def test_parse_rank_score():
rank_scores_info = "123:10"
variant_score = 10.0
family_id = '123'
parsed_rank_score = parse_rank_score(rank_scores_info, family_id)
assert variant_score == parsed_rank_score
def test_parse_rank_score_no_score():
## GIVEN a empty rank score string
rank_scores_info = ""
family_id = "123"
## WHEN parsing the rank score
parsed_rank_score = parse_rank_score(rank_scores_info, family_id)
## THEN assert that None is returned
assert parsed_rank_score == None
def test_parse_rank_score():
rank_scores_info = "123:10"
variant_score = 10.0
family_id = '123'
parsed_rank_score = parse_rank_score(rank_scores_info, family_id)
assert variant_score == parsed_rank_score
def test_parse_rank_score():
## GIVEN a rank score string on genmod format
rank_scores_info = "123:10"
variant_score = 10.0
family_id = "123"
## WHEN parsing the rank score
parsed_rank_score = parse_rank_score(rank_scores_info, family_id)
## THEN assert that the correct rank score is parsed
assert variant_score == parsed_rank_score
def test_parse_rank_score_no_score():
rank_scores_info = ""
family_id = '123'
parsed_rank_score = parse_rank_score(rank_scores_info, family_id)
assert parsed_rank_score == None
# def test_parse_rank_scores(variants, parsed_case):
# """docstring for test_parse_rank_score"""
# case_id = parsed_case['display_name']
# for variant in variants:
# rank_scores_dict = variant['rank_scores']
#
# rank_score = rank_scores_dict[case_id]
#
# assert float(rank_score) == parse_rank_score(variant, case_id)
def test_parse_rank_score_no_score():
rank_scores_info = ""
family_id = '123'
parsed_rank_score = parse_rank_score(rank_scores_info, family_id)
assert parsed_rank_score == None
# def test_parse_rank_scores(variants, parsed_case):
# """docstring for test_parse_rank_score"""
# case_id = parsed_case['display_name']
# for variant in variants:
# rank_scores_dict = variant['rank_scores']
#
# rank_score = rank_scores_dict[case_id]
#
# assert float(rank_score) == parse_rank_score(variant, case_id)
def _load_variants(
self,
variants,
variant_type,
case_obj,
individual_positions,
rank_threshold,
institute_id,
build=None,
rank_results_header=None,
vep_header=None,
category="snv",
sample_info=None,
):
"""Perform the loading of variants
This is the function that loops over the variants, parse them and build the variant
objects so they are ready to be inserted into the database.
Args:
variants(iterable(cyvcf2.Variant))
variant_type(str): ['clinical', 'research']
case_obj(dict)
individual_positions(dict): How individuals are positioned in vcf
rank_treshold(int): Only load variants with a rank score > than this
institute_id(str)
build(str): Genome build
rank_results_header(list): Rank score categories
vep_header(list)
category(str): ['snv','sv','cancer','str']
sample_info(dict): A dictionary with info about samples.
Strictly for cancer to tell which is tumor
Returns:
nr_inserted(int)
"""
build = build or "37"
genes = [gene_obj for gene_obj in self.all_genes(build=build)]
gene_to_panels = self.gene_to_panels(case_obj)
hgncid_to_gene = self.hgncid_to_gene(genes=genes, build=build)
genomic_intervals = self.get_coding_intervals(genes=genes)
LOG.info("Start inserting {0} {1} variants into database".format(
variant_type, category))
start_insertion = datetime.now()
start_five_thousand = datetime.now()
# These are the number of parsed varaints
nr_variants = 0
# These are the number of variants that meet the criteria and gets inserted
nr_inserted = 0
# This is to keep track of blocks of inserted variants
inserted = 1
nr_bulks = 0
# We want to load batches of variants to reduce the number of network round trips
bulk = {}
current_region = None
for nr_variants, variant in enumerate(variants):
# All MT variants are loaded
mt_variant = "MT" in variant.CHROM
rank_score = parse_rank_score(variant.INFO.get("RankScore"),
case_obj["_id"])
pathogenic = is_pathogenic(variant)
# Check if the variant should be loaded at all
# if rank score is None means there are no rank scores annotated, all variants will be loaded
# Otherwise we load all variants above a rank score treshold
# Except for MT variants where we load all variants
if ((rank_score is None) or (rank_score > rank_threshold)
or mt_variant or pathogenic):
nr_inserted += 1
# Parse the vcf variant
parsed_variant = parse_variant(
variant=variant,
case=case_obj,
variant_type=variant_type,
rank_results_header=rank_results_header,
vep_header=vep_header,
individual_positions=individual_positions,
category=category,
)
# Build the variant object
variant_obj = build_variant(
variant=parsed_variant,
institute_id=institute_id,
gene_to_panels=gene_to_panels,
hgncid_to_gene=hgncid_to_gene,
sample_info=sample_info,
)
# Check if the variant is in a genomic region
var_chrom = variant_obj["chromosome"]
var_start = variant_obj["position"]
# We need to make sure that the interval has a length > 0
var_end = variant_obj["end"] + 1
var_id = variant_obj["_id"]
# If the bulk should be loaded or not
load = True
new_region = None
intervals = genomic_intervals.get(var_chrom, IntervalTree())
genomic_regions = intervals.overlap(var_start, var_end)
# If the variant is in a coding region
if genomic_regions:
# We know there is data here so get the interval id
new_region = genomic_regions.pop().data
# If the variant is in the same region as previous
# we add it to the same bulk
if new_region == current_region:
load = False
# This is the case where the variant is intergenic
else:
# If the previous variant was also intergenic we add the variant to the bulk
if not current_region:
load = False
# We need to have a max size of the bulk
if len(bulk) > 10000:
load = True
# Load the variant object
if load:
# If the variant bulk contains coding variants we want to update the compounds
if current_region:
self.update_compounds(bulk)
try:
# Load the variants
self.load_variant_bulk(list(bulk.values()))
nr_bulks += 1
except IntegrityError as error:
pass
bulk = {}
current_region = new_region
bulk[var_id] = variant_obj
if nr_variants != 0 and nr_variants % 5000 == 0:
LOG.info("%s variants parsed", str(nr_variants))
LOG.info(
"Time to parse variants: %s",
(datetime.now() - start_five_thousand),
)
start_five_thousand = datetime.now()
if (nr_inserted != 0
and (nr_inserted * inserted) % (1000 * inserted) == 0):
LOG.info("%s variants inserted", nr_inserted)
inserted += 1
# If the variants are in a coding region we update the compounds
if current_region:
self.update_compounds(bulk)
# Load the final variant bulk
self.load_variant_bulk(list(bulk.values()))
nr_bulks += 1
LOG.info("All variants inserted, time to insert variants: {0}".format(
datetime.now() - start_insertion))
if nr_variants:
nr_variants += 1
LOG.info("Nr variants parsed: %s", nr_variants)
LOG.info("Nr variants inserted: %s", nr_inserted)
LOG.debug("Nr bulks inserted: %s", nr_bulks)
return nr_inserted
def _load_variants(self, variants, variant_type, case_obj, individual_positions, rank_threshold,
institute_id, build=None, rank_results_header=None, vep_header=None,
category='snv', sample_info = None):
"""Perform the loading of variants
This is the function that loops over the variants, parse them and build the variant
objects so they are ready to be inserted into the database.
"""
build = build or '37'
genes = [gene_obj for gene_obj in self.all_genes(build=build)]
gene_to_panels = self.gene_to_panels(case_obj)
hgncid_to_gene = self.hgncid_to_gene(genes=genes)
genomic_intervals = self.get_coding_intervals(genes=genes)
LOG.info("Start inserting {0} {1} variants into database".format(variant_type, category))
start_insertion = datetime.now()
start_five_thousand = datetime.now()
# These are the number of parsed varaints
nr_variants = 0
# These are the number of variants that meet the criteria and gets inserted
nr_inserted = 0
# This is to keep track of blocks of inserted variants
inserted = 1
nr_bulks = 0
# We want to load batches of variants to reduce the number of network round trips
bulk = {}
current_region = None
for nr_variants, variant in enumerate(variants):
# All MT variants are loaded
mt_variant = 'MT' in variant.CHROM
rank_score = parse_rank_score(variant.INFO.get('RankScore'), case_obj['_id'])
# Check if the variant should be loaded at all
# if rank score is None means there are no rank scores annotated, all variants will be loaded
# Otherwise we load all variants above a rank score treshold
# Except for MT variants where we load all variants
if (rank_score is None) or (rank_score > rank_threshold) or mt_variant:
nr_inserted += 1
# Parse the vcf variant
parsed_variant = parse_variant(
variant=variant,
case=case_obj,
variant_type=variant_type,
rank_results_header=rank_results_header,
vep_header=vep_header,
individual_positions=individual_positions,
category=category,
)
# Build the variant object
variant_obj = build_variant(
variant=parsed_variant,
institute_id=institute_id,
gene_to_panels=gene_to_panels,
hgncid_to_gene=hgncid_to_gene,
sample_info=sample_info
)
# Check if the variant is in a genomic region
var_chrom = variant_obj['chromosome']
var_start = variant_obj['position']
# We need to make sure that the interval has a length > 0
var_end = variant_obj['end'] + 1
var_id = variant_obj['_id']
# If the bulk should be loaded or not
load = True
new_region = None
genomic_regions = genomic_intervals.get(var_chrom, IntervalTree()).search(var_start, var_end)
# If the variant is in a coding region
if genomic_regions:
# We know there is data here so get the interval id
new_region = genomic_regions.pop().data
# If the variant is in the same region as previous
# we add it to the same bulk
if new_region == current_region:
load = False
# This is the case where the variant is intergenic
else:
# If the previous variant was also intergenic we add the variant to the bulk
if not current_region:
load = False
# We need to have a max size of the bulk
if len(bulk) > 10000:
load = True
# Load the variant object
if load:
# If the variant bulk contains coding variants we want to update the compounds
if current_region:
self.update_compounds(bulk)
try:
# Load the variants
self.load_variant_bulk(list(bulk.values()))
nr_bulks += 1
except IntegrityError as error:
pass
bulk = {}
current_region = new_region
bulk[var_id] = variant_obj
if (nr_variants != 0 and nr_variants % 5000 == 0):
LOG.info("%s variants parsed", str(nr_variants))
LOG.info("Time to parse variants: %s",
(datetime.now() - start_five_thousand))
start_five_thousand = datetime.now()
if (nr_inserted != 0 and (nr_inserted * inserted) % (1000 * inserted) == 0):
LOG.info("%s variants inserted", nr_inserted)
inserted += 1
# If the variants are in a coding region we update the compounds
if current_region:
self.update_compounds(bulk)
# Load the final variant bulk
self.load_variant_bulk(list(bulk.values()))
nr_bulks += 1
LOG.info("All variants inserted, time to insert variants: {0}".format(
datetime.now() - start_insertion))
if nr_variants:
nr_variants += 1
LOG.info("Nr variants parsed: %s", nr_variants)
LOG.info("Nr variants inserted: %s", nr_inserted)
LOG.debug("Nr bulks inserted: %s", nr_bulks)
return nr_inserted
def load_variants(adapter, variant_file, case_obj, variant_type='clinical',
category='snv', rank_threshold=5, chrom=None, start=None,
end=None):
"""Load all variant in variants
Args:
adapter(MongoAdapter)
variant_file(str): Path to variant file
case(Case)
variant_type(str)
category(str): 'snv' or 'sv'
rank_threshold(int)
chrom(str)
start(int)
end(int)
"""
institute_obj = adapter.institute(institute_id=case_obj['owner'])
if not institute_obj:
raise IntegrityError("Institute {0} does not exist in"
" database.".format(case_obj['owner']))
gene_to_panels = adapter.gene_to_panels()
hgncid_to_gene = adapter.hgncid_to_gene()
coordinates = {}
vcf_obj = VCF(variant_file)
rank_results_header = parse_rank_results_header(vcf_obj)
vep_header = parse_vep_header(vcf_obj)
# This is a dictionary to tell where ind are in vcf
individual_positions = {}
for i,ind in enumerate(vcf_obj.samples):
individual_positions[ind] = i
logger.info("Start inserting variants into database")
start_insertion = datetime.now()
start_five_thousand = datetime.now()
nr_variants = 0
nr_inserted = 0
inserted = 1
coordinates = False
if chrom:
coordinates = {
'chrom': chrom,
'start': start,
'end': end
}
try:
for nr_variants, variant in enumerate(vcf_obj):
rank_score = parse_rank_score(
variant.INFO.get('RankScore'),
case_obj['display_name']
)
variant_obj = None
add_variant = False
if coordinates or (rank_score > rank_threshold):
parsed_variant = parse_variant(
variant=variant,
case=case_obj,
variant_type=variant_type,
rank_results_header=rank_results_header,
vep_header = vep_header,
individual_positions = individual_positions
)
add_variant = True
# If there are coordinates the variant should be loaded
if coordinates:
if not check_coordinates(parsed_variant, coordinates):
add_variant = False
if add_variant:
variant_obj = build_variant(
variant=parsed_variant,
institute_id=institute_obj['_id'],
gene_to_panels=gene_to_panels,
hgncid_to_gene=hgncid_to_gene,
)
try:
load_variant(adapter, variant_obj)
nr_inserted += 1
except IntegrityError as error:
pass
if (nr_variants != 0 and nr_variants % 5000 == 0):
logger.info("%s variants parsed" % str(nr_variants))
logger.info("Time to parse variants: {} ".format(
datetime.now() - start_five_thousand))
start_five_thousand = datetime.now()
if (nr_inserted != 0 and (nr_inserted * inserted) % (1000 * inserted) == 0):
logger.info("%s variants inserted" % nr_inserted)
inserted += 1
except Exception as error:
if not coordinates:
logger.warning("Deleting inserted variants")
delete_variants(adapter, case_obj, variant_type)
raise error
logger.info("All variants inserted.")
logger.info("Number of variants in file: {0}".format(nr_variants + 1))
logger.info("Number of variants inserted: {0}".format(nr_inserted))
logger.info("Time to insert variants:{0}".format(datetime.now() - start_insertion))
def load_variants(self, case_obj, variant_type='clinical', category='snv',
rank_threshold=None, chrom=None, start=None, end=None,
gene_obj=None):
"""Load variants for a case into scout.
Load all variants for a specific analysis type and category into scout.
If no region is specified, load all variants above rank score threshold
If region or gene is specified, load all variants from that region
disregarding variant rank(if not specified)
Args:
case_obj(dict): A case from the scout database
variant_type(str): 'clinical' or 'research'. Default: 'clinical'
category(str): 'snv' or 'sv'. Default: 'snv'
rank_threshold(float): Only load variants above this score. Default: 5
chrom(str): Load variants from a certain chromosome
start(int): Specify the start position
end(int): Specify the end position
gene_obj(dict): A gene object from the database
Returns:
nr_inserted(int)
"""
institute_obj = self.institute(institute_id=case_obj['owner'])
if not institute_obj:
raise IntegrityError("Institute {0} does not exist in"
" database.".format(case_obj['owner']))
gene_to_panels = self.gene_to_panels()
hgncid_to_gene = self.hgncid_to_gene()
variant_file = None
if variant_type == 'clinical':
if category == 'snv':
variant_file = case_obj['vcf_files'].get('vcf_snv')
elif category == 'sv':
variant_file = case_obj['vcf_files'].get('vcf_sv')
elif variant_type == 'research':
if category == 'snv':
variant_file = case_obj['vcf_files'].get('vcf_snv_research')
elif category == 'sv':
variant_file = case_obj['vcf_files'].get('vcf_sv_research')
if not variant_file:
raise SyntaxError("Vcf file does not seem to exist")
vcf_obj = VCF(variant_file)
# Parse the neccessary headers from vcf file
rank_results_header = parse_rank_results_header(vcf_obj)
vep_header = parse_vep_header(vcf_obj)
# This is a dictionary to tell where ind are in vcf
individual_positions = {}
for i,ind in enumerate(vcf_obj.samples):
individual_positions[ind] = i
# Check if a region scould be uploaded
region = ""
if gene_obj:
chrom = gene_obj['chromosome']
start = gene_obj['start']
end = gene_obj['end']
if chrom:
rank_threshold = rank_threshold or -100
if not (start and end):
raise SyntaxError("Specify chrom start and end")
region = "{0}:{1}-{2}".format(chrom, start, end)
else:
rank_threshold = rank_threshold or 5
logger.info("Start inserting variants into database")
start_insertion = datetime.now()
start_five_thousand = datetime.now()
# These are the number of parsed varaints
nr_variants = 0
# These are the number of variants that meet the criteria and gets
# inserted
nr_inserted = 0
# This is to keep track of the inserted variants
inserted = 1
try:
for nr_variants, variant in enumerate(vcf_obj(region)):
rank_score = parse_rank_score(
variant.INFO.get('RankScore'),
case_obj['display_name']
)
if rank_score > rank_threshold:
# Parse the vcf variant
parsed_variant = parse_variant(
variant=variant,
case=case_obj,
variant_type=variant_type,
rank_results_header=rank_results_header,
vep_header = vep_header,
individual_positions = individual_positions
)
# Build the variant object
variant_obj = build_variant(
variant=parsed_variant,
institute_id=institute_obj['_id'],
gene_to_panels=gene_to_panels,
hgncid_to_gene=hgncid_to_gene,
)
try:
self.load_variant(variant_obj)
nr_inserted += 1
except IntegrityError as error:
pass
if (nr_variants != 0 and nr_variants % 5000 == 0):
logger.info("%s variants parsed" % str(nr_variants))
logger.info("Time to parse variants: {} ".format(
datetime.now() - start_five_thousand))
start_five_thousand = datetime.now()
if (nr_inserted != 0 and (nr_inserted * inserted) % (1000 * inserted) == 0):
logger.info("%s variants inserted" % nr_inserted)
inserted += 1
except Exception as error:
logger.warning("Deleting inserted variants")
self.delete_variants(case_obj['_id'], variant_type)
raise error
return nr_inserted
def load_variants(adapter,
variant_file,
case_obj,
variant_type='clinical',
category='snv',
rank_threshold=6,
chrom=None,
start=None,
end=None):
"""Load all variant in variants
Args:
adapter(MongoAdapter)
variant_file(str): Path to variant file
case(Case)
variant_type(str)
category(str): 'snv' or 'sv'
rank_threshold(int)
chrom(str)
start(int)
end(int)
"""
institute_obj = adapter.institute(institute_id=case_obj['owner'])
if not institute_obj:
raise IntegrityError("Institute {0} does not exist in"
" database.".format(case_obj['owner']))
gene_to_panels = adapter.gene_to_panels()
hgncid_to_gene = adapter.hgncid_to_gene()
coordinates = {}
vcf_obj = VCF(variant_file)
rank_results_header = parse_rank_results_header(vcf_obj)
vep_header = parse_vep_header(vcf_obj)
# This is a dictionary to tell where ind are in vcf
individual_positions = {}
for i, ind in enumerate(vcf_obj.samples):
individual_positions[ind] = i
LOG.info("Start inserting variants into database")
start_insertion = datetime.now()
start_five_thousand = datetime.now()
# To get it right if the file is empty
nr_variants = -1
nr_inserted = 0
inserted = 1
coordinates = False
if chrom:
coordinates = {'chrom': chrom, 'start': start, 'end': end}
try:
for nr_variants, variant in enumerate(vcf_obj):
# Get the neccesary coordinates
# Parse away any chr CHR prefix
chrom_match = CHR_PATTERN.match(variant.CHROM)
chrom = chrom_match.group(2)
position = variant.POS
add_variant = False
# If coordinates are specified we want to upload all variants that
# resides within the specified region
if coordinates:
if check_coordinates(chrom, position, coordinates):
add_variant = True
# If there are no coordinates we allways want to load MT variants
elif chrom == 'MT':
add_variant = True
# Otherwise we need to check is rank score requirement are fulfilled
else:
rank_score = parse_rank_score(variant.INFO.get('RankScore'),
case_obj['display_name'])
if rank_score >= rank_threshold:
add_variant = True
variant_obj = None
# Log the number of variants parsed
if (nr_variants != 0 and nr_variants % 5000 == 0):
LOG.info("%s variants parsed" % str(nr_variants))
LOG.info(
"Time to parse variants: {} ".format(datetime.now() -
start_five_thousand))
start_five_thousand = datetime.now()
if not add_variant:
continue
####### Here we know that the variant should be loaded #########
# We follow the scout paradigm of parse -> build -> load
# Parse the variant
parsed_variant = parse_variant(
variant=variant,
case=case_obj,
variant_type=variant_type,
rank_results_header=rank_results_header,
vep_header=vep_header,
individual_positions=individual_positions)
# Build the variant object
variant_obj = build_variant(
variant=parsed_variant,
institute_id=institute_obj['_id'],
gene_to_panels=gene_to_panels,
hgncid_to_gene=hgncid_to_gene,
)
# Load the variant abject
# We could get integrity error here since if we want to load all variants of a region
# there will likely already be variants from that region loaded
try:
load_variant(adapter, variant_obj)
nr_inserted += 1
except IntegrityError as error:
pass
# Log number of inserted variants
if (nr_inserted != 0
and (nr_inserted * inserted) % (1000 * inserted) == 0):
LOG.info("%s variants inserted" % nr_inserted)
inserted += 1
except Exception as error:
if not coordinates:
LOG.warning("Deleting inserted variants")
delete_variants(adapter, case_obj, variant_type)
raise error
LOG.info("All variants inserted.")
LOG.info("Number of variants in file: {0}".format(nr_variants + 1))
LOG.info("Number of variants inserted: {0}".format(nr_inserted))
LOG.info("Time to insert variants:{0}".format(datetime.now() -
start_insertion))
