All of Us diversity boosts polygenic risk scores, but not for every group
Researchers built polygenic risk scores for 32 traits and diseases using 245,388 whole-genome sequences from the All of Us program combined with UK Biobank data. Broader ancestral diversity in All of Us improved prediction accuracy overall, particularly for under-represented populations. But pooling the two datasets was not always the better choice: for less polygenic traits, training on All of Us data alone worked best for African ancestry participants, a pattern the authors attribute to ancestry-enriched genetic effects. Accuracy dropped in a straight line as a participant's ancestry diverged further from the discovery GWAS. Using multiancestry training data softened that decline.
Risk scores built from more ancestrally varied cohorts give clinicians and genetically diverse patients predictions they can trust more, though the right mix of data still depends on the trait and population involved.
Written by the Genomes desk from the primary source cited and linked above and checked against it. Research use only; not medical advice. Corrections: [email protected].