Alzheimer's Association International Conference 2026

INTRODUCTION

• Complex genetic risk scores for Alzheimer’s disease (AD) combine multiple polygenic risk scores (PRS)

• These scores have not yet provided additional predictive power beyond disease-specific PRSs

METHODS

• We calculated a composite genetic risk score for AD-type dementia (Meta-GRS) integrating 5 distinct PRSs specific for pathophysiological pathways

• We used machine learning models (elastic-net logistic regression with 10-fold cross-validation) to account for overlapping genetic contributions

• The Meta-GRS was compared with a disease-specific PRS specific for late-onset AD (PRS-AD)

• Data were drawn from HELIAD : a community-dwelling cohort of Greek individuals over 64 years of age

• 597 unrelated participants were included: without dementia at baseline (mean follow up of 2.9 years)

RESULTS

Discriminative Ability (AUC): The Meta-GRS showed superior discriminative ability (AUC=0.791) than the PRS-AD (AUC=0.742)

Explained Variability (pseudo-R2): Τhe Meta-GRS accounted for 9.8% of the variance in amnestic MCI (aMCI) /AD risk, which exceeded the proportion explained by the PRS-AD (R² = 6.1%)

CONCLUSIONS

• A Meta-GRS combining PRSs specific for biological pathways may improve risk prediction beyond disease-specific PRSs

• The selection of included PRSs is crucial for generating more accurate genetic risk scores

Alzheimer's Association International Conference 2026

Predicting Alzheimer’s Disease Through A Composite Genetic Risk Score Reflecting Multiple Pathophysiological Pathways

Poster Presenter

Authors

Affiliations

Part of Topic

Comments