Table 1.

Application of three summary statistics prediction methods in the Welcome Trust Case Control Consortium under three‐fold cross‐validation

Trait	LassoSum		LDPred		JAM
	AUC	r2	AUC	r2	AUC	r2
Bipolar disorder	0.67 (0.64, 0.69)	0.09	0.66 (0.63, 0.69)	0.08	0.70 (0.67, 0.73)	0.13
Coronary artery disease	0.59 (0.56, 0.62)	0.02	0.59 (0.56, 0.62)	0.02	0.65 (0.62, 0.67)	0.08
Crohn's disease	0.65 (0.62, 0.68)	0.07	0.69 (0.66, 0.72)	0.10	0.69 (0.66, 0.72)	0.12
Hypertension	0.61 (0.58, 0.64)	0.04	0.59 (0.56, 0.62)	0.03	0.58 (0.55, 0.61)	0.02
Rheumatoid arthritis	0.71 (0.68, 0.73)	0.12	0.72 (0.69, 0.75)	0.14	0.74 (0.71, 0.76)	0.16
Type 1 diabetes	0.83 (0.80, 0.85)	0.30	0.87 (0.85, 0.89)	0.39	0.86 (0.84, 0.88)	0.36
Type 2 diabetes	0.62 (0.60, 0.65)	0.04	0.60 (0.57, 0.63)	0.05	0.64 (0.61, 0.67)	0.08

Note: ROC AUCs and predictive r² are presented, with ROC AUC 95% confidence intervals calculated via 2,000 stratified bootstrap samples. For each method, performance is presented for the best performing sparsity.

Abbreviations: AUC, area under the curve; ROC, receiver operating characteristic.