Ancestry Matters in Multiple Sclerosis Risk Prediction: Evaluating Polygenic Risk Scores Across Diverse Populations
Multiple sclerosis is a chronic immune-mediated disease of the central nervous system in which inflammation and demyelination contribute to neurological disability. Although environmental factors are important, genetic susceptibility also plays a substantial role. The article, titled “Differential Results of Polygenic Risk Scoring for Multiple Sclerosis in European and African American Populations,” examines whether a polygenic risk score for multiple sclerosis can stratify disease risk equally across different ancestry groups. This question is scientifically and clinically important because polygenic risk scores are increasingly discussed as tools for early disease prediction, population stratification, and personalized medicine, yet many of these tools are derived primarily from genome-wide association studies conducted in individuals of European ancestry.
Study Objective and Scientific Rationale
The central objective of the study was to determine whether a multiple sclerosis polygenic risk score developed largely from known genetic risk variants could effectively classify individuals of non-European ancestries into meaningful risk categories. The investigators focused particularly on European, African, and Latino/admixed American genetic ancestry groups. This design directly addresses a major limitation in contemporary human genetics: the unequal representation of global populations in genomic discovery studies. If a risk model performs well in one ancestry group but poorly in another, its use in clinical or research settings could widen existing health disparities rather than reduce them.
Data Source and Cohort Design
The study used data from the All of Us Research Program, a large United States-based biomedical cohort designed to include participants historically underrepresented in biomedical research. The analysis included participants with both whole genome sequencing data and electronic health record data collected between 2018 and 2023. After applying inclusion criteria, the investigators analyzed 173,153 participants, with the largest ancestry groups being European, African, and Latino/admixed American. To support comparable analyses across groups, the authors randomly sampled equal-sized ancestry groups of 32,428 participants each from the European, African, and Latino/admixed American populations.
Construction of the Multiple Sclerosis Polygenic Risk Score
The exposure of interest was a polygenic risk score based on 282 independent single nucleotide variants previously associated with increased risk of multiple sclerosis. For each participant, the score was calculated by summing the number of risk alleles carried at each variant, weighted by the reported effect size of that allele. The authors then normalized the score within each ancestry group and divided participants into quintiles representing very low, low, intermediate, high, and very high polygenic risk. This structure allowed the investigators to test whether increasing genetic burden corresponded to increasing multiple sclerosis prevalence within each ancestry-defined population.
Main Findings Across Ancestry Groups
The results showed clear ancestry-dependent differences in polygenic risk score performance. Among participants of European ancestry, multiple sclerosis prevalence increased from 0.66% in the lowest polygenic risk quintile to 1.59% in the highest quintile. In adjusted logistic regression models, individuals in the highest quintile had 2.41-fold higher odds of multiple sclerosis compared with those in the lowest quintile. A similar pattern was observed in the Latino/admixed American group, where the highest quintile was associated with 2.56-fold higher odds of multiple sclerosis compared with the lowest quintile. In contrast, the score did not significantly stratify risk among participants of African ancestry, where the trend across quintiles was weaker and statistically non-significant.
Interpretation and Implications for Precision Medicine
These findings demonstrate that polygenic risk scores cannot be assumed to transfer uniformly across ancestry groups. The weaker performance in African ancestry participants may reflect underrepresentation of African populations in the genome-wide association studies used to identify multiple sclerosis risk variants. It may also reflect differences in linkage disequilibrium patterns, allele frequencies, gene-environment interactions, and the broader genetic diversity present in African ancestry populations. From a precision medicine perspective, the study highlights a critical point: a genetic prediction tool that is accurate in one population may be inadequate or misleading in another unless it has been appropriately validated.
Limitations and Future Directions
The study also emphasizes important limitations. Multiple sclerosis has a relatively low absolute prevalence, which may restrict the value of polygenic risk scores as stand-alone screening tools. In addition, the authors note that risk prediction should ultimately integrate genetic, environmental, lifestyle, and clinical variables rather than relying exclusively on inherited genetic burden. Nevertheless, the study provides a valuable demonstration of why inclusive genomic research is essential. Future work should expand genome-wide association studies in underrepresented populations, refine ancestry-aware polygenic risk models, and evaluate whether integrated risk prediction frameworks can support earlier diagnosis, better trial stratification, and more equitable implementation of genomic medicine.
Disclaimer: This blog post is based on the provided research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.
References:
Torabi-Rahvar, M., Talebi, S., Salehi, N. et al. Exome Sequencing and Molecular Modeling Reveal Novel Loci in Familial Multiple Sclerosis: The Importance of BTNL3 and BTNL8 in Disease Pathogenesis. Mol Neurobiol 63, 227 (2026). https://doi.org/10.1007/s12035-025-05436-w
