Table 4. Two-way gene-gene interaction models by using the GMDR method with adjustment for age and gender.

Phenotype	Two-way interaction model	Testing accuracy (%)	P value
MetS	ARNTL rs10832020, GSK3B rs2199503	52.92	0.039
	ARNTL rs10832020, PER3 rs10746473	53.33	0.021
	ARNTL rs10832020, RORA rs8034880	55.01	< 0.001
	ARNTL rs10832020, RORB rs972902	52.92	0.032
	GSK3B rs2199503, PER3 rs10746473	53.15	0.027
	GSK3B rs2199503, RORA rs8034880	55.09	< 0.001
	GSK3B rs2199503, RORB rs972902	54.04	0.002
	PER3 rs10746473, RORA rs8034880	54.78	< 0.001
	PER3 rs10746473, RORB rs972902	55.19	< 0.001
	RORA rs8034880, RORB rs972902	53.96	0.008

GMDR = generalized multifactor dimensionality reduction, MetS = metabolic syndrome. P value was based on 1,000 permutations. Analysis was obtained after adjustment for covariates including age and gender. P values of < 0.005 (Bonferroni correction: 0.05/10) are shown in bold.