Alzheimer's Imaging Consortium.

Area of Science:

• Neuroscience
• Biophysics
• Computational Biology

Background:

• Alzheimer's disease (AD) affects millions globally, characterized by tau protein accumulation with significant individual variability beyond traditional Braak staging.
• Current modeling methods for tau progression lack biophysical grounding or require unavailable longitudinal data.

Purpose of the Study:

• To develop a hybrid computational model integrating Event Based Models (EBMs) and Network Diffusion Models (NDMs) for predictive, individualized tau spread modeling in Alzheimer's disease.
• To overcome limitations of existing models by leveraging cross-sectional data for longitudinal predictions.

Main Methods:

• A hybrid approach combining EBMs and NDMs was developed.
• An EBM assigned disease stages to 650 ADNI subjects using cross-sectional biomarker data to generate population-level trajectories.
• An extended NDM (eNDM) modeled tau spread as a diffusion process on brain networks, fitted to EBM trajectories and individual tau patterns by optimizing seed and kinetic parameters.

Main Results:

• Individual seed optimization significantly improved model fit (mean R=0.85) compared to parameter optimization and prior benchmarks.
• Model predictions strongly correlated with longitudinal tau-PET data (mean R=0.81), validating the approach.
• Tau patterns exhibited maximal heterogeneity at disease onset, converging over time, and two distinct seeding archetypes (entorhinal-dominant and diffuse temporal lobe) were identified.

Conclusions:

• The hybrid EBM-NDM framework enables accurate, individualized prediction of tau spread from cross-sectional data, surpassing previous methods.
• This approach integrates longitudinal insights from EBMs with the biophysical basis of NDMs, addressing limitations of each.
• Findings suggest Alzheimer's disease involves diverse tau initiation pathways that converge, offering potential for personalized diagnostics and therapeutics.