Examining the Performance of Polygenic Risk Scores for Alzheimer Disease Within and Across Populations Using k-Fold...

Area of Science:

• Genetics
• Population Genetics
• Computational Biology

Background:

• Alzheimer's disease (AD) has a complex genetic basis, but most genetic studies focus on European ancestry populations.
• Genetic risk prediction tools, like polygenic risk scores (PRSs), may perform differently across diverse racial and ethnic groups.
• Understanding these variations is key to equitable genetic risk assessment for AD.

Purpose of the Study:

• To investigate the performance and transferability of AD-specific polygenic risk scores (PRSs) across different race and ethnicity groups.
• To evaluate how PRS accuracy varies when trained and tested within and across Hispanic, non-Hispanic Black, and non-Hispanic White individuals.
• To establish a framework for enhancing PRS utility in diverse populations.

Main Methods:

• Analyzed 11,254 individuals from the Alzheimer's Disease Sequencing Project (ADSP) Release 3, categorized into Hispanic, non-Hispanic Black, and non-Hispanic White groups.
• Employed a 5-fold cross-validation approach for genome-wide association studies (GWAS) and PRS construction within each group.
• Calculated the area under the curve (AUC) to assess PRS performance, examining various criteria including APOE region, sex-specific PRSs, and sample sizes.

Main Results:

• PRSs trained within the same race/ethnicity group consistently outperformed those trained in other groups, particularly benefiting the Hispanic group.
• The variability in AUC across iterations was greater when PRSs were applied across different racial/ethnic groups.
• These findings remained consistent even after including covariates like APOE, sex, and age, and with different PRS generation criteria.

Conclusions:

• AD polygenic risk scores exhibit variable performance and limited transferability across different race and ethnicity groups.
• The study underscores the critical need for diverse population data to improve the accuracy and applicability of genetic risk prediction for AD.
• This research provides a foundation for developing more equitable and effective PRS tools for Alzheimer's disease research and clinical applications.