def save_bin(self):
if self.args.gtf_bin:
sys.stderr.write(
"Error: Can't save since the binary provided at input")
sys.exit()
gene_models = pygenes.GeneModels()
gene_models.load_ensembl_gtf(args.gtf)
gene_models.save_binary(self.args.outfile + '.gene_models.binary')
def write_output(self):
gene_models = pygenes.GeneModels()
if (self.args.gtf_bin):
gene_models.load_binary(self.args.gtf_bin)
else:
gene_models.load_ensembl_gtf(self.args.gtf)
with open(args.infile, 'r') as titan_output:
while True:
line = titan_output.readline()
if line[0] == '#':
continue
header = line.rstrip()
self.outfile.write("%s\tPygenes(gene_id,gene_name;)\n" %
header)
break
##do something to.. data lines
for row in titan_output:
row = row.rstrip()
col = row.split('\t')
if self.args.demix:
chrom = col[2]
else:
chrom = col[1]
try:
if self.args.demix:
start = int(col[3])
end = int(col[4])
else:
start = int(col[2])
end = int(col[3])
except:
self.outfile.write(
"%s\t%s\n" %
(row, "[ERROR - position not of type int()]"))
continue
if (args.is_contained):
gene_ids = gene_models.find_contained_genes(
chrom, start, end)
else:
gene_ids = gene_models.find_overlapping_genes(
chrom, start, end)
pygenes_addition = ""
for gene_id in gene_ids:
gene_name = gene_models.get_gene(gene_id).name
pygenes_addition += "%s,%s;" % (gene_id, gene_name)
self.outfile.write("%s\t%s\n" % (row, pygenes_addition))
self.outfile.close()
def __init__(self,
infile,
outfile,
gtf_bin=None,
gtf=None,
is_contained=False):
self.infile = infile
self.outfile = outfile
self.is_contained = is_contained
if not gtf and not gtf_bin:
raise InputArgsException('Requires either gtf or gtf_bin files')
self.gene_models = pygenes.GeneModels()
if gtf_bin:
self.gene_models.load_binary(gtf_bin)
else:
self.gene_models.load_ensembl_gtf(gtf)
def tabulate_results(breakpoints_filename, likelihoods_filename, library_ids,
genome_fasta, gtf_filename, dgv_filename,
breakpoint_table, breakpoint_library_table):
lib_names = pd.DataFrame(library_ids.items(),
columns=['library', 'library_id'])
breakpoints = pd.read_csv(breakpoints_filename,
sep='\t',
names=destruct.predict_breaks.breakpoint_fields,
converters=converters)
breakpoints = breakpoints.drop(['breakpoint_id'], axis=1)
breakpoints = breakpoints.rename(columns={'count': 'num_split'})
breakpoints.loc[breakpoints['inserted'] == '.', 'inserted'] = ''
likelihoods = pd.read_csv(likelihoods_filename,
sep='\t',
names=destruct.predict_breaks.likelihoods_fields,
converters=converters)
likelihoods = likelihoods.drop(['breakpoint_id'], axis=1)
breakpoint_reads = (likelihoods.groupby(['cluster_id', 'library_id'
]).size().reset_index())
breakpoint_reads.columns = ['cluster_id', 'library_id', 'num_reads']
breakpoint_unique_reads = (likelihoods.drop_duplicates([
'cluster_id', 'library_id', 'template_length_1', 'template_length_2'
]).groupby(['cluster_id', 'library_id']).size().reset_index())
breakpoint_unique_reads.columns = [
'cluster_id', 'library_id', 'num_unique_reads'
]
breakpoint_library = (
breakpoint_reads.merge(breakpoint_unique_reads).merge(lib_names).drop(
['library_id'], axis=1))
agg_f = {
'log_likelihood': np.average,
'log_cdf': np.average,
'template_length_1': max,
'template_length_2': max,
}
breakpoint_stats = (
likelihoods.groupby('cluster_id').agg(agg_f).reset_index())
breakpoint_stats['template_length_min'] = breakpoint_stats[[
'template_length_1', 'template_length_2'
]].min(axis=1)
breakpoint_counts = (
likelihoods.groupby('cluster_id').size().reset_index())
breakpoint_counts.columns = ['cluster_id', 'num_reads']
breakpoint_unique_counts = (likelihoods.drop_duplicates(
['cluster_id', 'library_id', 'template_length_1',
'template_length_2']).groupby('cluster_id').size().reset_index())
breakpoint_unique_counts.columns = ['cluster_id', 'num_unique_reads']
breakpoints = breakpoints.merge(breakpoint_stats,
on='cluster_id',
how='inner')
breakpoints = breakpoints.merge(breakpoint_counts,
on='cluster_id',
how='inner')
breakpoints = breakpoints.merge(breakpoint_unique_counts,
on='cluster_id',
how='inner')
# Calculate breakpoint type
def breakpoint_type(row):
if row['chromosome_1'] != row['chromosome_2']:
return 'translocation'
if row['strand_1'] == row['strand_2']:
return 'inversion'
positions = sorted([(row['position_{0}'.format(side)],
row['strand_{0}'.format(side)])
for side in (1, 2)])
if positions[0][1] == '+':
return 'deletion'
else:
return 'duplication'
breakpoints['type'] = breakpoints.apply(breakpoint_type, axis=1)
# Calculate number inserted at the breakpoint
def calculate_num_inserted(row):
if row['inserted'] == '.':
return 0
else:
return len(row['inserted'])
breakpoints['num_inserted'] = breakpoints.apply(calculate_num_inserted,
axis=1)
# Annotate sequence
reference_sequences = dict()
for id, seq in destruct.utils.seq.read_sequences(open(genome_fasta, 'rt')):
reference_sequences[id] = seq
breakpoints['sequence'] = breakpoints.apply(
lambda row: create_sequence(row, reference_sequences), axis=1)
# Annotate gene information
gene_models = pygenes.GeneModels()
gene_models.load_ensembl_gtf(gtf_filename)
breakpoints = breakpoints.apply(
lambda row: annotate_genes(row, gene_models), axis=1)
# Annotate database of genomic variants
dgv = DGVDatabase(dgv_filename)
breakpoints['dgv_ids'] = breakpoints.apply(lambda row: query_dgv(row, dgv),
axis=1)
breakpoints = breakpoints.rename(columns={'cluster_id': 'prediction_id'})
breakpoints.to_csv(breakpoint_table,
sep='\t',
na_rep='NA',
header=True,
index=False)
breakpoint_library = breakpoint_library.rename(
columns={'cluster_id': 'prediction_id'})
breakpoint_library.to_csv(breakpoint_library_table,
sep='\t',
na_rep='NA',
header=True,
index=False)