An application of Random Forests to a genome-wide association dataset: Methodological considerations & new findings

Table 4 Top RF SNPs in the MS case-control dataset with LD pruning R² = 0.90

Chr	SNP	Gene	MAF	RF Rank	CHISQ	P-Value
7	rs6467970	SEMA3A (44.1 kb)*	0.19	1	19.71	9.00 * 10^-6
8	rs1466526	FAM164A (86.0 kb)*	0.25	2	15.60	7.84 * 10^-5
10	rs10823051	CTNNA3	0.16	3	17.61	2.71 * 10^-5
1	rs10754012	RGS1 (3.3 mb)*	0.23	4	22.24	2.41 * 10^-6
6	rs1040638	PHACTR2	0.48	5	13.82	2.01 * 10^-4
8	rs4739135	FAM164A (98.8 kb)*	0.19	6	16.33	5.34 * 10^-5
18	rs4798684	ADCYAP1 (19.8 kb)*	0.30	7	13.41	2.51 * 10^-4
10	rs7092549	PPAPDC1A (140.0 kb)*	0.23	8	17.06	3.62 * 10^-5
14	rs10483442	NPAS3	0.19	9	13.88	1.95 * 10^-4
20	rs2223712	BTBD3 (3.6 kb)*	0.19	10	11.86	2.71 * 10^-4
12	rs1106240	PITPNM2	0.20	11	18.71	1.52 * 10^-5
5	rs156823	ARL15	0.47	12	12.13	4.97 * 10^-4
9	rs10975130	KANK1	0.16	13	24.04	7.81 * 10^-6
12	rs12578774	AACS (1.3 mb)*	0.31	14	19.67	9.19 * 10^-6
2	rs7583622	ASB3	0.23	15	17.07	3.60 * 10^-5
6	rs6570578	PHACTR2	0.45	16	17.00	3.73 * 10^-5
1	rs282177	RPS6KA1	0.26	17	17.01	3.72 * 10^-5
5	rs11949767	MXD3 (59.7 kb)*	0.26	18	17.79	2.47 * 10^-5
10	rs7427	MSRB2	0.36	19	10.64	1.10 * 10^-3
8	rs1879818	TRAPPC9	0.30	20	7.35	6.72 * 10^-3
16	rs1974876	CCDC113	0.15	21	10.91	9.55 * 10^-3
17	rs11651517	GAS7	0.43	22	10.29	1.34 * 10^-3
20	rs6018946	BLCAP (581.3 kb)*	0.34	23	16.85	4.04 * 10^-5
2	rs11694785	ARHGAP25	0.40	24	10.15	1.44 * 10^-3
2	rs6746541	ATOH8	0.35	25	16.06	6.14 * 10^-5

The top 25 SNPs from RF analysis of the dataset without chromosome 6 SNPs are shown above in the R² = 0.90 runs after 3 sub-samplings. Results are similar to the analysis of full dataset, though there is more heterogenetity in the top findings, owing primarily to LD pruning. *Indicates that the gene is the closest gene with distance.

ISSN: 2730-6844