def test_run_simulations_and_get_percentile_allele_length_2(self):
"""test run_simulations and get_percentile"""
simulator = genotype_confidence_simulator.GenotypeConfidenceSimulator(
50, 300, 0.1, allele_length=2, iterations=5)
simulator.run_simulations()
expected_confidence_scores_percentiles = {
26: 20.0,
31: 40.0,
37: 60.0,
46: 80.0,
51: 100.0,
}
self.assertEqual(
expected_confidence_scores_percentiles,
simulator.confidence_scores_percentiles,
)
self.assertEqual(20.00, simulator.get_percentile(26))
self.assertEqual(40.00, simulator.get_percentile(31))
# Try getting numbers that are not in the dict and will have to be inferred
self.assertEqual(28.00, simulator.get_percentile(28))
self.assertEqual(84.00, simulator.get_percentile(47))
self.assertEqual(88.00, simulator.get_percentile(48))
# Try values outside the range of what we already have
self.assertEqual(0.00, simulator.get_percentile(25))
self.assertEqual(0.00, simulator.get_percentile(24))
self.assertEqual(100.00, simulator.get_percentile(51))
self.assertEqual(100.00, simulator.get_percentile(52))
def test_add_gt_conf_percentile_and_filters_to_vcf_file():
"""test _add_gt_conf_percentile_and_filters_to_vcf_file"""
original_file = os.path.join(data_dir,
"add_gt_conf_percentile_to_vcf_file.in.vcf")
tmp_file = "tmp.adjudicator.add_gt_conf_percentile_to_vcf_file.vcf"
expect_file = os.path.join(
data_dir, "add_gt_conf_percentile_to_vcf_file.expect.vcf")
shutil.copyfile(original_file, tmp_file)
error_rate = 0.00026045894282438386
simulations = genotype_confidence_simulator.GenotypeConfidenceSimulator(
60,
100,
error_rate,
iterations=1000,
call_hets=False,
)
simulations.run_simulations()
adjudicator.Adjudicator._add_gt_conf_percentile_and_filters_to_vcf_file(
tmp_file, simulations, min_dp=2, min_gcp=2.5, min_frs=0.9)
assert filecmp.cmp(tmp_file, expect_file, shallow=False)
os.unlink(tmp_file)
def run_gt_conf(self):
if self.mean_depth > 0:
simulations = genotype_confidence_simulator.GenotypeConfidenceSimulator(
self.mean_depth,
self.variance_depth,
self.read_error_rate,
allele_length=1,
iterations=self.genotype_simulation_iterations,
call_hets=self.call_hets,
)
simulations.run_simulations(
conf_scores_file=self.sim_conf_scores_file)
else:
simulations = None
logging.info(
f"Adding GT_CONF_PERCENTILE to final VCF file {self.final_vcf} & its debug counterpart, "
f"using mean depth {self.mean_depth}, variance depth {self.variance_depth}, error rate {self.read_error_rate}, "
f"and {self.genotype_simulation_iterations} simulation iterations")
for f in [self.unfiltered_vcf_file, self.final_vcf]:
if self.debug and f == self.final_vcf:
scores_file = self.real_conf_scores_file
else:
scores_file = None
Adjudicator._add_gt_conf_percentile_and_filters_to_vcf_file(
f,
simulations,
min_dp=self.filter_min_dp,
min_gcp=self.filter_min_gcp,
min_frs=self.filter_min_frs,
conf_scores_file=scores_file,
)
if scores_file is not None:
Adjudicator._plot_gt_conf_hists(
scores_file,
self.sim_conf_scores_file,
self.genotype_hist_pdf,
)
def _add_gt_conf_percentile_to_vcf_file(cls, vcf_file, mean_depth,
depth_variance, error_rate,
iterations):
'''Overwrites vcf_file, with new version that has GT_CONF_PERCENTILE added'''
simulations = genotype_confidence_simulator.GenotypeConfidenceSimulator(
mean_depth,
depth_variance,
error_rate,
allele_length=1,
iterations=iterations)
simulations.run_simulations()
vcf_header, vcf_lines = vcf_file_read.vcf_file_to_list(vcf_file)
for i, line in enumerate(vcf_header):
if line.startswith('##FORMAT=<ID=GT_CONF'):
break
else:
raise Exception(
f'No GT_CONF description found in header of VCF file {vcf_file}. Cannot continue'
)
vcf_header.insert(
i + 1,
r'''##FORMAT=<ID=GT_CONF_PERCENTILE,Number=1,Type=Float,Description="Percentile of GT_CONF"'''
)
with open(vcf_file, 'w') as f:
print(*vcf_header, sep='\n', file=f)
for vcf_record in vcf_lines:
if 'GT_CONF' in vcf_record.FORMAT:
conf = int(round(float(vcf_record.FORMAT['GT_CONF'])))
if 'GT' in vcf_record.FORMAT and '.' not in vcf_record.FORMAT[
'GT']:
vcf_record.set_format_key_value(
'GT_CONF_PERCENTILE',
str(simulations.get_percentile(conf)))
print(vcf_record, file=f)
def test_run_simulations_and_get_percentile_allele_length_2():
"""test run_simulations and get_percentile"""
simulator = genotype_confidence_simulator.GenotypeConfidenceSimulator(
50, 300, 0.01, allele_length=2, iterations=5)
simulator.run_simulations()
expected_confidence_scores_percentiles = {
193: 20.0,
221: 40.0,
271: 60.0,
278: 80.0,
303: 100.0
}
assert (simulator.confidence_scores_percentiles ==
expected_confidence_scores_percentiles)
assert simulator.get_percentile(193) == 20.00
assert simulator.get_percentile(221) == 40.00
# Try getting number that is not in the dict and will have to be inferred
assert simulator.get_percentile(207) == 30.0
# Try values outside the range of what we already have
simulator.get_percentile(192) == 0.00
simulator.get_percentile(191) == 0.00
simulator.get_percentile(304) == 100.00
simulator.get_percentile(305) == 100.00
def _add_gt_conf_percentile_and_filters_to_vcf_file(
cls,
vcf_file,
mean_depth,
depth_variance,
error_rate,
iterations,
min_dp=2,
min_gt_conf_percentile=2.5):
'''Overwrites vcf_file, with new version that has GT_CONF_PERCENTILE added,
and filter for DP and GT_CONF_PERCENTILE'''
simulations = genotype_confidence_simulator.GenotypeConfidenceSimulator(
mean_depth,
depth_variance,
error_rate,
allele_length=1,
iterations=iterations)
simulations.run_simulations()
vcf_header, vcf_lines = vcf_file_read.vcf_file_to_list(vcf_file)
for i, line in enumerate(vcf_header):
if line.startswith('##FORMAT=<ID=GT_CONF'):
break
else:
raise Exception(
f'No GT_CONF description found in header of VCF file {vcf_file}. Cannot continue'
)
vcf_header.insert(
i + 1,
r'''##FORMAT=<ID=GT_CONF_PERCENTILE,Number=1,Type=Float,Description="Percentile of GT_CONF">'''
)
vcf_header.insert(
i + 1,
f'##FILTER=<ID=MIN_DP,Description="Minimum DP of {min_dp}">')
vcf_header.insert(
i + 1,
f'##FILTER=<ID=MIN_GCP,Description="Minimum GT_CONF_PERCENTILE of {min_gt_conf_percentile}">'
)
with open(vcf_file, 'w') as f:
print(*vcf_header, sep='\n', file=f)
for vcf_record in vcf_lines:
vcf_record.FILTER = set()
if 'GT_CONF' in vcf_record.FORMAT:
conf = int(round(float(vcf_record.FORMAT['GT_CONF'])))
if 'GT' in vcf_record.FORMAT and '.' not in vcf_record.FORMAT[
'GT']:
vcf_record.set_format_key_value(
'GT_CONF_PERCENTILE',
str(simulations.get_percentile(conf)))
if 'DP' in vcf_record.FORMAT and float(
vcf_record.FORMAT['DP']) < min_dp:
vcf_record.FILTER.add('MIN_DP')
if float(vcf_record.FORMAT['GT_CONF_PERCENTILE']
) < min_gt_conf_percentile:
vcf_record.FILTER.add('MIN_GCP')
if len(vcf_record.FILTER) == 0:
vcf_record.FILTER.add('PASS')
print(vcf_record, file=f)
def _add_gt_conf_percentile_and_filters_to_vcf_file(
cls,
vcf_file,
mean_depth,
depth_variance,
error_rate,
iterations,
min_dp=5,
min_gcp=5,
):
"""Overwrites vcf_file, with new version that has GT_CONF_PERCENTILE added,
and filter for DP and GT_CONF_PERCENTILE"""
if mean_depth > 0:
simulations = genotype_confidence_simulator.GenotypeConfidenceSimulator(
mean_depth,
depth_variance,
error_rate,
allele_length=1,
iterations=iterations,
)
simulations.run_simulations()
vcf_header, vcf_lines = vcf_file_read.vcf_file_to_list(vcf_file)
for i, line in enumerate(vcf_header):
if line.startswith("##FORMAT=<ID=GT_CONF"):
break
else:
raise Exception(
f"No GT_CONF description found in header of VCF file {vcf_file}. Cannot continue"
)
vcf_header.insert(
i + 1,
r"""##FORMAT=<ID=GT_CONF_PERCENTILE,Number=1,Type=Float,Description="Percentile of GT_CONF">""",
)
vcf_header.insert(
i + 1,
f'##FILTER=<ID=MIN_DP,Description="Minimum DP of {min_dp}">')
vcf_header.insert(
i + 1,
f'##FILTER=<ID=MIN_GCP,Description="Minimum GT_CONF_PERCENTILE of {min_gcp}">',
)
with open(vcf_file, "w") as f:
print(*vcf_header, sep="\n", file=f)
for vcf_record in vcf_lines:
vcf_record.FILTER = set()
if "GT" in vcf_record.FORMAT and "GT_CONF" in vcf_record.FORMAT:
if "." not in vcf_record.FORMAT["GT"]:
conf = int(round(float(vcf_record.FORMAT["GT_CONF"])))
vcf_record.set_format_key_value(
"GT_CONF_PERCENTILE",
str(simulations.get_percentile(conf)))
if ("DP" in vcf_record.FORMAT
and float(vcf_record.FORMAT["DP"]) < min_dp):
vcf_record.FILTER.add("MIN_DP")
if float(vcf_record.FORMAT["GT_CONF_PERCENTILE"]
) < min_gcp:
vcf_record.FILTER.add("MIN_GCP")
if len(vcf_record.FILTER) == 0:
vcf_record.FILTER.add("PASS")
else:
# Add a default null percentile
vcf_record.set_format_key_value(
"GT_CONF_PERCENTILE", "0.0")
print(vcf_record, file=f)