Table 1.

Summary of footprinting tools, including software category, programming language, algorithm or statistical method, bias correction for DNase-seq or ATAC-seq, and output statistics. In addition, the second last column exemplifies the application of tools in ATAC-seq data

Tool	Category	Language	Algorithm	Bias correction?	Statistics	Used for ATAC in literature?	Reference
Neph	De novo	C++	Slide window	N	Footprint occupancy score (FOS)	N	[47]
HINT		Python	HMM	N	Probability	N	[133]
HINT-BC		Python	HMM	Y (DNase-seq)	Probability	Y [48]	[130]
HINT-ATAC		Python	HMM	Y (ATAC-seq)	Probability	Y [134]	[134]
Boyle		NA	HMM	N	Probability	N	[135]
Wellington		Python	Binomial test	N (visualize bias)	P value, FDR	Y [48]	[136]
Wellington-bootstrap		Python	Bootstrap DE analysis	N (visualize bias)	P value, FDR	Y [48]	[137]
DNase2TF		R	Binomial test, iteratively merge	Y (DNase-seq)	FDR	Y [134]	[129]
CENTIPEDE	Motif-centric	R	Bayesian mixture model, unsupervised	N	Posterior probability	Y [138]	[139]
msCentipede		Python and Cython	Bayesian multiscale model, unsupervised	Y (can extend to ATAC-seq)	Posterior probability	Y [140]	[140]
Romulus		R	Bayesian mixture model, unsupervised	N	Posterior probability	N	[141]
PIQ		R	Gaussian process model, unsupervised	N	Probability of binding times local chromatin accessibility	Y [134]	[147]
BinDNase		R	Logistic regression, supervised	N	Probability	N	[142]
MILLIPEDE		R	Logistic regression, supervised	N (robust to bias)	Probability	N	[143]
DeFCoM		Python	SVM, supervised	N	Ranking	Y [131, 134]	[131]
BPAC		Python	Random forest, supervised	N	Probability	N	[144]
BaGFoot		R	Differential motif activity	Y	P value	Y [132]	[132]

FDR false discovery rate, HMM hidden Markov model, SVM support vector machine