A new polygenic risk score (PRS) improves disease prediction for people with African or Asian ancestry, offering the chance to redress healthcare inequalities and widen the scope of personalized medicine.
The BridgePRS, published in Nature Genetics, predicts disease risk from DNA better than other similar risk scores, which are mostly created using data from European populations.
The powerful, trans-ancestry Bayesian method leverages shared genetic effects across ancestries to improve PRS portability outside European descent.
“We hope that our method opens up scientific investigation of disease risk in diverse populations worldwide,” said researcher Paul O’Reilly, PhD, associate professor of genetics and genomic sciences at New York’s Icahn school of medicine.
“Disease prevalence and the importance of different biological pathways can vary globally. Understanding these differences is crucial for advancing disease prediction and treatment.”
PRSs have typically been derived using data from genome-wide association studies (GWASs) that involve people of European ancestry.
This results in substantially lower predictive power for other populations, in particular those of African ancestry.
Another risk score, PRS-CSx, has been previously developed to tackle the PRS portability problem, making cross-population inference on the inclusion of each single-nucleotide polymorphism across the genome.
Unlike this fine-mapping approach, BridgePRS integrates trans-ancestry GWAS summary statistics to retain all variants within loci to best tag causal variants shared across ancestries.
The focus is on correctly estimating causal effect sizes, which the authors note is key when the goal is prediction rather than on estimating variant location. The approach is thereby less reliant on the inclusion and identification of causal variants.
The researchers found BridgePRS performed favourably compared with the leading alternative of PRS-CSx as well as two other single-ancestry PRS methods adapted to use GWAS data from multiple ancestries.
The team used simulations and GWAS data from UK and Japanese Biobanks to construct PRSs for 19 traits in individuals African, South Asian, and East Asian ancestry.
The resultant scores were then validated in unseen UK Biobank samples and in the entirely independent New York-based Mount Sinai BioMe biobank, producing results consistent with the simulations.
O’Reilly noted that the field of optimizing disease prediction through PRS is highly competitive, fostering rapid advancements.
He added: “Our BridgePRS method is particularly promising for predicting disease in individuals of African ancestry, a group with rich genetic diversity that can offer novel insights into human diseases.”