Prediction of regulatory elements in mammalian genomes using chromatin signatures

BMC Bioinformatics

Table 5 Comparison of active promoter predictions.

untreated cell^a
	Active promoters			Inactive promoters
	Total Prediction	Expression Supported Prediction	PPV
Heintzman et al. [15]^b	197	127	64.47%	31
	229	127	55.46%	32
HMM (c₁ = 1.95)	197	128	64.97%	25
HMM (c₁ = 1.6)	229	135	58.95%	31
HMM (c₁ = 0.5)	309	143	46.28%	40
treated cell^a
	Active promoters			Inactive promoters
	Total Prediction	Expression Supported Prediction	PPV
Heintzman et al. [15]	204	128	62.75%	23
	213	128	60.09%	23
HMM (c₁ = 1.853)	204	128	62.75%	19
HMM (c₁ = 1.367)	247	139	56.27%	22
HMM (c₁ = 0.5)	328	145	44.21%	30

^aThe total numbers of predictions in Table 5 are slightly different from Table 4 because when multiple predicted sites were supported by the same TSS or any enhancer evidence, we merged these predictions (see Methods).
^bThe number of correctly predicted active promoters did not change using a lower cut-off in the profile-based method.

ISSN: 1471-2105