Skip to main content

Table 1 The comparison of the evaluated TAD caller features

From: A comparison of topologically associating domain callers over mammals at high resolution

TAD Caller Runtime complexity Hierarchical TADs Gaps Allowed Parameters Input Details Implementation Language
3DNetMod [76] O(n) 18 Sparse 3-column matrix Python
Armatus [56] \(O(tn^2)\) 1 \(n \times n\) matrix C++
Arrowhead [3] \(O(n^2)\) 1 .hic file Java
CaTCH [60] NA 1 Sparse 4-column matrix R
Constrained HAC [78] \(O(n(h+\log (n)))\) 1 \(n \times n\) matrix R
CHDF [64] NA 1 \(n \times n\) matrix C++
chromoR [62] NA 2 \(n \times n\) matrix R
ClusterTAD [67] NA 1 \(n \times n\) matrix MATLAB
deDoc [74] \(O(n\log ^{2}n)\) 0 Sparse 3-column matrix Java
DI [1] NA 3 \(n \times (n + 3)\) matrix MATLAB, Perl
EAST [68] \(O(n^2)\) 3 \(n \times n\) matrix Python
GMAP [73] NA 10 \(n \times n\) matrix R, C++
GRiNCH [85] \(O(kn^2)\) 3 \(n \times n\) matrix C++
HiCExplorer [77] NA 4 h5 file Python
HiCKey [84] \(O(n^3)\) 3 \(n \times n\) matrix C++
HiCseg [61] \(O(Kn^{2})\) 3 \(n \times n\) matrix R, C
HiTAD [71] NA 1 .cool file Python
IC-Finder [66] NA 2 \(n \times n\) matrix MATLAB
InsulationScore [63] NA 5 \((n + 1) \times (n + 1)\) matrix Perl
Matryoshka [75] \(O(tl^{2})\) 1 \(n \times (n + 3)\) matrix C++
MrTADFinder [72] \(O(n^3)\) 1 Sparse 3-column matrix Julia
MSTD [80] NA 1 \(n \times n\) matrix Python
OnTAD [79] \(O(md^{2})\) 5 \(n \times n\) matrix C++
PSYCHIC [69] NA 1 \(n \times n\) matrix MATLAB, Python,Perl
Spectral [65] NA 2 \(n \times n\) matrix MATLAB
SpectralTAD [81] O(n) 11 \(n \times n\) matrix R
TADBD [82] NA 2 \(n \times n\) matrix R
TADbit [70] NA 1 \(n \times n\) matrix Python, C
TADpole [83] NA 3 \(n \times n\) matrix R
TADtree [59] \(O(nS^{5})\) 6 \(n \times n\) matrix Python
TopDom [57] NA 1 \(n \times (n + 3)\) matrix R
  1. The parameters below are as follows: Hi-C/Micro-C interaction matrix size is referred by n. TADtree’s S parameter refers to the maximum size of inferred TADs. OnTAD’s d parameter refers to the maximum size of inferred TADs, whereas m refers to the expected count of possible boundaries. Matryoshka’s t parameter defines the number of resolutions to be inferred, and l refers to interval frequency while clustering the inferred t resolutions. HiCseg’s K parameter defines the maximum number of diagonal TAD partitions. h parameter in Constrained HAC is the bandwidth. k in GRiNCH is the rank of low-dimensional matrices. Lastly, Armatus’ t parameter defines the number of resolutions to be inferred