Table 2.
Performance of optimized polygenic scores for cardiometabolic traits in Arabs
| Traits | European-derived polygenic scores | Arab optimized polygenic scores | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Categorical traits | PGS Catalog ID | Derivation Method | No. of variants in PRS | OR per SD (95% CI) | AUC (95% CI) | Derivation Method | No. of variants in PRS | OR per SD (95% CI) | AUC (95% CI) |
| Coronary artery disease | PGS000013 | LDpred | 5,706,928 | 1.41 (1.31–1.50) | 0.7909 (0.7726-0.8091) | lassosum2 | 10,440 | 1.51 (1.42–1.61) | 0.7950 (0.7768–0.8132) |
| Type 2 diabetes | PGS000014 | LDpred | 5,786,938 | 1.41 (1.32–1.49) | 0.7054 (0.6854-0.7255) | PRS-CS | 1,068,166 | 1.83 (1.74–1.92) | 0.7384 (0.7194–0.7574) |
| Cardiomyopathy | PGS002051 | LDpred2 | 621,802 | 1.01 (0.88–1.16) | 0.6277 (0.5890-0.6665) | LDpred2 | 1,010,014 | 1.34 (1.13–1.64) | 0.6453 (0.6086–0.6819) |
| Continuous traits | PGS Catalog ID | Derivation Method | No. of variants in PRS | Effect size per SD (SE) | Adjusted R-sq | Derivation Method | No. of variants in PRS | Effect size per SD (SE) | Adjusted R-sq |
|---|---|---|---|---|---|---|---|---|---|
| LDL Cholesterol [mg/dL] | PGS000892 | PRS-CS | 1,068,974 | 10.06 (1.10) | 0.0405 | PRS-CSx | 1,069,677 |
9.40 (1.10) |
0.0358 |
| HDL Cholesterol [mg/dL] | PGS002781 | PRS-CS | 1,112,500 |
3.49 (0.27) |
0.1351 | PRS-CSx | 1,069,677 |
3.67 (0.27) |
0.1424 |
|
Triglyceride [mg/dL] |
PGS002784 | Pruning and Thresholding | 12,709 | 25.19 (2.34) | 0.0682 | lassosum2 | 54,623 | 29.32 (2.31) | 0.0857 |
| Systolic blood pressure [mmHg] | PGS002238 | PRS-CS | 1,069,203 |
1.08 (0.45) |
0.0961 | PRS-CS | 1,056,790 |
3.10 (0.44) |
0.1108 |
| Diastolic blood pressure [mmHg] | PGS002239 | PRS-CS | 1,069,056 |
0.71 (0.26) |
0.0249 | lassosum2 | 25,857 |
1.80 (0.26) |
0.0397 |
| Body mass index [kg/m2] | PGS000027 | LDpred | 2,015,065 |
0.97 (0.11) |
0.0790 | PRS-CSx | 1,067,771 |
1.18 (0.11) |
0.0919 |
| Height [m] | PGS002804 | SBayesC | 1,054,056 | 0.027 (0.0013) | 0.5299 | PRS-CSx | 1,067,771 | 0.026 (0.0013) | 0.5202 |
Performance of European-derived polygenic scores and Arab-optimized polygenic scores for each trait are shown in the validation dataset of Arabs (N = 2699). European-derived polygenic scores were calculated using scoring files from The Polygenic Score (PGS) Catalog with datasets consisting primarily of European ancestry and not derived from the UK Biobank. Further information about the scoring files, including their GWAS sources, is included in Supplementary Table 1. Arab-optimized polygenic scores were derived in this study. The number of missing values varies across the disease or trait. Additional details are shown in Supplementary Table 5. OR per SD and AUC were determined using a logistic regression model adjusted for age, sex, array version, and the first 10 principal components of ancestry. Effect size per SD and adjusted R2 values for continuous traits were determined using a linear regression model with similar covariates. (OR per SD odd ratio per standard deviation, AUC area under the receiver operating characteristic curve, CI confidence interval, SE standard error).