def output_boundary_npa(boundaries, ref_genome_file, segment_size, data_file, label_file, n_features=5, isreverse=True):
'''
output boundary data to files
isreverse: output sequence of the reverse strand as well
'''
chrom_list = get_chromlist(ref_genome_file)
hdf5_file = None
hdf5_label_file = None
try:
hdf5_file = h5py.File(data_file, mode='w')
dt = hdf5_file.create_dataset('data', (len(boundaries) * 2, segment_size, n_features), compression="gzip", compression_opts=5)
if label_file:
hdf5_label_file = h5py.File(label_file, mode='w')
label = hdf5_label_file.create_dataset('label', (len(boundaries) * 2))
k = 0
for b in boundaries:
# print(b.chrom, b.start, b.end)
#print('start: {}, end: {}'.format(b.start, b.end))
seq = mutation_function.infer_true_seq(b, chrom_list[b.chrom], segment_size)
if len(seq) != segment_size:
raise Exception('length of sequences is wrong: %d' % (len(seq)))
seq = line_to_tensor(str(seq).upper())
dt[k,] = seq
if hdf5_label_file:
label[k] = b.label
k += 1
# complement sequences
if isreverse:
if type(seq) is MutableSeq:
seq2 = seq.toseq().reverse_complement()
else:
# print(type(seq1))
seq2 = seq2.reverse_complement()
dt[k,] = seq2
if hdf5_label_file:
label[k] = b.label
k += 1
finally:
if hdf5_file is not None:
hdf5_file.close()
if hdf5_label_file is not None:
hdf5_label_file.close()
def output_boundary(boundaries, ref_genome_file, segment_size, data_file, label_file, isreverse=True):
'''
output boundary data to files
isreverse: output sequence of the reverse strand as well
'''
chrom_list = get_chromlist(ref_genome_file)
hdf5_file = None
hdf5_label_file = None
try:
hdf5_file = h5py.File(data_file, mode='w')
if label_file:
hdf5_label_file = h5py.File(label_file, mode='w')
for b in boundaries:
# print(b.chrom, b.start, b.end)
#print('start: {}, end: {}'.format(b.start, b.end))
seq = mutation_function.infer_true_seq(b, chrom_list[b.chrom], segment_size)
if len(seq) != segment_size:
raise Exception('length of sequences is wrong: %d' % (len(seq)))
output_seq(hdf5_file, b.chrom + "_" + str(b.start) + "_" + str(b.end) + "_" + b.suffix + "_1", seq)
# complement sequences
if isreverse:
if type(seq) is MutableSeq:
seq2 = seq.toseq().reverse_complement()
else:
# print(type(seq1))
seq2 = seq2.reverse_complement()
output_seq(hdf5_file, b.chrom + "_" + str(b.start) + "_" + str(b.end) + "_" + b.suffix + "_2", seq2)
if hdf5_label_file:
hdf5_label_file.create_dataset(b.chrom + "_" + str(b.start) + "_" + str(b.end) + "_" + b.suffix + "_1", data=b.label)
if isreverse:
hdf5_label_file.create_dataset(b.chrom + "_" + str(b.start) + "_" + str(b.end) + "_" + b.suffix + "_2", data=b.label)
finally:
if hdf5_file is not None:
hdf5_file.close()
if hdf5_label_file is not None:
hdf5_label_file.close()
def output_loop_npa(data,
ref_genome_file,
segment_size,
data_file,
label_file,
n_feature=5):
'''
output loop data (with 2 boundaries) to files
data_file: contain IDs of 2 boundaries of loops, boundaries refer to common_data_file for sequence data
lable_file: 0: no loop, 1: loop
'''
chrom_list = get_chromlist(ref_genome_file)
lb = None
hdf5_file = None
hdf5_label_file = None
try:
hdf5_file = h5py.File(data_file, mode='w')
# loop, boundary1 or 2, 4000 x 5
dt = hdf5_file.create_dataset(
'data', (len(data) * 2, 2, segment_size, n_feature),
dtype='b1',
compression="gzip",
compression_opts=5)
if label_file:
hdf5_label_file = h5py.File(label_file, mode='w')
lb = hdf5_label_file.create_dataset('label', (len(data) * 2),
dtype='b1',
compression="gzip",
compression_opts=5)
# len(data) * 2 loops -- including reverse, each has 2 boundaries, each boundary is of size: segment_size * n_feature
#dt = np.zeros((len(data) * 2, segment_size * n_feature * 2))
#lb = np.zeros((len(data) * 2))
k = 0
for loop in data:
# print(b.chrom, b.start, b.end)
b1, b2 = loop.b1, loop.b2
seq1 = mutation_function.infer_true_seq(b1, chrom_list[b1.chrom],
segment_size)
seq2 = mutation_function.infer_true_seq(b2, chrom_list[b2.chrom],
segment_size)
if len(seq1) != segment_size or len(seq2) != segment_size:
raise Exception('length of sequences is wrong: %d, %d' %
(len(seq1), len(seq2)))
seq1_code = line_to_tensor(str(seq1).upper())
seq2_code = line_to_tensor(str(seq2).upper())
dt[k, 0, ] = seq1_code
dt[k, 1, ] = seq2_code
if lb:
lb[k] = loop.label
k += 1
# complement sequences
seq5 = seq1.toseq().reverse_complement(
) if type(seq1) is MutableSeq else seq1.reverse_complement()
seq6 = seq2.toseq().reverse_complement(
) if type(seq2) is MutableSeq else seq2.reverse_complement()
if seq5 is None or seq6 is None:
raise Exception('type sequences is wrong {}'.format(
type(seq1)))
if len(seq5) != segment_size or len(seq6) != segment_size:
raise Exception('length of sequences is wrong: %d, %d' %
(len(seq5), len(seq6)))
seq5_code = line_to_tensor(str(seq5).upper())
seq6_code = line_to_tensor(str(seq6).upper())
dt[k, 0, ] = seq6_code
dt[k, 1, ] = seq5_code
if lb:
lb[k] = loop.label
k += 1
finally:
if hdf5_file:
hdf5_file.close()
if hdf5_label_file:
hdf5_label_file.close()
def output_loop(data, ref_genome_file, segment_size, data_file, label_file):
"""Output loop data (with 2 boundaries) to files
data_file: contain IDs of 2 boundaries of loops, boundaries refer
to common_data_file for sequence data
lable_file: 0: no loop, 1: loop
"""
chrom_list = get_chromlist(ref_genome_file)
hdf5_file = None
hdf5_label_file = None
try:
hdf5_file = h5py.File(data_file, mode='w')
if label_file:
hdf5_label_file = h5py.File(label_file, mode='w')
for loop in data:
# print(b.chrom, b.start, b.end)
b1, b2 = loop.b1, loop.b2
seq1 = mutation_function.infer_true_seq(b1, chrom_list[b1.chrom],
segment_size)
seq2 = mutation_function.infer_true_seq(b2, chrom_list[b2.chrom],
segment_size)
if len(seq1) != segment_size or len(seq2) != segment_size:
raise Exception('length of sequences is wrong: %d, %d' %
(len(seq1), len(seq2)))
seq1_code = line_to_tensor(str(seq1).upper())
seq2_code = line_to_tensor(str(seq2).upper())
seq_forward = np.hstack((seq1_code, seq2_code))
hdf5_file.create_dataset(b1.chrom + "_" + str(b1.start) + "_" +
str(b1.end) + "_" + str(b2.start) + "_" +
str(b2.end) + "_1",
data=seq_forward,
compression="gzip",
compression_opts=5)
# complement sequences
seq5 = seq1.toseq().reverse_complement(
) if type(seq1) is MutableSeq else seq1.reverse_complement()
seq6 = seq2.toseq().reverse_complement(
) if type(seq2) is MutableSeq else seq2.reverse_complement()
if seq5 is None or seq6 is None:
raise Exception('type sequences is wrong {}'.format(
type(seq1)))
if len(seq5) != segment_size or len(seq6) != segment_size:
raise Exception('length of sequences is wrong: %d, %d' %
(len(seq5), len(seq6)))
seq5_code = line_to_tensor(str(seq5).upper())
seq6_code = line_to_tensor(str(seq6).upper())
seq_reverse = np.hstack((seq6_code, seq5_code))
# b.seq = seq
hdf5_file.create_dataset(b1.chrom + "_" + str(b1.start) + "_" +
str(b1.end) + "_" + str(b2.start) + "_" +
str(b2.end) + "_2",
data=seq_reverse,
compression="gzip",
compression_opts=5)
if hdf5_label_file:
hdf5_label_file.create_dataset(
b1.chrom + "_" + str(b1.start) + "_" + str(b1.end) + "_" +
str(b2.start) + "_" + str(b2.end) + "_1",
data=loop.label)
hdf5_label_file.create_dataset(
b1.chrom + "_" + str(b1.start) + "_" + str(b1.end) + "_" +
str(b2.start) + "_" + str(b2.end) + "_2",
data=loop.label)
finally:
if hdf5_file is not None:
hdf5_file.close()
if hdf5_label_file is not None:
hdf5_label_file.close()