Cardiovascular disease (CVD) is the No. 1 cause of death in the United States, killing about 610,000 people every year. Biomarkers are important tools to identify vulnerable individuals at high risk of CVD. Investigation of the genetic architecture for biomarkers and other risk factors related to CVD is of critical importance in the prevention and treatment of CVD. For my first chapter, I conducted genome-wide admixture and association studies for iron-related traits in 2347 African Americans (AAs) participants from the Jackson Heart Study (JHS). I identified, for the first time, a second independent genome-wide significant signal in the TF region associated with total iron binding capacity levels. I also identified a novel functional missense variant in the G6PD-GAB3 region significantly associated with ferritin levels. Both results were replicated in a second AA cohort with iron measures. For my second chapter, I conducted genome-wide admixture and association studies, and gene-based exome-wide association studies of rare variants, to identify variants or genes, harboring a high burden of rare functional variants, associated with lipoprotein(a) [Lp(a)] cholesterol levels in 2895 AAs participating in the JHS. I observed significant evidence for association between Lp(a) and both local ancestry and hundreds variants spanning ~10Mb the LPA gene region on chromosome 6q. Of note, the region containing associated variants became much narrower, centered over the LPA gene (<1Mb), after adjusting for local ancestry. I also observed a single significant non-synonymous SNP in APOE and a high burden of coding variants in LPA and APOE significantly associated with Lp(a) levels For my third chapter, I investigated the genetic association of four candidate variants with blood pressure and tested the modifying effects of environmental factors in 7,319 Chinese adults from the China Nutrition and Health Survey (CHNS). I observed that rs1458038 exhibited a significant genotype-by-BMI interaction affecting blood pressure measures, with the strongest variant effects in those with the highest BMI. Finally, for my last chapter, I described a multistage GWAS study design that uses selective phenotyping to increase power for studies with existing genome-wide genotypic data and to-be-measured quantitative phenotypes that are under a sample-size constraint. The approach uses simulated annealing to identify the optimal subset of subjects to be phenotyped in Stage 2 of a two-stage GWAS. I showed that our approach has greater statistical power than the conventional approach of randomly selecting a subset of subjects for phenotyping. We demonstrate the gains in power for both directly genotyped and imputed genetic variants. Together, these studies further our understanding of the genetic architecture of risk factors for CVD, suggest some candidates for future genetic and molecular studies, and also shed some light on the study design of future large-scale genetic association studies where the cost constraints will be determined by the expense of measuring new biomarkers in studies that have existing genetic data.