Journal: Nature genetics
This study addresses the long‑standing problem that existing breast cancer polygenic risk scores (PRS), largely derived from European-ancestry cohorts, perform poorly in women of African ancestry—especially for aggressive subtypes like triple‑negative breast cancer (TNBC).
Using data from the African Ancestry Breast Cancer Genetics consortium (17,391 cases; 18,800 controls), the investigators:
- Developed ancestry-informed PRS for:
- Overall breast cancer
- ER‑positive disease
- ER‑negative disease
- TNBC
- Applied multiple PRS construction methods and explicitly integrated information across both:
- Different ancestries
- Different breast cancer subtypes
Model performance (AUC in the development dataset):
- Overall breast cancer: 0.612
- ER‑positive: 0.621
- ER‑negative: 0.611
- TNBC: 0.639
These AUCs were replicated in independent external validation cohorts:
- Overall breast cancer: 0.612
- ER‑positive: 0.640
- ER‑negative: 0.605
- TNBC: 0.652
Notably, they also constructed a simplified 162‑variant PRS for TNBC that retained similar predictive performance (AUC 0.626), suggesting feasibility for practical implementation without requiring very large genome-wide panels.
From a clinical perspective, the key takeaway is that these PRS models substantially improve risk prediction for breast cancer, including TNBC, in women of African ancestry compared with earlier tools. This work supports movement toward more equitable, ancestry-appropriate risk-stratified screening and prevention strategies in breast oncology.