A Nature Genetics report from the Polygenic Risk Methods Development (PRIMED) Consortium and Children’s Hospital of Philadelphia (CHOP) calls for rethinking how researchers label human populations when developing polygenic risk scores (PRS), or the risk of developing certain diseases based on variations in a person’s genome. Researchers in the Center for Computational and Genomic Medicine (CCGM) found that population descriptors such as race, ethnicity and ancestry are woven into PRS development, validation and use, and early choices about these labels can affect PRS performance, equity and reliability in genomic medicine.
PRSs work well when training and test groups match genetically, but they often perform worse in people from underrepresented or admixed populations. Previous guidance by the National Academies of Sciences, Engineering and Medicine (NASEM) emphasized that population descriptors should be detailed and tailored in genomic studies. But concrete actionable guidance has been lacking – a gap this new study aimed to fill.
To provide practical recommendations and reporting standards, the authors reviewed how population descriptors such as self‑reported race/ethnicity, geography and genetic similarity are used throughout the PRS pipeline. Using examples from the PRIMED Consortium and public datasets, they showed how grouping decisions alter PRS results and address both technical and ethical or interpretative issues. As a result, the authors proposed a set of recommendations for researchers developing and applying PRSs.
“As polygenic risk scores move closer to clinical implementation, the risk of exacerbating health disparities is real,” said Quan Sun, PhD, a senior author and research assistant professor in the CCGM and Department of Biomedical and Health Informatics (DBHI) at CHOP. “By standardizing and justifying how population groups are defined and used in genetic studies, our recommended framework aims to improve trust and fairness and help to ensure more accurate risk predictions for diverse communities globally.”
Key Recommendations include:
- Researchers must explain why they select a specific descriptor and how it maps to genetic, social and environmental variation.
- From data collection through analysis and reporting, researchers should remain consistent and transparent to avoid undermining the validity of their outcomes.
- Labels based purely on social categories (i.e. race) may not correspond meaningfully to genetic diversity and using them interchangeably can be misleading.
- Whenever possible, it’s best to incorporate genetic ancestry information alongside or instead of social/population labels, especially for diverse populations, to improve accuracy and fairness of PRS performance across subgroups.
- Research studies should document and justify descriptor decisions to allow future researchers and clinicians to interpret risk scores appropriately.
The authors also emphasized that adopting these recommendations is not trivial. It requires genomic researchers and data analysts to rethink longstanding practices – from how they collect and code demographic data to how they interpret and report results. They encourage an ongoing community dialogue among geneticists, clinicians, ethicists and representatives of diverse populations to refine and adopt responsible practices.
The research was supported by the National Institutes of Health with grant funding for the study sites (CARDINAL (U01HG011717), CAPE (U01HG011715), D-PRISM (U01HG011723), EPIC-PRS (U01HG011720), FFAIRR-PRS (U01HG011719), PRIMED-Cancer (U01CA261339), PREVENT (U01HG011710)) and the PRIMED Coordinating Center (U01HG011697).
Smith, J.L et al. “Recommendations for responsible use of population descriptors in polygenic risk score development.” Nat Genet. Online November 24, 2025. DOI: 10.1038/s41588-025-02395-9.
Featured in this article
Specialties & Programs
Research
A Nature Genetics report from the Polygenic Risk Methods Development (PRIMED) Consortium and Children’s Hospital of Philadelphia (CHOP) calls for rethinking how researchers label human populations when developing polygenic risk scores (PRS), or the risk of developing certain diseases based on variations in a person’s genome. Researchers in the Center for Computational and Genomic Medicine (CCGM) found that population descriptors such as race, ethnicity and ancestry are woven into PRS development, validation and use, and early choices about these labels can affect PRS performance, equity and reliability in genomic medicine.
PRSs work well when training and test groups match genetically, but they often perform worse in people from underrepresented or admixed populations. Previous guidance by the National Academies of Sciences, Engineering and Medicine (NASEM) emphasized that population descriptors should be detailed and tailored in genomic studies. But concrete actionable guidance has been lacking – a gap this new study aimed to fill.
To provide practical recommendations and reporting standards, the authors reviewed how population descriptors such as self‑reported race/ethnicity, geography and genetic similarity are used throughout the PRS pipeline. Using examples from the PRIMED Consortium and public datasets, they showed how grouping decisions alter PRS results and address both technical and ethical or interpretative issues. As a result, the authors proposed a set of recommendations for researchers developing and applying PRSs.
“As polygenic risk scores move closer to clinical implementation, the risk of exacerbating health disparities is real,” said Quan Sun, PhD, a senior author and research assistant professor in the CCGM and Department of Biomedical and Health Informatics (DBHI) at CHOP. “By standardizing and justifying how population groups are defined and used in genetic studies, our recommended framework aims to improve trust and fairness and help to ensure more accurate risk predictions for diverse communities globally.”
Key Recommendations include:
- Researchers must explain why they select a specific descriptor and how it maps to genetic, social and environmental variation.
- From data collection through analysis and reporting, researchers should remain consistent and transparent to avoid undermining the validity of their outcomes.
- Labels based purely on social categories (i.e. race) may not correspond meaningfully to genetic diversity and using them interchangeably can be misleading.
- Whenever possible, it’s best to incorporate genetic ancestry information alongside or instead of social/population labels, especially for diverse populations, to improve accuracy and fairness of PRS performance across subgroups.
- Research studies should document and justify descriptor decisions to allow future researchers and clinicians to interpret risk scores appropriately.
The authors also emphasized that adopting these recommendations is not trivial. It requires genomic researchers and data analysts to rethink longstanding practices – from how they collect and code demographic data to how they interpret and report results. They encourage an ongoing community dialogue among geneticists, clinicians, ethicists and representatives of diverse populations to refine and adopt responsible practices.
The research was supported by the National Institutes of Health with grant funding for the study sites (CARDINAL (U01HG011717), CAPE (U01HG011715), D-PRISM (U01HG011723), EPIC-PRS (U01HG011720), FFAIRR-PRS (U01HG011719), PRIMED-Cancer (U01CA261339), PREVENT (U01HG011710)) and the PRIMED Coordinating Center (U01HG011697).
Smith, J.L et al. “Recommendations for responsible use of population descriptors in polygenic risk score development.” Nat Genet. Online November 24, 2025. DOI: 10.1038/s41588-025-02395-9.
Contact us
Karen Pasternack
Department of Biomedical and Health Informatics (DBHi)