Differential Performance of Genetic Risk Scores for Multiple Sclerosis Across European, Latino/Admixed, and African Ancestry: Evidence From
Rivier and colleagues (2025) examine a central challenge in translational human genetics: the portability of genetic risk scores (GRSs) across populations with different ancestral backgrounds. Multiple sclerosis (MS) is a complex autoimmune demyelinating disorder in which common genetic variants—each typically conferring modest individual effects—collectively contribute to measurable differences in disease susceptibility. Because most large genome-wide association studies (GWAS) for MS have been conducted in European-ancestry cohorts, the resulting variant weights and linkage disequilibrium (LD) structures embedded in published GRSs may not generalize to individuals of African ancestry. This is not a theoretical concern: if a GRS fails to stratify risk outside the discovery ancestry, it can exacerbate inequities in precision medicine by delivering clinically useful tools to some groups while providing noisy or misleading estimates to others.
Study Objective: Testing Stratification and Improving Portability
The authors pursue two tightly linked objectives. First, they evaluate whether an established MS GRS can stratify MS risk in African-ancestry and Latino/admixed-ancestry participants, alongside a European-ancestry comparator group. Second, recognizing that naïve score transfer often underperforms in African-ancestry datasets, they test whether an approach designed to improve cross-ancestry prediction—JointPRS—can enhance risk stratification specifically in the African-ancestry cohort. Framed this way, the study is not merely descriptive; it is a practical assessment of whether methodological advances in polygenic prediction can reduce ancestry-related performance gaps for a clinically relevant autoimmune disease.
Data Source and Design: A Cross-Sectional Analysis in “All of Us”
Using the U.S. “All of Us” Research Program dataset (2018–2022), the investigators conduct a cross-sectional case–control style analysis at scale. They construct a GRS from 232 MS-associated variants and apply it to three ancestry-defined groups: European, African, and Latino/admixed participants, with 32,428 individuals in each group. To evaluate stratification in a way that is interpretable and clinically adjacent, they partition each ancestry group into quintiles of the GRS distribution and compare MS prevalence and odds across these strata. MS case status is ascertained using electronic health record phenotyping based on ICD-9/10 and SNOMED codes—an approach that enables large sample sizes but also requires cautious interpretation because diagnostic coding is an imperfect proxy for gold-standard clinical adjudication.
Primary Results: Strong Stratification in European and Latino/Admixed Groups
The baseline MS prevalence differs across groups in this dataset—approximately 1.0% in European-ancestry participants, 0.56% in African-ancestry participants, and 0.46% in Latino/admixed participants—underscoring that population prevalence, ascertainment patterns, and healthcare access can influence observed rates in EHR-based studies. Critically, the 232-variant GRS demonstrates clear stratification in the European cohort, with substantially higher odds of MS in the top GRS quintile relative to the bottom (reported odds ratio about 2.30 with a significant trend). The Latino/admixed cohort shows a similarly strong or stronger gradient (reported odds ratio about 2.53 with a significant trend). These results support the clinical intuition that, when the score’s discovery ancestry aligns reasonably well with the target population’s genetic architecture and LD correlations, a compact variant set can still provide meaningful enrichment of risk at the distributional extremes.
The Portability Problem: Failure to Partition Risk in African-Ancestry Participants
In contrast, the same GRS does not significantly partition MS risk in the African-ancestry cohort (reported odds ratio about 1.30 with a non-significant trend). This outcome is highly consistent with broader polygenic risk literature: scores trained primarily in European cohorts often lose discriminative power in African-ancestry populations due to differences in allele frequencies, LD patterns (which alter how well tag variants capture causal variation), effect-size heterogeneity, and the underrepresentation of African-ancestry individuals in GWAS discovery pipelines. Importantly, the study’s negative result is not merely “lower performance”; it is a failure to achieve statistically credible risk stratification across quintiles, which is precisely the threshold at which a GRS becomes unreliable for downstream tasks such as cohort enrichment, trial stratification, or individualized counseling.
Methodological Remedy: JointPRS Improves Stratification in African Ancestry
To address this gap, the authors apply JointPRS, a method intended to improve polygenic prediction by incorporating ancestry-relevant information rather than relying solely on European-derived effect estimates. After this adjustment, stratification in the African-ancestry cohort improves substantially, with the highest quintile showing notably higher odds of MS compared with the lowest (reported odds ratio about 3.02 with a significant trend). While the confidence interval is wide—suggesting limited effective case counts and/or residual uncertainty—the directional shift is scientifically important: it indicates that performance deficits are not an immutable property of African-ancestry genetics, but rather a correctable artifact of data imbalance and model transfer. In practical terms, the result argues that integrating African-specific genetic information (either through multi-ancestry training, reweighting, or methods like JointPRS) can restore meaningful discriminative capacity for MS risk scoring.
Implications: Toward Equitable, Clinically Useful MS Risk Scores
The central message of Rivier et al. (2025) is operational and policy-relevant: an MS GRS built from predominantly European evidence can function well in European and some admixed populations yet fail in African-ancestry individuals, and this inequity can be mitigated when methods explicitly incorporate ancestry-matched data. For clinical translation, the study cautions against deploying a “one-score-fits-all” approach in healthcare systems that serve diverse populations, especially for applications like clinical trial enrichment or risk-based screening where miscalibration can produce systematic exclusion or misclassification. For research strategy, it strengthens the argument that expanding ancestry diversity in MS GWAS and biobank-scale phenotyping is not only ethically necessary but scientifically enabling—because it directly improves the validity and utility of genetic prediction tools. Ultimately, the work reframes GRS portability as a solvable engineering problem conditioned on inclusive datasets and appropriate multi-ancestry methodology, rather than as a limitation of any single population.
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:
Rivier, C. A., Xu, L., Clocchiatti-Tuozzo, S., Zhao, H., Ohno-Machado, L., Hafler, D. A., ... & Longbrake, E. E. (2025). Differential results of genetic risk scoring for multiple sclerosis in European and African American populations. Multiple Sclerosis Journal, 31(11), 1304-1313.
