def task_create_riskreport(
risk_report_id, genome_id
): # NOTE: arguments for celery task should be JSON serializable
risk_report = RiskReport.objects.get(id=risk_report_id)
genome = Genome.objects.get(id=genome_id)
log.info('Creating riskreport ...')
# TODO: Check for updates
latest_date = GwasCatalogSnp.objects.aggregate(
Max('date_downloaded'))['date_downloaded__max']
phenotypes = GwasCatalogPhenotype.objects.all()
log.info('#phenotypes: {}'.format(len(phenotypes)))
population = [genome.population]
for phenotype in phenotypes:
assert type(phenotype) == GwasCatalogPhenotype
gwas_snps = GwasCatalogSnp.objects.filter(
phenotype=phenotype,
population__contains=population,
date_downloaded=str(latest_date))
if not gwas_snps:
continue
# Select only one article for one phenotype
#
# TODO: add conditions
# - risk alleles are present
# - odds ratios are present
# - (beta-coeeff is not present)
# - lower than minimum p-value
evidence_article_1st = gwas_snps.exclude(
pubmed_id__isnull=True).order_by('reliability_rank').values_list(
'pubmed_id', flat=True).distinct().first()
evidence_snps = gwas_snps.filter(pubmed_id=evidence_article_1st)
evidence_snp_ids = evidence_snps.values_list('snp_id_current',
flat=True)
freqs = get_freqs(evidence_snp_ids, population=population)
genotypes = Genotype.objects.filter(genome__id=genome.id,
rs_id_current__in=evidence_snp_ids)
phenotype_risk_report, _ = PhenotypeRiskReport.objects.get_or_create(
risk_report=risk_report, phenotype=phenotype)
# Calculate cumulative risk
estimated_snp_risks = []
# Genotype specific risks for each SNP
with transaction.atomic():
for evidence_snp in evidence_snps:
# Risk allele and its frequency
risk_allele_forward = evidence_snp.risk_allele_forward
risk_allele_freq = freqs.get(evidence_snp.snp_id_current,
{}).get(risk_allele_forward)
odds_ratio = evidence_snp.odds_ratio
# My genotype
try:
genotype = ''.join(
genotypes.get(rs_id_current=evidence_snp.snp_id_current
).genotype)
zygosities = zyg(genotype, risk_allele_forward)
except Genotype.DoesNotExist:
zygosities = None
# Genotype specific risks
if None not in (risk_allele_freq, odds_ratio, zygosities):
genotype_specific_risks = genotype_specific_risks_relative_to_population(
risk_allele_freq, odds_ratio)
my_estimated_risk = estimated_risk(genotype_specific_risks,
zygosities)
else:
my_estimated_risk = None
SnpRiskReport(phenotype_risk_report=phenotype_risk_report,
evidence_snp=evidence_snp,
estimated_risk=my_estimated_risk).save()
estimated_snp_risks.append(my_estimated_risk)
phenotype_risk_report.estimated_risk = cumulative_risk(
estimated_snp_risks)
phenotype_risk_report.save()
log.info('Done')
def handle(self, *args, **options):
current_tz = timezone.get_current_timezone()
if not os.path.exists(settings.GWASCATALOG_DIR):
os.makedirs(settings.GWASCATALOG_DIR)
# TODO: automatically choose latest version
log.info('Fetching latest gwascatalog...')
catalog_path = os.path.join(settings.GWASCATALOG_DIR, 'dbsnp-pg-min-0.5.2-b144-GRCh37.gwascatalog-cleaned.tsv.gz')
get_url_content(url='https://github.com/knmkr/dbsnp-pg-min/releases/download/0.5.2/dbsnp-pg-min-0.5.2-b144-GRCh37.gwascatalog-cleaned.tsv.gz',
dst=catalog_path,
if_not_exists=True)
log.info('Fetching latest gwascatalog allele freq...')
catalog_freq_path = os.path.join(settings.GWASCATALOG_DIR, 'dbsnp-pg-min-0.5.2-b144-GRCh37.gwascatalog-snps-allele-freq.tsv.gz')
get_url_content(url='https://github.com/knmkr/dbsnp-pg-min/releases/download/0.5.2/dbsnp-pg-min-0.5.2-b144-GRCh37.gwascatalog-snps-allele-freq.tsv.gz',
dst=catalog_freq_path,
if_not_exists=True)
# - Gwas Catalog Allele Freq
log.info('Updating snp allele freq records for gwascatalog...')
num_created = 0
num_updated = 0
with transaction.atomic():
for record in csv.DictReader(gzip.open(catalog_freq_path, 'rb'), delimiter='\t',
fieldnames=['snp_id_current', 'allele', 'freq', 'populations']):
snp, created = Snp.objects.update_or_create(
snp_id_current=record['snp_id_current'],
population=record['populations'],
defaults={'allele': text2pg_array(record['allele']),
'freq': text2pg_array(record['freq'])}
)
if created:
num_created += 1
else:
num_updated += 1
log.info('updated: {} records'.format(num_updated))
log.info('created: {} records'.format(num_created))
# - Gwas Catalog
log.info('Importing gwascatalog...')
model_fields = [field for field in GwasCatalogSnp._meta.get_fields() if field.name not in ('id', 'created_at')]
model_field_names = [field.name for field in model_fields]
model_fields_map = dict(zip(model_field_names, model_fields))
gwascatalog_snps = []
num_phenotype_created = 0
for record in csv.DictReader(gzip.open(catalog_path, 'rb'), delimiter='\t'):
data = {}
# If date_downloaded is already imported, abort.
date_downloaded = record['date_downloaded']
if GwasCatalogSnp.objects.filter(date_downloaded=date_downloaded).exists():
raise GwasCatalogParseError('Already imported date_downloaded: {}'.format(date_downloaded))
# Import only pre-defined model fields
for k,v in record.items():
if k in model_field_names:
# Set blank or null
if v == '':
if type(model_fields_map[k]) in (models.fields.CharField, models.fields.TextField):
v = ''
else:
v = None
# Set datetime with timezone
if type(model_fields_map[k]) == models.DateTimeField:
v = current_tz.localize(datetime(*(parse_date(v).timetuple()[:5])))
data[k] = v
# Parse population
population = get_population(record['initial_sample'])
# Calculate reliability rank
reliability_rank = 1.0 # TODO: get_reliability_rank()
try:
# Parse and validate odds_ratio, beta_coeff
_, unit = get_ci_and_unit(record['confidence_interval_95_percent'])
odds_ratio, beta_coeff = get_odds_ratio_or_beta_coeff(record['odds_ratio_or_beta_coeff'], unit)
# Validate risk_allele
#
# Strands of risk alleles in GWAS Catalog are not set to forward strands with respect to
# the human reference genome b37. So we get forward strand alleles by checking consistences of
# allele frequencies between reported risk alleles and 1000 Genomes Project alleles.
if data['snp_id_current']:
snp_id = int(data['snp_id_current'])
database_freq = get_freqs([snp_id], population).get(snp_id)
risk_allele_forward = get_database_strand_allele(record['risk_allele'], record['risk_allele_freq_reported'],
database_freq, freq_diff_thrs=settings.GWASCATALOG_FREQ_DIFF_THRS)
else:
risk_allele_forward = AMBIGOUS
is_active = True
except GwasCatalogParseError as e:
log.error(e)
odds_ratio, beta_coeff = None, None
is_active = False
# - Phenotype
phenotype, phenotype_created = GwasCatalogPhenotype.objects.get_or_create(name=record['disease_or_trait'])
if phenotype_created:
num_phenotype_created += 1
data.update({'population': population,
'reliability_rank': reliability_rank,
'odds_ratio': odds_ratio,
'beta_coeff': beta_coeff,
'beta_coeff_unit': unit,
'risk_allele_forward': risk_allele_forward,
'phenotype': phenotype,
'is_active': is_active})
gwascatalog_snps.append(GwasCatalogSnp(**data))
# GwasCatalogSnp.objects.create(**data)
with transaction.atomic():
GwasCatalogSnp.objects.bulk_create(gwascatalog_snps)
log.info('GWAS Catalog snps processed: {} records'.format(len(gwascatalog_snps)))
log.info('GWAS Catalog phenotypes newly created: {} records'.format(num_phenotype_created))
log.info('Done.')
def task_create_riskreport(risk_report_id, genome_id): # NOTE: arguments for celery task should be JSON serializable
risk_report = RiskReport.objects.get(id=risk_report_id)
genome = Genome.objects.get(id=genome_id)
log.info('Creating riskreport ...')
# TODO: Check for updates
latest_date = GwasCatalogSnp.objects.aggregate(Max('date_downloaded'))['date_downloaded__max']
phenotypes = GwasCatalogPhenotype.objects.all()
log.info('#phenotypes: {}'.format(len(phenotypes)))
population = [genome.population]
for phenotype in phenotypes:
assert type(phenotype) == GwasCatalogPhenotype
gwas_snps = GwasCatalogSnp.objects.filter(phenotype=phenotype,
population__contains=population,
date_downloaded=str(latest_date))
if not gwas_snps:
continue
# Select only one article for one phenotype
#
# TODO: add conditions
# - risk alleles are present
# - odds ratios are present
# - (beta-coeeff is not present)
# - lower than minimum p-value
evidence_article_1st = gwas_snps.exclude(pubmed_id__isnull=True).order_by('reliability_rank').values_list('pubmed_id', flat=True).distinct().first()
evidence_snps = gwas_snps.filter(pubmed_id=evidence_article_1st)
evidence_snp_ids = evidence_snps.values_list('snp_id_current', flat=True)
freqs = get_freqs(evidence_snp_ids, population=population)
genotypes = Genotype.objects.filter(genome__id=genome.id, rs_id_current__in=evidence_snp_ids)
phenotype_risk_report, _ = PhenotypeRiskReport.objects.get_or_create(risk_report=risk_report, phenotype=phenotype)
# Calculate cumulative risk
estimated_snp_risks = []
# Genotype specific risks for each SNP
with transaction.atomic():
for evidence_snp in evidence_snps:
# Risk allele and its frequency
risk_allele_forward = evidence_snp.risk_allele_forward
risk_allele_freq = freqs.get(evidence_snp.snp_id_current, {}).get(risk_allele_forward)
odds_ratio = evidence_snp.odds_ratio
# My genotype
try:
genotype = ''.join(genotypes.get(rs_id_current=evidence_snp.snp_id_current).genotype)
zygosities = zyg(genotype, risk_allele_forward)
except Genotype.DoesNotExist:
zygosities = None
# Genotype specific risks
if None not in (risk_allele_freq, odds_ratio, zygosities):
genotype_specific_risks = genotype_specific_risks_relative_to_population(risk_allele_freq, odds_ratio)
my_estimated_risk = estimated_risk(genotype_specific_risks, zygosities)
else:
my_estimated_risk = None
SnpRiskReport(phenotype_risk_report=phenotype_risk_report,
evidence_snp=evidence_snp,
estimated_risk=my_estimated_risk).save()
estimated_snp_risks.append(my_estimated_risk)
phenotype_risk_report.estimated_risk = cumulative_risk(estimated_snp_risks)
phenotype_risk_report.save()
log.info('Done')