New Alzheimer’s genetic score improves risk prediction across ancestry groups

A single genetic variant cannot explain most people’s risk of late-onset Alzheimer’s disease. Thousands of smaller genetic differences also matter, but turning those differences into a useful risk score has produced an enduring problem: the scores often work best for people of European ancestry.

A new multiancestry approach aims to narrow that gap. Researchers developed an Alzheimer’s polygenic risk score using genetic data from European, African American, Caribbean Hispanic and East Asian populations, then tested it across several independent and ancestrally diverse groups.

The score predicted Alzheimer’s even after researchers removed genetic variants surrounding APOE, whose ε4 allele remains the strongest common genetic risk factor for late-onset disease. The findings also connected higher scores with cognitive performance, brain structure, Alzheimer’s biomarkers and brain pathology.

“A critical challenge in the application of PRS lies in the underrepresentation of diverse genetic ancestries in AD GWAS datasets,” said Lindsay A. Farrer of Boston University Chobanian & Avedisian School of Medicine.

Polygenic risk scores combine the effects of many genetic variants into one estimate of inherited susceptibility. Their accuracy depends heavily on the populations represented in the genome-wide association studies used to construct them.

The researchers drew on summarized genetic information from more than 63,000 Alzheimer’s cases and 484,000 age-matched controls. They then evaluated the score in an independent, diverse dataset containing 10,612 Alzheimer’s cases and 16,625 elderly controls, followed by additional validation.

One important step was separating the score from APOE. Variants within 1 million DNA bases of the APOE coding region were removed, while ε2 and ε4 status entered statistical models separately. This allowed the team to ask how much information the rest of the genome carried.

Across the main Alzheimer’s Disease Sequencing Project analysis, each one-standard-deviation increase in the score was associated with 1.39 times the odds of Alzheimer’s disease.

The association was strongest among people of European ancestry, where each standard-deviation increase corresponded to an odds ratio of 1.52. Significant associations also appeared among Caribbean Hispanic, African American and Native American Hispanic participants.

Independent datasets supported the broader pattern. Each standard-deviation increase was associated with an Alzheimer’s odds ratio of 1.21 among 71,549 All of Us participants, 1.35 among 897 African American Rush University participants, 1.65 among 425 Korean participants and 1.35 in 3,687 Framingham Heart Study participants.

The score did not show significant associations in South Asian or East Asian participants within the main sequencing dataset. Researchers noted that those groups were much smaller, particularly the East Asian sample.

That limitation reflects the problem the team is trying to address. Non-European populations remain underrepresented in Alzheimer’s genetic research, leaving fewer data for constructing ancestry-specific models and testing whether findings transfer across populations.

The investigators went beyond clinical diagnosis to ask whether the score tracked biological features of Alzheimer’s disease.

Among 1,393 people with cerebrospinal fluid measurements, higher genetic risk was associated with lower amyloid-beta42 and higher total tau and phosphorylated tau181. The relationship between the score and phosphorylated tau181 appeared stronger in women.

Blood biomarkers showed a similar pattern. Higher scores were associated with increased plasma phosphorylated tau181 among 1,679 Framingham participants and with higher phosphorylated tau217 in another group of 301 people.

Brain tissue findings pointed in the same direction. Increasing genetic risk was associated with greater neuritic plaque burden, greater amyloid deposition, more advanced Braak stage and greater overall Alzheimer’s Disease Neuropathologic Change.

MRI scans added another connection. Higher scores were associated with smaller hippocampal volume in both the Alzheimer’s Disease Sequencing Project and Framingham Heart Study samples. The hippocampus is among the brain regions strongly affected early in Alzheimer’s disease.

Genetic risk also tracked cognition. Higher scores were associated with poorer memory, executive function and language performance across repeated assessments.

Memory showed the strongest association. When participants were divided by genetic risk and their predicted performance was examined between ages 60 and 80, people in the highest 10% already showed lower memory scores around age 60. The gap widened as they aged.

The pattern changed after Alzheimer’s developed. Among people who eventually received a diagnosis, those at higher genetic risk showed somewhat faster memory decline beforehand, but differences became smaller after diagnosis.

That suggests the score may be more informative for identifying risk before clinical disease than for predicting how rapidly someone will deteriorate after Alzheimer’s has developed.

Sex differences also appeared in several analyses. Associations involving memory, hippocampal volume and phosphorylated tau were sometimes stronger among women, although researchers said the biological reasons remain uncertain.

The findings do not mean someone’s Alzheimer’s future can be read directly from a genetic score. Performance continued to vary by ancestry, and no standardized clinical thresholds exist for interpreting these scores.

Some datasets also identified Alzheimer’s through diagnostic and medical billing codes rather than detailed biological confirmation. A clinical diagnosis does not guarantee that Alzheimer’s pathology caused a person’s symptoms.

Smaller ancestry and sex subgroups reduced statistical power. Genome-wide association data were also unavailable for some populations, including Native American Hispanic and Caribbean Hispanic groups, limiting how precisely the score could be tailored to them.

The strongest potential use for an ancestry-aware score may be as one piece of a larger Alzheimer’s risk picture rather than a stand-alone answer.

Genetic risk could eventually be considered alongside memory testing, brain imaging and amyloid or tau biomarkers. Because differences appeared before diagnosis, the score may also help researchers identify people at elevated long-term risk for prevention studies or clinical trials.

“The associations we observed with early biological and cognitive changes and potential sex-specific differences support the value of this ancestry-aware PRS for long-range risk prediction,” said co-corresponding author Xiaoling Zhang.

More diverse genetic datasets remain essential. Improving representation could make future scores more accurate across populations and clarify which genetic effects are broadly shared and which vary with ancestry.

For now, the work shows that useful Alzheimer’s genetic information extends well beyond APOE, and that combining ancestry groups can make that information more broadly informative.

These resources examine how ancestry, genetic diversity and polygenic risk influence Alzheimer’s prediction and the broader path toward clinical use.

Transferability of European-derived Alzheimer’s disease polygenic risk scores across multiancestry populations: This study tested Alzheimer’s polygenic scores across populations in Europe, Africa, Asia and the Americas, finding broad transferability but continued limits caused by small diverse-ancestry datasets. (Nature Genetics, 2025)

Multi-ancestry genome-wide meta-analysis of 56,241 individuals identifies known and novel cross-population and ancestry-specific associations as novel risk loci for Alzheimer’s disease: A large multiancestry GWAS identified shared and population-specific Alzheimer’s susceptibility loci, illustrating how broader representation can uncover genetic signals missed by predominantly European datasets. (Alzheimer’s Research & Therapy, 2025)

Biobank-scale genetic characterization of Alzheimer’s disease and related dementias across diverse ancestries: This biobank-scale analysis examined genetic risk and resilience across diverse ancestral backgrounds, supporting more inclusive approaches to Alzheimer’s genetics and personalized risk research. (Nature Communications, 2025)

Multi-ancestry meta-analysis identifies genetic modifiers of age-at-onset of Alzheimer's disease at known and novel loci: Researchers examined genetic influences on when Alzheimer’s begins across diverse populations and identified both established and novel signals, including effects that remained after accounting for APOE. (Alzheimer’s & Dementia, 2025)

Clinical use of polygenic risk scores: current status, barriers and future directions: This review examines the promise and current limitations of bringing polygenic scores into medical care, including ancestry transferability, precision and interpretation challenges. (Nature Reviews Genetics, 2026)

Research findings are available online in the journal Nature Genetics.

The original story "New Alzheimer’s genetic score improves risk prediction across ancestry groups" is published in The Brighter Side of News.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.