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 = [
_build_ena_file_url(run_accession, "_1"),
_build_ena_file_url(run_accession, "_2"),
]
else:
files_urls = [_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
SraSurveyor._apply_metadata_to_experiment(experiment_object,
metadata)
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"
SraSurveyor._apply_harmonized_metadata_to_sample(
sample_object, metadata)
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]
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
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 _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_from_api(
self, experiment_accession_code: str) -> (Experiment, Dict):
"""Given an experiment accession code, create an Experiment object.
Also returns a dictionary of additional information about the
platform discovered for the experiment.
Will raise an UnsupportedPlatformException if this experiment was
conducted using a platform which we don't support.
See an example at: https://www.ebi.ac.uk/arrayexpress/json/v3/experiments/E-MTAB-3050/sample
"""
request_url = EXPERIMENTS_URL + experiment_accession_code
experiment_request = utils.requests_retry_session().get(request_url,
timeout=60)
try:
parsed_json = experiment_request.json(
)["experiments"]["experiment"][0]
except KeyError:
logger.error("Remote experiment has no Experiment data!",
experiment_accession_code=experiment_accession_code,
survey_job=self.survey_job.id)
raise
experiment = {}
experiment["name"] = parsed_json["name"]
experiment["experiment_accession_code"] = experiment_accession_code
# This experiment has no platform at all, and is therefore useless.
if 'arraydesign' not in parsed_json or len(
parsed_json["arraydesign"]) == 0:
logger.warn("Remote experiment has no arraydesign listed.",
experiment_accession_code=experiment_accession_code,
survey_job=self.survey_job.id)
raise UnsupportedPlatformException
# If there is more than one arraydesign listed in the experiment
# then there is no other way to determine which array was used
# for which sample other than looking at the header of the CEL
# file. That obviously cannot happen until the CEL file has been
# downloaded so we can just mark it as UNKNOWN and let the
# downloader inspect the downloaded file to determine the
# array then.
elif len(parsed_json["arraydesign"]
) != 1 or "accession" not in parsed_json["arraydesign"][0]:
experiment["platform_accession_code"] = UNKNOWN
experiment["platform_accession_name"] = UNKNOWN
experiment["manufacturer"] = UNKNOWN
else:
external_accession = parsed_json["arraydesign"][0]["accession"]
for platform in get_supported_microarray_platforms():
if platform["external_accession"] == external_accession:
experiment[
"platform_accession_code"] = get_normalized_platform(
platform["platform_accession"])
# Illumina appears in the accession codes for
# platforms manufactured by Illumina
if "ILLUMINA" in experiment[
"platform_accession_code"].upper():
experiment["manufacturer"] = "ILLUMINA"
experiment["platform_accession_name"] = platform[
"platform_accession"]
else:
# It's not Illumina, the only other supported Microarray platform is
# Affy. As our list of supported platforms grows this logic will
# need to get more sophisticated.
experiment["manufacturer"] = "AFFYMETRIX"
platform_mapping = get_readable_affymetrix_names()
experiment[
"platform_accession_name"] = platform_mapping[
platform["platform_accession"]]
if "platform_accession_code" not in experiment:
# We don't know what platform this accession corresponds to.
experiment["platform_accession_code"] = external_accession
experiment["platform_accession_name"] = UNKNOWN
experiment["manufacturer"] = UNKNOWN
experiment["release_date"] = parsed_json["releasedate"]
if "lastupdatedate" in parsed_json:
experiment["last_update_date"] = parsed_json["lastupdatedate"]
else:
experiment["last_update_date"] = parsed_json["releasedate"]
# Create the experiment object
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:
# We aren't sure these fields will be populated, or how many there will be.
# Try to join them all together, or set a sensible default.
experiment_descripton = ""
if "description" in parsed_json and len(
parsed_json["description"]) > 0:
for description_item in parsed_json["description"]:
if "text" in description_item:
experiment_descripton = experiment_descripton + description_item[
"text"] + "\n"
if experiment_descripton == "":
experiment_descripton = "Description not available.\n"
experiment_object = Experiment()
experiment_object.accession_code = experiment_accession_code
experiment_object.source_url = request_url
experiment_object.source_database = "ARRAY_EXPRESS"
experiment_object.title = parsed_json["name"]
# This will need to be updated if we ever use Array
# Express to get other kinds of data.
experiment_object.technology = "MICROARRAY"
experiment_object.description = experiment_descripton
experiment_object.source_first_published = parse_datetime(
experiment["release_date"])
experiment_object.source_last_modified = parse_datetime(
experiment["last_update_date"])
experiment_object.save()
json_xa = ExperimentAnnotation()
json_xa.experiment = experiment_object
json_xa.data = parsed_json
json_xa.is_ccdl = False
json_xa.save()
## Fetch and parse the IDF/SDRF file for any other fields
IDF_URL_TEMPLATE = "https://www.ebi.ac.uk/arrayexpress/files/{code}/{code}.idf.txt"
idf_url = IDF_URL_TEMPLATE.format(code=experiment_accession_code)
idf_text = utils.requests_retry_session().get(idf_url,
timeout=60).text
lines = idf_text.split('\n')
idf_dict = {}
for line in lines:
keyval = line.strip().split('\t')
if len(keyval) == 2:
idf_dict[keyval[0]] = keyval[1]
elif len(keyval) > 2:
idf_dict[keyval[0]] = keyval[1:]
idf_xa = ExperimentAnnotation()
idf_xa.data = idf_dict
idf_xa.experiment = experiment_object
idf_xa.is_ccdl = False
idf_xa.save()
if 'Investigation Title' in idf_dict:
experiment_object.title = idf_dict['Investigation Title']
if 'Person Affiliation' in idf_dict:
# This is very rare, ex: E-MEXP-32
if isinstance(idf_dict['Person Affiliation'], list):
unique_people = list(set(idf_dict['Person Affiliation']))
experiment_object.submitter_institution = ", ".join(
unique_people)[:255]
else:
experiment_object.submitter_institution = idf_dict[
'Person Affiliation']
# Get protocol_description from "<experiment_url>/protocols"
# instead of from idf_dict, because the former provides more
# details.
protocol_url = request_url + '/protocols'
protocol_request = utils.requests_retry_session().get(protocol_url,
timeout=60)
try:
experiment_object.protocol_description = protocol_request.json(
)['protocols']
except KeyError:
logger.warning(
"Remote experiment has no protocol data!",
experiment_accession_code=experiment_accession_code,
survey_job=self.survey_job.id)
if 'Publication Title' in idf_dict:
# This will happen for some superseries.
# Ex: E-GEOD-29536
# Assume most recent is "best:, store the rest in experiment annotation.
if isinstance(idf_dict['Publication Title'], list):
experiment_object.publication_title = "; ".join(
idf_dict['Publication Title'])
else:
experiment_object.publication_title = idf_dict[
'Publication Title']
experiment_object.has_publication = True
if 'Publication DOI' in idf_dict:
if isinstance(idf_dict['Publication DOI'], list):
experiment_object.publication_doi = ", ".join(
idf_dict['Publication DOI'])
else:
experiment_object.publication_doi = idf_dict[
'Publication DOI']
experiment_object.has_publication = True
if 'PubMed ID' in idf_dict:
if isinstance(idf_dict['PubMed ID'], list):
experiment_object.pubmed_id = ", ".join(
idf_dict['PubMed ID'])
else:
experiment_object.pubmed_id = idf_dict['PubMed ID']
experiment_object.has_publication = True
# 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()
platform_dict = {}
for k in ('platform_accession_code', 'platform_accession_name',
'manufacturer'):
platform_dict[k] = experiment[k]
return experiment_object, platform_dict
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