The case-control design is one of the most commonly used designs in genome- wide asociation studies. When we increase the sample size of either the controls or, more importantly, the cases, the power of whatever test we use will certainly increase. However increasing the sample size, means that addi- tional individuals need to be genotyped and this implies extra financial costs. However, nowadays with the emergence of genetic studies, a large number of genetic data are available at low or no extra cost. Even though those data may not be completely relevant to the current study, they can still be used to increase the probability to identify true associations. Furthermore, additional information, non-necessarily genetic, can also be used to improve the power of a method. In this thesis we extend the case-control design in order to take ad- vantage of such types of additional data and/or information. We discuss three designs; the case-cohort-control, the kin-cohort and the super-case– case–control–super-control designs. For each of these, we present methods that are adjusted or modified versions of standard case-control methods but we also propose novel ones developed with those extended designs in mind. Ultimately, we describe how those methods can be used in order to increase the power of association tests, especially compared to similar methods of the case-control design.

The aims of this study were to analyse known and potential biomarkers of common and genetically complex human disorders and to identify genetic and environmental variation associated with plasma biomarker concentrations. Two groups of protein biomarkers were analysed. First, plasma complement factor H (CFH) was selected as a potential biomarker for age-related macular degeneration (AMD), since common variants in the CFH gene show strong association with this disorder. Secondly, two isoforms of amyloid-β (Aβ40 and Aβ42) were selected as biomarkers for Alzheimer disease (AD) since Aβ deposits are major constituents of the amyloid plaques characteristic of this disorder. Physiological and anthropometric measurements and samples of human and genomic DNA were collected from a population sample of 1,021 individuals from the Croatian island of Vis. Quantitative determination of plasma Aβ40 and Aβ42 concentrations was performed using enzyme-linked immunosorbent assays. Heritabilities and significant covariate effects were estimated for each trait in the Croatian data set. Genome-wide linkage and association analyses were conducted for the biomarker traits. A novel finding was the genome-wide significant association between a CFH and several polymorphisms close to and within the CFH gene. The strongest association was with an intronic SNP within CFH, which explained 28% of the total trait variance (P < 10-50). The association was also replicated in a Dutch sample set. A SNP haplotype was identified which accounted for a higher proportion of the phenotypic variance. Conditional haplotype analysis showed that the effect of this haplotype on plasma CFH concentration was independent of the CFH Y402H variant, and significantly stronger than a deletion of the adjacent CFHR3/CFHR1 which was already known to affect AMD susceptibility. Genetic analysis of 382 AMD cases and 201 controls was consistent with the CFH Y402H variant being the strongest AMD susceptibility locus. Variation in plasma CFH concentration was found to explain up to 1.8% of the variation in susceptibility to AMD with an odds 2.1 (95% C.I. 1.3-3.4, P = 0.003). SNPs that were strongly associated with a CFH concentration also influenced AMD susceptibility (P < 0.05) independently of the CFH Y402H polymorphism. Functional analysis of genomic regions associated with plasma CFH is needed to identify the causal variants. Associations were observed between plasma Aβ40 concentration and several novel candidate loci, spanning regions of approximately 0.2 Mb, on chromosomes 9 and X. Similarly, novel associations with plasma Aβ42 were found in several regions, each spanning 0.2-0.4 Mb, on chromosomes 2, 5, 9, 15 and 20. The proportion of the phenotypic variance in plasma Aβ42 explained by these putative associations ranged between 1.8 and 2.8%. However, none of the associated SNPs was significant after correction for multiple testing, therefore replication is required. Finally, attempts were made to identify and quantitate new protein biomarkers of disease in human plasma using mass spectrometry. Development and optimisation of techniques was initially undertaken to deplete high-abundance plasma proteins and improve signal:noise ratio. This allowed the assessment of downstream proteomic approaches including MALDI-TOF mass spectrometry (MS), capillary electrophoresis (CE) and ion exchange chromatography (IEC), each with the potential for large-scale quantitation of plasma proteins. Although the analysis of single protein analytes, using CE and IEC proved promising, the results highlighted the difficulty associated with MALDI-TOF and protein ionisation techniques in analysing complex mixtures such as plasma.

Sleep disturbances and insomnia are prevalent, with around 33% of adults indicating that they experience at least one main symptom of insomnia, and bidirectional relationships exist with common psychopathology, particularly major depressive disorder (MDD). However, genetic and environmental (e.g., traumatic event exposure) contributions to the etiology of these phenotypes are not yet well understood. A genetically informative sample of approximately 12,000 Han Chinese women aged 30-60 (50% with recurrent MDD) was used to address several gaps within the sleep literature. Sleep disturbances were assessed in all individuals using a general item addressing sleeplessness (GS). A sleep within depression sum score (SDS) was also created in MDD cases, combining information from the GS and two insomnia items within MDD. A total of 11 traumatic events were assessed and additional information on childhood sexual abuse (CSA) was also obtained. First, factor analyses were conducted to determine trauma factor structure. The best-fit solution included 3 factors: interpersonal, child interpersonal, and non-assaultive, and composite variables were constructed accordingly. A series of hierarchical regressions were run to examine differential effects of trauma type and timing on sleeplessness. All traumatic events predicted sleeplessness at similar magnitudes, although population models indicated that childhood interpersonal trauma may be particularly potent. An association between CSA and sleeplessness was also replicated. A series of genetic analyses demonstrated that the single nucleotide polymorphism-based heritability of sleep phenotypes did not differ significantly from zero. Further, association analyses did not identify any genome-wide significant loci. However, using a liberal false discovery rate threshold of 0.5, two genes of interest, KCNK9 and ALDH1A2, emerged for the SDS. Polygenic risk score (PRS) analyses demonstrated genetic overlap between the SDS in MDD cases and GS in MDD controls, with PRSs explaining 0.2-0.3% of the variance. A final combined model of both genetic and environmental risk indicated that both PRS and traumatic events were significant predictors of sleeplessness. While genetic results should be interpreted with caution given the lack of heritability, additional research into the genetic and environmental contributions to insomnia, utilizing more standardized phenotypes and properly ascertained samples, is clearly warranted.

Worldwide obesity has more than doubled since 1980 and at least 2.8 million people die each year as a result of being overweight or obese. An elevated body weight is the result of the interplay between susceptibility gene variants and an obesogenic environment, and recent evidence shows that epigenetic processes are likely involved. The growing availability of high-throughput technologies has made it possible to assess quickly the entire epigenome of large samples at a relatively low cost. As a result, vast amounts of data have been generated and researchers are now confronted to both bioinformatic and biostatistic challenges to make sense of such data in the context of obesity and its complications. In this doctoral thesis, we explored associations between the human blood methylome and obesity-associated gene variants as well as dietary fat quality and quantity. We used well described preprocessing techniques and statistical methods, along with publicly available data from consortiums and other research groups, as well as tools for pathway enrichment and chromatin state inference. We found associations between obesityassociated SNPs and methylation levels at proximal promoters and enhancers, and some of these associations were replicated in multiple tissues. We also found that contrary to dietary fat quantity, dietary fat quality associates with methylation levels in the promoter of genes involved in metabolic pathways. Then, using a gene-targeted approach, we looked at the impact of an acute environmental stress (sleep loss) on the methylation and transcription levels of circadian clock genes in skeletal muscle and adipose tissue of healthy men. We found that a single night of wakefulness can alter the epigenetic and transcriptional profile of core circadian clock genes in a tissue-specific manner. Finally, we looked at the effects of chronic maternal obesity and subsequent weight loss on the transcription of epigenetic machinery genes in the fetus and placenta of mice. We found that the transcription of epigenetic machinery genes is highly sensitive to maternal weight trajectories, and particularly those of the histone acetylation pathway. Overall, this thesis demonstrated that genetics, obesogenic environment stimuli and maternal programming impact epigenetic marks at genomic locations relevant in the pathogenesis of obesity.

Lowering LDL-cholesterol with statin therapy produces substantial reductions in cardiovascular events, and larger cholesterol reductions may produce larger benefits. Rarely, myopathy occurs with statins, especially at higher doses and in combination with certain medications. Similarly strong associations might exist between myopathy with high-dose statin regimens and genetic variants, especially those affecting blood statin levels. This study aimed to find genetic variants associated with statin-induced myopathy. A feasibility study was completed to assess whether plausible effect sizes of 5 to10-fold higher risks per genetic variants could be detected among 50-100 cases with statin-induced myopathy and to consider the best study design. A genome-wide association study was then carried out using approximately 300,000 genetic markers (and additional fine-mapping) in 85 people with definite or incipient myopathy and 90 controls, who were all taking 80mg simvastatin daily in a 12,000 participant trial of 80mg vs 20mg simvastatin daily. The cases were also compared to 2,300 additional controls who had not been exposed to intensive-dose statin therapy. Replication of the myopathy result and lipid-lowering associations were tested in a 20,000 participant trial of 40mg simvastatin daily versus placebo. The genome-wide scan yielded a single strong association (p = 4x10^-9) of myopathy with the rs4363657 single nucleotide polymorphism (SNP) located within the SLCO1B1 gene on chromosome 12. This non-coding SNP was in nearly complete linkage disequilibrium (r²=0.97) with the non-synonymous rs4149056 SNP. The population prevalence of the rs4149056 C allele was 15%, and the odds ratio for myopathy was 4.5 (95% confidence interval 2.6 to 7.7) for each copy of the C allele and 16.9 (4.7 to 61.1) for CC vs TT homozygotes. Over 60% of these myopathy cases could be attributed to the C variant. The SLCO1B1 gene encodes the organic anion transport polypeptide OATP1B1, which has been shown to regulate hepatic uptake of statins. In literature reports, rs4149056 reduced statin transport and was associated with 37% (31% to 44%) higher systemic statin acid levels per C allele. The association of rs4149056 with myopathy was replicated in the trial of 40mg simvastatin daily, which also showed that it was associated with the cholesterol-lowering effects of simvastatin. No SNPs in any other region were clearly associated with myopathy (although comparison of the myopathy cases with the 2,300 controls identified a region of chromosome 1p12 that warrants further study). This study identified common variants in SLCO1B1 that influence the risks of statin-induced myopathy substantially. Genotyping these variants may be useful for tailoring both the statin dose and safety monitoring. More generally, such studies of the genetic determinants of serious adverse reactions with other drug classes may help to improve the balance between treatment efficacy and safety.

Decades of work has led researchers to believe that risk for complex behavioral phenotypes, such as alcohol use disorders, is likely influenced by multiple genes of small effect acting in conjunction with each other and the environment. Currently, the field of psychiatric genetics is developing methodologies for the identification of genetic risk variants that predispose individuals to the development of complex behavioral disorders. Several challenges related to the complex and polygenic nature of these phenotypes, must be considered. This dissertation study attempts to address these important challenges in the context of alcohol use disorders and related phenotypes. A rich twin and family study literature has indicated that 40-70% of the variance in alcohol use disorders (AUDs) is influenced by genetics. Recent attempts to identify specific x genetic risk variants associated with AUDs have been met with limited success. Meanwhile, evidence of the moderating effects of the environment on AUDs has been mounting, providing a strong rationale for examining gene-environment interaction. In the following chapters several studies will be described that integrate established twin methodologies into gene identification projects in an effort to reduce heterogeneity (both phenotypic and genotypic), elucidate environmental constructs that moderate genetic influences, and to enhance statistical power to detect the subtle genetic influences on alcohol related phenotypes.

O grande potencial da genômica em benefício do melhoramento genético aplicado é a utilização direta das informações de marcadores de DNA na seleção, de forma a permitir igual ou maior eficiência seletiva, maior rapidez na obtenção de ganhos genéticos e redução de custos de fenotipagem, em comparação com a seleção tradicional. Esta era genômica está trazendo novas oportunidades para os melhoristas florestais, porém desafios ainda existem para o uso operacional da Seleção Genômica Ampla. Sendo assim, este trabalho teve por objetivo avaliar a capacidade preditiva de características de crescimento volumétrico de modelos de SGA previamente construídos para características de crescimento volumétrico e qualidade da madeira com uma população elite de melhoramento genético da International Paper do Brasil que atuou como população de treinamento. Os modelos preditivos foram aplicados em quatro populações de candidatos à seleção constituídos por testes de progênies e teste clonal de Eucalyptus com características contrastantes do ponto de vista de relacionamento genético com a população na qual os modelos foram desenvolvidos: (1) três populações compostas por indivíduos geneticamente relacionados com a população de descoberta e (2) uma população composta por indivíduos geneticamente não relacionados com a população de descoberta. Posteriormente, as quatro populações de avaliação, compostas por 100 indivíduos genotipados em cada uma, foram utilizadas para construir novos modelos preditivos e validações cruzadas realizadas entre estas populações. Foram utilizados para as análises SNPs com frequência de declaração de genótipo (call rate) &ge; 0.90 e MAF (menor frequência alélica) &ge; 0.01, totalizando 29.090 marcadores. O modelo preditivo previamente elaborado para população elite da International Paper apresentou capacidades preditivas que variaram de -0,296 para o caráter Altura em população geneticamente não relacionada até 0,440 para o caráter DAP em uma população geneticamente relacionada. Com os modelos desenvolvidos com as quatro populações genotipadas, e realizando seleção de marcas com base nos seus efeitos, as maiores capacidades preditivas foram obtidas por um número de SNPs médio que variou de 1371 para volume e IMA a 1467 para DAP. Usando esta abordagem, a capacidade preditiva maximizada para DAP foi de 0,744, 0,727 para altura, 0,751 para volume e 0,752 para IMA. Acurácias preditivas maximizadas, foram em seguida estimadas ao utilizar o menor número de SNPs selecionados. Com 237 SNPs acurácias da ordem de 0,660 para DAP, 0,555 para altura, 0,743 para volume e 0,743 para IMA foram obtidas. Embora estes resultados sugerem que a SGA teria bom resultado somente entre indivíduos geneticamente relacionados, uma análise conjunta dos dados utilizados para o desenvolvimento dos modelos preditivos anteriormente gerados com os dados das quarto populações aqui avaliadas, se faz necessária para alcançar resultados mais conclusivos. Adicionalmente, a abordagem de seleção de marcas para maximizar a capacidade preditiva ou acurácia deverá ser melhor avaliada à luz do seu impacto na medida que a SGA venha a ser praticada em futuras gerações de seleção. / The potential of genomics to the benefit of applied breeding is the direct use of DNA markers information for selection, allowing equal or higher selection efficiency, higher genetic gains per unit time and cost reduction in phenotyping, when compared with traditional selection based on phenotypic data. Researchers have proposed a new selection method called genome-wide selection (GWS), which has been successfully applied in livestock genetics and promises to revolutionize the improvement of perennial species with long life cycles. This genomic era is bringing new opportunities to forest breeders, but major challenges still need to be overcome for the operational use of GWS. Thus, this study aimed to assess the predictive ability of GWS models previously built for volume growth and wood quality traits based on an elite breeding population of International Paper in Brazil who served as discovery or training population. Predictive models were applied to four populations composed by progeny and clonal trials with contrasting characteristics from the standpoint of relatedness to the training population: (1) a population of individuals genetically related to the training population and (2) a population of individuals genetically unrelated to the training population. Subsequently, the four populations evaluated, with 100 genotyped individuals each, were used to build new models which were cross validated among them. Only SNPs with call rate &ge; 0.90 and MAF (minor allele frequency) &ge; 0.01 were used totaling 29,090 markers. Model previously developed yielded predictive abilities ranging from a lower -0.296 for height growth in genetically unrelated population to 0.440 for DBH and 0.219 for height in a genetically related population. With the GWS models built with the four genotyped populations and applying SNP marker selection based on their estimated effect the highest predictive capabilities were obtained when using an average number of SNPs ranging from 1371 for volume and MAI, to 1467 for DBH. The average predictive ability was maximized at 0.744 for DBH, 0.727 for height growth, 0.751 for volume and 0.752 for MAI. Maximized accuracies were obtained using the smallest number of SNPs with highest effects. With 237 selected SNPs accuracies around 0.660 for DBH, 0.555 for height, 0.743 for volume and 0.743 for MAI were obtained. While these results suggest that GWS should work well only across related individuals, a combined analysis of the data from the previous models with those of the four tested populations is necessary to reach more conclusive results. Furthermore, the approach of SNP marker selection to maximize predictive ability or accuracy is one that is still controversial and should be better evaluated in light of its impact as GWS is practiced in further generations of selection.

L’analyse des régulations de l’expression des gènes chez les bactéries permet de comprendre l’adaptation des bactéries à leur environnement et dans un contexte de biologie de synthèse d’optimiser la production microbienne de molécules d’intérêt. Notre objectif a été d’étudier la traduction au niveau du génome et ses relations avec les autres processus cellulaires par une approche de biologie des systèmes. Le traductome a été mesuré : pour chacun des ARN messagers, son pourcentage de copies en traduction et sa densité en ribosomes. Pour la première fois, une image complète de l’état traductionnel de E. coli en croissance rapide a été obtenue, caractérisée par une majorité de transcrits avec un très fort pourcentage de copies en traduction mais faiblement chargés en ribosomes. Notre modèle statistique a identifié des facteurs liés à la séquence comme déterminants de la traduction et le rôle important d’un paramètre physiologique : la concentration en ARNm. Pour la première fois, cet effet de la transcription sur la traduction a été validé à l’échelle moléculaire sur plusieurs gènes. Nous avons montré qu’une augmentation de la concentration d’un ARNm par induction transcriptionnelle entrainait une augmentation du pourcentage de copies en traduction et de la charge en ribosomes. / The analysis of gene expression regulation is necessary to better understand bacterial adaptation to environment and to be able in a context of synthetic biology to optimize the production of molecules of interest. The goal of this thesis was to study translation at the genome-wide level and its relationship to other cellular processes using a systems biology approach. First, translation activity at the -omic scale (called the traductome) was measured : for each messenger RNAs, its percentage of copies in translation and ribosome density. For the first time, a complete picture of the translational state in fast growing E. coli cells was obtained, characterized by a majority of transcripts with a very high percentage of copies in translation but a low ribosome density. Our model identified sequence-related factors as determinants of translation but, more surprisingly, the model predicted the important role of a physiological parameter: the mRNA concentration. Thus, more concentrated mRNA would have higher percentage of copies in translation and higher ribosome density. For the first time, this effect of transcription on translation has been validated at the molecular level on several genes. We showed that an increase in mRNA concentration by transcriptional induction results in increases in percentage of copies in translation and in ribosome load.

O grande potencial da genômica em benefício do melhoramento genético aplicado é a utilização direta das informações de marcadores de DNA na seleção, de forma a permitir igual ou maior eficiência seletiva, maior rapidez na obtenção de ganhos genéticos e redução de custos de fenotipagem, em comparação com a seleção tradicional. Esta era genômica está trazendo novas oportunidades para os melhoristas florestais, porém desafios ainda existem para o uso operacional da Seleção Genômica Ampla. Sendo assim, este trabalho teve por objetivo avaliar a capacidade preditiva de características de crescimento volumétrico de modelos de SGA previamente construídos para características de crescimento volumétrico e qualidade da madeira com uma população elite de melhoramento genético da International Paper do Brasil que atuou como população de treinamento. Os modelos preditivos foram aplicados em quatro populações de candidatos à seleção constituídos por testes de progênies e teste clonal de Eucalyptus com características contrastantes do ponto de vista de relacionamento genético com a população na qual os modelos foram desenvolvidos: (1) três populações compostas por indivíduos geneticamente relacionados com a população de descoberta e (2) uma população composta por indivíduos geneticamente não relacionados com a população de descoberta. Posteriormente, as quatro populações de avaliação, compostas por 100 indivíduos genotipados em cada uma, foram utilizadas para construir novos modelos preditivos e validações cruzadas realizadas entre estas populações. Foram utilizados para as análises SNPs com frequência de declaração de genótipo (call rate) &ge; 0.90 e MAF (menor frequência alélica) &ge; 0.01, totalizando 29.090 marcadores. O modelo preditivo previamente elaborado para população elite da International Paper apresentou capacidades preditivas que variaram de -0,296 para o caráter Altura em população geneticamente não relacionada até 0,440 para o caráter DAP em uma população geneticamente relacionada. Com os modelos desenvolvidos com as quatro populações genotipadas, e realizando seleção de marcas com base nos seus efeitos, as maiores capacidades preditivas foram obtidas por um número de SNPs médio que variou de 1371 para volume e IMA a 1467 para DAP. Usando esta abordagem, a capacidade preditiva maximizada para DAP foi de 0,744, 0,727 para altura, 0,751 para volume e 0,752 para IMA. Acurácias preditivas maximizadas, foram em seguida estimadas ao utilizar o menor número de SNPs selecionados. Com 237 SNPs acurácias da ordem de 0,660 para DAP, 0,555 para altura, 0,743 para volume e 0,743 para IMA foram obtidas. Embora estes resultados sugerem que a SGA teria bom resultado somente entre indivíduos geneticamente relacionados, uma análise conjunta dos dados utilizados para o desenvolvimento dos modelos preditivos anteriormente gerados com os dados das quarto populações aqui avaliadas, se faz necessária para alcançar resultados mais conclusivos. Adicionalmente, a abordagem de seleção de marcas para maximizar a capacidade preditiva ou acurácia deverá ser melhor avaliada à luz do seu impacto na medida que a SGA venha a ser praticada em futuras gerações de seleção. / The potential of genomics to the benefit of applied breeding is the direct use of DNA markers information for selection, allowing equal or higher selection efficiency, higher genetic gains per unit time and cost reduction in phenotyping, when compared with traditional selection based on phenotypic data. Researchers have proposed a new selection method called genome-wide selection (GWS), which has been successfully applied in livestock genetics and promises to revolutionize the improvement of perennial species with long life cycles. This genomic era is bringing new opportunities to forest breeders, but major challenges still need to be overcome for the operational use of GWS. Thus, this study aimed to assess the predictive ability of GWS models previously built for volume growth and wood quality traits based on an elite breeding population of International Paper in Brazil who served as discovery or training population. Predictive models were applied to four populations composed by progeny and clonal trials with contrasting characteristics from the standpoint of relatedness to the training population: (1) a population of individuals genetically related to the training population and (2) a population of individuals genetically unrelated to the training population. Subsequently, the four populations evaluated, with 100 genotyped individuals each, were used to build new models which were cross validated among them. Only SNPs with call rate &ge; 0.90 and MAF (minor allele frequency) &ge; 0.01 were used totaling 29,090 markers. Model previously developed yielded predictive abilities ranging from a lower -0.296 for height growth in genetically unrelated population to 0.440 for DBH and 0.219 for height in a genetically related population. With the GWS models built with the four genotyped populations and applying SNP marker selection based on their estimated effect the highest predictive capabilities were obtained when using an average number of SNPs ranging from 1371 for volume and MAI, to 1467 for DBH. The average predictive ability was maximized at 0.744 for DBH, 0.727 for height growth, 0.751 for volume and 0.752 for MAI. Maximized accuracies were obtained using the smallest number of SNPs with highest effects. With 237 selected SNPs accuracies around 0.660 for DBH, 0.555 for height, 0.743 for volume and 0.743 for MAI were obtained. While these results suggest that GWS should work well only across related individuals, a combined analysis of the data from the previous models with those of the four tested populations is necessary to reach more conclusive results. Furthermore, the approach of SNP marker selection to maximize predictive ability or accuracy is one that is still controversial and should be better evaluated in light of its impact as GWS is practiced in further generations of selection.

Background: Coronary heart disease (CHD) is one of the leading causes of death worldwide, and global mortality rates are expected to continue to rise over the coming decades. In Pakistan in particular, chronic diseases are responsible for 50% of the total disease burden. Circulating lipids are strongly and linearly associated with risk of CHD; however, despite considerable efforts to demonstrate causality, available evidence is conflicting and insufficient. Study of the underlying metabolic pathways implicated in the association between lipids and CHD would help to disentangle and elucidate these complex relationships. Objectives: The primary objectives of this dissertation were to (1) identify the genetic determinants of lipid metabolites and (2) advance understanding of the effect of perturbations in lipid metabolite levels on CHD and its risk factors. Methods: Direct infusion high-resolution mass spectrometry was performed on 5662 participants from the Pakistan Risk of Myocardial Infarction Study to obtain signals for 444 known lipid metabolites. Correlations and associations of the lipids with smoking, physical activity, circulating biomarkers, and other CHD risk factors were assessed. Genome-wide analyses were conducted to analyse the association of each lipid with over 6.7 million imputed single nucleotide polymorphisms. Functional annotation and Gaussian Graphical Modelling were used to link the variants associated with each lipid to the most likely mediating gene, discern the underlying metabolic pathways, and provide a visual representation of the genetic determinants of human metabolism. Mendelian randomisation was also implemented to examine the causal effect of lipids on risk of CHD. Results: The lipids were highly correlated with each other and with levels of major circulating lipids, and they exhibited significant associations with several CHD risk factors. There were 254 lipids that had significant associations with one or more genetic variants and 355 associations between lipids and variants, with a total of 89 sentinel variants from 23 independent loci. The analyses described in this dissertation resulted in the discovery of four novel loci, identified novel relationships between genetic variants and lipids, and revealed new biological insights into lipid metabolism. Conclusion: Analyses of lipid metabolites in large epidemiological studies can contribute to enhanced understanding of mechanisms for CHD development and identification of novel causal pathways and new therapeutic targets.