Table 3.
Comparison of PRS Methods on Breast Cancer PRS Performance in MGI and UKB
| Method Tuning Parameter | # SNPs | Pseudo-R2 | Brier Score | AAUC (95% CI) | Odds Ratio Continuous PRS (95% CI) | Odds Ratio Top 1% (95% CI) |
|---|---|---|---|---|---|---|
| MGI Cohort | ||||||
| Lassosum: s = 0.5, λ = 0.0043 | 118,388 | 0.0592 | 0.134 | 0.641 (0.625,0.656) | 1.70 (1.59,1.81) | 2.48 (1.63,3.77) |
| P&T: p ≤ 4 × 10−4 | 3,038 | 0.0532 | 0.135 | 0.635 (0.618,0.651) | 1.64 (1.54,1.75) | 2.84 (1.88,4.28) |
| Fixed threshold: p ≤ 5 × 10−5 | 1,307 | 0.0521 | 0.135 | 0.634 (0.618,0.65) | 1.63 (1.53,1.74) | 2.75 (1.81,4.17) |
| Fixed threshold: p ≤ 5 × 10−6 | 712 | 0.0484 | 0.135 | 0.629 (0.612,0.645) | 1.60 (1.50,1.70) | 3.06 (2.04,4.59) |
| Fixed threshold: p ≤ 5 × 10−7 | 464 | 0.0476 | 0.135 | 0.627 (0.611,0.644) | 1.60 (1.50,1.70) | 3.38 (2.28,5.02) |
| Fixed threshold: p ≤ 5 × 10−8 | 334 | 0.0462 | 0.136 | 0.625 (0.609,0.641) | 1.58 (1.49,1.69) | 3.32 (2.24,4.93) |
| Fixed threshold: p ≤ 5 × 10−9 | 264 | 0.0455 | 0.136 | 0.624 (0.608,0.64) | 1.58 (1.48,1.68) | 2.56 (1.68,3.90) |
| UKB Cohort | ||||||
| Lassosum: s = 0.9, λ = 0.0043 | 286,144 | 0.0487 | 0.0807 | 0.643 (0.637,0.65) | 1.70 (1.65,1.75) | 4.02 (3.46,4.67) |
| P&T: p ≤ 1 × 10−4 | 1,682 | 0.0401 | 0.0811 | 0.63 (0.623,0.637) | 1.61 (1.57,1.66) | 3.69 (3.16,4.31) |
| Fixed threshold: p ≤ 5 × 10−6 | 712 | 0.0402 | 0.0811 | 0.628 (0.62,0.635) | 1.61 (1.57,1.65) | 3.32 (2.83,3.90) |
| Fixed threshold: p ≤ 5 × 10−5 | 1,307 | 0.0392 | 0.0812 | 0.627 (0.62,0.634) | 1.60 (1.56,1.64) | 3.49 (2.98,4.08) |
| Fixed threshold: p ≤ 5 × 10−7 | 464 | 0.0384 | 0.0812 | 0.626 (0.618,0.633) | 1.59 (1.55,1.63) | 3.81 (3.27,4.44) |
| Fixed threshold: p ≤ 5 × 10−8 | 334 | 0.0361 | 0.0813 | 0.622 (0.615,0.63) | 1.57 (1.53,1.61) | 3.69 (3.16,4.31) |
| Fixed threshold: p ≤ 5 × 10−9 | 264 | 0.0347 | 0.0813 | 0.62 (0.612,0.627) | 1.55 (1.51,1.59) | 3.28 (2.79,3.86) |
PRSs are based on the BCAC Consortium GWAS on overall breast cancer.11 Italic values indicate best performing PRSs according to the corresponding metrics for MGI or UKB.