def load_unique_vntrs_data():
vntrs = []
db_file = settings.TRAINED_MODELS_DB
db = sqlite3.connect(db_file)
cursor = db.cursor()
cursor.execute(
'''SELECT id, nonoverlapping, chromosome, ref_start, gene_name, annotation, pattern, left_flanking,
right_flanking, repeats, scaled_score FROM vntrs''')
for row in cursor:
new_row = []
for element in row:
if type(element) != int and type(element) != float:
new_row.append(str(element))
else:
new_row.append(element)
vntr_id, overlap, chrom, start, gene, annotation, pattern, left_flank, right_flank, segments, score = new_row
repeat_segments = segments.split(',')
repeats = len(repeat_segments)
vntr = ReferenceVNTR(int(vntr_id),
pattern,
int(start),
chrom,
gene,
annotation,
repeats,
scaled_score=score)
vntr.init_from_xml(repeat_segments, left_flank, right_flank)
vntr.non_overlapping = True if overlap == 'True' else False
vntrs.append(vntr)
return vntrs
def load_unprocessed_vntrseek_data(vntrseek_output, chromosome=None):
vntrs = []
genes_info = get_genes_info()
with open(vntrseek_output) as input_file:
input_lines = [
line.strip() for line in input_file.readlines()
if line.strip() != ''
]
for vntr_id, line in enumerate(input_lines):
vntrseek_repeat, _, pattern, chromosome_number, start = line.split(
)
if len(pattern) > 100:
continue
start = int(start) - 1
estimated_repeats = int(float(vntrseek_repeat) + 5)
if chromosome is not None and chromosome_number != chromosome:
continue
estimated_end = estimated_repeats * len(pattern) + start
if not is_vntr_close_to_gene(genes_info, chromosome_number, start,
estimated_end):
continue
vntrs.append(
ReferenceVNTR(vntr_id, pattern, start, chromosome_number, None,
None, estimated_repeats))
print('%s VNTRs are close to a gene' % len(vntrs))
return vntrs
def add_model(args, addmodel_parser):
if not args.reference:
print_error(addmodel_parser, '--reference is required')
if not args.chromosome:
print_error(addmodel_parser, '--chromosome is required')
if not args.pattern:
print_error(addmodel_parser, '--pattern is required')
if not args.start:
print_error(addmodel_parser, '--start is required')
if not args.end:
print_error(addmodel_parser, '--end is required')
chromosome = args.chromosome
chr_sequence = ''
fasta_sequences = SeqIO.parse(open(args.reference), 'fasta')
for fasta in fasta_sequences:
name, sequence = fasta.id, str(fasta.seq)
if name == chromosome:
chr_sequence = sequence
if args.models is not None:
settings.TRAINED_MODELS_DB = args.models
if not os.path.exists(settings.TRAINED_MODELS_DB):
create_vntrs_database(settings.TRAINED_MODELS_DB)
vntr_id = get_largest_id_in_database() + 1
estimated_repeats = int((args.end - args.start) / len(args.pattern) + 5)
ref_vntr = ReferenceVNTR(vntr_id, args.pattern, args.start, chromosome,
None, None, estimated_repeats, chr_sequence)
ref_vntr.init_from_vntrseek_data()
vntr_finder = VNTRFinder(ref_vntr)
print(
'Searching reference genome for regions with shared kmers with VNTR. It takes a few hours for human genome'
)
scaled_recruitment_score = vntr_finder.train_classifier_threshold(
args.reference)
ref_vntr.scaled_score = scaled_recruitment_score
save_reference_vntr_to_database(ref_vntr)
print('Training completed. VNTR saved with ID: %s to the database' %
vntr_id)
def get_reference_vntr(self, ru_count=10):
pattern = 'ACGTACGT'
ref_vntr = ReferenceVNTR(1, pattern, 1000, 'chr1', None, None)
ref_vntr.repeat_segments = [pattern] * ru_count
return ref_vntr
def get_reference_vntr(self):
ref_vntr = ReferenceVNTR(1, 'CACA', 1000, 'chr1', None, None)
return ref_vntr