NEW YORK – New research suggests that while polygenic risk scores (PRS) show promise for predicting several common conditions in individuals of European ancestry, they appear to have relatively poor transferability to other ancestry groups, particularly those with African ancestry.
"Our findings indicate that in the populations investigated, the current genome-wide polygenic scores for common diseases have potential for clinical utility within different healthcare settings for individuals of European ancestry, but that the utility in individuals of African ancestry is currently much lower," senior and corresponding author Samuli Ripatti, a researcher affiliated with the University of Helsinki's Institute for Molecular Medicine and the Broad Institute, and his colleagues wrote in Cell Genomics on Wednesday.
In its publication, the team outlined findings from PRS analyses done with genome-wide association study summary statistics from half a dozen biobank cohorts: the BioBank Japan, the Estonian Biobank, FinnGen, the Trondelag Health Study, the Mass General Brigham Biobank, and the UK Biobank.
Using data for nearly 807,800 European ancestry individuals and more than 195,500 participants with non-European ancestry from these cohorts, the researchers assessed the performance and cross-ancestry transferability of established risk scores for four common conditions: coronary artery disease, type 2 diabetes, breast cancer, and prostate cancer.
Although the PRSs showed some transferability among individuals of European descent, including structured populations within Finland, the team saw much more modest transferability to individuals with South Asian or East Asian ancestry, and particularly poor PRS transferability for individuals with African ancestry.
"What our results now show is that PRSs in these diseases identify well high-risk individuals with European ancestry in Finland, Estonia, UK, Norway, or Massachusetts, but also high-risk individuals with South or East Asian ancestry," Ripatti explained in an email. "However, we need to work hard to advance risk prediction … for African ancestry individuals to ensure equitable use of genetic prediction for everybody."
PRS transferability was somewhat better for risk scores developed with many disease-related variants across the genome compared to PRSs based on more limited variant sets, the investigators noted. Even so, they explained, the results underscore the importance of expanding the ancestral and population diversity of genetic studies that are being used to develop genetic risk scores.
"Without prioritizing diversity in PRS evaluations and translation efforts," the authors wrote, "widely adopting PRSs to clinical care may exacerbate health disparities, and efforts to overcome the lack of diversity have great potential to improve health outcomes across ancestries."
In particular, Ripatti noted that there is a need for "global networks of genetic studies" — such as artificial intelligence-aided human genetics work by the international INTERVENE consortium that is coordinated by the University of Helsinki — to ensure comprehensive representation for diverse and admixed individuals in genetic studies, including those used to develop and test PRS scores and algorithms.