Improved ChIP-chip analysis by a mixture model approach

BMC Bioinformatics

Table 1 Comparison of different methods for the original data set

	Mixer (Lowess)		Mixer (MA2C)		MA2C		TileMap (p = 0.01, q = 0.05)		TileMap (p = 0.002, q = 0.02)		HGMM
Sample	D	FDR	D	FDR	D	FDR	D	FDR	D	FDR	D	FDR
GSM254930	108	0.28	503	0.84	241	0.66	85	0.08	84	0.07	111	0.20
GSM254971	100	0.28	113	0.37	227	0.64	86	0.09	85	0.09	N/A	N/A
GSM254972	98	0.29	195	0.61	178	0.53	84	0.07	88	0.08	N/A	N/A
GSM254973	98	0.24	92	0.23	146	0.45	71	0.07	73	0.07	N/A	N/A
GSM254805	66	0.20	153	0.56	116	0.43	81	0.22	52	0.09	N/A	N/A
GSM254806	89	0.19	184	0.61	85	0.19	236	0.66	143	0.42	89	0.18
GSM254807	97	0.24	102	0.26	100	0.21	76	0.08	91	0.13	123	0.32

The first four samples, GSM254930, GSM254971, GSM254972, and GSM254973 were spiked with unamplified DNA, while the last three samples GSM254805, GSM254806, and GSM254807 were spiked with amplified DNA. Among the total of 385,149 probes, about 820 (~0.2%) of them are from spike-in regions. We did not obtain results of HGMM for some arrays (N/A) due to failure of function HGMM.

ISSN: 1471-2105