Seeing the forest through the trees: prioritising potentially functional interactions from Hi-C

Table 2 Methods for identification of statistically significant interactions for Hi-C data

Method name	Type	Base model	Specific features	Reference
Duan et al. 2010	Global background	Binomial	Specifically designed for yeast genome	[121]
Fit-Hi-C/FitHiC2	Global background	Binomial	Spline fitting procedure, compatible with different formats	[122, 123]
HOMER	Global background	Binomial	Highly compatible with the HOMER Hi-C analysis pipeline	[124]
GOTHiC	Global background	Binomial	Use relative coverage to estimate biases	[125]
FitHiChIP	Global background	Binomial	Specifically designed for HiChIP data	[126]
HIPPIE	Global background	Negative binomial	Account for fragment length and distance biases	[72, 127]
HiC-DC	Global background	Negative binomial	Use zero-inflated model	[128]
HMRFBayesHiC	Global background	Negative binomial	Use hidden Markov random field model	[129]
FastHiC	Global background	Negative binomial	An updated version of HMRFBayesHi, with improved computing speed	[130]
MaxHiC	Global background	Negative binomial	Use ADAM algorithm, identify interactions with enrichment for regulatory elements	[131]
CHiCAGO	Global background	Negative binomial	Specifically designed for CHi-C data	[132]
ChiCMaxima	Global background	Local maxima	Specifically designed for CHi-C data, more stringent and robust when comparing biological replicates	[133]
HICCUP	Local background	Local enrichment	Robust for finding chromatin loops	[3]
cLoops	Local background	DBSCAN	Loop detection with less computational resource	[134]
Automated identification of stripes	Local background	Local enrichment	Specifically designed to identify architectural stripes	[135]

ISSN: 1756-8935