Juicer module

class fanc.compatibility.juicer.JuicerHic(hic_file, resolution=None, mode='r', tmpdir=None, norm=None)

Bases: fanc.matrix.RegionMatrixContainer

add_contact(contact, *args, **kwargs)

Alias for add_edge()

Parameters:
  • contactEdge
  • args – Positional arguments passed to _add_edge()
  • kwargs – Keyword arguments passed to _add_edge()
add_contacts(contacts, *args, **kwargs)

Alias for add_edges()

add_edge(edge, check_nodes_exist=True, *args, **kwargs)

Add an edge / contact between two regions to this object.

Parameters:
  • edgeEdge, dict with at least the attributes source and sink, optionally weight, or a list of length 2 (source, sink) or 3 (source, sink, weight).
  • check_nodes_exist – Make sure that there are nodes that match source and sink indexes
  • args – Positional arguments passed to _add_edge()
  • kwargs – Keyword arguments passed to _add_edge()
add_edge_from_dict(edge, *args, **kwargs)

Direct method to add an edge from dict input.

Parameters:edge – dict with at least the keys “source” and “sink”. Additional keys will be loaded as edge attributes
add_edge_from_edge(edge, *args, **kwargs)

Direct method to add an edge from Edge input.

Parameters:edgeEdge
add_edge_from_list(edge, *args, **kwargs)

Direct method to add an edge from list or tuple input.

Parameters:edge – List or tuple. Should be of length 2 (source, sink) or 3 (source, sink, weight)
add_edge_simple(source, sink, weight=None, *args, **kwargs)

Direct method to add an edge from Edge input.

Parameters:
  • source – Source region index
  • sink – Sink region index
  • weight – Weight of the edge
add_edges(edges, *args, **kwargs)

Bulk-add edges from a list.

List items can be any of the supported edge types, list, tuple, dict, or Edge. Repeatedly calls add_edge(), so may be inefficient for large amounts of data.

Parameters:edges – List (or iterator) of edges. See add_edge() for details
add_region(region, *args, **kwargs)

Add a genomic region to this object.

This method offers some flexibility in the types of objects that can be loaded. See parameters for details.

Parameters:region – Can be a GenomicRegion, a str in the form ‘<chromosome>:<start>-<end>[:<strand>], a dict with at least the fields ‘chromosome’, ‘start’, and ‘end’, optionally ‘ix’, or a list of length 3 (chromosome, start, end) or 4 (ix, chromosome, start, end).
static bin_intervals(intervals, bins, interval_range=None, smoothing_window=None, nan_replacement=None, zero_to_nan=False)

Bin a given set of intervals into a fixed number of bins.

Parameters:
  • intervals – iterator of tuples (start, end, score)
  • bins – Number of bins to divide the region into
  • interval_range – Optional. Tuple (start, end) in base pairs of range of interval to be binned. Useful if intervals argument does not cover to exact genomic range to be binned.
  • smoothing_window – Size of window (in bins) to smooth scores over
  • nan_replacement – NaN values in the scores will be replaced with this value
  • zero_to_nan – If True, will convert bins with score 0 to NaN
Returns:

iterator of tuples: (start, end, score)

static bin_intervals_equidistant(intervals, bin_size, interval_range=None, smoothing_window=None, nan_replacement=None, zero_to_nan=False)

Bin a given set of intervals into bins with a fixed size.

Parameters:
  • intervals – iterator of tuples (start, end, score)
  • bin_size – Size of each bin in base pairs
  • interval_range – Optional. Tuple (start, end) in base pairs of range of interval to be binned. Useful if intervals argument does not cover to exact genomic range to be binned.
  • smoothing_window – Size of window (in bins) to smooth scores over
  • nan_replacement – NaN values in the scores will be replaced with this value
  • zero_to_nan – If True, will convert bins with score 0 to NaN
Returns:

iterator of tuples: (start, end, score)

bin_size

Return the length of the first region in the dataset.

Assumes all bins have equal size.

Returns:int
binned_regions(region=None, bins=None, bin_size=None, smoothing_window=None, nan_replacement=None, zero_to_nan=False, *args, **kwargs)

Same as region_intervals, but returns GenomicRegion objects instead of tuples.

Parameters:
  • region – String or class:~GenomicRegion object denoting the region to be binned
  • bins – Number of bins to divide the region into
  • bin_size – Size of each bin (alternative to bins argument)
  • smoothing_window – Size of window (in bins) to smooth scores over
  • nan_replacement – NaN values in the scores will be replaced with this value
  • zero_to_nan – If True, will convert bins with score 0 to NaN
  • args – Arguments passed to _region_intervals
  • kwargs – Keyword arguments passed to _region_intervals
Returns:

iterator of GenomicRegion objects

bins_to_distance(bins)

Convert fraction of bins to base pairs

Parameters:bins – float, fraction of bins
Returns:int, base pairs
chromosome_bins

Returns a dictionary of chromosomes and the start and end index of the bins they cover.

Returned list is range-compatible, i.e. chromosome bins [0,5] cover chromosomes 1, 2, 3, and 4, not 5.

chromosome_lengths

Returns a dictionary of chromosomes and their length in bp.

chromosomes()

Get a list of chromosome names.

close(remove_tmp=True)

Close this Juicer file and run exit operations.

If file was opened with tmpdir in read-only mode: close file and delete temporary copy.

Parameters:remove_tmp – If False, does not delete temporary copy of file.
distance_to_bins(distance)

Convert base pairs to fraction of bins.

Parameters:distance – distance in base pairs
Returns:float, distance as fraction of bin size
edge_data(attribute, *args, **kwargs)

Iterate over specific edge attribute.

Parameters:
  • attribute – Name of the attribute, e.g. “weight”
  • args – Positional arguments passed to edges()
  • kwargs – Keyword arguments passed to edges()
Returns:

iterator over edge attribute

edge_subset(key=None, *args, **kwargs)

Get a subset of edges.

This is an alias for edges().

Returns:generator (Edge)
edges

Iterate over contacts / edges.

edges() is the central function of RegionPairsContainer. Here, we will use the Hic implementation for demonstration purposes, but the usage is exactly the same for all compatible objects implementing RegionPairsContainer, including JuicerHic and CoolerHic.

import fanc

# file from FAN-C examples
hic = fanc.load("output/hic/binned/fanc_example_1mb.hic")

We can easily find the number of edges in the sample Hic object:

len(hic.edges)  # 8695

When used in an iterator context, edges() iterates over all edges in the RegionPairsContainer:

for edge in hic.edges:
    # do something with edge
    print(edge)
    # 42--42; bias: 5.797788472650082e-05; sink_node: chr18:42000001-43000000; source_node: chr18:42000001-43000000; weight: 0.12291311562018173
    # 24--28; bias: 6.496381719803623e-05; sink_node: chr18:28000001-29000000; source_node: chr18:24000001-25000000; weight: 0.025205961072838057
    # 5--76; bias: 0.00010230955745211447; sink_node: chr18:76000001-77000000; source_node: chr18:5000001-6000000; weight: 0.00961709840049876
    # 66--68; bias: 8.248432587969082e-05; sink_node: chr18:68000001-69000000; source_node: chr18:66000001-67000000; weight: 0.03876763316345468
    # ...

Calling edges() as a method has the same effect:

# note the '()'
for edge in hic.edges():
    # do something with edge
    print(edge)
    # 42--42; bias: 5.797788472650082e-05; sink_node: chr18:42000001-43000000; source_node: chr18:42000001-43000000; weight: 0.12291311562018173
    # 24--28; bias: 6.496381719803623e-05; sink_node: chr18:28000001-29000000; source_node: chr18:24000001-25000000; weight: 0.025205961072838057
    # 5--76; bias: 0.00010230955745211447; sink_node: chr18:76000001-77000000; source_node: chr18:5000001-6000000; weight: 0.00961709840049876
    # 66--68; bias: 8.248432587969082e-05; sink_node: chr18:68000001-69000000; source_node: chr18:66000001-67000000; weight: 0.03876763316345468
    # ...

Rather than iterate over all edges in the object, we can select only a subset. If the key is a string or a GenomicRegion, all non-zero edges connecting the region described by the key to any other region are returned. If the key is a tuple of strings or GenomicRegion, only edges between the two regions are returned.

# select all edges between chromosome 19
# and any other region:
for edge in hic.edges("chr19"):
    print(edge)
    # 49--106; bias: 0.00026372303696871666; sink_node: chr19:27000001-28000000; source_node: chr18:49000001-50000000; weight: 0.003692122517562033
    # 6--82; bias: 0.00021923129703834945; sink_node: chr19:3000001-4000000; source_node: chr18:6000001-7000000; weight: 0.0008769251881533978
    # 47--107; bias: 0.00012820949175399097; sink_node: chr19:28000001-29000000; source_node: chr18:47000001-48000000; weight: 0.0015385139010478917
    # 38--112; bias: 0.0001493344481069762; sink_node: chr19:33000001-34000000; source_node: chr18:38000001-39000000; weight: 0.0005973377924279048
    # ...

# select all edges that are only on
# chromosome 19
for edge in hic.edges(('chr19', 'chr19')):
    print(edge)
    # 90--116; bias: 0.00021173151730025176; sink_node: chr19:37000001-38000000; source_node: chr19:11000001-12000000; weight: 0.009104455243910825
    # 135--135; bias: 0.00018003890596887822; sink_node: chr19:56000001-57000000; source_node: chr19:56000001-57000000; weight: 0.10028167062466517
    # 123--123; bias: 0.00011063368998965993; sink_node: chr19:44000001-45000000; source_node: chr19:44000001-45000000; weight: 0.1386240135570439
    # 92--93; bias: 0.00040851066434864896; sink_node: chr19:14000001-15000000; source_node: chr19:13000001-14000000; weight: 0.10090213409411629
    # ...

# select inter-chromosomal edges
# between chromosomes 18 and 19
for edge in hic.edges(('chr18', 'chr19')):
    print(edge)
    # 49--106; bias: 0.00026372303696871666; sink_node: chr19:27000001-28000000; source_node: chr18:49000001-50000000; weight: 0.003692122517562033
    # 6--82; bias: 0.00021923129703834945; sink_node: chr19:3000001-4000000; source_node: chr18:6000001-7000000; weight: 0.0008769251881533978
    # 47--107; bias: 0.00012820949175399097; sink_node: chr19:28000001-29000000; source_node: chr18:47000001-48000000; weight: 0.0015385139010478917
    # 38--112; bias: 0.0001493344481069762; sink_node: chr19:33000001-34000000; source_node: chr18:38000001-39000000; weight: 0.0005973377924279048
    # ...

By default, edges() will retrieve all edge attributes, which can be slow when iterating over a lot of edges. This is why all file-based FAN-C RegionPairsContainer objects support lazy loading, where attributes are only read on demand.

for edge in hic.edges('chr18', lazy=True):
    print(edge.source, edge.sink, edge.weight, edge)
    # 42 42 0.12291311562018173 <fanc.matrix.LazyEdge for row /edges/chrpair_0_0.row (Row), pointing to row #0>
    # 24 28 0.025205961072838057 <fanc.matrix.LazyEdge for row /edges/chrpair_0_0.row (Row), pointing to row #1>
    # 5 76 0.00961709840049876 <fanc.matrix.LazyEdge for row /edges/chrpair_0_0.row (Row), pointing to row #2>
    # 66 68 0.03876763316345468 <fanc.matrix.LazyEdge for row /edges/chrpair_0_0.row (Row), pointing to row #3>
    # ...

Warning

The lazy iterator reuses the LazyEdge object in every iteration, and overwrites the LazyEdge attributes. Therefore do not use lazy iterators if you need to store edge objects for later access. For example, the following code works as expected list(hic.edges()), with all Edge objects stored in the list, while this code list(hic.edges(lazy=True)) will result in a list of identical LazyEdge objects. Always ensure you do all edge processing in the loop when working with lazy iterators!

When working with normalised contact frequencies, such as obtained through matrix balancing in the example above, edges() automatically returns normalised edge weights. In addition, the bias attribute will (typically) have a value different from 1.

When you are interested in the raw contact frequency, use the norm=False parameter:

for edge in hic.edges('chr18', lazy=True, norm=False):
    print(edge.source, edge.sink, edge.weight)
    # 42 42 2120.0
    # 24 28 388.0
    # 5 76 94.0
    # 66 68 470.0
    # ...

You can also choose to omit all intra- or inter-chromosomal edges using intra_chromosomal=False or inter_chromosomal=False, respectively.

Returns:Iterator over Edge or equivalent.
edges_dict(*args, **kwargs)

Edges iterator with access by bracket notation.

This iterator always returns unnormalised edges.

Returns:dict or dict-like iterator
expected_values(selected_chromosome=None, norm=True, *args, **kwargs)

Calculate the expected values for genomic contacts at all distances.

This calculates the expected values between genomic regions separated by a specific distance. Expected values are calculated as the average weight of edges between region pairs with the same genomic separation, taking into account unmappable regions.

It will return a tuple with three values: a list of genome-wide intra-chromosomal expected values (list index corresponds to number of separating bins), a dict with chromosome names as keys and intra-chromosomal expected values specific to each chromosome, and a float for inter-chromosomal expected value.

Parameters:
  • selected_chromosome – (optional) Chromosome name. If provided, will only return expected values for this chromosome.
  • norm – If False, will calculate the expected values on the unnormalised matrix.
  • args – Not used in this context
  • kwargs – Not used in this context
Returns:

list of intra-chromosomal expected values, dict of intra-chromosomal expected values by chromosome, inter-chromosomal expected value

expected_values_and_marginals(selected_chromosome=None, norm=True, *args, **kwargs)

Calculate the expected values for genomic contacts at all distances and the whole matrix marginals.

This calculates the expected values between genomic regions separated by a specific distance. Expected values are calculated as the average weight of edges between region pairs with the same genomic separation, taking into account unmappable regions.

It will return a tuple with three values: a list of genome-wide intra-chromosomal expected values (list index corresponds to number of separating bins), a dict with chromosome names as keys and intra-chromosomal expected values specific to each chromosome, and a float for inter-chromosomal expected value.

Parameters:
  • selected_chromosome – (optional) Chromosome name. If provided, will only return expected values for this chromosome.
  • norm – If False, will calculate the expected values on the unnormalised matrix.
  • args – Not used in this context
  • kwargs – Not used in this context
Returns:

list of intra-chromosomal expected values, dict of intra-chromosomal expected values by chromosome, inter-chromosomal expected value

find_region(query_regions, _regions_dict=None, _region_ends=None, _chromosomes=None)

Find the region that is at the center of a region.

Parameters:query_regions – Region selector string, :class:~GenomicRegion, or list of the former
Returns:index (or list of indexes) of the region at the center of the query region
intervals(*args, **kwargs)

Alias for region_intervals.

mappable(region=None)

Get the mappability vector of this matrix.

marginals(masked=True, *args, **kwargs)

Get the marginals vector of this Hic matrix.

Sums up all contacts for each bin of the Hi-C matrix. Unmappable regoins will be masked in the returned vector unless the masked parameter is set to False.

By default, corrected matrix entries are summed up. To get uncorrected matrix marginals use norm=False. Generally, all parameters accepted by edges() are supported.

Parameters:
  • masked – Use a numpy masked array to mask entries corresponding to unmappable regions
  • kwargs – Keyword arguments passed to edges()
matrix(key=None, log=False, default_value=None, mask=True, log_base=2, *args, **kwargs)

Assemble a RegionMatrix from region pairs.

Parameters:
  • key – Matrix selector. See edges() for all supported key types
  • log – If True, log-transform the matrix entries. Also see log_base
  • log_base – Base of the log transformation. Default: 2; only used when log=True
  • default_value – (optional) set the default value of matrix entries that have no associated edge/contact
  • mask – If False, do not mask unmappable regions
  • args – Positional arguments passed to regions_and_matrix_entries()
  • kwargs – Keyword arguments passed to regions_and_matrix_entries()
Returns:

RegionMatrix

classmethod merge(pairs, *args, **kwargs)

Merge two or more RegionPairsContainer objects.

Parameters:
  • pairslist of RegionPairsContainer
  • args – Positional arguments passed to constructor of this class
  • kwargs – Keyword arguments passed to constructor of this class
possible_contacts()

Calculate the possible number of contacts in the genome.

This calculates the number of potential region pairs in a genome for any possible separation distance, taking into account the existence of unmappable regions.

It will calculate one number for inter-chromosomal pairs, return a list with the number of possible pairs where the list index corresponds to the number of bins separating two regions, and a dictionary of lists for each chromosome.

Returns:possible intra-chromosomal pairs, possible intra-chromosomal pairs by chromosome, possible inter-chromosomal pairs
region_bins(*args, **kwargs)

Return slice of start and end indices spanned by a region.

Parameters:args – provide a GenomicRegion here to get the slice of start and end bins of onlythis region. To get the slice over all regions leave this blank.
Returns:
region_intervals(region, bins=None, bin_size=None, smoothing_window=None, nan_replacement=None, zero_to_nan=False, score_field='score', *args, **kwargs)

Return equally-sized genomic intervals and associated scores.

Use either bins or bin_size argument to control binning.

Parameters:
  • region – String or class:~GenomicRegion object denoting the region to be binned
  • bins – Number of bins to divide the region into
  • bin_size – Size of each bin (alternative to bins argument)
  • smoothing_window – Size of window (in bins) to smooth scores over
  • nan_replacement – NaN values in the scores will be replaced with this value
  • zero_to_nan – If True, will convert bins with score 0 to NaN
  • args – Arguments passed to _region_intervals
  • kwargs – Keyword arguments passed to _region_intervals
Returns:

iterator of tuples: (start, end, score)

region_subset(region, *args, **kwargs)

Takes a class:~GenomicRegion and returns all regions that overlap with the supplied region.

Parameters:region – String or class:~GenomicRegion object for which covered bins will be returned.
regions

Iterate over genomic regions in this object.

Will return a GenomicRegion object in every iteration. Can also be used to get the number of regions by calling len() on the object returned by this method.

Returns:RegionIter
regions_and_edges(key, *args, **kwargs)

Convenient access to regions and edges selected by key.

Parameters:
  • key – Edge selector, see edges()
  • args – Positional arguments passed to edges()
  • kwargs – Keyword arguments passed to edges()
Returns:

list of row regions, list of col regions, iterator over edges

regions_and_matrix_entries(key=None, score_field=None, *args, **kwargs)

Convenient access to non-zero matrix entries and associated regions.

Parameters:
  • key – Edge key, see edges()
  • oe – If True, will divide observed values by their expected value at the given distance. False by default
  • oe_per_chromosome – If True (default), will do a per-chromosome O/E calculation rather than using the whole matrix to obtain expected values
  • score_field – (optional) any edge attribute that returns a number can be specified here for filling the matrix. Usually this is defined by the _default_score_field attribute of the matrix class.
  • args – Positional arguments passed to edges()
  • kwargs – Keyword arguments passed to edges()
Returns:

list of row regions, list of col regions, iterator over (i, j, weight) tuples

regions_dict

Return a dictionary with region index as keys and regions as values.

Returns:dict {region.ix: region, …}
static regions_identical(pairs)

Check if the regions in all objects in the list are identical.

Parameters:pairslist of RegionBased objects
Returns:True if chromosome, start, and end are identical between all regions in the same list positions.
scaling_factor(matrix, weight_column=None)

Compute the scaling factor to another matrix.

Calculates the ratio between the number of contacts in this Hic object to the number of contacts in another Hic object.

Parameters:
  • matrix – A Hic object
  • weight_column – Name of the column to calculate the scaling factor on
Returns:

float

to_bed(file_name, subset=None, **kwargs)

Export regions as BED file

Parameters:
  • file_name – Path of file to write regions to
  • subset – optional GenomicRegion or str to write only regions overlapping this region
  • kwargs – Passed to write_bed()
to_bigwig(file_name, subset=None, **kwargs)

Export regions as BigWig file.

Parameters:
  • file_name – Path of file to write regions to
  • subset – optional GenomicRegion or str to write only regions overlapping this region
  • kwargs – Passed to write_bigwig()
to_gff(file_name, subset=None, **kwargs)

Export regions as GFF file

Parameters:
  • file_name – Path of file to write regions to
  • subset – optional GenomicRegion or str to write only regions overlapping this region
  • kwargs – Passed to write_gff()