Table 6 Estimated and real group mean and standard deviation of PRS for depressive episode cases and controls in UK biobank population.

P-thres	#SNPs	Trait	Our method (ReACt)		PRSice2
			t-test		t-test		Reg. w/o covatiate		Reg. w/top 5PCs
			Mean PRS	p-val	Mean PRS	p-val	\(r^2\)	p-val	\(r^2\)	p-val
0.1	4236	Cases	\(-\)0.0023	5.50\(\cdot 10^{-3}\)	\(-\)0.0023	\(3.97 \cdot 10^{-3}\)	\(2.48\cdot 10^{-5}\)	\(4.18 \cdot 10^{-3}\)	\(3.54 \cdot 10^{-5}\)	\(4.14 \cdot 10^{-3}\)
0.1	4236	Controls	\(-\)0.0023	5.50\(\cdot 10^{-3}\)	\(-\)0.0024	\(3.97 \cdot 10^{-3}\)	\(2.48\cdot 10^{-5}\)	\(4.18 \cdot 10^{-3}\)	\(3.54 \cdot 10^{-5}\)	\(4.14 \cdot 10^{-3}\)
0.01	594	Cases	\(-\)0.0036	1.47\(\cdot 10^{-3}\)	\(-\)0.0032	\(1.42 \cdot 10^{-3}\)	\(3.06 \cdot 10^{-5}\)	1.45 \(\cdot 10^{-3}\)	\(4.35 \cdot 10^{-5}\)	1.44 \(\cdot 10^{-3}\)
0.01	594	Controls	\(-\)0.0036	1.47\(\cdot 10^{-3}\)	\(-\)0.0032	\(1.42 \cdot 10^{-3}\)	\(3.06 \cdot 10^{-5}\)	1.45 \(\cdot 10^{-3}\)	\(4.35 \cdot 10^{-5}\)	1.44 \(\cdot 10^{-3}\)
0.001	82	Cases	0.0112	1.09\(\cdot 10^{-1}\)	0.0147	\(1.54 \cdot 10^{-1}\)	\(6.17 \cdot 10^{-6}\)	1.53 \(\cdot 10^{-1}\)	\(3.19 \cdot 10^{-5}\)	\( 1.51 \cdot 10^{-1}\)
0.001	82	Controls	0.0112	1.09\(\cdot 10^{-1}\)	0.0146	\(1.54 \cdot 10^{-1}\)	\(6.17 \cdot 10^{-6}\)	1.53 \(\cdot 10^{-1}\)	\(3.19 \cdot 10^{-5}\)	\( 1.51 \cdot 10^{-1}\)
\(10^{-4}\)	10	Cases	\(-\)0.0244	9.36\(\cdot 10^{-2}\)	\(-\)0.0247	\(1.16 \cdot 10^{-1}\)	\(7.57 \cdot 10^{-6}\)	1.13 \(\cdot 10^{-1}\)	\(2.96 \cdot 10^{-5}\)	\(1.12 \cdot 10^{-1}\)
\(10^{-4}\)	10	Controls	\(-\)0.0246	9.36\(\cdot 10^{-2}\)	\(-\)0.0249	\(1.16 \cdot 10^{-1}\)	\(7.57 \cdot 10^{-6}\)	1.13 \(\cdot 10^{-1}\)	\(2.96 \cdot 10^{-5}\)	\(1.12 \cdot 10^{-1}\)

We assessed the performance of our method using the summary statistics of an independent MDD GWAS as the base study, and the UK biobank samples, including 18,368 cases with depressive episode and 312,849 controls, as the target population. We generated summary statistics for the target populations and estimated group mean PRS and standard deviation of target PRS using ReACt. We computed the individual level PRS for the target study using PRSice2. For both methods, we computed PRS using independent SNPs from the base summary statistics with p-values below various thresholds (P-thres) and compared the performances under each threshold. For ReACt, mean PRS represents the estimated group mean PRS for cases and controls; p-val are the t-test p-values comparing PRS distribution in cases and in controls. For PRSice2, mean PRS represents real group mean PRS computed from individual level data and p-val are the t-test p-values comparing real PRS distribution in cases and in controls; reg. w/o covariate indicates regression results without covariates, which include the regression \(r^2\) value (reg. \(r^2\)) and the p-value for the PRS predictor (p-val); reg. w/top 5PCs indicates the regression results including the top five PCs as covariates, which also included the regression \(r^2\) value (reg. \(r^2\)) and the p-value for the PRS predictor (p-val).

Search