Association of APP gene polymorphisms and promoter methylation with essential hypertension in Guizhou: a case–control study

Table 8 MDR analysis of SNPs and CpGs

Nation	Locus No	Best model	Training Bal. Acc	Testing Bal. Acc	Cross-validation Consistency	χ²	P^a
	SNP-SNP interaction
Total populations	1	rs438031	0.531	0.507	8/10	3.384	0.066
	2	rs467021, rs364051	0.554	0.522	7/10	7.633	0.006
	3	rs2040273, rs2211772, rs438031	0.579	0.513	4/10	15.877	< 0.001
Miao population	1	rs2040273	0.570	0.570	10/10	4.695	0.030
	2	rs2040273, rs2830077	0.588	0.480	5/10	5.602	0.018
	3	rs2040273, rs2211772, rs463946	0.625	0.497	8/10	12.995	< 0.001
Buyi population	1	rs467021	0.557	0.479	7/10	2.432	0.119
	2	rs467021, rs463946	0.593	0.484	3/10	7.765	0.005
	3	rs2040273, rs467021, rs364048	0.626	0.471	3/10	14.706	< 0.001
Han population	1	rs2211772	0.533	0.533	10/10	6.153	0.013
	2	rs2211772, rs368196	0.560	0.525	8/10	9.156	0.003
	3	rs2211772, rs2830077, rs368196	0.572	0.499	2/10	9.156	0.003
	CpG-CpG interaction
Total populations	1	CpG_19	0.611	0.611	10/10	8.837	0.003
	2	CpG_19, CpG_20	0.654	0.586	8/10	13.769	< 0.001
	3	CpG_11, CpG_19, CpG_20	0.672	0.571	10/10	15.575	< 0.001
	CpG-SNP interaction
Total populations	1	CpG_19	0.611	0.611	10/10	8.837	0.003
	2	CpG_19, CpG_20	0.658	0.545	5/10	13.769	< 0.001
	3	CpG_11, CpG_19, rs364051	0.703	0.545	8/10	19.874	< 0.001

Bold indicates P < 0.05
Training Bal. Acc., Training Balanced accuracy; Testing Bal. Acc, Testing Balanced accuracy; The training/testing balance accuracy represented the accuracy of the training set and the test set, and was used to evaluate the prediction error of the interaction model. Cross-validation consistency represented ten-fold cross-validation, comparing the number of times the same factor combination was determined, N/10 meant that N out of 10 cross-validations were significant. ^aP value of permutation test, P < 0.05 means that the model is statistically significant

ISSN: 1479-7364