All of Us diversity and scale yield context-dependent improvements in polygenic prediction.

Multiancestry polygenic risk scores (PRSs) trained on diverse datasets improve prediction accuracy, especially in under-represented populations, with gains varying by trait and ancestry.

• Why it matters: Current genetic datasets lack sufficient diversity, limiting the equitable application of PRSs across different populations. Enhancing diversity is crucial for improving prediction accuracy and reducing health disparities.
• What they did: Using 245,388 whole-genome sequences from the All of Us program combined with UK Biobank data, researchers developed PRSs for 32 traits and evaluated how ancestry, methodology, and genetic architecture affected performance across diverse groups.
• The result: Increased diversity in training data improved PRS accuracy for several traits, particularly in African ancestry groups, and multiancestry training mitigated accuracy decay caused by ancestry divergence, highlighting the importance of representative biobanks.