Aladyn Individual: Bayesian Hierarchical Dynamic Genetic Modeling of Comorbidity Progression
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Early identification of high-risk individuals through the analysis of their unique disease trajectories has a strong potential to support efficient prevention and clinical management across a range of chronic conditions. In this paper we present a novel approach for dynamic modeling of the evolution of chronic disease risks over time, incorporating individual genetic predispositions. Our approach uses a hierarchical Bayesian topic model including Gaussian Processes to capture age effects. It accounts for genetic predisposition through a time-warping function and topic-dependent genetic scores, enabling both simultaneous learning and updated predictions of complex comorbidity patterns, inclusive of genomic and non-genomic effects. We systematically compare to previous approaches and provide detailed simulations at https://bookdown.org/sarahmurbut/dynamic_ehr/ and https://surbut.shinyapps.io/dynamic_ehr .
Impact Statement
Our model significantly advances healthcare for aging populations by facilitating the early identification of high-risk individuals through the analysis of their unique disease trajectories within complex comorbidity patterns. Existing models are limited in their capacity to manage diverse comorbidity patterns, particularly those that are time-dependent. We introduce an approach leveraging Bayesian hierarchical modeling to concurrently learn population-level patterns and provide updated real-time predictions across 350 diseases, thereby uncovering and forecasting intricate comorbidity patterns. This methodology paves the way for preventive measures and targeted interventions that enhance health outcomes, mitigate late-stage disease burdens, and foster healthier aging. Furthermore, our model incorporates genetic influences via a genetic predisposition parameter to estimate the lifetime risk of specific diseases and disorders, alongside a time-warping function to facilitate personalized predictions of disease trajectories.
