BGWAS: Bayesian variable selection in linear mixed models with nonlocal priors for genome-wide association studies

Table 2 Results for GWAS data simulated from LMM with \(n = 2772, p = 800{,}000, \kappa = 0.1\), and \(\sigma ^2 = 0.1\)

Nominal FDR	Method	\(\beta ^{(1)} = 0.1\)				\(\beta ^{(1)} = 0.4\)				\(\beta ^{(1)} = 1.6\)
Nominal FDR	Method	TP	FP	F1	Time (s)	TP	FP	F1	Time (s)	TP	FP	F1	Time (s)
0.05	SMA-Approx.	14.1	169.5	0.14	139	18.7	223.9	0.14	95	9.0	260.2	0.06	106
	SMA-Exact	14.1	169.8	0.14	1137	18.7	223.9	0.14	1093	9.0	260.2	0.06	1968
	NP, \(\tau = 0.348\)	13.0	1.1	0.76	259	16.6	2.2	0.85	281	8.4	0.7	0.58	208
	NP, \(\tau = 0.022\)	13.8	1.6	0.78	279	16.9	2.4	0.86	321	10.1	1.7	0.64	242
	NP, \(\tau\) estimated	14.0	1.6	0.79	283	16.8	2.7	0.85	339	10.9	1.5	0.67	254
0.1	SMA-Approx.	14.2	176.6	0.14	139	18.8	234.5	0.14	95	9.1	274.4	0.06	106
	SMA-Exact	14.2	177.0	0.14	1137	18.8	234.5	0.14	1093	9.1	274.4	0.06	1968
	NP, \(\tau = 0.348\)	13.1	1.4	0.76	265	16.9	2.1	0.87	293	8.4	0.9	0.58	210
	NP, \(\tau = 0.022\)	14.0	1.6	0.79	289	17.1	2.4	0.87	313	11.2	1.4	0.69	252
	NP, \(\tau\) estimated	14.2	1.8	0.79	291	16.9	2.8	0.85	340	11.7	1.4	0.71	267

In this table, there are 20 causal SNPs. The regression coefficients of the 20 causal SNPs are \(\varvec{\beta }= (\beta ^{(1)}, 0.4, 0.4, 0.4, \beta ^{(1)}, 0.4, 0.4, 0.4, \beta ^{(1)}, 0.4, 0.4, 0.4, \beta ^{(1)}, 0.4, 0.4, 0.4, \beta ^{(1)}, 0.4, 0.4, 0.4)^\top\). TP indicates Average number of True Positives, FP is Average number of False Positives, and F1 is the Average F1 score. Average Performance of each method over 50 datasets for each setting

ISSN: 1471-2105