Highlights
- Multi-ancestry polygenic risk scores (PRS) outperform single-ancestry models and APOE genotyping in predicting Alzheimer’s disease (AD) risk among Hispanics.
- Method-focused PRS models, integrating GWAS summary statistics from African, European, and Hispanic populations, explain more variance in clinical AD and cognitive outcomes.
- APOE and multi-ancestry PRS capture distinct biological facets of AD, as reflected in different plasma biomarker associations.
- The study addresses a critical gap by evaluating genomic risk prediction in a historically understudied population.
Study background and disease burden
Alzheimer’s disease (AD) represents a major public health challenge, with rapidly increasing prevalence in aging populations worldwide. The Hispanic community in the United States and the Americas faces a particularly steep rise in AD incidence, yet remains underrepresented in genetic research. Most polygenic risk scores (PRS)—tools that aggregate the effects of many genetic variants to predict disease risk—have been developed and validated largely in populations of European descent. This Eurocentric focus has limited the clinical utility of PRS for non-European groups, including Hispanics, potentially exacerbating health disparities. Recognizing the need for more precise genomic risk assessment in diverse populations, recent advances in multi-ancestry genome-wide association studies (GWAS) and novel PRS methodologies aim to improve the accuracy and equity of AD risk prediction across ancestries.
Study design
This study by Xu et al. (Lancet Reg Health Am. 2025) systematically assessed the predictive performance of both traditional and innovative PRS models for Alzheimer’s disease in a Hispanic cohort. The research utilized data from 2,961 Hispanic individuals drawn from two large population-based studies. The investigators compared the following genetic predictors:
- The well-established Apolipoprotein-E (APOE) locus, especially the ε4 allele
- Single-ancestry PRS derived from GWAS in homogenous populations
- Multi-ancestry PRS constructed by two approaches:
- GWAS-focused: PRS derived from multi-ancestry GWAS summary statistics
- Method-focused: PRS constructed using advanced statistical methods (e.g., PROSPER/PRS-CSx) integrating GWAS summary statistics across African, European, and Hispanic populations
The study endpoints included clinical AD diagnosis, incident AD, cognitive performance (quantified by standardized testing), and plasma AD biomarkers (Aβ and P-tau). The genetic predictors were evaluated using robust statistical techniques, including ten repetitions of five-fold cross-validation and tuning-validation splits for biomarker analysis.
Key findings
The study’s major findings provide critical insights into the evolving landscape of genomic risk prediction in diverse populations:
1. APOE Remains a Strong but Incomplete Predictor
The APOE-ε4 allele showed a robust association with clinical AD (pooled OR: 1.83 [1.72–1.94], mean ΔR2 = 0.014 [0.011–0.016], mean AUC = 0.60 [0.59–0.62]), incident AD (pooled HR: 1.43 [1.34–1.54], mean ΔR2 = 0.021 [0.016–0.027]), and cognitive impairment (pooled beta: −0.78 [−0.85 to −0.71], mean ΔR2 = 0.015 [0.013–0.017]). The APOE genetic score explained slightly more variance than the simple presence/absence of ε4, but the difference was marginal once ancestral background was considered.
2. Multi-Ancestry PRS Methodology Outperforms Traditional Models
The method-focused multi-ancestry PRS, which aggregated GWAS summary statistics from African, European, and Hispanic cohorts using state-of-the-art statistical approaches, explained up to 1.6% of the variance in clinical AD, 3.9% in incident AD, and 1.7% in cognition. These figures are comparable to, and in some endpoints greater than, the variance explained by APOE alone. Notably, this approach consistently outperformed both single-ancestry PRS and the GWAS-focused multi-ancestry PRS, underscoring the importance of methodological innovation and data integration in polygenic prediction, especially for admixed populations.
3. Biomarker Analyses Reveal Distinct Genetic Contributions
In analyses of plasma biomarkers, APOE explained a greater proportion of variation in phosphorylated tau (P-tau) levels—implicating its role in tau pathology—while the multi-ancestry PRS explained more variance in amyloid-beta (Aβ) levels. These findings support the hypothesis that APOE and broader polygenic risk capture partially distinct biological mechanisms underlying AD pathogenesis.
4. Clinical Implications and Statistical Robustness
The use of repeated cross-validation and independent validation for biomarker outcomes enhances the robustness and generalizability of these findings. While absolute variance explained remains modest, the incremental improvement in risk prediction—particularly in a population where traditional PRS models underperform—signals a meaningful advance.
Expert Commentary
The findings from Xu et al. represent an important step toward equitable precision medicine in Alzheimer’s disease. By demonstrating that multi-ancestry PRS, constructed with advanced statistical models, can outperform traditional approaches in Hispanics, the study addresses a pressing gap in genomic risk stratification. As highlighted by recent expert commentaries (Martin AR et al., Nat Genet. 2019; Majara L et al., Nat Med. 2021), PRS portability across ancestries is a major challenge for clinical genomics. This work provides empirical evidence that combining GWAS data across ancestries and leveraging statistical innovation can provide more accurate and inclusive risk prediction.
However, several limitations warrant consideration. Despite improvements, the variance explained by PRS remains well below thresholds required for individual clinical decision-making. Environmental, sociocultural, and nongenetic risk factors—often undermeasured in genetic studies—likely play substantial roles in AD risk among Hispanics. Additionally, the study’s sample size, while among the largest for Hispanic cohorts, is still modest compared to large European datasets, which may limit the discovery of population-specific risk loci.
Mechanistically, the distinct biomarker associations suggest that APOE and polygenic risk aggregate different sets of biological pathways, reinforcing the need for multimodal risk assessment in AD. The growing availability of multi-ethnic GWAS and biobank data, coupled with methodological advances such as PRS-CSx, offer a promising path forward.
Conclusion
Multi-ancestry polygenic risk scores, particularly those employing advanced integrative methodologies, substantially improve Alzheimer’s disease risk prediction in Hispanic populations compared to single-ancestry models and APOE genotyping. While not yet ready for standalone clinical use, these findings mark a significant advance in the equitable application of genomic medicine. Future research should focus on increasing sample sizes in diverse populations, integrating genetic with environmental and lifestyle risk factors, and moving toward actionable risk stratification for prevention and intervention in Alzheimer’s disease.
References
1. Xu Y, Qiao M, Gunasekaran TI, et al. Evaluating polygenic risk score prediction performance for Alzheimer’s disease in a population-based Hispanic cohort using single- and multi-ancestry models. Lancet Reg Health Am. 2025 Jul 25;49:101198. doi: 10.1016/j.lana.2025.101198.
2. Martin AR, Kanai M, Kamatani Y, Okada Y, Neale BM, Daly MJ. Clinical use of current polygenic risk scores may exacerbate health disparities. Nat Genet. 2019;51(4):584-591.
3. Majara L, Kalungi A, Koen N, et al. Low generalizability of polygenic scores in African populations due to genetic and environmental diversity. Nat Med. 2021;27(9):1530-1536.