arcfish.tl.LoopCaller#

class arcfish.tl.LoopCaller(fdr_cutoff: float = 0.1, pval_cutoff: float = 1e-05, cut_lo: float = 100000.0, cut_up: float = 1000000.0, gap: float = 50000.0, outer_cut: float = 50000.0, ltclass: ~arcfish.tools.loop.LoopTestAbstract = <class 'arcfish.tools.loop.AxisWiseF'>)#

Class for chromatin loop calling.

Parameters:

  • fdr_cutoff (float) – FDR cut-off for chromatin loops, by default 0.1.

  • pval_cutoff (float) – P-value cut-off to filter loop summits, by default 1e-5.

  • cut_lo (float, optional) – Minimum loop size (1D genomic distance between the first locus and the second locus), by default 1e5.

  • cut_up (float, optional) – Maximum loop size, by default 1e6.

  • gap (float, optional) – Loop candidates gap away from each other are considered candidates for the same summit, by default 50e3.

  • outer_cut (float, optional) – Loci with 1D genomic distance within outer_cut from the target locus is included in the local background, by default 50e3. Must be larger than 25kb, which is the exclusion cutoff.

  • ltclass (LoopTestAbstract) – Test method used, by default AxisWiseF.

__init__(fdr_cutoff: float = 0.1, pval_cutoff: float = 1e-05, cut_lo: float = 100000.0, cut_up: float = 1000000.0, gap: float = 50000.0, outer_cut: float = 50000.0, ltclass: ~arcfish.tools.loop.LoopTestAbstract = <class 'arcfish.tools.loop.AxisWiseF'>)#

Methods

__init__([fdr_cutoff, pval_cutoff, cut_lo, ...])

call_loops(adata)

Call chromatin loops from a single chromosome.

spread_label(candidates, d1d)

Group loop candidates.

to_bedpe(result, adata)

Convert loop calling result to a data frame with columns c1, s1, e1, c2, s2, e2, stat, pval, fdr, candidate, label, summit.

Attributes

fdr_cutoff

FDR cut-off for chromatin loops.

gap

Loop candidates gap away from each other are considered candidates for the same summit.

loop_range

Loop size considered.

outer_cut

Loci with 1D genomic distance within outer_cut from the target locus is included in the local background.

pval_cutoff

P-value cut-off to filter loop summits.
property fdr_cutoff: float#

FDR cut-off for chromatin loops.

Type:

float

property pval_cutoff: float#

P-value cut-off to filter loop summits.

Type:

float

property loop_range: Tuple[int, int]#

Loop size considered.

Type:

Tuple[int, int]

property gap: int#

Loop candidates gap away from each other are considered candidates for the same summit.

Type:

int

property outer_cut: int#

Loci with 1D genomic distance within outer_cut from the target locus is included in the local background.

Type:

int

call_loops(adata: AnnData) → dict#

Call chromatin loops from a single chromosome.

Parameters:

adata (AnnData) – adata of a single chromosome, created by arcfish.pp.FOF_CT_Loader.create_adata().

Returns:

A dictionary with keys stat, pval, fdr, candidate, label, summit. Values are (p,p) matrices. Might also contain other values depending on the test class used.

Return type:

dict

spread_label(candidates: ndarray, d1d: ndarray) → ndarray#

Group loop candidates. Each group corresponds to one summit.

Parameters:

  • candidates ((p,p) np.ndarray) – Boolean matrix with candidate entries being True.

  • d1d ((p,) np.ndarray) – Array of 1D genomic locations of imaging loci.

Returns:

Integer valued matrix. Entries with the same value define the same loop summit.

Return type:

(p, p) np.ndarray

to_bedpe(result: dict, adata: AnnData) → DataFrame#

Convert loop calling result to a data frame with columns c1, s1, e1, c2, s2, e2, stat, pval, fdr, candidate, label, summit. c1, s1, e1 are the chromosome, starting position, and ending position of the first locus of the pair, c2, s2, e2 are the values for the second locus. stat, pval, fdr are test results for the pair. Candidate is boolean valued, label is integer valued, and summit is boolean valued.

Parameters:
Returns:

Output dataframe in bedpe format.

Return type:

pd.DataFrame