def split_fasta(input_fasta,
output_dir,
num_of_recs_per_file=None,
num_of_files=None,
output_prefix=None):
"""
by default splits input files into files with num_of_recs_per_file.
if num_of_files is set num_of_recs_per_file is ignored.
"""
FileRoutines.save_mkdir(output_dir)
out_prefix = FileRoutines.split_filename(
input_fasta)[1] if output_prefix is None else output_prefix
sequence_dict = SeqIO.index_db("temp.idx", input_fasta, "fasta")
split_index = 1
records_written = 0
record_ids_list = list(sequence_dict.keys())
number_of_records = len(record_ids_list)
num_of_recs = int(
number_of_records /
num_of_files) + 1 if num_of_files else num_of_recs_per_file
while (records_written + num_of_recs) <= number_of_records:
SeqIO.write(SequenceRoutines.record_by_id_generator(
sequence_dict,
record_ids_list[records_written:records_written +
num_of_recs]),
"%s/%s_%i.fasta" %
(output_dir, out_prefix, split_index),
format="fasta")
split_index += 1
records_written += num_of_recs
if records_written != number_of_records:
SeqIO.write(SequenceRoutines.record_by_id_generator(
sequence_dict, record_ids_list[records_written:]),
"%s/%s_%i.fasta" %
(output_dir, out_prefix, split_index),
format="fasta")
os.remove("temp.idx")
def split_fasta_by_seq_len(self,
input_fasta,
output_dir,
max_len_per_file=None,
output_prefix=None):
"""
by default splits input files into files with num_of_recs_per_file.
if num_of_files is set num_of_recs_per_file is ignored.
"""
FileRoutines.save_mkdir(output_dir)
out_prefix = FileRoutines.split_filename(
input_fasta)[1] if output_prefix is None else output_prefix
sequence_dict = SeqIO.index_db("temp.idx", input_fasta, "fasta")
length = 0
for record_id in sequence_dict:
length += len(sequence_dict[record_id].seq)
max_len = max_len_per_file if max_len_per_file else int(length /
self.threads)
split_index = 1
id_list = []
total_length = 0
for record_id in sequence_dict:
record_length = len(sequence_dict[record_id].seq)
if record_length >= max_len:
SeqIO.write(sequence_dict[record_id],
"%s/%s_%i.fasta" %
(output_dir, out_prefix, split_index),
format="fasta")
elif total_length + record_length > max_len:
SeqIO.write(SequenceRoutines.record_by_id_generator(
sequence_dict, id_list),
"%s/%s_%i.fasta" %
(output_dir, out_prefix, split_index),
format="fasta")
total_length = record_length
id_list = [record_id]
elif total_length + record_length == max_len:
id_list.append(record_id)
SeqIO.write(SequenceRoutines.record_by_id_generator(
sequence_dict, id_list),
"%s/%s_%i.fasta" %
(output_dir, out_prefix, split_index),
format="fasta")
total_length = 0
id_list = []
elif total_length + record_length < max_len:
id_list.append(record_id)
total_length += record_length
continue
split_index += 1
if id_list:
SeqIO.write(SequenceRoutines.record_by_id_generator(
sequence_dict, id_list),
"%s/%s_%i.fasta" %
(output_dir, out_prefix, split_index),
format="fasta")
os.remove("temp.idx")