Phenotype prediction based on genome-wide DNA methylation data

BMC Bioinformatics

Table 3 Numbers of joint CpGs amongst the 500 most significant ones

	Normal vs CIN2+a	Normal vs CIN2+b	Normal vs Cancer	Normal HPV+ vs N.HPV-	Normal HPV+ vs CIN2+a	Normal HPV+ vs CIN2+b	Normal HPV+ vs Cancer	Normal HPV- vs CIN2+a	Normal HPV- vs CIN2+b	Normal HPV- vs Cancer	CIN2+a vs CIN2+b	CIN2+a vs Cancer	CIN2+b vs Cancer
t	32	10	19	15	44	12	20	9	12	14	16	140	22
	39	15	28	11	55	17	35	9	7	16	32	190	26
	18	11	24	6	22	9	19	7	15	17	10	39	18
MWU	17	10	16	8	28	10	16	7	9	15	19	119	22
	25	13	17	7	42	13	29	9	8	11	29	171	23
	12	10	27	7	18	8	21	7	9	13	11	50	16
Bartlett	34	22	23	28	26	19	21	37	36	27	54	45	45
	75	53	37	32	71	53	26	39	39	25	98	62	57
	35	12	16	28	41	15	20	37	10	15	24	26	18
Levene	60	22	19	12	40	17	19	12	6	8	49	70	17
	90	41	32	14	59	30	29	12	14	17	68	79	34
	11	2	9	9	18	16	16	14	6	14	17	13	7
Age-corr.	15	32	10	107	16	30	9	13	18	11	19	7	9
Age-corr.	12	43	8	138	12	37	10	10	30	5	18	8	15
	32	38	22	49	29	30	23	29	29	16	13	19	12

CpGs were ordered according to five different tests (t-test, Mann–Whitney U test, Bartlett’s test, Levene’s test and test for methylation-age-correlation) and the number of overlapping CpGs between the first 500 of two datasets determined. The three rows per cell correspond to all, hyper- and hypo-CpGs. Significant overlaps (p < 0.01) are shown in bold.

ISSN: 1471-2105