Table 2 Number of SNPs selected through cross-validation for the PRS and XGBoost Model.

From: Non-linear machine learning models incorporating SNPs and PRS improve polygenic prediction in diverse human populations

Phenotype

XGBoost alone

LASSO

XGBoost with PRS

PRS

Lassosum2

Sleep duration

35

35

36

1M

140,507

Diastolic blood pressure

297

297

298

1M

197,039

Systolic blood pressure

38

38

27

1M

249,700

Triglycerides

186

186

109

6799

8035

LDL cholesterol

727

727

429

5825

2056

HDL cholesterol

6746

6746

168

7694

10,651

Total cholesterol

1181

1181

84

6258

4340

Body mass index

44

44

51

1M

51,133

Height

4807

4807

559

1M

59,714

  1. Displayed are number of SNPs selected for each of the phenotypes in the four models in this study: PRS (best-performing PRS from PRSice, LDpred2, or lassosum2), XGBoost alone, LASSO (which has the same number of variants as in the XGBoost alone model, because the LASSO selected the variants used by XGBoost), and XGBoost with PRS.
Back to article page