Bayesian Modeling of Multivariate Lifetime Disease Trajectories in Cardiovascular Disease

About This Grant

Project Summary / Abstract Developing statistical frameworks for understanding how genetic effects vary across phenotypes and over time remains a critical challenge in genomics. While advances in electronic health records and genomics have provided unprecedented data, existing models fail to capture individual-level dynamics and miss critical gene- environment interactions. In her previously published work, Dr. Urbut developed MSGene for modeling time-varying genetic effects on cardiovascular risk, demonstrating that genomic influences can vary substantially across life stages. She extended this work through Aladyn, a dynamic topic modeling framework applied to over 400,000 individuals in the UK Biobank, which established the computational feasibility of large-scale Bayesian inference across heterogeneous disease types. Building on these successes, her preliminary work for the current project with a novel survival-based framework has identified 20 unique disease signatures, capturing previously unrecognized patterns of disease progression and enabling both genomic discovery and individualized prediction. These results demonstrate the feasibility of Bayesian hierarchical models for understanding diverse phenotypes across the life course. First, Dr. Urbut proposes to develop and validate a novel Bayesian multivariate model integrating time-varying genetic and clinical factors across over 300 diagnostic phenotypes. This model has the dual objective of both identification of latent signatures for genomic discovery and producing calibrated estimates for prediction. Second, she will investigate how genetic factors influence disease trajectory timing through a novel "genetic warping" framework. Third, she will leverage identified disease trajectories to optimize therapeutic strategies by analyzing treatment response variability across patient subgroups. This work will take place in the Division of Cardiology at Massachusetts General Hospital. Dr. Urbut will perform this research under the mentorship of Dr. Pradeep Natarajan, Director of Preventive Cardiology and Associate Professor of Medicine, and Professor Giovanni Parmigiani, an expert in Bayesian methodology at the Harvard T.H. Chan School of Public Health and Dana Farber Cancer Institute. Dr. Urbut's goal is to become an independent investigator developing novel statistical methods to understand complex disease evolution and improve therapeutic targeting. She aims to use this K08 research as a foundation for future R01 applications in computational approaches to precision medicine.