Biomarkers

Area of Science:

• Neuroscience
• Biomarkers
• Machine Learning

Background:

• Distinguishing cognitive resilience from typical Alzheimer's disease (AD) tau pathology is challenging due to overlapping biomarker profiles.
• Cognitive resilience to AD pathology is a critical area of research for understanding disease mechanisms.
• Machine learning (ML) approaches offer potential for novel biomarker development in AD.

Purpose of the Study:

• To evaluate the tau heterogeneity score (THETA) as an effective marker of cognitive resilience to Alzheimer's disease.
• To assess THETA's ability to provide a better estimation of tau's temporal progression compared to existing biomarkers.
• To differentiate between typical AD, resilient, and resistant groups based on tau distribution and progression.

Main Methods:

• Longitudinal imaging data from the Mayo Clinic Study of Aging (MCSA) and ADRC were analyzed.
• Participants were categorized into typical AD, resilient, and resistant groups using amyloid PiB-PET, Flortaucipir (FTP-PET), and MMSE.
• Regional THETA scores were generated using an ML model to capture spatial tau heterogeneity; progression was modeled using SILA with plasma biomarkers (p-tau181, GFAP).

Main Results:

• THETA identified regions of reduced tau uptake and preserved cortical thickness in resilient individuals compared to the typical AD group.
• The Sampled Iterative Local Approximation (SILA) model demonstrated THETA's ability to distinguish global tau temporal progressions across different groups.
• Earlier onset of GFAP (inflammation marker) was observed in resilient and typical AD groups, suggesting early inflammatory processes.

Conclusions:

• THETA effectively identifies reduced tau pathology and preserved cortical thickness in resilient individuals, differentiating them from typical AD.
• THETA shows potential in uncovering pathological tau mechanisms underlying resilience and enhancing understanding of AD.
• The findings suggest THETA can provide valuable insights into resilience mechanisms and improve AD progression modeling.