def get_ru_count_with_coverage_method(self, pattern_occurrences,
total_counted_vntr_bp,
average_coverage):
haplotypes = 1 if self.is_haploid else 2
estimate = [
int(pattern_occurrences / (float(average_coverage) * haplotypes))
] * 2
return estimate
pattern_occurrences = total_counted_vntr_bp / float(
len(self.reference_vntr.pattern))
read_mode = 'r' if alignment_file.endswith('sam') else 'rb'
samfile = pysam.AlignmentFile(alignment_file, read_mode)
reference = get_reference_genome_of_alignment_file(samfile)
bias_detector = CoverageBiasDetector(alignment_file,
self.reference_vntr.chromosome,
reference)
coverage_corrector = CoverageCorrector(
bias_detector.get_gc_content_coverage_map())
logging.info('Sequencing mean coverage: %s' %
coverage_corrector.get_sequencing_mean_coverage())
observed_copy_number = pattern_occurrences / coverage_corrector.get_sequencing_mean_coverage(
)
scaled_copy_number = coverage_corrector.get_scaled_coverage(
self.reference_vntr, observed_copy_number)
logging.info('scaled copy number and observed copy number: %s, %s' %
(scaled_copy_number, observed_copy_number))
return [scaled_copy_number]
def get_spanning_reads_of_aligned_pacbio_reads(self, alignment_file):
sema = Semaphore(settings.CORES)
manager = Manager()
length_distribution = manager.list()
mapped_spanning_reads = manager.list()
vntr_start = self.reference_vntr.start_point
vntr_end = self.reference_vntr.start_point + self.reference_vntr.get_length(
)
region_start = vntr_start
region_end = vntr_end
read_mode = 'r' if alignment_file.endswith('sam') else 'rb'
samfile = pysam.AlignmentFile(alignment_file, read_mode)
reference = get_reference_genome_of_alignment_file(samfile)
chromosome = self.reference_vntr.chromosome if reference == 'HG19' else self.reference_vntr.chromosome[
3:]
process_list = []
for read in samfile.fetch(chromosome, region_start, region_end):
sema.acquire()
p = Process(target=self.check_if_pacbio_read_spans_vntr,
args=(sema, read, length_distribution,
mapped_spanning_reads))
process_list.append(p)
p.start()
for p in process_list:
p.join()
logging.info('length_distribution of mapped spanning reads: %s' %
list(length_distribution))
return list(mapped_spanning_reads)
def select_illumina_reads(self,
alignment_file,
unmapped_filtered_reads,
update=False,
hmm=None):
recruitment_score = None
sema = Semaphore(settings.CORES)
manager = Manager()
selected_reads = manager.list()
vntr_bp_in_unmapped_reads = Value('d', 0.0)
number_of_reads = 0
read_length = 150
process_list = []
for read_segment in unmapped_filtered_reads:
if number_of_reads == 0:
read_length = len(str(read_segment.seq))
number_of_reads += 1
if not hmm:
hmm = self.get_vntr_matcher_hmm(read_length=read_length)
if not recruitment_score:
recruitment_score = self.get_min_score_to_select_a_read(
read_length)
if len(read_segment.seq) < read_length:
continue
sema.acquire()
p = Process(target=self.process_unmapped_read,
args=(sema, str(read_segment.seq), hmm,
recruitment_score, vntr_bp_in_unmapped_reads,
selected_reads))
process_list.append(p)
p.start()
for p in process_list:
p.join()
logging.debug('vntr base pairs in unmapped reads: %s' %
vntr_bp_in_unmapped_reads.value)
vntr_bp_in_mapped_reads = 0
vntr_start = self.reference_vntr.start_point
vntr_end = self.reference_vntr.start_point + self.reference_vntr.get_length(
)
read_mode = 'r' if alignment_file.endswith('sam') else 'rb'
samfile = pysam.AlignmentFile(alignment_file, read_mode)
reference = get_reference_genome_of_alignment_file(samfile)
chromosome = self.reference_vntr.chromosome if reference == 'HG19' else self.reference_vntr.chromosome[
3:]
for read in samfile.fetch(chromosome, vntr_start, vntr_end):
if not recruitment_score:
read_length = len(read.seq)
recruitment_score = self.get_min_score_to_select_a_read(
read_length)
if not hmm:
hmm = self.get_vntr_matcher_hmm(read_length=read_length)
if read.is_unmapped:
continue
if len(read.seq) < int(read_length * 0.9):
logging.debug('Rejecting read for short length: %s' % read.seq)
continue
read_end = read.reference_end if read.reference_end else read.reference_start + len(
read.seq)
if vntr_start - read_length < read.reference_start < vntr_end or vntr_start < read_end < vntr_end:
if read.seq.count('N') <= 0:
sequence = str(read.seq).upper()
logp, vpath = hmm.viterbi(sequence)
rev_logp, rev_vpath = hmm.viterbi(
str(Seq(read.seq).reverse_complement()).upper())
if logp < rev_logp:
sequence = str(Seq(
read.seq).reverse_complement()).upper()
logp = rev_logp
vpath = rev_vpath
length = len(sequence)
if is_low_quality_read(read) and not self.recruit_read(
logp, vpath, recruitment_score, length):
logging.debug('Rejected Read: %s' % sequence)
continue
selected_reads.append(
SelectedRead(sequence, logp, vpath, read.mapq,
read.reference_start))
end = min(read_end, vntr_end)
start = max(read.reference_start, vntr_start)
vntr_bp_in_mapped_reads += end - start
logging.debug('vntr base pairs in mapped reads: %s' %
vntr_bp_in_mapped_reads)
if update:
selected_reads = self.iteratively_update_model(
alignment_file, unmapped_filtered_reads, selected_reads, hmm)
return selected_reads
def select_illumina_reads(self,
alignment_file,
unmapped_filtered_reads,
update=False,
hmm=None):
recruitment_score = None
dnn_model = None
selected_reads = []
vntr_bp_in_unmapped_reads = Value('d', 0.0)
number_of_reads = 0
read_length = 150
for read_segment in unmapped_filtered_reads:
if number_of_reads == 0:
read_length = len(str(read_segment.seq))
number_of_reads += 1
if not hmm:
hmm = self.get_vntr_matcher_hmm(read_length=read_length)
if not recruitment_score:
recruitment_score = self.get_min_score_to_select_a_read(
read_length)
model_file = settings.DNN_MODELS_DIR + '/%s.hd5' % self.reference_vntr.id
if os.path.exists(model_file):
dnn_model = load_model(model_file)
if len(read_segment.seq) < read_length:
continue
if dnn_model is None:
self.process_unmapped_read(None, str(read_segment.seq), hmm,
recruitment_score,
vntr_bp_in_unmapped_reads,
selected_reads)
else:
self.process_unmapped_read_with_dnn(str(read_segment.seq), hmm,
recruitment_score,
vntr_bp_in_unmapped_reads,
selected_reads, True,
dnn_model)
logging.debug('vntr base pairs in unmapped reads: %s' %
vntr_bp_in_unmapped_reads.value)
vntr_bp_in_mapped_reads = 0
vntr_start = self.reference_vntr.start_point
vntr_end = self.reference_vntr.start_point + self.reference_vntr.get_length(
)
read_mode = self.get_alignment_file_read_mode(alignment_file)
samfile = pysam.AlignmentFile(
alignment_file,
read_mode,
reference_filename=self.reference_filename)
reference = get_reference_genome_of_alignment_file(samfile)
chromosome = self.reference_vntr.chromosome if reference == 'HG19' else self.reference_vntr.chromosome[
3:]
for read in samfile.fetch(chromosome, vntr_start, vntr_end):
if not recruitment_score:
read_length = len(read.seq)
recruitment_score = self.get_min_score_to_select_a_read(
read_length)
if not hmm:
hmm = self.get_vntr_matcher_hmm(read_length=read_length)
if read.is_unmapped:
continue
if len(read.seq) < int(read_length * 0.9):
logging.debug('Rejecting read for short length: %s' % read.seq)
continue
read_end = read.reference_end if read.reference_end else read.reference_start + len(
read.seq)
if vntr_start - read_length < read.reference_start < vntr_end or vntr_start < read_end < vntr_end:
if read.seq.count('N') <= 0:
sequence = str(read.seq).upper()
logp, vpath = hmm.viterbi(sequence)
rev_logp, rev_vpath = hmm.viterbi(
str(Seq(read.seq).reverse_complement()).upper())
if logp < rev_logp:
sequence = str(Seq(
read.seq).reverse_complement()).upper()
logp = rev_logp
vpath = rev_vpath
if is_low_quality_read(read) or not self.recruit_read(
logp, vpath, recruitment_score, sequence):
logging.debug('Rejected Aligned Read: %s' % sequence)
continue
selected_reads.append(
SelectedRead(sequence, logp, vpath, read.mapq,
read.reference_start))
end = min(read_end, vntr_end)
start = max(read.reference_start, vntr_start)
vntr_bp_in_mapped_reads += end - start
logging.debug('vntr base pairs in mapped reads: %s' %
vntr_bp_in_mapped_reads)
if update:
selected_reads = self.iteratively_update_model(
alignment_file, unmapped_filtered_reads, selected_reads, hmm)
return selected_reads