Bayesian Integrative Analysis and Prediction with Application to Atherosclerosis Cardiovascular Disease
Cardiovascular diseases (CVD), including atherosclerosis CVD (ASCVD), are multifactorial diseases that present a major economic and social burden worldwide. Tremendous efforts have been made to understand traditional risk factors for ASCVD, but these risk factors account for only about half of all cases of ASCVD. It remains a critical need to identify nontraditional risk factors (e.g., genetic variants, genes) contributing to ASCVD. Further, incorporating functional knowledge in prediction models have the potential to reveal pathways associated with disease risk. We propose Bayesian hierarchical factor analysis models that associate multiple omics data, predict a clinical outcome, allow for prior functional information, and can accommodate clinical covariates. The models, motivated by available data and the need for other risk factors of ASCVD, are used for the integrative analysis of clinical, demographic, and multi-omics data to identify genetic variants, genes, and gene pathways potentially contributing to 10-year ASCVD risk in healthy adults. Our findings revealed several genetic variants, genes and gene pathways that were highly associated with ASCVD risk. Interestingly, some of these have been implicated in CVD risk. The others could be explored for their potential roles in CVD. Our findings underscore the merit in joint association and prediction models.