Emergence of an Alzheimer's Early Risk Prediction Model Based on Multiancestry Genome-Wide Analysis, Moving Beyond European Bias
Background
Diagnosis and risk prediction of Alzheimer's disease have traditionally been designed based on genetic data from European populations. This is largely due to the overrepresentation of Western populations in genome-wide association studies (GWAS). As a result, the predictive accuracy of existing polygenic risk scores (PRS) has been limited in non-European populations, leading to criticism that the benefits of precision medicine are confined to specific racial groups.
The key genetic factor in Alzheimer's disease, the apolipoprotein E (APOE) gene variant, has also been found to function differently across races. For example, the APOE ε4 allele significantly increases risk in European populations but has a relatively weaker effect in African populations. Therefore, a new risk evaluation tool that comprehensively integrates whole-genome variations and is universally applicable across diverse populations is required. Only by incorporating racial diversity can the risk of disease onset in non-European patients be predicted and early intervention be initiated.
Key Findings
A research team led by Nuzulul Kurniansyah and Xiaoling Zhang from the Boston University School of Medicine developed a new PRS model that transcends racial boundaries by utilizing multiancestry genomic data. The team first analyzed GWAS summary statistics from 63,000 Alzheimer's patients and 484,000 age-matched controls to derive an algorithm for calculating risk scores. The model's performance was then validated in an independent cohort consisting of 16,120 patients and 16,625 controls. Additional validation was conducted using another group of 1,500 patients and 75,500 controls to further reinforce reliability.
Validation results showed that the APOE-independent PRS developed by the research team consistently correlated with Alzheimer's risk across diverse populations, including African Americans, Hispanics, and East Asians. Higher score groups exhibited significant functional decline in memory, executive function, and language abilities. Notably, the score showed a clear correlation with hippocampal volume reduction in magnetic resonance imaging (MRI) analysis. Abnormal levels of amyloid-beta and phosphorylated Tau (pTau), core pathological markers of Alzheimer's disease, were also confirmed through cerebrospinal fluid testing.
Differences in the manifestation of genetic risk by gender are also a notable achievement. The team found that female patients showed a significantly greater deviation in pTau accumulation compared to males, even with the same genetic risk score. This suggests the possibility that sex hormones or sex-specific immune responses may interact with genetic vulnerabilities during the disease process. Longitudinal analysis also revealed that the top group with the highest genetic risk scores showed the most rapid cognitive decline even before clinical symptoms fully emerged.
Significance and Outlook
This study marks a pivotal shift in genomic analysis technology, moving away from race-centric bias toward a more inclusive precision medicine approach for all of humanity. Previous models lacked reliability when applied to non-European populations, limiting their clinical utility. The new predictive tool developed by the research team has demonstrated consistent accuracy across diverse global populations, ensuring fairness and practicality in genome-based disease prediction. It can now serve as a diagnostic aid to secure the critical treatment window.
The model also has positive implications for improving the efficiency of new drug development clinical trials. By applying the new risk score, high-risk groups among multiracial clinical participants can be precisely selected, thereby increasing the reliability of trials. However, the model alone cannot fully explain the complex influence of environmental factors such as education level, dietary habits, and cardiovascular health on Alzheimer's onset. Integrated predictive models that combine genetic information with lifestyle data should follow.
Nature Genetics, Published online: 27 August 2026; doi:10.1038/s41588-026-02722-8An APOE-independent polygenic risk score for Alzheimer’s disease derived from multiancestry GWAS summary statistics shows consistent associations with cognitive, imaging, biomarker and neuropathological features in diverse populations.
The new PRS prediction model can contribute to increasing the precision of early dementia screening in primary care settings. For example, if a patient undergoes a simple genetic test, healthcare providers can numerically present the likelihood of a sharp decline in cognitive function within the next five years based on the score. This test is expected to serve as a useful first-line screening tool in medically underserved populations and multicultural societies where access to expensive positron emission tomography (PET) scans or cerebrospinal fluid testing is limited.
In the digital healthcare industry, scenarios are possible where this risk score information is integrated with personalized prevention solution apps. By providing brain function activation cognitive training programs or dietary guides that promote vascular health to users classified as high-risk, the onset of the disease can be delayed as much as possible. In the pharmaceutical industry, it can also be used as a tool to significantly reduce the cost and duration of subject screening in multinational, multicenter clinical trials for preventive drugs, in addition to amyloid-targeting new drugs.