinct methods: 1) a regression-based model using intensities and population allele frequencies and 2) a multivariate normal mixture model that looks at the clustering of intensities. For NGS data, we model the reads using a mixture model to determine the proportion of reads from the true sample and the contaminating sample. Chapter 3 outlines a method to make accurate genotype calls with contaminated NGS data. Given an estimated level of contamination, we propose a likelihood that can be maximized to call genotypes and estimate allele frequencies for samples with no previous genotype data. We investigate the method from data from two common sequencing strategies: 1) low-pass (2-4x depth) genome-wide sequencing and 2) high-depth (50-100x depth) exome sequencing. Chapter 4 looks at contamination in the context of RNA sequencing (RNA-Seq) data. While the technology to generate RNA-Seq data is similar to exome sequencing, the difference in expression between the contaminating and true sample makes it more difficult to accurately estimate the contamination proportion. We propose methods to improve the quality of these estimates.

When an epidemic moves through a population of hosts, the process of transmission may leave a signature in the genetic sequences of the pathogen. Patterns in pathogen sequences may therefore be a rich source of information on disease dynamics. Genetic sequences may replace or supplement other epidemiological observations. Furthermore, sequences may contain information not present in other datatypes, opening the possibility of inferences inaccessible by other means. The field of phylodynamic inference aims to reconstruct disease dynamics from pathogen genetic sequences. Although a wide variety of phylodynamic inference methods have been proposed, most methods for fitting mechanistic models of disease operate in two disjoint steps, first estimating the phylogeny of the pathogen and then fitting models of disease dynamics to properties of the estimated phylogeny. Logical inconsistency in demographic assumptions underlying the two stages of inference may create bias in resulting parameter estimates. Joint inference of disease dynamics and phylogeny ensures consistent assumptions, but few methods for joint inference are currently available. The central work of this thesis is a new method for joint inference of disease dynamics and phylogeny from pathogen genetic sequences. This likelihood-based method, which we call genPomp, allows for fitting mechanistic models of arbitrary complexity to genetic sequences. The organization of this thesis is as follows. In Chapter I, we present background on the field of phylodynamic inference. In Chapter II, we use simulation to study a two-stage inference approach proposed by Rasmussen et al. (2011). We find that errors in phylogenetic reconstruction may drive bias in two-stage phylodynamic inference. This result underscores the need for methodology for joint inference of the transmission model and the pathogen phylogeny. In Chapter III, we propose a flexible method for joint inference and demonstrate the feasibility of this method through simulation and a study on stage-specific infectiousness of HIV in Detroit, MI. This method is comprised of a class of algorithms that use sequential Monte Carlo to estimate and maximize likelihoods. In Appendix A we show theoretical support for our algorithms. In Chapter IV, we demonstrate the flexibility of our approach by developing a model of transmission of Vancomycin-resistant enterococcus in a hospital setting. To allow for fitting this model to patient-level data we developed a targeted proposal, detailed in Appendix B. We present exploratory analysis of a hospital outbreak at NIH that motivates the form of the model, and carry out a study on simulated data. Although some assumptions of the simulated example are unrealistic, these initial results will inform future efforts at fitting real data. In Chapter V, we summarize the progress represented in this thesis and consider possibilities for future work.

With the availability of genotyping data of very large samples, there is an increasing need for tools that can efficiently identify genetic relationships among all individuals in the sample. Modern biobanks cover genotypes up to 0.1%-1% of an entire large population. At this scale, genetic relatedness among samples is ubiquitous. However, current methods are not efficient for uncovering genetic relatedness at such a scale. We developed a new method, Random Projection for IBD Detection (RaPID), for detecting Identical-by-Descent (IBD) segments, a fundamental concept in genetics in large panels. RaPID detects all IBD segments over a certain length in time linear to the sample size. We take advantage of an efficient population genotype index, Positional BWT (PBWT), by Richard Durbin. PBWT achieves linear time query of perfectly identical subsequences among all samples. However, the original PBWT is not tolerant to genotyping errors which often interrupt long IBD segments into short fragments. The key idea of RaPID is that the problem of approximate high-resolution matching over a long range can be mapped to the problem of exact matching of low-resolution subsampled sequences with high probability. PBWT provides an appropriate data structure for bi-allelic data. With the increasing sample sizes, more multi-allelic sites are expected to be observed. Hence, there is a necessity to handle multi-allelic genotype data. We also introduce a multi-allelic version of the original Positional Burrows-Wheeler Transform (mPBWT). The increasingly large cohorts of whole genome genotype data present an opportunity for searching genetically related people within a large cohort to an individual. At the same time, doing so efficiently presents a challenge. The PBWT algorithm offers constant time matching between one haplotype and an arbitrarily large panel at each position, but only for the maximal matches. We used the PBWT data structure to develop a method to search for all matches of a given query in a panel. The matches larger than a given length correspond to the all shared IBD segments of certain lengths between the query and other individuals in the panel. The time complexity of the proposed method is independent from the number of individuals in the panel. In order to achieve a time complexity independent from the number of haplotypes, additional data structures are introduced. Some regions of genome may be shared by multiple individuals rather than only a pair. Clusters of identical haplotypes could reveal information about the history of intermarriage, isolation of a population and also be medically important. We propose an efficient method to find clusters of identical segments among individuals in a large panel, called cPBWT, using PBWT data structure. The time complexity of finding all clusters of identical matches is linear to the sample size. Human genome harbors several runs of homozygous sites (ROHs) where identical haplotypes are inherited from each parent. We applied cPBWT on UK-Biobank and searched for clusters of ROH region that are shared among multiple. We discovered strong associations between ROH regions and some non-cancerous diseases, specifically auto-immune disorders.

More than 26 million individuals have submitted to direct-to-consumer genetic testing (DTCGT) services since the personal genomics industry gained traction in the late 2010s. Private firms AncestryDNA and 23andMe dominate the market, and as a result, hold ownership of some of the largest collections of human genetic data in the world. As a novel technology that analyzes consumer disease risk or genetic genealogy, DTCGT has emerged alongside a number of ethical, legal, and social implications that require scholarly investigation. This study places a critical eye on AncestryDNA and 23andMe to examine how DTCGT is contextualized in the news media. Using a Foucauldian discourse analysis, 50 articles from The New York Times, The Wall Street Journal, and USA Today were analyzed to capture present discourses and ethical critiques of DTCGT. Results of this study cast the genetic testing consumer as self-managing, datafied, and valuable. Consumers’ quests to develop their genetic knowledge form the foundation of a bioeconomy that private firms and government entities use to advance research and innovation while implicitly reiterating neoliberal notions of autonomy, productivity, control, and individual responsibility. Issues of third-party sharing of genetic data, informed consent, risks and uncertainties about testing, genetic privacy, and more are discussed as they are presented in the news media. In light of these critiques, this analysis demonstrates a need for regulatory action and scholarly analysis on how DTCGT is contextualized as society becomes increasingly datafied and risk-aware.

Background: Cigarette smoking is the leading preventable risk factor of morbidity and mortality in the world. Many studies have examined the association between smoking and health outcomes. Observational, cross-sectional studies can be confounded, and hence the casualty of associations between smoking and health outcomes cannot be established. A genetic epidemiological approach such as Mendelian randomization (MR) can be informative concerning potential causal associations between smoking and health outcomes. MR leverages the availability of genetic data on smoking and health outcomes to estimate confounder-free associations. The current thesis was carried out to investigate the observational and causal associations between smoking behaviour and cardiometabolic diseases, stroke, and lipid biomarkers. Methods: Firstly, detailed reviews of prior research were conducted, highlighting that the majority of previous research was observational in nature. The thesis utilised the relatively large sample of UK Biobank (N=~502k) to conduct observational as well as MR-based analyses. The observational approach was based on self-report for multiple smoking phenotypes (smoking status, smoking intensity, and age at smoking initiation), clinical diagnoses for cardiometabolic diseases (CMDs; coronary heart disease (CHD), hypertension (HTN), and diabetes mellitus (DM)), stroke, and lipid biomarkers (total cholesterol, low-density lipoproteins, triglycerides, and high-density lipoproteins). The genetic analysis was based on 14 single-nucleotide polymorphisms (SNPs) for smoking intensity (cigarettes smoked per day: CperD) and 15 SNPs for smoking history. The genetic analysis was conducted in the UK Biobank sample (one-sample MR) as well as publicly available 'summary statistic' genetic data (two-sample MR). The analyses were conducted using R software and the MR-Base platform. Results: Observationally (analysis: chapter four), current smokers had a higher risk of CHD (odds ratio [OR]: 1.61, P<0.001), stroke (OR: 1.64, P<0.001), and DM (OR: 1.12, P<0.001), and lower risk for HTN (OR=0.89, P<0.001) compared to never smokers. Additionally, as individuals smoke one more cigarette per day on average (smoking intensity), the risk for all CMDs increases (CHD, stroke, and HTN: OR=1.01, DM: 1.02, all P<0.001 per average daily cigarette). Finally, as an individual initiates smoking one year later in life, the risk of all CMDs decreases except for HTN (CHD and stroke: OR = 0.96, P<0.001, DM: OR=0.99, P > 0.05, and HTN: 1.01, P<0.001). For lipid biomarkers (analysis: chapter five), current smokers showed higher levels of cholesterol (β: 0.05 mmol/L, P<0.001), LDL (β: 0.06 mmol/L, P<0.001), and TG (β: 0.09 mmol/L, P<0.001), and lower level of HDL (β=-0.14 mmol/L, P<0.001) compared to never smokers. Similarly, as individuals smoke one more cigarette per day (smoking intensity), the levels of cholesterol, LDL, and TG increase, and the level of HDL decreases (cholesterol: β=0.02 mmol/L, LDL: β=0.03 mmol/L, TG: β=0.02 mmol/L, and HDL: β=-0.04 mmol/L, all P<0.001). Lastly, as an individual starts to smoke one year later in life, the levels of all lipid biomarkers increase except for TG (cholesterol: β=0.01 mmol/L, P=0.026, LDL: β=0.001 mmol/L, P > 0.05, TG: β=-0.01 mmol/L, P<0.001, HDL: β=0.04 mmol/L, P<0.001). In terms of MR-based causal estimates (analysis: chapter four), there was no evidence of any causal relationship between smoking behaviour variables with CHD, stroke, and lipid biomarkers (analysis: chapter five) in the UK Biobank sample (one-sample MR) nor in other samples or approaches (summary-level in MR-Base platform or R). The only significant causal associations were observed in two isolated MR analyses; one between smoking status (ever) and HTN in one sample MR in the UKB sample and the other was between smoking intensity (CperD) and DM in two sample MR in R. Conclusion: The observational findings indicated that cigarette smoking increases the risk of CHD, stroke, DM, and levels of total cholesterol, LDL, and TG observationally, but this was not supported by 'causal' genetic evidence. Smoking behaviour seems to be associated with lower blood pressure (observationally and genetically) and HDL levels (observationally, not genetically). Finally, findings on HTN, cholesterol, LDL, and HDL have varied depending on the smoking variable. These ambiguous findings point toward some of smoking's association with poorer health perhaps being due to poor lifestyle generally and not smoking itself in isolation. Evidence of potentially protective findings of smoking is likely to be driven by instrumentation or attrition bias. More research is needed to meticulously determine the impact of each smoking variable on health outcomes, both observationally and genetically.

Major depressive disorder (MDD) is a common psychiatric disorder and one of the leading causes of disability worldwide. MDD is moderately heritable (~30­40%), suggesting both genetic and environmental effects are influential. Several strands of evidence point to the possible presence of gene­environment correlations and gene­environment interaction effects in MDD, although findings to date have been relatively inconsistent. Recently, research using available genetic and environmental data on biological parents and offspring (trios), have shown that parental genetic nurturing effects are detectable using polygenic scores (PGSs) in more heritable traits such as educational attainment. Research findings also point to potential gene­by­trauma exposure interaction effects involved in MDD. It is evident that data limitations may result in power issues, confounding effects and biases, which impact reliable and accurate quantification of these effects, highlighting the need for methods that maximise statistical power and minimise bias and confounding when exploring these effects. This thesis aims to adapt existing statistical frameworks and use of data to explore gene­environment interplay effects, which are robust to the limitations of the available data. Here, two large population­scale datasets, the UK Biobank (N~150,000) and Generation Scotland: Scottish Family Health Study (N~2680 trios), were utilised to explore genome­by­trauma exposure interaction effects in depression, as well as parental genetic nurturing effects in a range of traits including MDD. The research aims of this thesis included (1) implementing models exploring parental genetic nurturing effects using available trio PGSs; (2) expanding these models to explore mechanisms of genetic nurturing effects with available parental phenotypic data, both addressed in chapter 2. Here, the quantification of parental genetic nurturing effects using trio PGSs was found to be reliable and robust to data limitations. However, expanding these models by including parental phenotypes resulted in confounded effects. Simulation analyses demonstrated that the confounding was induced by power issues associated with PGSs, highlighting the need for improved measures of genetic variance. The final research aim (3) was to explore genome­by­trauma exposure interaction effects using relationship matrices capturing genetic, trauma exposure and genome­by­trauma exposure interaction similarity between participants. Genomic relationship matrices utilised all available genetic data, and thus, served as an improved representation of genetic variance. Environmental relationship matrices utilised principal components, capturing underlying dimensions of trauma exposure. A substantial proportion of MDD variance was found to be attributed to genetic, trauma exposure and genome-­by-trauma interaction effects. However, little insight was inferred regarding the specificity and direction of trauma exposure involved in MDD manifestation, due to the difficulty in interpreting the underlying dimensions of trauma exposure. The two studies provide a strong rationale for the use of improved measures of genetic and environmental components of MDD. Specifically, findings from chapter 2 highlight the need for measures that capture a substantial proportion of genetic variance to explore these complex gene­environment interplay effects. Chapter 3 results demonstrated how leveraging all available genetic data can uncover substantial effects that were previously missed. Evaluations of study designs highlight how future work can incorporate omics (e.g. methylation) data to improve measures of environmental factors, which would aid translational interpretation of results obtained from these models. Methylation data could help identify additional trait associated genetic loci as well as their mode of action, and hence, specific drug targets; which may not have been previously identified by traditional analyses such as genome­wide association studies (GWAS) not containing the relevant genetic variants.

Morphometric traits are some of the most extensively-studied quantitative traits, with easy-to-measure exemplars-such as height and weight-collected in a variety of studies as a matter of routine. Recent efforts by national population cohorts, such as the UK Biobank, have-in some cases for the first time-made it possible for researchers to obtain information on other, harder-to-measure, morphometric traits, in a large population. The genetic data that is also available for large cohorts, such as the UK Biobank, has enabled the study of how such traits are genetically determined. Many morphometric traits have complex architectures, involving hundreds-or thousands-of genetic variants, most of which have small effects; it is generally not possible to detect these small effects in smaller cohorts, which lack sufficient power. The project described herein considers several morphometric aspects of the human skeleton in the context of variance in skeletal dimensions-e.g., the lengths of the long bones, hip (acetabular) width, shoulder width, and torso length-after correcting for overall body height. The aims of the project are: (1) to investigate the genetic components of the phenotypic architecture of skeletal dimensions relative to height; (2) to determine whether the same portions of the genome that affect skeletal structures during development are responsible for such relative variation; (3) to investigate the role sex-specific effects might play in such phenotypes; and (4) to demonstrate that there are novel ways to use existing data from cohorts in order to enable the investigation of heretofore unconsidered phenotypes, e.g., using image analysis methods with medical images to obtain morphometric measures of skeletal structures that can not be directly measured. Using image data from two population cohorts, Orcades and the UK Biobank, and a novel application of image processing, I have extracted the aforementioned skeletal measures from dual-energy, x-ray absorptiometry (DXA) images-typically used to measure bone mineral density. I then combined this phenotypic data with the cohorts' genetic data to investigate the genetic contribution to variation in these traits. My results show that there is a genetic basis for residual variation in skeletal measurements after accounting for height-with heritabilities ranging from 33% for radius length, to 58% for tibia length. They also show that this is driven by different variants than those affecting height-genetic correlations of the residual skeletal measurements with height range from −.0391 for fibula length, to .1621 for acetabular width-and that are located in genomic regions other than those involved in limb morphogenesis. There are also indications of associations with a particular trait that exist for one sex, and not the other, or that are much stronger for one sex. These may indicate an autosomal basis for sex differences in these traits-which may be due to a different set of loci than those affecting trait variation-although replication should be sought in another cohort. Pending replication of significant genetic associations, further study looking into potential confounding factors may be necessary to fully understand these differences. Finally, the work herein shows the benefit of collaboration with experts in other fields, such as image analysis, for genetic research. In this case it has made possible the derivation of novel phenotypes-that are of interest for research-from existing data that is otherwise not widely used. This kind of cross-disciplinary collaboration could help cohort maintainers increase the value of their resources, without having to expend more funds on additional equipment, or lab work.

The evolutionary ecology of a species is driven by a combination of random events, ecological and environmental mechanisms, and social behavior. Gene-culture coevolutionary theory attempts to understand the evolutionary trajectory of a species by examining the interactions between these potential drivers. Further, our choice of data type will affect the patterns we observe, therefore by integrating several types of data we achieve a holistic understanding of the various aspects of evolutionary ecology within a species.In order to understand population structure in short-finned pilot whales, I use a combination of genetic and acoustic data to examine structure on evolutionary (genetic) and cultural (acoustic) timescales. I first examine structure among geographic populationsin the Pacific Ocean. Using genetic sequences from the mitochondrial control region, I show that two genetically and morphologically distinct types of short-finned pilot whale, described off the coast of Japan, have non-overlapping distributions throughout their rangein the Pacific Ocean. Analysis of the acoustic features of their social calls indicates that they are acoustically differentiated, possibly due to limited communication between the two types. This evidence supports the hypothesis that the two types may be separate species orsubspecies.Next, I examine structure among island communities and social groups within the Hawaiian Island population of short-finned pilot whales. Using a combination of mitochondrial and nuclear DNA, I showed that the hierarchical social structure in Hawaiianpilot whales is driven by genetic relatedness; individuals remain in groups with their immediate family members, and preferentially associate with relatives. Similarly, social structure affects genetic differentiation, likely by restricting access to mates. Acousticdifferentiation among social groups indicates that social structure may also restrict the flow of cultural information, such as vocal repertoire or dialect. The qualitative correlation between social structure, cultural information transfer, and genetic structure suggest that gene-culture coevolution may be an important mechanism to the evolutionary ecology of short-finned pilot whales. Further research may reveal asimilar structure in the transmission of ecological behaviors, such as diet preference, habitat use, or movements. The results of this research underscore the applicability of gene-culture coevolutionary theory to non-human taxa.

Humans are driving the sixth mass extinction causing biodiversity to decline at an unprecedented rate. To halt these declines effective conservation strategies are vital, if possible, these should include genetic data to ensure that the extinction risk of a population is not being underestimated. This thesis presents resources and tools that have been used to integrate genetic data into the management of the endangered pink pigeon (Nesoenas mayeri). A pseudo-chromosome assembly enabled the first whole genome analyses of the pink pigeon (Chapter 2), these analyses provided insight into their past demography and revealed a surprisingly large amount of variation within its genome. This challenges previous results obtained using restriction site associated DNA sequencing (RAD-Seq) data from wild pink pigeons but current methods for processing RAD data produce biased data sets that under-estimate diversity. A novel tool, RADiKal is presented (Chapter 3), which extracts information directly from raw RAD reads and avoids biasing data sets with complex parameterisation. The painted chromosomes produced by RADiKal provide an overview about the levels of variation present in the wild population which are comparable to those observed from whole genome analyses. Despite increasing access to genetic data one of the greatest challenges is its integration into management, one solution is the use of population viability analysis (PVA). An updated PVA was produced for pink pigeons showing that without genetic rescue they could face extinction within 100 years (Chapter 4). Selecting which individuals are most valuable for a genetic rescue is not trivial, especially in the absence of empirical data. I Choose You (I.C.Y) is an easy-to-use tool that allows practitioners to select individuals for genetic rescue based on their genetic diversity measured using founder equivalents calculated from studbook data (Chapter 5). Overall this thesis aims to demonstrate how better tools can lead to better resources and better conservation.

Genetic diversity is considered one of the three main pillars of biodiversity, yet to date conservation policy has largely overlooked monitoring and protecting it in favour of species and ecosystem-based targets. Populations may experience considerable genetic erosion without extinction, and this loss can take considerably longer to recover from than population size alone. Low genetic diversity can have detrimental effects on the evolutionary potential of species and the viability of populations, yet population recovery at the genetic level remains poorly understood in the wild. In this thesis, I use large scale spatio-temporal sampling to explore the recovery of the otter population across the UK at a genetic level, after a well-documented decline in the latter half of the 20th century. Successive national surveys have documented increased occupancy rates, from small fragmented stronghold areas to a near continuous distribution in the 21st century. Firstly, I show that despite increased gene flow between sub-populations over the last 20 years, significant population structure persists, and intra-regional genetic diversity has not increased. Secondly, I use a landscape genetics approach, finding that favourable habitat is the most significant environmental factor affecting functional connectivity of otters across the Wales and Borders region, with implications for conservation management. Thirdly, I show that despite a near contiguous distribution and assumed population recovery, effective population sizes remain significantly below those required for long-term population viability in regional otter populations, and in some cases even raise concerns over short-term viability. This thesis serves as an important demonstration of the practical value of genetic data, even for populations where a significant (non-genetic) monitoring program has existed for several decades. The differing conclusions regarding population status, when based on genetic data rather than occupancy data, highlight the inherent dangers in assuming that spatial connectivity is an indicator of genetic connectivity.

Polymyalgia rheumatica and giant cell arteritis are two inflammatory disorders that are more common among female individuals with Scandinavian ancestry. The conditions co-occur in the same individual more often than expected by chance, share common symptoms and are difficult to diagnose. Earlier genetic studies identified a genetic association for both diseases with the HLA-DRB1*04 allele. Non-genetic data from UK Biobank were used in order to explore association with factors related to the susceptibility and severity of these diseases. Additionally, genetic data from UK GCA consortium were used in order to explore for potential genetic differences among clinically diagnosed GCA cases with different results from temporal artery biopsy (TAB), a gold standard method of diagnosis. Finally, genetic data from UK Biobank were used for the genome-wide association study of PMR cases. The results from the genetic analysis of PMR and GCA cases were further analysed for the identification of potential cross-phenotype associations. Cases with PMR and GCA were more likely to report a previous diagnosis of hypothyroidism, transient ischaemic attack and hypertension. Female cases with PMR were more likely to have ever received hormone replacement treatment and female cases with GCA to have an early menopause (<43 years). GCA cases were found to have a strong association with HLA-DRB1*0401 and PMR cases with HLA-DRB1*0404. The estimated proportion of patients misdiagnosed as GCA was 67% in the negative-TAB group and 33% in the group without TAB result. Meta-analysis results reveal a novel association with rs148462291 in chromosome 3. Pair-wise analysis of PMR and GCA data suggest this variant is associated with both PMR and GCA. There was no evidence that shared genetic risk factors were due to subgroup heterogeneity within PMR cases, although this analysis lacked statistical power.

The analyses contained herein focus on making comparisons between model inferences obtained using different scales of pathogen identification, with a particular focus on respiratory syncytial virus (RSV). A significant proportion of lower respiratory tract infections in children has been attributed to infection by RSV and as such, there has been global interest in understanding its transmission characteristics in order to plan for effective control. Mathematical models have often been used to explore potential mechanisms that drive the patterns observed in data collected at different scales. Several models have been used to explore how immunity to RSV is acquired and maintained, vaccination strategies and potential drivers of seasonality. However, most of these models do not make a distinction between the two antigenically and genetically distinct RSV groups (RSV A and RSV B), neither do they consider its ecological environment, in particular, potential interactions between RSV and other viral pathogens. This thesis therefore presents work done aimed at understanding the transmission characteristics of viral respiratory pathogens spreading in a group of households using a dynamic model of transmission The data analysed is cohort data collected between December 2009 and June 2010 from 493 individual distributed across 47 households from a rural coastal community in Kenya. Individuals in the study had nasopharyngeal swab samples collected twice weekly irrespective of symptom status. Infecting viral pathogens were identified using RT-PCR resulting in the identification of 4 main pathogens: RSV, human coronavirus, rhinovirus and adenovirus. RSV and coronavirus were further classified according to genetically distinct subgroups. Some of the RSV samples were sequenced to obtain whole genome sequences (WGS) and further classified into genetic clades/clusters. I first conducted a review of methods to identify the best way to integrate socialtemporal data and WGS genetic data into a single modelling framework for RSV. Given that the social-temporal data and genetic data were available at different sampling densities, I decided to use a model that focused on the data with the highest density. The results in this thesis are thus presented in three main chapters; the first focuses on analysing social-temporal shedding patterns of RSV identified at the group level (i.e. distinguish between RSV A and RSV B); the second incorporates the available genetic data into the model used to analyse the social-temporal data (i.e. separating RSV-A into 5 clusters, and RSV-B into 7 clusters); the third is an analysis of the interaction of two pathogens, RSV and coronavirus, identified at two different scales. One of the main findings in this thesis is that the household setting plays an important role in the spread of RSV, a finding that is made clearer with added detail on pathogen type. In the case of the data analysed here, and the social structuring from which it was collected, RSV clades appeared to mimic household structure as such identification at this level did not drastically change the transmission characteristic observed with identification at the group level. However, the combination of epidemiological and genetic data elucidated transmission chains within the household enabling the identification of the sources of infant RSV infections. For this particular study, it was inferred that the sources of infant RSV infections were both in the same household as the infant and from external sources. Where infant infections occurred in the household, the source of infection was often a child between the ages of 2-13 years. It was inferred that previous infection with one RSV group type reduced susceptibility to re-infection by heterologous group type within the same epidemic. Interactions were also observed between RSV and human coronavirus groups. In particular, previous infection with RSV B was estimated to increase susceptibility to corona OC43 by 81% (95% CrI: 40%, 134%). Detailed data of infection events in individual hosts can provide a wealth of knowledge. The inferences made from this study should be explored at larger spatial and temporal scales to determine the population level impact, and hence public-health significance, of pathogen interactions, whether these interactions are between strains of the same pathogen of between different pathogens. In planning for, and assessing the impact of, an intervention against a particular pathogen, investigators should not ignore the preexisting ecological balance and should make efforts to understand how this will be disrupted by an intervention against one or more pathogens.

Recently, collection of sequence data has become increasingly rapid and cost-efficient, prompting much research into using this kind of data in the analysis of infectious diseases. There is currently substantial interest in developing epidemic model frameworks which can incorporate this new abundance of data. Whole-genome sequence (WGS) data reveal to us the unique construction- the 'fingerprint'- of the DNA of a sample pathogen. These high resolution data introduce the possibility that we may be able to discover who infected whom in an epidemic outbreak, allowing us to better understand transmission dynamics and therefore design improved preventative and intervention measures. WGS data may prove useful in understanding how levels of infectiousness and susceptibility vary between individuals in a population, or patients on a hospital ward. Genetic data are becoming increasingly widely available, and it is now possible to sequence isolates of some pathogens in real-time in the field with mobile sequencing technologies. Therefore, developing the models and methods to best exploit this is of considerable importance. The first focus of the research presented here is on antibiotic-resistant nosocomial infections, or 'hospital superbugs', as these still pose a significant problem in hospitals, especially in developing countries. Antibiotic resistance is estimated to kill 700 thousand people globally every year. Current public focus on the threat of an 'antibiotic apocalypse' focuses on the need to reduce the overuse of antibiotics, but another important strategy is to better understand the transmission of such pathogens in order that better prevention and intervention strategies can be designed. Hospital wards present a unique environment, data from which require their own models and methods to analyse outbreaks of infectious disease. Initial research in this thesis has concentrated on outbreaks of methicillin-resistant Staphylococcus aureus (MRSA), as it is the most widespread and most common antibiotic-resistant nosocomial infection. In this thesis, discrete-time stochastic epidemic models are developed which can be used to analyse both epidemiological and genetic data from an outbreak of MRSA on a hospital ward. These new models can be used to estimate routes of transmission through the hospital ward on the level of individual transmission events by harnessing the information available in the genetic distances between isolate sequences taken from colonised patients. The unobserved transmission dynamics in the models can be inferred using Bayesian inference in a data-augmented MCMC algorithm. Although techniques have been developed to assess the goodness-of-fit of epidemic models in Bayesian settings, they do not assess how well a model fits the genetic data. Methods for doing so are developed in this thesis. An outbreak of MRSA is analysed using the presented models, and the new goodness-of-fit techniques are used to suggest ways to improve the fit of models. The ideas behind the models for genetic data from MRSA outbreaks are also applicable to other epidemic outbreaks for which genetic data are available. In this thesis we present continuous-time stochastic epidemic models for the spread of avian influenza. These models have a spatial aspect and can be used to estimate the transmission events between farms by analysing genetic and epidemiological data from each farm. Avian influenza is carried endemically by wild birds, so it is very difficult to prevent outbreaks entirely. Therefore, it is very useful to better understand the transmission dynamics of outbreaks and to be able to make predictions about the course of a future epidemic. The combined analysis of both epidemiological and genetic data through novel models and methods allows transmission of pathogens in epidemic outbreaks to be investigated on the level of individuals in the population. This can have a great public health impact, as results about the routes of infection can inform prevention and control measures.

Background: In the era of large-scale biobanks and detailed multi-omic and clinical phenotyping, it is now possible to simultaneously measure several thousands of biological traits and test their associations with health outcomes. The unprecedented scale of the available epidemiological data has led to a surge in the number of proposed risk factors for cardiometabolic and other diseases identified through observational research. However, prominent failures of randomised controlled trials to replicate observational findings highlight that new approaches are urgently required to prioritise risk factors for interventional studies based on their likelihood of causality. Aim: Genetic evidence provides the opportunity to obtain unconfounded associations in an observational setting and can help to efficiently identify causal, aetiological pathways to cardiometabolic diseases. This work aims to identify and prioritise causal pathways to cardiometabolic diseases by integrating genetic data with detailed metabolic phenotypes, including blood metabolites and objective measures of body size and composition. Methods: The first chapter summarises and critically evaluates existing genetic strategies for assessing causality in an observational setting and reviews the literature on reported associations between blood metabolites and incident type 2 diabetes (T2D). As a proof-of-concept, I then adopt a genetic approach to a) generate genetic predictors and b) assess evidence of causality for T2D and coronary heart disease (CHD) risk for a selected metabolite (glycine), for which consistent observational evidence suggests protective associations with T2D and CHD development. The approach is then extended to move beyond studying simple indices of obesity and enable causal assessment of refined anthropometric traits. For this, I have led collaborative efforts to develop genetic instruments for overall and regional fat and lean mass. Findings: Genetic approaches to studying causality in an observational setting rely on important assumptions that are sensitive to violations, particularly in the context of highly correlated 'omic' measures. Existing methods generally consider overall risk factor "levels", rather than assessing the causality of the distinct mechanisms contributing to levels. I provide genetic evidence for a causal cardio-protective effect of the glycine pathway in men and women, with blood pressure as a potential mediating factor. In contrast, no strong evidence for a causal link between glycine and T2D was found, with evidence suggesting that the inverse glycine-to-T2D association is the consequence of a glycine-lowering effect of insulin resistance. Despite total body fat percentage (BF%) and fat-free mass index (FFMI) being strongly observationally and genetically correlated with BMI, I identify 16 loci associated with higher BF% and lower FFMI but not BMI. Based on observational and genetic studies of regional fat and lean mass, I identify patterns of fat distribution which may not be captured by traditional anthropometric phenotypes, such as fat mass in the arms and in the subcutaneous android region. Fat mass in the arms and subcutaneous android region were observationally only weakly correlated with BMI, WHR and other fat compartments, and genetic loci specific to arm and subcutaneous android fat mass were identified. Conversely, no evidence was found that genetic loci associated with lean mass have heterogeneous effects on lean mass in different regions of the body. Conclusion: This thesis demonstrates how the integration of large-scale genetic, metabolomic and clinical data can not only prioritise novel aetiological pathways to cardiometabolic conditions, but also formulate hypotheses regarding the underlying physiological mechanisms. The genetic factors identified for refined anthropometric traits, such as a high relative body fat mass in the absence of overweight, allow for a causal assessment of novel and specific body size and composition traits. In summary, this thesis demonstrates and utilises the opportunities that arise from integrating genetic data with refined phenotypes at scale to identify novel targets for the treatment and prevention of cardiometabolic diseases.

Major depressive disorder (MDD) is a highly prevalent and disabling condition with a heritability of around 37%. Key symptoms of MDD include low mood and psychological distress, but the mechanisms underlying MDD and its symptoms are unclear. Genetic and neuroimaging techniques are important methods with which to better understand the aetiology and mechanisms of depression. Recently, through the availability of the UK Biobank and ENIGMA datasets, it has been possible to conduct well-powered imaging studies of heterogeneous traits like MDD, with genome-wide genetic data. These genetic data can act as causal instruments and can be utilised to identify differences in neurobiological mechanisms. The current thesis presents neurobiological associations with depressive symptoms and genetic risk for MDD using data from the UK Biobank imaging project (N range from 5,000 to 12,000). My overall aims were to investigate the neurobiological basis of MDD status, depressive symptoms and MDD polygenic risk. First, MDD case-control differences in subcortical volumes and white matter microstructure indexed by fractional anisotropy and mean diffusivity, are presented using the largest structural neuroimaging samples to date. MDD was associated with worse white matter microstructure in the thalamic-radiation subset and forceps major (posterior corpus callosum). No group difference was found for the volume of any subcortical structure. Next, associations between depressive symptom severity (including longitudinal and cross-sectional measures) with white matter microstructure were tested. Over 8,000 participants had repeated measure of depressive symptoms assessed on 2-4 occasions across 5.89 to 10.69 years. I found several novel associations between measures of depressive symptom severity (at the time of imaging, their variance within individuals over time, and with longitudinal increasing depression severity) all associated with lower white matter microstructure in the thalamic radiations. This was the first study of this size looking at imaging associations with longitudinal symptom measures and demonstrates consistent findings implicating thalamocortical connections. The third study presents results of phenotype wide association ('PheWAS') analysis of polygenic risk for MDD, including imaging and other available phenotypes. In total, 1,744 phenotypes were tested, covering sociodemographic, physical health, mental health, subcortical volumes, white matter microstructure assessed with FA and MD (mean diffusivity) and resting-state connectivity. I found that MDD polygenic risk was associated with MDD-related phenotypes including severity of depression and neuroticism, sleep, smoking, subjective well-being as well as neurobiological phenotypes including white matter microstructure and resting-state connectivity. In my final data chapter, neurobiological associations with cognition, as an important risk factor of major depressive disorder, were also reported. I found that higher connectivity related to the default mode network was associated with better cognitive performance. These studies suggest two features of neurobiology related to MDD traits and genetic risk. First, they implicate microstructure of thalamic white matter connections as an important biomarker for MDD risk, psychological distress and genetic risk, as reflected by its consistent associations with depressive status, depressive symptoms, within-subject variability of depression and MDD polygenic risk. Secondly, the aberrant connections within the default mode network were related to MDD phenotypes and polygenic risk. These findings, therefore, provide evidence that these features may play a key role in MDD-related neuroarchitecture.

Epigenetic factors can contribute to phenotypic diversity and to ecological processes. For instance, DNA methylation can influence gene regulation, and thus phenotypic plasticity. However, little is yet known about how and why methylation varies in the wild. In this dissertation, I build on this knowledge by combining ecological, genetic and DNA methylation data from natural and experimental populations of the stick insect Timema cristinae. This species is an important system to ecological genetics studies, which provides good starting point for the investigation of the patterns, drivers, and the possible ecological consequences of natural methylation variation. I obtained methylation data using whole-genome bisulfite sequencing (BS-seq) and genetic data from restriction site associated DNA sequencing (RAD-seq). From a population survey, I found natural methylation variation in T. cristinae (1) is characteristic of "Hemimetabola" insects; (2) is structured in geographical space; and (3) is strongly correlated to genetic variation. In addition, an experiment simulating a host shift was carried out to test for the direct effects of host plant species on T. cristinae methylation levels. In both the population survey and in the experiment, binomial mixed models were used to perform a methylome scan in search of candidate single methylation polymorphisms (SMPs) associated with host plant use. This analysis is analogous to genome-wide analysis studies, but applied to methylation levels. They use genetic data to estimate random effects arising from relatedness. The results suggest (4) an association between methylation levels and host plant in specific regions and that (5) some of them could be responsive to host shift treatment. Finally, the model suggested (6) significant mean heritability of methylation status, estimated based on the genetic relatedness. My results collectively indicate methylation variation could be ecologically relevant to T. cristinae, and adds to the general understanding of the importance of epigenetic variation.

Genome-wide association studies (GWASs) have identifified many genetic variants associated with type 2 diabetes (T2D) risk and glycaemic traits, most of which are noncoding. Although existing islet RNA sequencing (RNA-seq) studies have identified genes regulated by T2D-associated variants, sample sizes are small, and most known regulatory targets are messenger RNAs RNAs). The work presented in this thesis aims to characterise long noncoding RNA (lncRNA) expression in islets, identify candidate lncRNA regulatory mechanisms, and through incorporation of genetic data, identify genes with potential causal mediation of T2D-associated genetic variation. Using data from 43 GTEx tissues, I first showed that lncRNAs are more tissue specific than mRNAs. Using 38 human islet samples with paired ribosomal rRNA depletion (RD) and poly-A-tailed RNA selection (PA) RNA-seq, I then showed that RD more comprehensively isolates lncRNAs, but the relative effective sequencing depth is approximately half that of PA. Using RD RNA-seq from 54 islet samples, I identified previously unannotated expressed regions, but comparison to lncRNAs from recently published islet-specific annotations showed high discordance between studies. Thus, although RD better captures lncRNAs, its use in transcriptome-wide islet analyses is not practical until sample sizes increase. By analysing differential expression (DE) by T2D status in islets (n = 207) and 48 tissues from GTEx, I found 89 unique lncRNAs with DE in at least one GTEx tissue, and 75 with DE in islets. To give these functional context, co-expression modules were constructed in each tissue with lncRNAs and mRNAs. One common gene module between pancreas and islet was enriched for lncRNAs with DE, and contained mRNAs associated with insulin secretion regulation, ultimately pointing to lncRNA CTB-12O2.1 as a candidate regulator of T2D-relevant gene expression. I next incorporated genetic data with islet RNA-seq (n = 426) to demonstrate a modelling approach for identifying sparse, robust genetic models of gene expression. My approach is replicable, robust, and provides models that are sparser and explain more expression variation than LASSO or elastic net models. Using these models along with expression quantitative trait locus models, I identified 33 candidate expression mediators of genetic associations with T2D and glycaemic traits. Amongst these were two lncRNAs that were transcriptionally associated with biologically relevant mRNAs, making them candidate regulatory targets of corresponding T2D-associated noncoding genetic variants. The data presented throughout has therefore implicated previously unstudied lncRNAs in T2D risk and the regulation of insulin secretion, and has demonstrated how integration of disease-focussed islet RNA-seq and large-scale public datasets such as GTEx and GWAS meta-analyses can effectively identify novel regulatory mechanisms underlying genetic associations with T2D.

Many of the patterns of genetic variation we observe today have arisen via the complex dynamics of interactions and isolation of historic human populations. In this thesis, we focus on two important features of the genetics of populations that can be used to learn about human history: population structure and admixture. The Iberian peninsula has a complex demographic history, as well as rich linguistic and cultural diversity. However, previous studies using small genomic regions (such as Y-chromosome and mtDNA) as well as genome-wide data have so far detected limited genetic structure in Iberia. Larger datasets and powerful new statistical methods that exploit information in the correlation structure of nearby genetic markers have made it possible to detect and characterise genetic differentiation at fine geographic scales. We performed the largest and most comprehensive study of Spanish population structure to date by analysing genotyping array data for ~1,400 Spanish individuals genotyped at ~700,000 polymorphic loci. We show that at broad scales, the major axis of genetic differentiation in Spain runs from west to east, while there is remarkable genetic similarity in the north-south direction. Our analysis also reveals striking patterns of geographically-localised and subtle population structure within Spain at scales down to tens of kilometres. We developed and applied new approaches to show how this structure has arisen from a complex and regionally-varying mix of genetic isolation and recent gene-flow within and from outside of Iberia. To further explore the genetic impact of historical migrations and invasions of Iberia, we assembled a data set of 2,920 individuals (~300,000 markers) from Iberia and the surrounding regions of north Africa, Europe, and sub-Saharan Africa. Our admixture analysis implies that north African-like DNA in Iberia was mainly introduced in the earlier half (860 - 1120 CE) of the period of Muslim rule in Iberia, and we estimate that the closest modern-day equivalents to the initial migrants are located in Western Sahara. We also find that north African-like DNA in Iberia shows striking regional variation, with near-zero contributions in the Basque regions, low amounts (~3%) in the north east of Iberia, and as high as (~11%) in Galicia and Portugal. The UK Biobank project is a large prospective cohort study of ~500,000 individuals from across the United Kingdom, aged between 40-69 at recruitment. A rich variety of phenotypic and health-related information is available on each participant, making the resource unprecedented in its size and scope. Understanding the role that genetics plays in phenotypic variation, and its potential interactions with other factors, provides a critical route to a better understanding of human biology and population health. As such, a key component of the UK Biobank resource has been the collection of genome-wide genetic data (~805,000 markers) on every participant using purpose-designed genotyping arrays. These data are the focus of the second part of this thesis. In particular, we designed and implemented a quality control (QC) pipeline on behalf of the current and future use of this multi-purpose resource. Genotype data on this scale offers novel opportunities for assessing quality issues, although the wide range of ancestral backgrounds in the cohort also creates particular challenges. We also conducted a set of analyses that reveal properties of the genetic data, including population structure and familial relatedness, that can be important for downstream analyses. We find that cryptic relatedness is common among UK Biobank participants (~30% have at least one first cousin relative or closer), and a full range of human population structure is present in this cohort: from world-wide ancestral diversity to subtle population structure at sub-national geographic scales. Finally, we performed a genome-wide association scan on a well-studied and highly polygenic phenotype: standing height. This provided a further test of the effectiveness of our QC, as well as highlighting the potential of the resource to uncover novel regions of association.

Introduction and Aims: In recent years, there has been a growing interest in studying genetic data so as to answer specific medical questions; for example, indicative biomarkers that can accurately predict (classify) outcomes (e.g. healthy and disease or different categories of patients' response to treatment). In genome-wide data analysis, a typical procedure is to use a variable selection approach, often univariable, where the primary aim is to select the most important genetic variants, particularly Single Nucleotide Polymorphisms (SNPs), associated with an outcome of interest. This thesis proposes a novel variable selection method by considering the multivariate nature of the genetic data. The aim of this thesis is threefold: (i) to develop a quantitative variable selection method for classification which can be used in the multivariate setting, computationally inexpensive and easy to understand and to apply, (ii) to propose a multi-step approach that selects SNPs and evaluates the classification performance of the resulting models in a cross-validation framework, and (iii) to jointly model the longitudinal clinical and SNP data for classification using the Standard and New Antiepileptic Drugs (SANAD) dataset. Methods: A literature search was conducted to study the different approaches of variable selection and their relationship with classification performance. A novel variable selection method, tSNR within a logistic regression framework was developed to select the most informative SNPs. In addition, a multi-step framework that involved univariable and multivariable selection in a cross-validation setting was proposed. Then, the filter metric tSNR and the multi-step framework were assessed using simulated datasets. The methods were further examined using an epilepsy pharmacogenomics dataset (EpiPGX) in which the phenotype of interest is the remission from seizures status after receiving first well-tolerated antiepileptic drugs (AEDs). A second epilepsy dataset from the SANAD trial was used as the validation dataset. Within the SANAD dataset, the longitudinal clinical and SNP data were jointly modelled using a longitudinal discriminant analysis (LoDA) approach with multivariate generalised linear mixed model (MGLMM). The classification performance was measured by calculating the probability of correct classification (PCC) and area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). Results: The literature review suggested the need for variable selection methods, which could potentially aid better classification accuracy. In the simulation study, the univariable tSNR ranking was able to capture the causal SNPs in the top ten ranked SNPs. In addition, within the proposed framework, the results using simulated datasets suggested that the classification performance using SNPs selected by cumulative tSNR (multivariate) are better than the SNPs based on univariable tSNR ranking. The results were further confirmed using the real clinical datasets. The addition of SNP data to the longitudinal model based on clinical data improved the mean prediction time at which patients who will not achieve remission from seizures within five years of commencing treatment are identified. However, it did not provide an improvement to the classification performance. Conclusions: The developed approach using a tSNR filter metric proved to be effective in ranking and selecting subset of SNPs that are associated with the outcome of interest. The SNPs selected by tSNR were shown to give good classification accuracy. Also, by jointly modelling the longitudinal clinical data and SNP data (selected using tSNR) in a longitudinal model the prediction time at which patients can be classified was improved.

A rapid increase in the amount of available pathogen genetic data, which is ongoing and likely to continue for the foreseeable future, presents new opportunities and challenges in molecular epidemiology, and in the emerging field of “phylodynamics”, which seeks to unify the study of the evolutionary and epidemiological dynamics of pathogen populations. This thesis explores some of these challenges and opportunities, with a focus on pathogens infecting livestock and poultry. I conducted analyses of sequences from two serotypes of foot-and-mouth-disease virus (FMDV) in order to investigate the global population dynamics of the virus. For serotype SAT 2, the amount of publicly available genomic data is still small enough that all of it could be included in a single analysis. A particular focus was the origins of historical outbreaks occurring in North Africa and the Middle East, outside the endemic area for the serotype. The results suggested sources for these in countries just south of the Sahara, and that the viruses responsible for three outbreaks occurring in 2012 were the result of separate introductions. For serotype O, including every available sequence was not feasible and the data had to be sub-sampled. Little research has been conducted on how to design a sampling strategy for sequence analysis of pathogens, an issue of increasing importance, so a simulation study was conducted to identify one. This suggested that, when reconstructing the temporal and spatial dynamics of a structured population of pathogens or infected individuals, it is preferable to stratify by subpopulation and by time period. The type O analysis itself showed that the south-east Asian topotype moves between countries according to cattle trade networks, but that geographic proximity is also important for strains from southern Asia and the Middle East. With genetic data available at an epidemiological resolution that was previously inconceivable, there are opportunities for new types of inference. For example, if we can acquire a sequence from all or most infected cases in an epidemic, they can inform inference of who infected who, complementing traditional contact-tracing approaches. I introduce a novel phylodynamic method for the simultaneous reconstruction of phylogeny and transmission tree for an epidemic in a situation where every infected host or premises can be identified and a sequence acquired from most of them. The performance of this method was demonstrated using simulated data, and then it was applied to reconstruct both trees from the 2003 H7N7 avian influenza outbreak in the Netherlands.

Excessive fibrogenic response is a pathological hallmark of chronic complex diseases, including cardiovascular disease. To date, very few gene targets for cardiac fibrosis that led to effective treatments have been identified in humans. In this thesis I study and dissect the genetic component underlying cardiac fibrosis. This study integrates histomorphometric measurements of fibrosis in the rat left ventricle (LV) with gene expression (RNA-Seq from LV) and genetic data in a panel of recombinant inbred (RI) rat strains (n=30). In addition, I integrated RNA-seq LV and genetic data in humans (n=187, healthy and dilated cardiomyopathy (DCM) patients), as well as DCM genome-wide association studies (GWAS) data. I started by carrying out an unbiased co-expression network analysis in the rat heart. The reconstructed cardiac transcriptional modules were associated with quantitative levels of fibrosis. Co-expression networks were also independently built in the heart of DCM patients and by using the rat data, co-expression networks associated with fibrosis, conserved across rats and humans and not present in control human heart were prioritised. In the prioritised networks, I also analysed their cardiac cell type specificity, differential expression after TGFβ induction, potential driving transcription factors and conservation in other fibrotic diseases by analysing human data collected from other organs. Furthermore, I aimed to identify common genetic regulators of the networks (also called master genetic regulators) by using Bayesian multivariate regression approaches. Finally, I integrated GWAS data in DCM (n=2,287) to dissect the genetic basis of DCM. This systems genetics study evidences that there are transcriptional processes involved in the human cardiac fibrogenic response that are conserved across rats and humans, some of them also underlying DCM aetiology. In an attempt to suggest new gene targets for cardiac fibrosis, I also identified the WWP2 gene as a novel trans-acting genetic regulator of cardiac fibrosis.

Infectious diseases pose a significant threat to animal and human health across the globe, with much of the burden falling on low-income countries. Despite efforts to control many of these diseases, very few have ever been eradicated. Their dynamics are often embedded in complex, heterogeneous landscapes defined by interacting population and landscape level processes. As such, landscape heterogeneity plays a key role in driving disease transmission and persistence. Incorporating landscape heterogeneity in studies of pathogen dynamics is challenging but the accessibility of data, particularly next generation sequencing data, has opened new avenues of research. Landscape epidemiology involves using an integrated approach to understand spatial patterns of disease, using methods that combine landscape genetics, ecology and epidemiology. in this thesis I use these integrative methods to determine the underlying mechanisms facilitating the spread and persistence of canine rabies virus in Tanzania. Whole genome level characterisation of rabies virus samples was achieved and used in combination with cutting-edge inference techniques to explore spatial patterns of rabies at different spatial scales. Phylogeographic patterns were able to characterise spatial scales of endemic rabies transmission in Tanzania, uncovering strong viral population structure at sub-continental levels with evidence of a more fluid dispersal dynamic at local ( less than 100km2 area) spatial scales . Within-country phylogeographic patterns revealed large regional movements within Tanzania that could be attributed to human-mediated movements and revealed the presence of multiple co-circulating lineages within a single administrative district. Finely resolved incidence data from the Serengeti District complemented with whole genome sequences enabled the exploration of local scales of transmission in more detail. By extending phylogeographic diffusion models to incorporate landscape heterogeneity I was able to uncover evidence supporting landscape predictors of rabies diffusion. While much of the spatial structure was attributable to the effects of isolation by distance, landscape predictors had discernible effects on diffusion. In particular, rivers appeared to act as a barrier to dispersal and road networks facilitated diffusion and I found evidence to support vaccination as an effective control measure for canine rabies in the Serengeti District. Importantly, I also found evidence to support vaccination as resistance to diffusion and therefore an effective control measure for dog rabies. As a complementary approach a space-time-genetic algorithm was used to determine who-infected-whom in the Serengeti District. The model explicitly accounted for the possibility of exogenous sources of infection and how to incorporate genetic data available for only a proportion of samples. Direct transmission events were estimated between 42% of observed cases and highlighted the co-circulation of two major lineages in both time and space. Direct transmission events predominantly occurred over very small distances, less than 1km, but a large proportion of cases had unobserved sources that could represent transmission from dogs in neighbouring regions or larger indirect transmission events. A future development of the model is to delineate between these possibilities to assess the true contribution of exogenous sources to the system dynamic. Ultimately these integrative models are at an early stage of development but highlight the power of genetic data to delineate fine-scale transmission patterns. The results from this thesis suggest that landscape features such as rivers could be exploited as barriers in step-wise vaccination campaigns and highlight the utility of genetic surveillance to monitor control and elimination as rabies management progresses.

Over the last two decades many in the archaeological community have developed a degree of scepticism and suspicion towards using genetics to study the past. My project aims to dispel some of these concerns, highlighting the inferences about demographic and social history which can be made from genetic data gathered from modern populations. Centred around the British-Irish Isles and in particular, the Isle of Man, this study investigates for the first time, differing demographic histories in the male and female gene-pools. The focus of this research is the 400 years of Norwegian rule in the Irish Sea region following the Viking visitations of 900CE, when the Isle of Man was an important centre in the politics of the Irish Sea kingdom. DNA samples were collected from male volunteers, followed by the extraction of high resolution mtDNA and Y chromosome data. Volunteers were enlisted on the basis of surnames taken from 16th-century documentation to ensure deep historical roots on the Island. The high quality genetic data generated for the Isle of Man not only gave greater distinguishing power for looking at closely related populations, but also allowed separate male and female population histories to be explored. Approximate Bayesian Computation (ABC) modelling was used to explore the demographic history of the Isle of Man and how the impact of the Norwegian diaspora differed for men and women, providing a powerful statistical and probabilistic approach to admixture analysis. Whilst drawing evidence from history, archaeology and incorporating genetic data to provide indications of where the Island draws its genetic influences from, this project provided a case study for how the fields of archaeology and genetics can be better integrated for the exploration of the past.

Genetic diversity is not only necessary for evolutionary processes but also for ecosystem functioning and population stability. This is particularly true for ecosystem engineers that contribute broadly to function and food webs in the habitats in which they are found. In this study, we assessed the global patterns of extinction risk, range size, and genetic diversity, specifically focusing on two large invertebrate clades: Mollusca and Arthropoda. We collected data from the International Union for Conservation of Nature (IUCN) and also conducted a systematic, quantitative review of existing Mollusca and Arthropoda nuclear and mitochondrial data. In total we considered 4075 species evaluated by the IUCN as well as 2452 populations with genetic data. We found that range size is significantly smaller in critically endangered species compared to least concern species, but that genetic diversity did not vary significantly between extinction risk categories. However, when we considered the variability in genetic diversity between populations within a species, we found that mollusks are significantly more variable than arthropods in the nuclear genome but not the mitochondrial genome. Combined, this suggests that future efforts aimed at identifying extinction risk using genetic data will need to consider taxonomic biases when applying population-level assessments to species-level determinations.

Humans have been modifying the Earth’s land surface for millennia. In the last 300 years these changes have increase in intensity and spatial extent. In tropical regions these anthropogenic changes are dominated by the expansion of agriculture. This has led the majority of remaining tropical forests to exist as fragments embedded in a matrix of agricultural production. This production varies in type and diversity of crops and management practices, through which organisms must navigating through or surviving within the matrix. For this reason, understanding the effects of these agricultural landscapes on species dispersal is crucial in the development of successful conservation planning. The purpose of this dissertation was to explore the effects of varying coffee production management practices on the population structure and connectivity of tropical rodents. This study was conducted in the coffee growing region of Soconusco and the El Triunfo Biosphere Reserve in Chiapas, Mexico. We used genetic and landscape data to study the population structure and connectivity of two common rodent species, Heteromys desmarestianus goldmani and Peromysucs gymnotis, in this landscape. We found that levels of population connectivity and genetic diversity vary between the two sampled species, which is supported by their differences in ecological specialization. Heteromys in the coffee farms were characterized by subtle genetic structure, which correlates with high management intensity coffee production and high genetic diversity. On the other hand, P. gymnotis individuals showed no signal of population structure and lower degrees of genetic diversity. When comparing H. d. goldmani populations from the continuous forest (El Triunfo Biosphere Reserve) and the coffee production region we found similar levels of genetic diversity, suggesting that high levels of migration and gene flow can be maintained in the coffee agroecosystem. This study highlights the potential of integrating molecular and landscape data to explore population connectivity of elusive species, such as terrestrial small mammals. It also shows the importance of studying the responses to environmental change for species with different levels of ecological specialization within a group, since these responses can vary. Additionally, it identifies coffee production as an important refuge for rodent species within anthropogenic landscapes. This work adds to the growing body of literature in landscape genetics by demonstrating that rodents can show population structure at small scales and that this structure can be driven by landscape factors linked to agricultural management.