Dissecting Alzheimer's disease heterogeneity by cross-trait polygenic prediction

Area of Science:

• Genetics
• Neuroscience
• Computational Biology

Background:

• Alzheimer's disease (AD) exhibits significant clinical and pathological heterogeneity.
• Genetic dissection of AD heterogeneity is challenging due to limited large cohorts with deep phenotyping.
• Polygenic score (PGS) analysis offers a strategy to leverage pleiotropy for dissecting complex trait genetics.

Purpose of the Study:

• To develop and apply a PGS analysis strategy to investigate the genetic underpinnings of AD heterogeneity.
• To identify genetic associations between pre-trained PGS and deep AD phenotypes.
• To explore the potential of PGS for predicting AD-related traits and stratifying patient subtypes.

Main Methods:

• Integrated 713 UK Biobank-derived PGS with 36 deep AD phenotypes from the ROSMAP cohort (n=1678).
• Performed cross-cohort, cross-trait PGS analysis to identify significant associations (FDR<0.1).
• Developed predictive models using prioritized PGS and assessed their performance against existing AD PGS and APOE status.

Main Results:

• Identified 268 significant associations between 12 prioritized PGS (including lipid, inflammatory, and cancer traits) and 36 AD phenotypes (cognition, amyloid, tau).
• 49 associations remained significant after excluding APOE-related PGS.
• Predictive models incorporating multiple PGS outperformed APOE or AD PGS alone in predicting amyloid burden and cognitive status.
• Discovered six distinct individual-level AD polygenic subtypes characterized by unique pathological patterns.

Conclusions:

• PGS analysis is a powerful approach to dissecting genetic heterogeneity in complex diseases like AD.
• Identified novel genetic factors associated with AD heterogeneity beyond APOE.
• The developed strategy provides a framework for stratifying disease cohorts and advancing personalized medicine in AD.