BIOLOGICAL AND CLINICAL IMPLICATIONS OF OBESITY GENOMICS IN ANCESTRALLY DIVERSE POPULATIONS

Obesity, a major risk factor for numerous health outcomes, particularly cardiovascular diseases (CVD), is a highly polygenic trait. Thousands of obesity-associated genetic loci have been identified, facilitating more accurate risk prediction through polygenic risk scores (PRS). Nonetheless, significant research gaps in obesity genomics exist, notably regarding two key aspects: (1) Heterogeneities in PRS prediction across different PRS estimation methods, self-reported race/ethnicity, and different individual-level contexts, and (2) Heterogeneities in shared genetic underpinnings between obesity and dyslipidemia, a major contributor to CVD risk. This dissertation had two specific aims that addressed these research gaps as follows: to characterize the prediction performance of PRS for obesity traits across different PRS estimation methods and diverse settings, including self-reported race/ethnicity, demographic factors, lifestyle factors, and comorbidities (Aim 1); and to identify shared genetic underpinnings in obesity and lipid traits that increased the risk of obesity but were protective for dyslipidemia, as a means to understand why not all obese populations have high risk of CVD (Aim 2). To achieve these goals, we leveraged data from the Population Architecture Using Genomics and Epidemiology (PAGE) study. Our findings reveal notable differences in PRS prediction across different PRS estimation methods, self-reported racial/ethnic groups, age, gender, smoking status, hypertension, and type 2 diabetes. We also identified 966 genomic regions (among a total of 2,495 partitioned genomic regions) with shared genetic signals between obesity-related traits and lipid traits, with 16 genomic regions of these loci exhibiting counterintuitive directions (associated with increased body mass index (BMI) but decreased dyslipidemia). In PAGE, we observed significant associations of the PRS constructed from variants within these counterintuitive BMI-HDL bivariate loci with lower levels of CVD risk factors. These results enhance our understanding of the heterogeneous underpinnings of obesity susceptibility.Doctor of Philosoph