Table 2.
Performance of ancestry-specific and joint prediction PRS models in the validation set
Weight ( ) for each predictora
|
OR (95% CI)a | P | AUC (95% CI)a | |
|---|---|---|---|---|
| Overall breast cancer | ||||
| PRSAFR (genome-wide threshold P < 0.05) | 1.13 (1.07–1.19) | 7.8 × 10−06 | 0.535 (0.519–0.551) | |
| PRS from European ancestry (PRSEUR)b | 1.30 (1.23–1.37) | 2.8 × 10−21 | 0.571 (0.557–0.585) | |
 PRSAFR +  PRSEUR
|
=0.31,  =0.69 |
1.34 (1.27–1.41) | 3.4 × 10−25 | 0.577 (0.561–0.593) |
| PRShybridc | 1.34 (1.27–1.42) | 3.0 × 10−26 | 0.581 (0.566–0.597) | |
| ER-positive | ||||
| PRSAFR.ERp (genome-wide threshold P < 0.05) | 1.22 (1.13–1.31) | 2.7 × 10−7 | 0.546 (0.527–0.566) | |
| PRS from European ancestry (PRSEUR.ERp)b | 1.43 (1.33–1.53) | 6.1 × 10−24 | 0.597 (0.577–0.617) | |
 PRSAFR.ERp +  PRSEUR.ERp
|
=0.35,  =0.65 |
1.49 (1.39–1.60) | 1.1 × 10−28 | 0.608 (0.588–0.627) |
| ER-negative | ||||
| PRSAFR.ERn (genome-wide threshold P < 0.05) | 1.20 (1.11–1.31) | 1.1 × 10−5 | 0.548 (0.525–0.572) | |
| PRS from European ancestry (PRSEUR.ERn)b | 1.23 (1.13–1.34) | 8.7 × 10−7 | 0.557 (0.534–0.581) | |
 PRSAFR.ERn +  PRSEUR.ERn
|
= 0.47,  = 0.53 |
1.31 (1.21–1.43) | 1.1 × 10−10 | 0.576 (0.553–0.598) |
aWeight (
) in the joint PRSs was estimated in validation set with a logistic regression model, including two-component PRSs (PRSAFR and PRSEUR) as predictors, and adjusting for age, consortium/study and 10 PCs; OR per 1 SD. AUC of PRSs was calculated under the covariate-adjusted ROC model adjusting for age, consortium/study and 10 PCs of genotype data.
bFor the 313 SNPs reported by Mavaddat et al. (5) for PRS in women of European ancestry, 307 SNPs appeared in our data of African ancestry.
cPRShybrid for overall cancer risk is a linear combination of the two joint PRSs for ER-positive and ER-negative breast cancer, with a weight of 0.62 for ER-positive and 0.38 for ER-negative cancer.