Integrating Event-Based and Biophysical Models to Predict Individual Tau Progression in Alzheimer's Disease

Motivation: Current modeling approaches fail to capture individual longitudinal variation in tau protein neurofibrillary tangle spread in Alzheimer's Disease (AD). Goal(s): Our goal was to develop a modeling approach capable of predicting tau's origin as well as its future spread for individual subjects. Approach: We paired event-based and biophysical modeling techniques: first statistically extracting longitudinal tau trajectories from cross-sectional data, then optimizing network diffusion models to fit each subject's regional tau distributions. Results: Our combined modeling approach achieved strong correlations with empirical baseline and longitudinal tau data across subjects, revealing previously unobserved patterns of inter-subject convergence over time as well as distinct seeding archetypes. Impact: Our method can be applied to explore inter-subject heterogeneity of protein spread patterns across a range of neurodegenerative conditions, enabling the development of precision therapeutic treatments that target individuals' unique pathology.