In recent years, the United States has observed a significant increase in the Hispanic population within its borders through immigration. Since the 1980s, there has been a shift in the immigration and movement of the Hispanic population from border states to states in the southeastern United States. Many Hispanics, attracted by employment opportunities, have moved into growing metropolitan areas in the American South. In some of these cities, new Hispanic immigrants have created distinct enclaves. These enclaves provide almost all needs for the Hispanic community, such as shopping, healthcare, legal assistance, dining, employment, entertainment, and religion. This study examines the creation and functionality of the burgeoning Hispanic enclave within Davidson County, Tennessee through fieldwork, in-depth interviews with immigrants, volunteer work with Hispanic organizations, and census data. It suggests that the rise of this enclave, complete with Hispanic cultures, businesses, organizations, and churches, is interfering with the assimilation of Hispanics into the local community. The results from this study indicate that many Hispanics immigrants choose to function almost entirely within this enclave and that the enclave creates an environment in which immigrants do not need to intermingle with the host society. While Hispanic enclaves can serve as an important transition tool for many newly arrived immigrants, these findings suggest that ethnic enclaves can also have negative impacts on assimilation into the larger host community.

Preterm birth (PTB), defined as live birth before 37 weeks completed gestation, is the leading cause of infant mortality worldwide. Despite this major public health concern, little is known about the pathogenesis of PTB. The limited insight into PTB is contributed to by the fact that the mechanism for normal parturition is not known in humans. A number of lines of evidence suggest that PTB has a genetic component such as PTB aggregating in families, segregation analysis and genetic modeling. Primarily through candidate gene studies, there have been a number of single nucleotide polymorphisms (SNPs) associated with PTB; however, contradictory evidence from replication studies exists, and none of these have large effect sizes or have implicated novel insight into the mechanism of birth timing and parturition. The primary objective of this dissertation was to identify unidentified or unappreciated genetic variants responsible for the predisposition or pathogenesis of PTB. In order to investigate this hypothesis, we performed studies using a number of complementary approaches, which included genome-wide association studies (GWAS), exome array association, whole-exome sequence and pathway analysis. In addition to using a group of complementary methods, we also interrogated both maternal and fetal genomes for their potential role. Our GWAS in mothers identified several SNPs associated with the dichotomous trait, PTB, and the quantitative trait gestational age. Additionally, when we used whole-exome sequencing and exome array association in mothers we identified the Kyoto Encyclopedia of Genes and Genomes (KEGG) complement and coagulation cascade to be implicated in PTB and found a robust association with coding region SNPs in complement receptor 1 (CR1). Finally, when we switched our interrogation to infant samples we observed a handful of robust associations with PTB and the quantitative traits gestational age and birth weight z-scores.

With the exception of ApoE gene, no universally accepted genetic association has been identified with the complex Late-onset Alzheimer Disease (LOAD). A broad region of chromosome 10 has engendered continued interest generated from both preliminary genetic linkage and candidate gene studies. To better examine this region, we applied the genomic convergence approach by combining unbiased genetic linkage with candidate gene association studies. We genotyped 36 SNPs across 80.2 Mb in 567 multiplex families to narrow the peak region of linkage using both covariate and subset analyses. Simultaneously, we examined seven functional candidate genes that also fell within the broad area of linkage. Although a two point LOD score of 2.69 was obtained in the linkage analysis, the associated candidate genes were not under the linkage peak, suggesting a more extensive heterogeneity on chromosome10 than previously expected. We then converged linkage analysis and gene expression data to identify genes that were under linkage peaks and also differentially expressed in AD cases and controls based on the rationale that genes showing positive results in multiple studies are more likelihood to be involved in AD. We identified and examined 28 genes on chromosome 10 for the association with AD. Both single marker and haplotypic associations were tested in overall and eight subsets that were stratified by age, gender, ApoE status and clinical diagnosis. Gene-gene interaction was tested to detect important genes in this complex disease. PTPLA gene showed allelic, genotypic and haplotypic association in the overall dataset. The SORCS1 gene showed very significant association in the female dataset (allelic association p=0.00002, a 3-locus haplotype has p=0.00098). Two SNPs in CACNB2 gene showed gene-gene interaction in overall dataset using Multifactor Dimensionality Reduction (MDR). The work presented in this dissertation applied a multifactorial, multistep approach, genomic convergence, which combined linkage analysis, gene expression data, and candidate gene association analysis to identify and prioritize candidate susceptibility genes for AD. This study suggests that genetic variations in PTPLA, SORCS1 and CACNB2 genes might alter the risk for Alzheimer disease by affecting multiple pathways.

Low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and triglyceride (TG) levels are well known independent risk factors for cardiovascular disease. Other lipoproteins, such as Lp(a), are also emerging, independent risk factors as increasing epidemiologic evidence suggests. Lipid-associated single-nucleotide polymorphisms (SNPs) are being discovered in genome-wide association studies (GWAS) in samples of European descent, but little data exist in other populations. Therefore, there is a strong need to characterize the effect sizes and allele frequencies of these GWAS-identified variants in a diverse, population-based cohort. Also, despite the ever-growing number of loci detected by GWAS, the proportion of trait variation explained is collectively small. To investigate this missing heritability, it is important to continue to identify novel variants that are associated with lipid levels and to explore gene-environment interactions, which may also contribute to trait variation. The primary objective of this work was to identify and characterize common genetic variants that explain a proportion of the inter-individual variability in lipids levels, including LDL-C, HDL-C, TG, and Lp(a) levels. To achieve this goal, I selected a set of SNPs associated with lipid levels from the literature and demonstrated that the majority of associations replicate and generalize in a diverse, independent cohort. An additional GWAS of children was used to discover a novel variants associated with LDL-C, HDL-C, and TG. I also performed a candidate gene study and determined that common variants in LPA were associated with Lp(a) levels. Lastly, I identified several environmental modifiers of replicated variants associated with LDL-C, HDL-C, and TG.

Each year in the United States about one million (17%) of all pregnancies experience complications that result in fetal loss. Of the five million pregnancies that end in a live birth approximately 12% are born prematurely. Vaginal disorders and infections during pregnancy, particularly bacterial vaginosis (BV), are known risk factors for both miscarriage and preterm birth (PTB). Identifying the environmental and genetic risk factors for these complex reproductive disorders is important for improving health through better diagnostic methods and treatments. The purpose of this project is to examine the genetics and epidemiology of BV and PTB as examples of complex reproductive disorders. Cervical immunity clearly plays an important role in the pathogenesis and progression of BV. Therefore, cervical cytokine concentrations were measured during the first trimester of pregnancy in BV positive (BV+) and BV negative (BV-) women. Genetic associations between cytokine receptor genes and cervical cytokine concentrations were identified and discovered to differ by both BV status and race. The genetics of PTB was examined by genotyping approximately 1500 single nucleotide polymorphisms (SNPs) from 160 genes involved in preterm pathways, including decidual hemorrhage, infection and inflammation, activation of the hypothalamic-pituitary-adrenal axis, and uterine distention. SNPs in both the decidual hemorrhage and infection/inflammation pathways were associated with PTB. Additionally, many of these associations corroborate results from a previous association study.

Growing interest and burgeoning technology for discovering genetic mechanisms that influence disease processes have ushered in a flood of genetic association studies over the last decade, yet little heritability in highly studied complex traits has been explained by genetic variation. Non-additive gene-gene interactions, which are not often explored, are thought to be one source of this "missing" heritability. I begin this dissertation with a review of study designs and analytical methods for genetic association studies. Next, I characterize and present a series of improvements in using grammatical evolution to train neural networks for discovering gene-gene interactions in disease gene association studies. I then present an analysis of cis-epistasis - nonadditive multi-SNP interactions that influence gene expression. Next, I present a cohesive set of quality control procedures to be used for genome-wide association studies. Finally, I conclude by presenting results from a knowledge-driven gene-gene interaction analysis of HDL level in two clinical practice-based population biobanks.

Evaluating epistasis in whole-genome association studies is an important challenge in human genetics, as many common diseases are thought to have complex underlying genetic architectures that include small independent effects and interactions between many genes. In this project, I applied a simple bioinformatics approach for generating and ranking biologically supported multi-locus models of multiple sclerosis (MS) susceptibility, using data sources implying interaction of molecules, sources implying gene relationship to disease, and literature-based information. Putative gene-gene interaction models were constructed based on these relationships. These models were evaluated in whole-genome association dataset consisting of 931 MS case/pseudo-control pairs, 2,950 population-based controls, and a replication sample of 808 MS cases and 1,720 controls. Using this approach, I highlight the potential utility of this knowledge-driven analysis technique, and propose a potential role for inositol-based signaling molecules in multiple sclerosis susceptibility.

Biological organisms are complex systems that dynamically integrate inputs from a multitude of physiological and environmental factors. Complex clinical outcomes arise from the concerted interactions among the myriad components of a biological system. Therefore, in addressing questions concerning the etiology of phenotypes as complex as common human disease or adverse reaction to vaccination, it is essential that the systemic nature of biology be taken into account. Analysis methods must integrate the information provided by each data type in a manner analogous to the operation of the body itself. It is hypothesized that such integrated approaches will provide a more comprehensive portrayal of the mechanisms underlying complex phenotypes and lend confidence to the biological interpretation of analytical conclusions.<p>This dissertation concerns the development of a comprehensive analysis paradigm wherein experimental data of multiple types were analyzed jointly in the study of complex phenotypes. Flexible machine learning methods were used to integrate information that is insensitive to spatial and temporal flux (genetic polymorphisms) with information subject to dynamic changes (protein concentrations measured at multiple time points). This strategy was applied to genetic and proteomic data in both simulated and real analysis situations. Results of studies using simulated data indicated that utilizing multiple data types is beneficial when the disease model is complex and the phenotypic outcome-associated data type is unknown. The successful application to combined genetic and proteomic data from smallpox vaccine studies supported the hypothesis that such integrated approaches are analytically beneficial. <p>Considering the rapid progress in experimental technologies able to reliably generate vast quantities of data, as well as continual improvements in cost efficiency, it is expected that datasets including multiple types of experimental information will become commonplace in the near future. It is hoped that the positive conclusions from this dissertation will help spur the adoption of an analytical approach that rightfully takes the broader physiological context of complex biological systems into account.

Multiple sclerosis (MS) is characterized as a neurodegenerative autoimmune disease. This clinically complex disease has provided great challenges for geneticists over the years. With the advent of genome-wide association studies (GWAS), the strong genetic component associated with MS is finally beginning to be characterized. One of the first discoveries to emerge in this new era was the association with rs6897932 in the interleukin-7 receptor alpha chain (IL7RA) gene. The goal of the work presented in this dissertation was to identify additional genes that increase ones susceptibility to MS. Our studies involved examining genes in the extended biological pathway related to IL7RA to identify novel associations. Through this approach, we identified two additional novel gene regions that are likely associated with MS. These results help to further delineate the genetic architecture of MS and validate our pathway approach as an effective method to identify novel associations associated in a complex disease. We began our investigation with a discovery screen containing SNPs from 73 genes with putative functional relationships to IL7RA and subsequently genotyped 7,865 single nucleotide polymorphism (SNPs) in and around these genes. Two of the gene regions examined, IL7 and SOCS1, had significantly associated SNPs that further replicated in an independent case-control dataset with joint p-values reaching 8.29x10-5 and 3.48x10-7, respectively, exceeding the threshold for experiment-wide significance. Our results also implicated two additional novel gene regions that are likely to be associated with MS: PRKCE with p-values reaching 3.47x10-4 and BCL2 with p-values reaching 4.32x10-4. The TYK2 gene, which emerged in our analysis, also has recently been associated with MS in other studies. The work presented in this dissertation confirmed two novel regions and implicated several others that need further examination as MS disease loci. Thus, using the pathway approach in conjunction with large datasets and dense genotyping, the etiology of MS is finally starting to be dissected. By building on the knowledge of these gene effects, these studies will hopefully result in further understanding of the pathogenesis of MS.

Spontaneous preterm birth (PTB), gestational age less than 37 weeks, is a major public health problem worldwide. Almost 500,000 preterm births occur annually in the United States (U.S.), accounting for ~12% of deliveries. The outcomes associated with PTB include increased risk of infant morbidity and mortality. In addition to the effects on the health of the child, there is a significant disparity in incidence and clinical correlates of PTB between ethnic groups in the U.S. PTB, for example, among African Americans is ~18% compared to ~12% in Caucasians. Also, African American women who experience PTB are also more likely to have future PTB and early PTB associated with infection (less than 32 weeks gestation) than Caucasian women. Both the disparity in PTB rates and clinical correlates between ethnic groups in the U.S. may be explained by genetic causes. An important aspect of PTB that distinguishes it from other phenotypes is that both maternal and fetal genes can affect pregnancy outcome. Because of the complex relationship between mother and fetus, the numerous factors affecting pregnancy and substantial selection for term delivery to ensure survival, it is unlikely that any single factor will be sufficient to explain PTB. To better understand the genetic contribution to PTB and the rate disparity between ethnic groups, we performed a comprehensive genetic analysis. A high-throughput candidate gene screen and an analysis of amniotic fluid (AF) cytokine protein levels were performed to identify PTB risk factors. We selected candidate genes based on genes showing positive results in previous studies and/or being involved in hypothesized pathophysiological pathways. We selected 130 genes and 1536 single nucleotide polymorphisms (SNPs) from the nuclear genome and three candidate genes from the mitochondrial genome. We also examined AF protein concentration of seven well-established biomarkers of PTB for association with PTB. Our study indicates that several common risk factors in the infection and inflammatory response pathways may explain a portion of the risk in both Caucasians and African Americans; however, different genes lead to increased risk in Caucasians and African Americans, with the maternal influence being greater in Caucasians than in African Americans.