Despite improvements in medical treatments, the global burden of cancer has steadily increased over time. Breast cancer remains the most frequently diagnosed malignancy in women worldwide. Disparities in the incidence and mortality of breast cancer tend to be pronounced among women of African ancestry, driven in part by a higher prevalence of aggressive subtypes and a general lack of effective screening tools. To improve the assessment of cancer risk in diverse populations, as well as our understanding of the underlying etiology of these diseases, it is paramount to study both genetic and environmental factors that contribute to the development of cancer. This dissertation examines two distinct but complementary research foci: the genetic epidemiology of breast cancer and the molecular epidemiology of environmental carcinogen exposure. In Chapter 1, we characterize the genetic architecture of breast cancer by conducting a cross-ancestry genome-wide association study of intrinsic subtypes, along with a cell type-aware transcriptome-wide association study, to identify susceptibility loci and genes for each subtype. Following this, in Chapter 2, we develop and validate polygenic risk score (PRS) models for several breast cancer phenotypes including overall, estrogen receptor-positive, estrogen receptor-negative, and triple-negative breast cancer among African-ancestry women, leveraging information across ancestries, subtypes, and PRS methodologies to improve predictive performance. Chapter 3 then shifts to assessing the molecular epidemiology of arsenic (As) exposure, a well-established carcinogen affecting millions of individuals worldwide. Utilizing DNA methylation (DNAm) data from Bangladeshi adults who are chronically exposed to a wide range of As levels, we identify CpG sites associated with As exposure and then apply Mendelian Randomization to demonstrate that genetically predicted As metabolism efficiency causally impacts DNAm at these sites. We then develop a DNAm-based biomarker of As exposure that robustly predicts exposures across multiple diverse populations. This biomarker also accurately predicts As-associated toxicities including arsenical skin lesions with high sensitivity and specificity, and associates with overall mortality even in an external population in the United States with lower levels of As exposure. Together, these three chapters demonstrate that leveraging both genetic and epigenetic data can improve our ability to characterize disease risk and improve risk-stratification at the population-level.