BayMeth: improved DNA methylation quantification for affinity capture sequencing data using a flexible Bayesian approach

Table 3 Performance assessment for LNCaP analysis by copy number

Copy number	Number of bins	Method	Mean bias	Mean of squared differences	Spearman correlation
2	18,010	BayMeth	0.04	0.04	0.78
		BayMeth (SssI-free)	0.08	0.05	0.79
		BayMeth (CNV-unaware)	-0.11	0.06	0.78
		BayMeth (SssI-free, CNV-unaware)	-0.05	0.05	0.79
		Batman	0.03	0.06	0.74
		MEDIPS	-0.23	0.11	0.76
		BALM	-0.29	0.16	0.78
3	65,982	BayMeth	0.05	0.04	0.80
		BayMeth (SssI-free)	0.09	0.05	0.80
		BayMeth (CNV-unaware)	-0.01	0.04	0.80
		BayMeth (SssI-free, CNV-unaware)	0.05	0.04	0.80
		Batman	0.11	0.06	0.77
		MEDIPS	-0.19	0.09	0.76
		BALM	-0.20	0.10	0.79
4	256,074	BayMeth	0.05	0.04	0.81
		BayMeth (SssI-free)	0.10	0.05	0.81
		BayMeth (CNV-unaware)	0.05	0.04	0.81
		BayMeth (SssI-free, CNV-unaware)	0.11	0.06	0.81
		Batman	0.16	0.08	0.79
		MEDIPS	-0.17	0.09	0.76
		BALM	-0.12	0.07	0.80
5	11,790	BayMeth	0.04	0.03	0.83
		BayMeth (SssI-free)	0.07	0.05	0.82
		BayMeth (CNV-unaware)	0.09	0.04	0.83
		BayMeth (SssI-free, CNV-unaware)	0.12	0.06	0.82
		Batman	0.18	0.08	0.80
		MEDIPS	-0.12	0.07	0.80
		BALM	-0.08	0.05	0.82

Results are shown for 100-bp bins with a mappability of at least 75% stratified into the four most frequent copy number states. A threshold of 13 was applied for the depth of the SssI-control. Taking SssI information into account a uniform prior for the methylation level was used. In the SssI-free version a Dirac-Beta-Dirac mixture with weights fixed to 0.1, 0.8 and 0.1 was used. CNV, copy number variation; MSE, mean of squared differences.

ISSN: 1474-760X