Pharmacogenomic variants are near-universal in a large, diverse US cohort

  • Whole-genome sequencing and health records from nearly 100,000 All of Us participants were used to profile variants in 16 pharmacogenes and infer predicted drug-response phenotypes.
  • Every participant carried at least one variant, and almost all had a predicted phenotype linked to prescribing recommendations.
  • About a quarter had both an actionable phenotype and a past exposure to an affected medication.
  • Thousands of novel variants and many atypical allele frequencies emerged, especially in Americas and African ancestry groups.
💡 Points Based on the abstract

Near-universal carriage is expected, since most people carry some variant in a pharmacogene, so the useful question is whether genotype changes outcomes. Predicted phenotype is not measured exposure: without drug levels or response data, the actionable-and-exposed fraction overstates clinical impact. The next step is linking the novel and atypical-frequency alleles to functional assays and EHR outcomes, particularly in under-represented ancestries.