Back to the Future: Predicting Individual Tau Progression in Alzheimer’s Disease

TL;DR

This paper introduces a new method to predict how tau spreads in Alzheimer’s patients, improving understanding of disease progression and variability.

Contribution

StaND combines staging and biophysical modeling to predict individual tau progression in Alzheimer’s Disease.

Findings

StaND outperforms existing models in predicting cross-sectional and longitudinal tau distributions.

Inferred tau seed patterns reveal spatial heterogeneity and rate parameters explain cognitive variability.

Two distinct tau seeding archetypes with unique clinical profiles were identified.

Abstract

Alzheimer’s Disease (AD) is characterized by the spread of tau neurofibrillary tangles along the brain’s structural network. The marked variability in pathology spread patterns across individuals necessitates a precision medicine approach. Here we introduce Stage-based Network Diffusion (StaND), a novel algorithm that combines statistical staging with biophysical modeling to predict patient-specific tau progression. Using data from 650 subjects in the Alzheimer’s Disease Neuroimaging Initiative, StaND first estimates each subject’s disease stage and then infers their individual tau seeding pattern, agglomeration rate, and transmission rate. The model is applied forward in time to predict regional tau distributions cross-sectionally and longitudinally. StaND outperforms benchmark models in both instances. Inferred tau seed patterns capture spatial heterogeneity, while rate parameters explain temporal and cognitive variability. Despite diverse initial seeding patterns, tau distributions converge across subjects over time. We also identify two distinct tau seeding archetypes with distinct clinical and demographic profiles. StaND offers a powerful new approach for understanding and forecasting the spatiotemporal dynamics of AD and is widely applicable to other neurodegenerative diseases.

Figures