Table 2.

Genome-wide correlation coefficients between experimental nucleosome occupancies and nucleosome occupancies predicted by different models

Prediction models	Features used in a model	Pearson correlation
		In vitro [9]	In vivo [45]
Our integrated model (this paper)	12 structural features in a linear model	0.88	0.42
Xi et al., 2010 [54]	Position-dependant k-mer preferences (k up to 5)	0.618	0.34
Kaplan et al., 2009 [9]	Position-dependant 5-mer preferences and periodic dinucleotide	0.89	0.34
Tillo and Hughes, 2009 [10]	A linear model combining G + C content, propeller twist, slide and several 4-mer occurrence	0.86	0.38
Yuan and Liu, 2008 [14]	Periodic dinucleotide signals of linker and nucleosomal sequence	0.35	0.27
Gabdank et al.,2010 [24]	Uses DNA bendability matrix	0.41	0.25
Miele et al., 2008 [20]	Sequence-dependant free energy of nucleosome formation	0.38	0.22
Field et al., 2008 [2]	Uses 5-mer preferences and periodic dinucleotide	0.74	0.39
Lee et al., 2007 [45]	G + C content, 4-mer occurrence, TFBSs and several structural features	0.63	0.42