Skip to main content

Table 1 Best-performing polygenic risk scores and testing for associations with measured lipid traits

From: Development of genome-wide polygenic risk scores for lipid traits and clinical applications for dyslipidemia, subclinical atherosclerosis, and diabetes cardiovascular complications among East Asians

Polygenic risk score Lipid trait No. of SNPs Derivation strategy and tuning parameter Cohort n Correlation with lipid trait Linear regression Proportion of variance
Pearson r Spearman r β SE P PQ explained by polygenic risk score
PRSTC TC 229 Pruning and thresholding Children 909 0.299 0.295 0.050 0.005 2.2 × 10−21 9.49%
(r2 = 0.2 and p = 10− 5) Adolescents 1973 0.327 0.322 0.056 0.004 8.9 × 10− 53 10.93%
Healthy adults 441 0.334 0.320 0.057 0.008 2.2 × 10−13 10.07%
Adult women 948 0.243 0.259 0.042 0.005 3.2 × 10−16 6.26%
Combined analysis in validation dataset 4271 0.304 0.302 0.052 0.002 7.5 × 10−103 0.1197
Adults 426 0.251 0.263 0.043 0.008 2.9 × 10−7 5.06%
T2D patients in HKDR study 4917 0.186 0.199 0.042 0.003 3.7 × 10−45 3.93%
T2D patients in HKDB phase 1 study 1941 0.187 0.186 0.036 0.004 2.3 × 10−16 3.32%
T2D patients in HKDB phase 2 study 865 0.173 0.176 0.034 0.006 1.8 × 10−7 2.98%
Combined analysis in T2D patients 7723 0.185 0.193 0.039 0.002 1.5 × 10−66 0.3721
PRSTG TG 142 Pruning and thresholding Children 909 0.243 0.245 0.106 0.012 3.7 × 10−17 7.19%
  (r2 = 0.2 and p = 5 × 10−8) Adolescents 1973 0.244 0.227 0.097 0.008 3.4 × 10−30 5.99%
  Healthy adults 441 0.257 0.245 0.139 0.022 9.5 × 10−10 5.59%
  Adult women 948 0.294 0.256 0.146 0.015 9.9 × 10−23 8.59%
  Combined analysis in validation dataset 4271 0.256 0.239 0.111 0.006 1.3 × 10−75 0.0150
  Adults 426 0.251 0.230 0.133 0.023 1.7 × 10−8 5.14%
  T2D patients in HKDR study 4917 0.200 0.251 0.165 0.008 1.3 × 10−87 7.31%
  T2D patients in HKDB phase 1 study 1941 0.209 0.191 0.133 0.013 1.6 × 10−25 5.00%
  T2D patients in HKDB phase 2 study 865 0.231 0.197 0.140 0.019 2.8 × 10−13 5.58%
  Combined analysis in T2D patients 7723 0.206 0.230 0.154 0.006 2.3 × 10−126 0.0721
PRSHDL HDL-C 549 Pruning and thresholding Children 909 0.263 0.250 0.052 0.007 4.0 × 10−15 6.37%
  (r2 = 0.2 and p = 10−4) Adolescents 1973 0.281 0.258 0.054 0.004 4.0 × 10− 38 7.27%
  Healthy adults 441 0.347 0.313 0.084 0.010 4.0 × 10−15 9.44%
  Adult women 948 0.271 0.257 0.058 0.007 1.7 × 10−17 6.35%
  Combined analysis in validation dataset 4271 0.282 0.262 0.057 0.003 9.3 × 10−83 0.0458
  Adults 426 0.272 0.269 0.066 0.012 2.3 × 10−8 5.58%
  T2D patients in HKDR study 4917 0.222 0.226 0.064 0.004 7.4 × 10−64 5.24%
  T2D patients in HKDB phase 1 study 1941 0.239 0.237 0.073 0.006 8.6 × 10−32 6.06%
  T2D patients in HKDB phase 2 study 865 0.267 0.252 0.077 0.008 6.1 × 10−21 8.27%
  Combined analysis in T2D patients 7723 0.231 0.232 0.068 0.003 3.2 × 10− 116 0.1906
PRSLDL LDL-C 84 Pruning and thresholding Children 909 0.239 0.243 0.068 0.008 1.1 × 10−15 6.81%
(r2 = 0.2 and p = 5 × 10−8) Adolescents 1973 0.309 0.294 0.093 0.006 2.6 × 10−53 10.88%
Healthy adults 441 0.274 0.274 0.077 0.012 9.8 × 10−10 7.22%
Adult women 948 0.262 0.275 0.071 0.008 1.0 × 10−16 6.33%
Combined analysis in validation dataset 4271 0.281 0.277 0.081 0.004 2.4 × 10−93 0.0439
Adults 426 0.255 0.288 0.072 0.012 5.8 × 10−9 6.72%
T2D patients in HKDR study 4917 0.178 0.175 0.059 0.004 1.8 × 10−38 3.51%
T2D patients in HKDB phase 1 study 1941 0.190 0.175 0.054 0.006 3.0 × 10−17 3.60%
T2D patients in HKDB phase 2 study 865 0.195 0.171 0.058 0.010 9.2 × 10−9 3.74%
     Combined analysis in T2D patients 7723 0.183 0.175 0.057 0.003 3.2 × 10−62 0.8303
  1. The optimal polygenic risk scores for four lipid traits were chosen based on the validation datasets which consisted of 909 children, 1973 adolescents, 441 health adults, and 948 adult women. These scores were further tested in 426 adults from the general population and 7723 individuals from three cohorts of T2D patients. Associations between polygenic risk scores and lipid traits were assessed by Pearson and Spearman correlations and a linear regression model. Within individual cohorts, p values were obtained from linear regression with adjustment of principal components, sex, age, and body mass index. Lipid traits were natural log-transformed for the linear regression analysis. Results from cohorts of children, adolescents, healthy adults, and adult women were meta-analyzed using a fixed effects model in the validation stage (n = 4271), while results from three cohorts of T2D patients were combined in the testing stage (n = 7723). PQ refers to the p value of Cochran’s Q-statistics in heterogeneity test. The proportion of variance for a lipid trait explained by the corresponding optimal PRS was computed as the R2 obtained from a full model including both PRS and covariates (PCs, sex, age, and BMI) minus the R2 obtained from a model including covariates alone