SNPs selection using support vector regression and genetic algorithms in GWAS

BMC Genomics

Table 8 Mean and standard deviation of the Pearson correlation for the SVR models constructed from subsets of SNPs selected by adjusted p-values of the simulated database 2.

Group	SNP selection with adjusted p-values	# SNPs	Linear**	RBF**	PUK**
1	< 10^-9	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
2	< 10^-8	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
3	< 10^-7	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
4	< 10^-6	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
5	< 10^-5	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
6	< 10^-4	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
7	< 10^-3	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
8	< 10^-2	1^a	0.53(0.07)	0.94(0.04)	0.95(0.04)
9	< 10^-1	3^b	0.53(0.07)	0.94(0.04)	0.95(0.04)
10	< 0.20	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)
11	< 0.30	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)
12	< 0.40	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)
13	< 0.50	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)
14	< 0.60	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)
15	< 0.70	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)
16	< 0.80	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)
17	< 0.90	4^c	0.53(0.07)	0.94(0.04)	0.95(0.04)

*Adjusted p-values of the Spearman correlation with Bonferroni correction.
** Mean and standard deviation of the Pearson correlation in 10-fold cross validation.
(a) Only the SNP3. (b) Only SNPs 3, 4 and 5. (c) Only SNPs 3, 4, 5 and 8964.
For each group defined after the first selection, a model SVR was constructed, evaluated for each kernel, through the Pearson correlation coefficient in 10-fold cross validation. The bold line indicates the best model from the first selection for PUK.

ISSN: 1471-2164