Polygenic Risk Scores for Prediction of Breast Cancer Risk in Women of African Ancestry: a Cross-Ancestry Approach

Abstract

Polygenic risk scores (PRSs) are useful to predict breast cancer risk, but the prediction accuracy of existing PRSs in women of African ancestry (AA) remain relatively low. We aim to develop optimal PRSs for prediction of overall and estrogen receptor (ER) subtype-specific breast cancer risk in women of African ancestry. The AA dataset comprised 9,235 cases and 10,184 controls from four genome-wide association study (GWAS) consortia and a GWAS study in Ghana. We randomly divided samples into training and validation sets. Genetic variants were selected by forward stepwise logistic regression or lasso penalized regression in the training set and the corresponding PRSs were evaluated in the validation set. To improve accuracy, we also developed joint PRSs that combined 1) the best PRSs built in the AA training dataset, 2) a previously-developed 313-variant PRS in women of European ancestry, and 3) PRSs using variants that were discovered in previous GWASs in women of European and African ancestry and were nominally significant the training set. For overall breast cancer, the odd ratio (OR) per standard deviation of the joint PRS in the validation set was 1.39 (95%CI: 1.31-1.46) with area under receiver operating characteristic curve (AUC) of 0.590. Compared to women with average risk (40th-60th PRS percentile), women in the top decile of the PRS had a 2.03-fold increased risk (95%CI: 1.68-2.44). For PRSs of ER-positive and ER-negative breast cancer, the AUCs were 0.609 and 0.597, respectively. The proposed PRS can improve prediction of breast cancer risk in women of African ancestry. Author SummaryPolygenic risk scores have been developed to predict breast cancer risk in non-Hispanic white American women, where polygenic risk score combines the effects of multiple single nucleotide polymorphisms. However, reliable polygenic risk scores do not exist for women of African ancestry, including African Americans, African Barbadians, and indigenous Africans. Due to distinct allele frequencies and linkage disequilibrium structures across populations, polygenic risk scores developed in European ancestry populations have an attenuated predictive value when applied to African ancestry populations. In this study, we constructed polygenic risk scores for African ancestry women by using African ancestry datasets. Since the sample sizes of existing African ancestry datasets are much smaller than those from European-ancestry studies, these polygenic risk scores using only African ancestry datasets may have limited accuracy. To increase the prediction accuracy, we constructed joint polygenic risk scores by combining polygenic risk scores trained in African ancestry datasets with polygenic risk scores that were previously developed using a large European ancestry dataset. Results showed that the joint polygenic risk scores could improve prediction of breast cancer risk in women of African ancestry.