def generate_feature_matrix(input_dir, input_files, input_files_id, format_,
caller, svtype_col_name, as_breakpoint,
definitions, output):
"""
Generate feature matrix from VCF or BEDPE files.
"""
if format_ == 'bedpe':
if (input_dir is None) & (input_files is None):
return
elif (input_files is None):
data = viola.read_bedpe_multi(input_dir,
svtype_col_name=svtype_col_name)
elif (input_dir is None):
ls_input = input_files.split(',')
ls_bedpe = [
viola.read_bedpe(path, svtype_col_name=svtype_col_name)
for path in ls_input
]
if input_files_id is None:
ls_names = range(len(ls_bedpe))
else:
ls_names = input_files_id.split(',')
data = viola.MultiBedpe(ls_bedpe, ls_names)
else:
return
else:
if (input_dir is None) & (input_files is None):
return
elif (input_files is None):
data = viola.read_vcf_multi(input_dir,
variant_caller=caller,
as_breakpoint=as_breakpoint)
elif (input_dir is None):
ls_input = input_files.split(',')
if as_breakpoint:
ls_vcf = [
viola.read_vcf(
path, variant_caller=caller).breakend2breakpoint()
for path in ls_input
]
else:
ls_vcf = [
viola.read_vcf(path, variant_caller=caller)
for path in ls_input
]
if input_files_id is None:
ls_names = range(len(ls_vcf))
else:
ls_names = input_files_id.split(',')
data = viola.MultiBedpe(ls_vcf, ls_names)
else:
return
result = data.classify_manual_svtype(definitions=definitions)
result.to_csv(output, sep='\t')
def test_classify_manual_svtype_from_file():
bedpe1 = viola.read_bedpe(StringIO(data))
bedpe2 = viola.read_bedpe(StringIO(data))
ls_conditions = [
small_del, large_del, small_dup, large_dup, small_inv, tra
]
ls_names = [
'small_del', 'large_del', 'small_dup', 'large_dup', 'small_inv', 'tra'
]
multibedpe = viola.MultiBedpe([bedpe1, bedpe2], ['bedpe1', 'bedpe2'])
path = os.path.join(HERE, '../bedpe/data/example_definition.txt')
result = multibedpe.classify_manual_svtype(definitions=path)
manual_sv_type = multibedpe.manual_sv_type
manual_sv_type.set_index('id', inplace=True)
manual_sv_type_expected = pd.read_csv(StringIO(data_expected),
sep='\t',
names=('id', 'value_idx',
'manual_sv_type'))
manual_sv_type_expected.set_index('id', inplace=True)
pd.testing.assert_frame_equal(manual_sv_type,
manual_sv_type_expected,
check_like=True)
result_expected = pd.DataFrame([[2, 3, 1, 0, 2, 2, 2],
[2, 3, 1, 0, 2, 2, 2]])
result_expected.columns = ls_names + ['others']
result_expected.columns.name = 'manual_sv_type'
result_expected.index = ['bedpe1', 'bedpe2']
result_expected.index.name = 'patients'
pd.testing.assert_frame_equal(result, result_expected)
def test_classify_manual_svtype_exclude_empty():
bedpe1 = viola.read_bedpe(StringIO(data))
bedpe2 = viola.read_bedpe(StringIO(data))
empty1 = viola.read_bedpe(StringIO(data_empty))
empty2 = viola.read_bedpe(StringIO(data_empty))
ls_conditions = [
small_del, large_del, small_dup, large_dup, small_inv, tra
]
ls_names = [
'small_del', 'large_del', 'small_dup', 'large_dup', 'small_inv', 'tra'
]
multibedpe = viola.MultiBedpe([bedpe1, empty1, bedpe2, empty2],
['bedpe1', 'empty1', 'bedpe2', 'empty2'])
result = multibedpe.classify_manual_svtype(ls_conditions=ls_conditions,
ls_names=ls_names,
exclude_empty_cases=True)
manual_sv_type = multibedpe.manual_sv_type
manual_sv_type.set_index('id', inplace=True)
manual_sv_type_expected = pd.read_csv(StringIO(data_expected),
sep='\t',
names=('id', 'value_idx',
'manual_sv_type'))
manual_sv_type_expected.set_index('id', inplace=True)
pd.testing.assert_frame_equal(manual_sv_type,
manual_sv_type_expected,
check_like=True)
result_expected = pd.DataFrame([[2, 3, 1, 0, 2, 2, 2],
[2, 3, 1, 0, 2, 2, 2]])
result_expected.columns = ls_names + ['others']
result_expected.columns.name = 'manual_sv_type'
result_expected.index = ['bedpe1', 'bedpe2']
result_expected.index.name = 'patients'
pd.testing.assert_frame_equal(result, result_expected)
def merge(self,
ls_bedpe=[],
ls_caller_names=None,
threshold=100,
linkage="complete",
str_missing=True):
"""
merge(ls_bedpe:list, ls_caller_names:list, threshold:float, linkage = "complete", str_missing=True)
Return a merged bedpe object from mulitple caller's bedpe objects in ls_bedpe
Parameters
----------
ls_bedpe:list
A list of bedpe objects to be merged, which are the same order with ls_caller_names
ls_caller_names:list
A list of names of bedpe objects to be merged, which should have self's name as the first element
threshold:float
Two SVs whose diference of positions is under this threshold are cosidered to be identical.
linkage:{‘complete’, ‘average’, ‘single’}, default=’complete’
The linkage of hierarchical clustering.
To keep the mutual distance of all SVs in each cluster below the threshold,
"complete" is recommended.
str_missing:boolean, default="True"
If True, all the missing strands are considered to be identical to the others.
Returns
----------
A merged bedpe object
"""
if self in ls_bedpe:
pass
else:
ls_bedpe = [self] + ls_bedpe
multibedpe = viola.MultiBedpe(ls_bedpe, ls_caller_names)
distance_matrix = self._generate_distance_matrix_by_distance(
multibedpe, penalty_length=3e9, str_missing=str_missing)
hcl_clustering_model = AgglomerativeClustering(
n_clusters=None,
affinity="precomputed",
linkage=linkage,
distance_threshold=threshold)
labels = hcl_clustering_model.fit_predict(X=distance_matrix)
positions_table = multibedpe.get_table("positions")
mergedid_dict = {labels[0]: 0}
ls_mergedid = []
idx_head = 0
for label in labels:
if label in mergedid_dict:
ls_mergedid.append(mergedid_dict[label])
else:
idx_head += 1
mergedid_dict[label] = idx_head
ls_mergedid.append(mergedid_dict[label])
N = len(positions_table)
value_idx = pd.Series(np.zeros(N, dtype=int))
df_mergedid = pd.DataFrame({
"id": positions_table["id"],
"value_idx": value_idx,
"mergedid": pd.Series(ls_mergedid)
})
originalid = multibedpe.get_table("global_id")["id"]
df_originalid = pd.DataFrame({
"id": positions_table["id"],
"value_idx": value_idx,
"originalid": originalid
})
############## Edited by Sugita ##################
df_id = multibedpe.get_table("global_id")
df_patients = multibedpe.get_table("patients")
df_id_patients = df_id.merge(df_patients,
left_on="patient_id",
right_on="id")
caller = df_id_patients["patients"]
df_caller = pd.DataFrame({
"id": positions_table["id"],
"value_idx": value_idx,
"caller": caller
})
############## /Edited by Sugita #################
df_svpos = multibedpe._df_svpos
odict_df_info = multibedpe._odict_df_info
merged_bedpe = viola.Bedpe(df_svpos=df_svpos,
odict_df_info=odict_df_info)
merged_bedpe.add_info_table(table_name="mergedid", df=df_mergedid)
merged_bedpe.add_info_table(table_name="originalid", df=df_originalid)
merged_bedpe.add_info_table(table_name="caller", df=df_caller)
return merged_bedpe
def test_read_bedpe_with_empty():
bedpe1 = viola.read_bedpe(StringIO(data))
bedpe2 = viola.read_bedpe(StringIO(data))
bedpe_empty = viola.read_bedpe(StringIO(data_empty))
multibedpe = viola.MultiBedpe([bedpe1, bedpe2, bedpe_empty], ['bedpe1', 'bedpe2', 'empty'])