Polygenic risk scores hold enormous promise for identifying individuals at elevated risk of complex diseases before symptoms appear — but that promise has been largely reserved for people of European ancestry. A new computational framework may meaningfully close that gap, with direct implications for equitable preventive medicine and early disease detection across diverse populations.
Published in Nature Genetics, MIXPRS is a statistical framework that combines multiple polygenic risk score (PRS) methods and multi-population data using only genome-wide association study (GWAS) summary statistics — bypassing the requirement for individual-level genotype data that has historically restricted these analyses. The method employs single nucleotide polymorphism pruning to reduce linkage disequilibrium mismatch, a persistent technical challenge when applying models across ancestrally distinct groups, and uses non-negative least squares regression to optimally weight the combined scores. Evaluated across simulations and real-data analyses spanning up to 26 traits, MIXPRS consistently outperformed individual existing methods. An extended version, MIXPRS+, which incorporates functional genomic annotations and established clinical PRS models, delivered further predictive gains specifically in non-European populations.
The equity dimension here is the crux of the contribution. Current PRS models, largely trained on UK Biobank and similar predominantly white European cohorts, perform substantially worse in African, South Asian, and East Asian populations — a disparity that, if translated into clinical tools, could worsen existing health inequities. MIXPRS addresses this by aggregating across methods rather than relying on any single approach, exploiting complementary strengths without requiring access to protected individual data. That summary-statistics-only design also significantly lowers the computational and data-access barriers for research institutions in lower-resource settings. The limitation worth tracking is that this is primarily a methodological advance evaluated in simulation and retrospective cohort data; prospective clinical validation across diverse real-world populations will be essential before MIXPRS informs patient-facing risk tools. Still, as a framework innovation rather than an incremental parameter tweak, this represents a meaningful step toward genuinely population-agnostic genomic risk assessment.