NEW YORK — Polygenic risk scores for cancer developed using data from European-ancestry populations could be adjusted to identify individuals from other ancestral backgrounds who are at increased risk of disease, according to a new analysis.
Genome-wide association studies, which are used to generate polygenic risk scores, have largely relied on populations of European ancestry. One recent analysis found that 67 percent of polygenic risk scores were developed using cohorts of only European ancestry, and other studies have revealed that polygenic risk scores are not always portable across ancestries, likely due to differences in causal variants, linkage disequilibrium, allele frequencies, and effect sizes.
But in some cases, cancer polygenic risk scores developed in European ancestry populations may be able to be adapted to gauge risk in other populations, a new study appearing in PLOS Genetics has found.
Researchers from the University of Michigan School of Public Health developed polygenic risk scores for breast and prostate cancer using data from a European ancestry cohort and applied the scores to African, East Asian, European, and South Asian ancestry groups. While the risk scores were not directly transferable, the researchers found that if they considered the top 10 percent of risk scores within each population group, they could identify individuals at increased disease risk, suggesting this approach could be used as a stopgap measure while more diverse GWAS are conducted.
"Though there are many efforts underway to increase diversity in genetic studies, the prevailing lack of diversity will continue for the foreseeable future," first author Lars Fritsche, an associate research scientist at UMich, wrote in an email. "But we need solutions that work with the data we have right now, even if they are not ideal."
He and his colleagues generated polygenic risk scores for breast and prostate cancer using the European subset of the UK Biobank. They then applied the scores to the four main ancestry groups represented in the UKB: African, East Asian, European, and South Asian. Overall, breast cancer polygenic risk scores were higher on average among non-European populations, while prostate cancer polygenic risk scores were higher among African but lower among East and South Asian populations. This indicates the scores cannot be directly transferred, the researchers noted, as a high risk score among Europeans for breast cancer, for instance, would then fall at the low end of the distribution for other populations.
But the researchers noticed that if the scores were scaled within each ancestry group, they could identify people with higher risk of cancer.
"We didn't expect to see this relatively consistent enrichment of cases in the top tails of the risk scores across the analyzed ancestry groups," Fritsche said. "On the contrary, we were originally interested in documenting the limited transferability of polygenic risk scores for cancers across ancestries, but then spotted the obvious differences in the risk score distributions between cases and controls within each ancestry group."
While this scaling approach could be useful for stratifying disease risk, it might not be applicable to all cancer types or diseases and is only a temporary solution. According to Fritsche, the approach might be best suited for traits that are common across ancestry groups, for which common genetic variants can explain a large portion of heritability, and for which there is ample data to generate a polygenic risk score as well as case-control data from various ancestry groups to place those scores into context.
Fritsche noted that for many complex common diseases, genetic factors influence risk alongside environmental and demographic factors, and that they need to be combined to predict disease outcomes. He and his colleagues are now using genetic biobanking efforts in combination with electronic health record, census, and other data to develop better prediction models and possibly uncover modifiable health disparities.