Asian-derived polygenic risk scores boost risk stratification for breast cancer

Drawing from the largest available studies of Asian breast cancer patients, a recent Singapore study has developed polygenic risk scores (PRS) that can help improve the risk stratification of women of Asian ancestry.

“We have shown that genome-wide PRS derived from trans-ancestry method had significantly higher predictive accuracy for women of Asian ancestry than existing breast cancer PRSs,” the researchers said. “We also showed that European-based PRS can be improved for use in Asian populations by integrating population-specific weights and combined with Asian-specific PRS.

“Importantly, the differences in distribution of the same PRS across different ethnic groups (among Asians, and between Asian and Europeans) emphasize the need for ethnic-specific calibration before translating PRS into practice for diverse Asian populations,” they added.

Using genome-wide sets of single-nucleotide variants, the best overall PRS for women of East Asian descent had an odds ratio per unit standard deviation (OR_perSD) of 1.62 (95 percent confidence interval [CI], 1.52–1.68). The score’s corresponding area under the curve (AUC) was 0.636 (95 percent CI, 0.622–0.649). [Genet Med 2021;doi:10.1016/j.gim.2021.11.008]

Notably, this PRS was obtained by generating several different PRSs using the linear combination of European- and Asian-specific posterior weights, and repeated across a range of global shrinkage parameters (optimal shrinkage parameter, 10^–4).

Moreover, the best PRS was found to be markedly better than all PRSs currently in use. In particular, one PRS developed from European-descent women, which has been traditionally applied to Asian patients, only has a hazard ratio per SD (HR_perSD) of 1.46.

Subsequent analyses found that PRSs derived from women of East Asian descent also had considerable predictive value in a South Asian sample, though to a lesser degree. Moreover, while linear combinations of European and Asian PRSs boosted performance in East Asian women, such an effect was largely attenuated in South Asian patients.

The obtained PRSs were then tested in prospective cohorts. Scores based on Asian data alone had HR_perSD values of 1.40–1.45. Those derived from European data had a larger estimate, with a value of 1.50. However, the combination of both Asian and European data continued to produce PRSs with superior performance, with HR_perSD values ranging from 1.53–1.62.

“Personalized risk stratification for prevention and early detection of breast cancer has gained increasing interest; however, it is important to recognize the need to study women representing diverse ancestries to lessen health disparities,” the researchers said. “Our study provides essential information about the utility of PRSs for breast cancer risk prediction in women of Asian ancestry.”

However, since the obtained PRSs demonstrated weaker predictive value for women of South Asian descent, the researchers noted that there is a “need for larger studies of women of South Asian ancestry both to optimize the PRS and validate in prospective cohorts.”