def _apply_harmonized_metadata_to_sample(sample: Sample, metadata: dict):
"""Harmonizes the metadata and applies it to `sample`"""
sample.title = harmony.extract_title(metadata)
harmonized_sample = harmony.harmonize([metadata])
for key, value in harmonized_sample[sample.title].items():
setattr(sample, key, value)
def create_experiment_and_samples_from_api(
self, experiment_accession_code) -> (Experiment, List[Sample]):
""" The main surveyor - find the Experiment and Samples from NCBI GEO.
Uses the GEOParse library, for which docs can be found here: https://geoparse.readthedocs.io/en/latest/usage.html#working-with-geo-objects
"""
# Cleaning up is tracked here: https://github.com/guma44/GEOparse/issues/41
gse = GEOparse.get_GEO(experiment_accession_code,
destdir=self.get_temp_path(),
how="brief",
silent=True)
preprocessed_samples = harmony.preprocess_geo(gse.gsms.items())
harmonized_samples = harmony.harmonize(preprocessed_samples)
# Create the experiment object
try:
experiment_object = Experiment.objects.get(
accession_code=experiment_accession_code)
logger.debug(
"Experiment %s already exists, skipping object creation.",
experiment_accession_code,
survey_job=self.survey_job.id,
)
except Experiment.DoesNotExist:
experiment_object = Experiment()
experiment_object.accession_code = experiment_accession_code
GeoSurveyor._apply_metadata_to_experiment(experiment_object, gse)
experiment_object.save()
experiment_annotation = ExperimentAnnotation()
experiment_annotation.data = gse.metadata
experiment_annotation.experiment = experiment_object
experiment_annotation.is_ccdl = False
experiment_annotation.save()
# Okay, here's the situation!
# Sometimes, samples have a direct single representation for themselves.
# Othertimes, there is a single file with references to every sample in it.
created_samples = []
for sample_accession_code, sample in gse.gsms.items():
try:
sample_object = Sample.objects.get(
accession_code=sample_accession_code)
logger.debug(
"Sample %s from experiment %s already exists, skipping object creation.",
sample_accession_code,
experiment_object.accession_code,
survey_job=self.survey_job.id,
)
# Associate it with the experiment, but since it
# already exists it already has original files
# associated with it and it's already been downloaded,
# so don't add it to created_samples.
ExperimentSampleAssociation.objects.get_or_create(
experiment=experiment_object, sample=sample_object)
ExperimentOrganismAssociation.objects.get_or_create(
experiment=experiment_object,
organism=sample_object.organism)
except Sample.DoesNotExist:
organism = Organism.get_object_for_name(
sample.metadata["organism_ch1"][0].upper())
sample_object = Sample()
sample_object.source_database = "GEO"
sample_object.accession_code = sample_accession_code
sample_object.organism = organism
# If data processing step, it isn't raw.
sample_object.has_raw = not sample.metadata.get(
"data_processing", None)
ExperimentOrganismAssociation.objects.get_or_create(
experiment=experiment_object, organism=organism)
sample_object.title = sample.metadata["title"][0]
self.set_platform_properties(sample_object, sample.metadata,
gse)
GeoSurveyor._apply_harmonized_metadata_to_sample(
sample_object, harmonized_samples[sample_object.title])
# Sample-level protocol_info
sample_object.protocol_info = self.get_sample_protocol_info(
sample.metadata, sample_accession_code)
sample_object.save()
logger.debug("Created Sample: " + str(sample_object))
sample_annotation = SampleAnnotation()
sample_annotation.sample = sample_object
sample_annotation.data = sample.metadata
sample_annotation.is_ccdl = False
sample_annotation.save()
sample_supplements = sample.metadata.get(
"supplementary_file", [])
for supplementary_file_url in sample_supplements:
# Why do they give us this?
if supplementary_file_url == "NONE":
break
# We never want these!
if "idat.gz" in supplementary_file_url.lower():
continue
if "chp.gz" in supplementary_file_url.lower():
continue
if "ndf.gz" in supplementary_file_url.lower():
continue
if "pos.gz" in supplementary_file_url.lower():
continue
if "pair.gz" in supplementary_file_url.lower():
continue
if "gff.gz" in supplementary_file_url.lower():
continue
# Sometimes, we are lied to about the data processing step.
lower_file_url = supplementary_file_url.lower()
if (".cel" in lower_file_url
or ("_non_normalized.txt" in lower_file_url)
or ("_non-normalized.txt" in lower_file_url)
or ("-non-normalized.txt" in lower_file_url)
or ("-non_normalized.txt" in lower_file_url)):
sample_object.has_raw = True
sample_object.save()
# filename and source_filename are the same for these
filename = FileUtils.get_filename(supplementary_file_url)
original_file = OriginalFile.objects.get_or_create(
source_url=supplementary_file_url,
filename=filename,
source_filename=filename,
has_raw=sample_object.has_raw,
is_archive=FileUtils.is_archive(filename),
)[0]
logger.debug("Created OriginalFile: " + str(original_file))
original_file_sample_association = OriginalFileSampleAssociation.objects.get_or_create(
original_file=original_file, sample=sample_object)
if original_file.is_affy_data():
# Only Affymetrix Microarrays produce .CEL files
sample_object.technology = "MICROARRAY"
sample_object.manufacturer = "AFFYMETRIX"
sample_object.save()
# It's okay to survey RNA-Seq samples from GEO, but we
# don't actually want to download/process any RNA-Seq
# data unless it comes from SRA.
if sample_object.technology != "RNA-SEQ":
created_samples.append(sample_object)
# Now that we've determined the technology at the
# sample level, we can set it at the experiment level,
# just gotta make sure to only do it once. There can
# be more than one technology, this should be changed
# as part of:
# https://github.com/AlexsLemonade/refinebio/issues/1099
if not experiment_object.technology:
experiment_object.technology = sample_object.technology
experiment_object.save()
ExperimentSampleAssociation.objects.get_or_create(
experiment=experiment_object, sample=sample_object)
# These supplementary files _may-or-may-not_ contain the type of raw data we can process.
for experiment_supplement_url in gse.metadata.get(
"supplementary_file", []):
# filename and source_filename are the same for these
filename = experiment_supplement_url.split("/")[-1]
original_file = OriginalFile.objects.get_or_create(
source_url=experiment_supplement_url,
filename=filename,
source_filename=filename,
has_raw=sample_object.has_raw,
is_archive=True,
)[0]
logger.debug("Created OriginalFile: " + str(original_file))
lower_supplement_url = experiment_supplement_url.lower()
if (("_non_normalized.txt" in lower_supplement_url)
or ("_non-normalized.txt" in lower_supplement_url)
or ("-non-normalized.txt" in lower_supplement_url)
or ("-non_normalized.txt" in lower_supplement_url)):
for sample_object in created_samples:
sample_object.has_raw = True
sample_object.save()
OriginalFileSampleAssociation.objects.get_or_create(
sample=sample_object, original_file=original_file)
# Delete this Original file if it isn't being used.
if (OriginalFileSampleAssociation.objects.filter(
original_file=original_file).count() == 0):
original_file.delete()
# These are the Miniml/Soft/Matrix URLs that are always(?) provided.
# GEO describes different types of data formatting as "families"
family_url = self.get_miniml_url(experiment_accession_code)
miniml_original_file = OriginalFile.objects.get_or_create(
source_url=family_url,
source_filename=family_url.split("/")[-1],
has_raw=sample_object.has_raw,
is_archive=True,
)[0]
for sample_object in created_samples:
# We don't need a .txt if we have a .CEL
if sample_object.has_raw:
continue
OriginalFileSampleAssociation.objects.get_or_create(
sample=sample_object, original_file=miniml_original_file)
# Delete this Original file if it isn't being used.
if (OriginalFileSampleAssociation.objects.filter(
original_file=miniml_original_file).count() == 0):
miniml_original_file.delete()
# Trash the temp path
try:
shutil.rmtree(self.get_temp_path())
except Exception:
# There was a problem during surveying so this didn't get created.
# It's not a big deal.
pass
return experiment_object, created_samples
def create_samples_from_api(self, experiment: Experiment,
platform_dict: Dict) -> List[Sample]:
"""Generates a Sample item for each sample in an AE experiment.
There are many possible data situations for a sample:
- If the sample only has raw data available:
- If it is on a platform that we support:
Download this raw data and process it
- If it is not on a platform we support:
Don't download anything, don't process anything
- If the sample has both raw and derived data:
- If the raw data is on a platform we support:
Download the raw data and process it, abandon the derived data
- If the raw data is not on a platform we support
Download the derived data and no-op it, abandon the raw data
- If the sample only has derived data:
Download the derived data and no-op it.
See an example at: https://www.ebi.ac.uk/arrayexpress/json/v3/experiments/E-MTAB-3050/samples
"""
created_samples = []
samples_endpoint = SAMPLES_URL.format(experiment.accession_code)
r = utils.requests_retry_session().get(samples_endpoint, timeout=60)
samples = r.json()["experiment"]["sample"]
# The SDRF is the complete metadata record on a sample/property basis.
# We run this through our harmonizer and then attach the properties
# to our created samples.
SDRF_URL_TEMPLATE = "https://www.ebi.ac.uk/arrayexpress/files/{code}/{code}.sdrf.txt"
sdrf_url = SDRF_URL_TEMPLATE.format(code=experiment.accession_code)
sdrf_samples = harmony.parse_sdrf(sdrf_url)
harmonized_samples = harmony.harmonize(sdrf_samples)
# An experiment can have many samples
for sample_data in samples:
# For some reason, this sample has no files associated with it.
if "file" not in sample_data or len(sample_data['file']) == 0:
continue
# Each sample is given an experimenatlly-unique title.
flat_sample = utils.flatten(sample_data)
title = harmony.extract_title(flat_sample)
# A sample may actually have many sub files.
# If there is raw data, take that.
# If not, take the derived.
has_raw = False
for sub_file in sample_data['file']:
# For ex: E-GEOD-15645
if isinstance(sub_file['comment'], list):
sub_file_mod = sub_file
sub_file_mod['comment'] = sub_file['comment'][0]
else:
sub_file_mod = sub_file
# Some have the 'data' field, but not the actual data
# Ex: E-GEOD-9656
if sub_file_mod['type'] == "data" and sub_file_mod[
'comment'].get('value', None) != None:
has_raw = True
if 'raw' in sub_file_mod['comment'].get('value', ''):
has_raw = True
skip_sample = False
for sub_file in sample_data['file']:
# Don't get the raw data if it's only a 1-color sample.
if 'Cy3' in str(sample_data) and 'Cy5' not in str(sample_data):
has_raw = False
# Skip derived data if we have it raw.
if has_raw and "derived data" in sub_file['type']:
continue
download_url = None
filename = sub_file["name"]
# sub_file["comment"] is only a list if there's
# more than one comment...
comments = sub_file["comment"]
if isinstance(comments, list):
# Could be: "Derived ArrayExpress Data Matrix FTP
# file" or: "ArrayExpress FTP file". If there is
# no comment with a name including "FTP file" then
# we don't know where to download it so we need to
# mark this job as an error. Therefore don't catch
# the potential exception where download_url
# doesn't get defined.
for comment in comments:
if "FTP file" in comment["name"]:
download_url = comment["value"]
break
else:
download_url = comments["value"]
if not download_url:
logger.error(
"Sample %s did not specify a download url, skipping.",
sample_accession_code,
experiment_accession_code=experiment.accession_code,
survey_job=self.survey_job.id,
sub_file=sub_file)
skip_sample = True
continue
if not filename:
logger.error(
"Sample %s did not specify a filename, skipping.",
sample_accession_code,
experiment_accession_code=experiment.accession_code,
survey_job=self.survey_job.id,
sub_file=sub_file)
skip_sample = True
continue
if skip_sample:
continue
# The accession code is not a simple matter to determine.
sample_source_name = sample_data["source"].get("name", "")
sample_assay_name = sample_data["assay"].get("name", "")
sample_accession_code = self.determine_sample_accession(
experiment.accession_code, sample_source_name,
sample_assay_name, filename)
# Figure out the Organism for this sample
organism_name = UNKNOWN
for characteristic in sample_data["characteristic"]:
if characteristic["category"].upper() == "ORGANISM":
organism_name = characteristic["value"].upper()
if organism_name == UNKNOWN:
logger.error(
"Sample %s did not specify the organism name.",
sample_accession_code,
experiment_accession_code=experiment.accession_code,
survey_job=self.survey_job.id)
organism = None
continue
else:
organism = Organism.get_object_for_name(organism_name)
# Create the sample object
try:
# Associate it with the experiment, but since it
# already exists it already has original files
# associated with it and it's already been downloaded,
# so don't add it to created_samples.
sample_object = Sample.objects.get(
accession_code=sample_accession_code)
# If input experiment includes new protocol information,
# update sample's protocol_info.
existing_protocols = sample_object.protocol_info
protocol_info, is_updated = self.update_sample_protocol_info(
existing_protocols, experiment.protocol_description,
experiment.source_url + '/protocols')
if is_updated:
sample_object.protocol_info = protocol_info
sample_obejct.save()
logger.debug(
"Sample %s already exists, skipping object creation.",
sample_accession_code,
experiment_accession_code=experiment.accession_code,
survey_job=self.survey_job.id)
except Sample.DoesNotExist:
sample_object = Sample()
# The basics
sample_object.source_database = "ARRAY_EXPRESS"
sample_object.title = title
sample_object.accession_code = sample_accession_code
sample_object.source_archive_url = samples_endpoint
sample_object.organism = organism
sample_object.platform_name = platform_dict[
"platform_accession_name"]
sample_object.platform_accession_code = platform_dict[
"platform_accession_code"]
sample_object.manufacturer = platform_dict["manufacturer"]
sample_object.technology = "MICROARRAY"
protocol_info, is_updated = self.update_sample_protocol_info(
existing_protocols=[],
experiment_protocol=experiment.protocol_description,
protocol_url=experiment.source_url + '/protocols')
# Do not check is_updated the first time because we must
# save a list so we can append to it later.
sample_object.protocol_info = protocol_info
sample_object.save()
# Directly assign the harmonized properties
harmonized_sample = harmonized_samples[title]
for key, value in harmonized_sample.items():
setattr(sample_object, key, value)
sample_object.save()
sample_annotation = SampleAnnotation()
sample_annotation.data = sample_data
sample_annotation.sample = sample_object
sample_annotation.is_ccdl = False
sample_annotation.save()
original_file = OriginalFile()
original_file.filename = filename
original_file.source_filename = filename
original_file.source_url = download_url
original_file.is_downloaded = False
original_file.is_archive = True
original_file.has_raw = has_raw
original_file.save()
original_file_sample_association = OriginalFileSampleAssociation(
)
original_file_sample_association.original_file = original_file
original_file_sample_association.sample = sample_object
original_file_sample_association.save()
created_samples.append(sample_object)
logger.debug(
"Created " + str(sample_object),
experiment_accession_code=experiment.accession_code,
survey_job=self.survey_job.id,
sample=sample_object.id)
# Create associations if they don't already exist
ExperimentSampleAssociation.objects.get_or_create(
experiment=experiment, sample=sample_object)
ExperimentOrganismAssociation.objects.get_or_create(
experiment=experiment, organism=organism)
return created_samples
def _generate_experiment_and_samples(
self,
run_accession: str,
study_accession: str = None) -> (Experiment, List[Sample]):
"""Generates Experiments and Samples for the provided run_accession."""
metadata = SraSurveyor.gather_all_metadata(run_accession)
if metadata == {}:
if study_accession:
logger.error("Could not discover any metadata for run.",
accession=run_accession,
study_accession=study_accession)
else:
logger.error("Could not discover any metadata for run.",
accession=run_accession)
return (None, None) # This will cascade properly
if DOWNLOAD_SOURCE == "ENA":
if metadata["library_layout"] == "PAIRED":
files_urls = [
SraSurveyor._build_ena_file_url(run_accession, "_1"),
SraSurveyor._build_ena_file_url(run_accession, "_2")
]
else:
files_urls = [SraSurveyor._build_ena_file_url(run_accession)]
else:
files_urls = [SraSurveyor._build_ncbi_file_url(run_accession)]
# Figure out the Organism for this sample
organism_name = metadata.pop("organism_name", None)
if not organism_name:
logger.error("Could not discover organism type for run.",
accession=run_accession)
return (None, None) # This will cascade properly
organism_name = organism_name.upper()
organism = Organism.get_object_for_name(organism_name)
##
# Experiment
##
experiment_accession_code = metadata.get('study_accession')
try:
experiment_object = Experiment.objects.get(
accession_code=experiment_accession_code)
logger.debug(
"Experiment already exists, skipping object creation.",
experiment_accession_code=experiment_accession_code,
survey_job=self.survey_job.id)
except Experiment.DoesNotExist:
experiment_object = Experiment()
experiment_object.accession_code = experiment_accession_code
experiment_object.source_url = ENA_URL_TEMPLATE.format(
experiment_accession_code)
experiment_object.source_database = "SRA"
experiment_object.technology = "RNA-SEQ"
# We don't get this value from the API, unfortunately.
# experiment_object.platform_accession_code = experiment["platform_accession_code"]
if not experiment_object.description:
experiment_object.description = "No description."
if "study_title" in metadata:
experiment_object.title = metadata["study_title"]
if "study_abstract" in metadata:
experiment_object.description = metadata["study_abstract"]
if "lab_name" in metadata:
experiment_object.submitter_institution = metadata["lab_name"]
if "experiment_design_description" in metadata:
experiment_object.protocol_description = metadata[
"experiment_design_description"]
if "pubmed_id" in metadata:
experiment_object.pubmed_id = metadata["pubmed_id"]
experiment_object.has_publication = True
if "study_ena_first_public" in metadata:
experiment_object.source_first_published = parse_datetime(
metadata["study_ena_first_public"])
if "study_ena_last_update" in metadata:
experiment_object.source_last_modified = parse_datetime(
metadata["study_ena_last_update"])
# Rare, but it happens.
if not experiment_object.protocol_description:
experiment_object.protocol_description = metadata.get(
"library_construction_protocol",
"Protocol was never provided.")
# Scrape publication title and authorship from Pubmed
if experiment_object.pubmed_id:
pubmed_metadata = utils.get_title_and_authors_for_pubmed_id(
experiment_object.pubmed_id)
experiment_object.publication_title = pubmed_metadata[0]
experiment_object.publication_authors = pubmed_metadata[1]
experiment_object.save()
##
# Experiment Metadata
##
json_xa = ExperimentAnnotation()
json_xa.experiment = experiment_object
json_xa.data = metadata
json_xa.is_ccdl = False
json_xa.save()
##
# Samples
##
sample_accession_code = metadata.pop('run_accession')
# Create the sample object
try:
sample_object = Sample.objects.get(
accession_code=sample_accession_code)
# If current experiment includes new protocol information,
# merge it into the sample's existing protocol_info.
protocol_info, is_updated = self.update_sample_protocol_info(
sample_object.protocol_info,
experiment_object.protocol_description,
experiment_object.source_url)
if is_updated:
sample_object.protocol_info = protocol_info
sample_object.save()
logger.debug(
"Sample %s already exists, skipping object creation.",
sample_accession_code,
experiment_accession_code=experiment_object.accession_code,
survey_job=self.survey_job.id)
except Sample.DoesNotExist:
sample_object = Sample()
sample_object.source_database = "SRA"
sample_object.accession_code = sample_accession_code
sample_object.organism = organism
sample_object.platform_name = metadata.get(
"platform_instrument_model", "UNKNOWN")
# The platform_name is human readable and contains spaces,
# accession codes shouldn't have spaces though:
sample_object.platform_accession_code = sample_object.platform_name.replace(
" ", "")
sample_object.technology = "RNA-SEQ"
if "ILLUMINA" in sample_object.platform_name.upper() \
or "NEXTSEQ" in sample_object.platform_name.upper():
sample_object.manufacturer = "ILLUMINA"
elif "ION TORRENT" in sample_object.platform_name.upper():
sample_object.manufacturer = "ION_TORRENT"
else:
sample_object.manufacturer = "UNKNOWN"
# Directly apply the harmonized values
sample_object.title = harmony.extract_title(metadata)
harmonized_sample = harmony.harmonize([metadata])
for key, value in harmonized_sample.items():
setattr(sample_object, key, value)
protocol_info, is_updated = self.update_sample_protocol_info(
existing_protocols=[],
experiment_protocol=experiment_object.protocol_description,
experiment_url=experiment_object.source_url)
# Do not check is_updated the first time because we must
# save a list so we can append to it later.
sample_object.protocol_info = protocol_info
sample_object.save()
for file_url in files_urls:
original_file = OriginalFile.objects.get_or_create(
source_url=file_url,
source_filename=file_url.split('/')[-1],
has_raw=True)[0]
original_file_sample_association = OriginalFileSampleAssociation.objects.get_or_create(
original_file=original_file, sample=sample_object)
# Create associations if they don't already exist
ExperimentSampleAssociation.objects.get_or_create(
experiment=experiment_object, sample=sample_object)
ExperimentOrganismAssociation.objects.get_or_create(
experiment=experiment_object, organism=organism)
return experiment_object, [sample_object]
def create_experiment_and_samples_from_api(
self, experiment_accession_code) -> (Experiment, List[Sample]):
""" The main surveyor - find the Experiment and Samples from NCBI GEO.
Uses the GEOParse library, for which docs can be found here: https://geoparse.readthedocs.io/en/latest/usage.html#working-with-geo-objects
"""
# Cleaning up is tracked here: https://github.com/guma44/GEOparse/issues/41
gse = GEOparse.get_GEO(experiment_accession_code,
destdir=self.get_temp_path(),
how="brief",
silent=True)
preprocessed_samples = harmony.preprocess_geo(gse.gsms.items())
harmonized_samples = harmony.harmonize(preprocessed_samples)
# Create the experiment object
try:
experiment_object = Experiment.objects.get(
accession_code=experiment_accession_code)
logger.debug(
"Experiment %s already exists, skipping object creation.",
experiment_accession_code,
survey_job=self.survey_job.id)
except Experiment.DoesNotExist:
experiment_object = Experiment()
experiment_object.accession_code = experiment_accession_code
experiment_object.source_url = (
"https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=" +
experiment_accession_code)
experiment_object.source_database = "GEO"
experiment_object.title = gse.metadata.get('title', [''])[0]
experiment_object.description = gse.metadata.get('summary',
[''])[0]
# Source doesn't provide time information, assume midnight.
submission_date = gse.metadata["submission_date"][
0] + " 00:00:00 UTC"
experiment_object.source_first_published = dateutil.parser.parse(
submission_date)
last_updated_date = gse.metadata["last_update_date"][
0] + " 00:00:00 UTC"
experiment_object.source_last_updated = dateutil.parser.parse(
last_updated_date)
unique_institutions = list(set(gse.metadata["contact_institute"]))
experiment_object.submitter_institution = ", ".join(
unique_institutions)
experiment_object.pubmed_id = gse.metadata.get("pubmed_id",
[""])[0]
# Scrape publication title and authorship from Pubmed
if experiment_object.pubmed_id:
pubmed_metadata = utils.get_title_and_authors_for_pubmed_id(
experiment_object.pubmed_id)
experiment_object.publication_title = pubmed_metadata[0]
experiment_object.publication_authors = pubmed_metadata[1]
experiment_object.save()
experiment_annotation = ExperimentAnnotation()
experiment_annotation.data = gse.metadata
experiment_annotation.experiment = experiment_object
experiment_annotation.is_ccdl = False
experiment_annotation.save()
# Okay, here's the situation!
# Sometimes, samples have a direct single representation for themselves.
# Othertimes, there is a single file with references to every sample in it.
created_samples = []
for sample_accession_code, sample in gse.gsms.items():
try:
sample_object = Sample.objects.get(
accession_code=sample_accession_code)
logger.debug(
"Sample %s from experiment %s already exists, skipping object creation.",
sample_accession_code,
experiment_object.accession_code,
survey_job=self.survey_job.id)
# Associate it with the experiment, but since it
# already exists it already has original files
# associated with it and it's already been downloaded,
# so don't add it to created_samples.
ExperimentSampleAssociation.objects.get_or_create(
experiment=experiment_object, sample=sample_object)
ExperimentOrganismAssociation.objects.get_or_create(
experiment=experiment_object,
organism=sample_object.organism)
except Sample.DoesNotExist:
organism = Organism.get_object_for_name(
sample.metadata['organism_ch1'][0].upper())
sample_object = Sample()
sample_object.source_database = "GEO"
sample_object.accession_code = sample_accession_code
sample_object.organism = organism
# If data processing step, it isn't raw.
sample_object.has_raw = not sample.metadata.get(
'data_processing', None)
ExperimentOrganismAssociation.objects.get_or_create(
experiment=experiment_object, organism=organism)
sample_object.title = sample.metadata['title'][0]
self.set_platform_properties(sample_object, sample.metadata,
gse)
# Directly assign the harmonized properties
harmonized_sample = harmonized_samples[sample_object.title]
for key, value in harmonized_sample.items():
setattr(sample_object, key, value)
# Sample-level protocol_info
sample_object.protocol_info = self.get_sample_protocol_info(
sample.metadata, sample_accession_code)
sample_object.save()
logger.debug("Created Sample: " + str(sample_object))
sample_annotation = SampleAnnotation()
sample_annotation.sample = sample_object
sample_annotation.data = sample.metadata
sample_annotation.is_ccdl = False
sample_annotation.save()
sample_supplements = sample.metadata.get(
'supplementary_file', [])
for supplementary_file_url in sample_supplements:
# Why do they give us this?
if supplementary_file_url == "NONE":
break
# We never want these!
if "idat.gz" in supplementary_file_url.lower():
continue
if "chp.gz" in supplementary_file_url.lower():
continue
if "ndf.gz" in supplementary_file_url.lower():
continue
if "pos.gz" in supplementary_file_url.lower():
continue
if "pair.gz" in supplementary_file_url.lower():
continue
if "gff.gz" in supplementary_file_url.lower():
continue
# Sometimes, we are lied to about the data processing step.
lower_file_url = supplementary_file_url.lower()
if '.cel' in lower_file_url \
or ('_non_normalized.txt' in lower_file_url) \
or ('_non-normalized.txt' in lower_file_url) \
or ('-non-normalized.txt' in lower_file_url) \
or ('-non_normalized.txt' in lower_file_url):
sample_object.has_raw = True
sample_object.save()
# filename and source_filename are the same for these
filename = supplementary_file_url.split('/')[-1]
original_file = OriginalFile.objects.get_or_create(
source_url=supplementary_file_url,
filename=filename,
source_filename=filename,
has_raw=sample_object.has_raw,
is_archive=True)[0]
logger.debug("Created OriginalFile: " + str(original_file))
original_file_sample_association = OriginalFileSampleAssociation.objects.get_or_create(
original_file=original_file, sample=sample_object)
if original_file.is_affy_data():
# Only Affymetrix Microarrays produce .CEL files
sample_object.technology = 'MICROARRAY'
sample_object.manufacturer = 'AFFYMETRTIX'
sample_object.save()
# It's okay to survey RNA-Seq samples from GEO, but we
# don't actually want to download/process any RNA-Seq
# data unless it comes from SRA.
if sample_object.technology != 'RNA-SEQ':
created_samples.append(sample_object)
# Now that we've determined the technology at the
# sample level, we can set it at the experiment level,
# just gotta make sure to only do it once. There can
# be more than one technology, this should be changed
# as part of:
# https://github.com/AlexsLemonade/refinebio/issues/1099
if not experiment_object.technology:
experiment_object.technology = sample_object.technology
experiment_object.save()
ExperimentSampleAssociation.objects.get_or_create(
experiment=experiment_object, sample=sample_object)
# These supplementary files _may-or-may-not_ contain the type of raw data we can process.
for experiment_supplement_url in gse.metadata.get(
'supplementary_file', []):
# filename and source_filename are the same for these
filename = experiment_supplement_url.split('/')[-1]
original_file = OriginalFile.objects.get_or_create(
source_url=experiment_supplement_url,
filename=filename,
source_filename=filename,
has_raw=sample_object.has_raw,
is_archive=True)[0]
logger.debug("Created OriginalFile: " + str(original_file))
lower_supplement_url = experiment_supplement_url.lower()
if ('_non_normalized.txt' in lower_supplement_url) \
or ('_non-normalized.txt' in lower_supplement_url) \
or ('-non-normalized.txt' in lower_supplement_url) \
or ('-non_normalized.txt' in lower_supplement_url):
for sample_object in created_samples:
sample_object.has_raw = True
sample_object.save()
OriginalFileSampleAssociation.objects.get_or_create(
sample=sample_object, original_file=original_file)
# Delete this Original file if it isn't being used.
if OriginalFileSampleAssociation.objects.filter(
original_file=original_file).count() == 0:
original_file.delete()
# These are the Miniml/Soft/Matrix URLs that are always(?) provided.
# GEO describes different types of data formatting as "families"
family_url = self.get_miniml_url(experiment_accession_code)
miniml_original_file = OriginalFile.objects.get_or_create(
source_url=family_url,
source_filename=family_url.split('/')[-1],
has_raw=sample_object.has_raw,
is_archive=True)[0]
for sample_object in created_samples:
# We don't need a .txt if we have a .CEL
if sample_object.has_raw:
continue
OriginalFileSampleAssociation.objects.get_or_create(
sample=sample_object, original_file=miniml_original_file)
# Delete this Original file if it isn't being used.
if OriginalFileSampleAssociation.objects.filter(
original_file=miniml_original_file).count() == 0:
miniml_original_file.delete()
# Trash the temp path
try:
shutil.rmtree(self.get_temp_path())
except Exception:
# There was a problem during surveying so this didn't get created.
# It's not a big deal.
pass
return experiment_object, created_samples
def handle(self, *args, **options):
"""Refreshes the metadata for all samples, or samples from a specific database
"""
possible_source_databases = ["ARRAY_EXPRESS", "GEO", "SRA"]
if options.get("source_database", None) is None:
samples = Sample.objects.all()
elif options["source_database"] in possible_source_databases:
source_database = options["source_database"]
samples = Sample.objects.filter(source_database=source_database)
else:
logger.error('Invalid source database "{}"'.format(
options["source_database"]) +
"\nPossible source databases: {}".format(", ".join(
possible_source_databases)))
sys.exit(1)
paginator = PerformantPaginator(samples, PAGE_SIZE)
page = paginator.page()
while True:
for sample in samples:
logger.debug("Refreshing metadata for a sample.",
sample=sample.accession_code)
if sample.source_database == "SRA":
metadata = SraSurveyor.gather_all_metadata(
sample.accession_code)
SraSurveyor._apply_harmonized_metadata_to_sample(
sample, metadata)
elif sample.source_database == "GEO":
gse = GEOparse.get_GEO(
sample.experiments.first().accession_code,
destdir="/tmp/management",
how="brief",
silent=True,
)
preprocessed_samples = harmony.preprocess_geo(
gse.gsms.items())
harmonized_samples = harmony.harmonize(
preprocessed_samples)
GeoSurveyor._apply_harmonized_metadata_to_sample(
sample, harmonized_samples[sample.title])
elif sample.source_database == "ARRAY_EXPRESS":
SDRF_URL_TEMPLATE = (
"https://www.ebi.ac.uk/arrayexpress/files/{code}/{code}.sdrf.txt"
)
sdrf_url = SDRF_URL_TEMPLATE.format(
code=sample.experiments.first().accession_code)
sdrf_samples = harmony.parse_sdrf(sdrf_url)
harmonized_samples = harmony.harmonize(sdrf_samples)
ArrayExpressSurveyor._apply_harmonized_metadata_to_sample(
sample, harmonized_samples[sample.title])
sample.save()
if not page.has_next():
break
else:
page = paginator.page(page.next_page_number())
# 2000 samples queued up every five minutes should be fast
# enough and also not thrash the DB.
time.sleep(60 * 5)
