def cis_sites():
return [
CisSite(id='CIS1', chromosome='1', position=182408172,
strand=1, metadata=frozendict()),
CisSite(id='CIS2', chromosome='4', position=132408091,
strand=1, metadata=frozendict())
] # yapf: disable
def cis_insertions():
return [
# 1000 bp upstream of Trp53bp2.
Insertion(id='INS1', chromosome='1', position=182408172, strand=1,
support=2, metadata=frozendict({'cis_id': 'CIS1'})),
# Different chromosome.
Insertion(id='INS2', chromosome='4', position=77843175, strand=1,
support=2, metadata=frozendict({'cis_id': 'CIS2'}))
] # yapf: disable
def insertions():
# Trp53bp2 location: 1: 182,409,172-182,462,432.
# Myh9 location: 15: 77,760,587-77,842,175.
return [
# 1000 bp upstream of Trp53bp2.
Insertion(id='INS1', chromosome='1', position=182408172,
strand=1, support=2, metadata=frozendict()),
# 2000 bp downstream of Myh9.
Insertion(id='INS2', chromosome='15', position=77758587,
strand=1, support=2, metadata=frozendict()),
# Different chromosome.
Insertion(id='INS3', chromosome='4', position=77843175,
strand=1, support=2, metadata=frozendict())
] # yapf: disable
def _annotate_insertion(self, insertion):
trees = self._trees
# Identify overlapping features.
hits = set()
for window in self._windows:
applied_window = window.apply(insertion.chromosome,
insertion.position, insertion.strand)
hits |= {(feature['gene_id'], feature['gene_name'], window.name)
for feature in applied_window.get_overlap(trees)}
# Filter for blacklist.
if self._blacklist is not None:
hits = {hit for hit in hits if hit[1] not in self._blacklist}
if len(hits) > 0:
# Annotate insertion with overlapping genes.
for gene_id, gene_name, window_name in hits:
metadata = {'gene_id': gene_id, 'gene_name': gene_name}
if window_name is not None:
metadata['window'] = window_name
metadata = toolz.merge(insertion.metadata, metadata)
yield insertion._replace(metadata=frozendict(metadata))
else:
# In case of no overlap, return original insertion.
yield insertion
def assign_strand(cis_sites, insertions, mapping, min_homogeneity=0.75):
"""Assigns CIS sites the average strand of their insertions."""
ins_lookup = {insertion.id: insertion for insertion in insertions}
for cis_site in cis_sites:
# Lookup strands of CIS insertions.
cis_strands = np.array(
[ins_lookup[ins_id].strand for ins_id in mapping[cis_site.id]])
# Calculate mean strand, strand and homogeneity.
mean_strand = np.mean(cis_strands)
strand = np.sign(mean_strand)
homogeneity = np.sum(cis_strands == strand) / len(cis_strands)
# If homogeneity is below the given threshold, then we don't
# assign a specific strand (signified by a nan).
if homogeneity < min_homogeneity:
strand = np.nan
# Merge strand metadata with existing metadata.
strand_metadata = {
'strand_mean': mean_strand,
'strand_homogeneity': homogeneity
}
metadata = toolz.merge(cis_site.metadata, strand_metadata)
yield cis_site._replace(strand=strand, metadata=frozendict(metadata))
def _to_insertion(ref, pos, strand, ends, id_=None, **kwargs):
metadata = toolz.merge({
'depth': len(ends),
'depth_unique': len(set(ends))
}, kwargs)
return Insertion(id=id_,
chromosome=ref,
position=pos,
strand=strand,
support=metadata['depth_unique'],
metadata=frozendict(metadata))
def _annotate_insertions(self, insertions, cis_map):
for insertion in insertions:
genes = cis_map.get(insertion.metadata['cis_id'], set())
if len(genes) > 0:
for gene_name, gene_id in genes:
metadata = {'gene_id': gene_id, 'gene_name': gene_name}
metadata = toolz.merge(insertion.metadata, metadata)
if self._drop_cis_id:
metadata.pop('cis_id')
yield insertion._replace(metadata=frozendict(metadata))
else:
if self._drop_cis_id:
metadata = dict(insertion.metadata)
metadata.pop('cis_id')
yield insertion._replace(metadata=frozendict(metadata))
else:
yield insertion
def from_frame(cls, df):
"""Converts dataframe into a list of objects."""
cls.check_frame(df)
basic_fields = cls._non_metadata_fields()
metadata_fields = list(set(df.columns) - set(basic_fields))
for row in df.itertuples():
row_dict = row._asdict()
metadata = {k: row_dict.pop(k) for k in metadata_fields}
metadata = frozendict(toolz.valfilter(_not_nan, metadata))
row_dict.pop('Index', None)
if not set(basic_fields) == set(row_dict.keys()):
missing_fields = set(basic_fields) - set(row_dict.keys())
raise ValueError('Missing required fields ({})'.format(
', '.join(missing_fields)))
yield cls(metadata=metadata, **row_dict)
