def toProtocolElement(self):
dataset = protocol.Dataset()
dataset.id = self.getId()
dataset.name = pb.string(self.getLocalId())
dataset.description = pb.string(self.getDescription())
self.serializeAttributes(dataset)
return dataset
def _gaFeatureForFeatureDbRecord(self, feature):
"""
:param feature: The DB Row representing a feature
:return: the corresponding GA4GH protocol.Feature object
"""
gaFeature = protocol.Feature()
gaFeature.id = self.getCompoundIdForFeatureId(feature['id'])
if feature.get('parent_id'):
gaFeature.parent_id = self.getCompoundIdForFeatureId(
feature['parent_id'])
else:
gaFeature.parent_id = ""
gaFeature.feature_set_id = self.getId()
gaFeature.reference_name = pb.string(feature.get('reference_name'))
gaFeature.start = pb.int(feature.get('start'))
gaFeature.end = pb.int(feature.get('end'))
gaFeature.name = pb.string(feature.get('name'))
if feature.get('strand', '') == '-':
gaFeature.strand = protocol.NEG_STRAND
else:
# default to positive strand
gaFeature.strand = protocol.POS_STRAND
gaFeature.child_ids.extend(
map(self.getCompoundIdForFeatureId,
json.loads(feature['child_ids'])))
gaFeature.feature_type.CopyFrom(
self._ontology.getGaTermByName(feature['type']))
attributes = json.loads(feature['attributes'])
# TODO: Identify which values are ExternalIdentifiers and OntologyTerms
for key in attributes:
for v in attributes[key]:
gaFeature.attributes.attr[key].values.add().string_value = v
if 'gene_name' in attributes and len(attributes['gene_name']) > 0:
gaFeature.gene_symbol = pb.string(attributes['gene_name'][0])
return gaFeature
def search_reference_sets(self,
accession=None,
md5checksum=None,
assembly_id=None):
"""
Returns an iterator over the ReferenceSets fulfilling the specified
conditions.
:param str accession: If not null, return the reference sets for which
the `accession` matches this string (case-sensitive, exact match).
:param str md5checksum: If not null, return the reference sets for
which the `md5checksum` matches this string (case-sensitive, exact
match). See :class:`ga4gh.protocol.ReferenceSet::md5checksum` for
details.
:param str assembly_id: If not null, return the reference sets for
which the `assembly_id` matches this string (case-sensitive,
exact match).
:return: An iterator over the :class:`ga4gh.protocol.ReferenceSet`
objects defined by the query parameters.
"""
request = protocol.SearchReferenceSetsRequest()
request.accession = pb.string(accession)
request.md5checksum = pb.string(md5checksum)
request.assembly_id = pb.string(assembly_id)
request.page_size = pb.int(self._page_size)
return self._run_search_request(request, "referencesets",
protocol.SearchReferenceSetsResponse)
def toProtocolElement(self):
dataset = protocol.Dataset()
dataset.id = self.getId()
dataset.name = pb.string(self.getLocalId())
dataset.description = pb.string(self.getDescription())
self.serializeAttributes(dataset)
return dataset
def _gaFeatureForFeatureDbRecord(self, feature):
"""
:param feature: The DB Row representing a feature
:return: the corresponding GA4GH protocol.Feature object
"""
gaFeature = protocol.Feature()
gaFeature.id = self.getCompoundIdForFeatureId(feature['id'])
if feature.get('parent_id'):
gaFeature.parent_id = self.getCompoundIdForFeatureId(
feature['parent_id'])
else:
gaFeature.parent_id = ""
gaFeature.feature_set_id = self.getId()
gaFeature.reference_name = pb.string(feature.get('reference_name'))
gaFeature.start = pb.int(feature.get('start'))
gaFeature.end = pb.int(feature.get('end'))
gaFeature.name = pb.string(feature.get('name'))
if feature.get('strand', '') == '-':
gaFeature.strand = protocol.NEG_STRAND
else:
# default to positive strand
gaFeature.strand = protocol.POS_STRAND
gaFeature.child_ids.extend(map(
self.getCompoundIdForFeatureId,
json.loads(feature['child_ids'])))
gaFeature.feature_type.CopyFrom(
self._ontology.getGaTermByName(feature['type']))
attributes = json.loads(feature['attributes'])
# TODO: Identify which values are ExternalIdentifiers and OntologyTerms
for key in attributes:
for v in attributes[key]:
gaFeature.attributes.attr[key].values.add().string_value = v
if 'gene_name' in attributes and len(attributes['gene_name']) > 0:
gaFeature.gene_symbol = pb.string(attributes['gene_name'][0])
return gaFeature
def toProtocolElement(self):
"""
Returns the GA4GH protocol representation of this ReadGroup.
"""
# TODO this is very incomplete, but we don't have the
# implementation to fill out the rest of the fields currently
readGroup = protocol.ReadGroup()
readGroup.id = self.getId()
readGroup.created = self._creationTime
readGroup.updated = self._updateTime
dataset = self.getParentContainer().getParentContainer()
readGroup.dataset_id = dataset.getId()
readGroup.name = self.getLocalId()
readGroup.predicted_insert_size = pb.int(self.getPredictedInsertSize())
referenceSet = self._parentContainer.getReferenceSet()
readGroup.sample_name = pb.string(self.getSampleName())
readGroup.biosample_id = pb.string(self.getBiosampleId())
if referenceSet is not None:
readGroup.reference_set_id = referenceSet.getId()
readGroup.stats.CopyFrom(self.getStats())
readGroup.programs.extend(self.getPrograms())
readGroup.description = pb.string(self.getDescription())
readGroup.experiment.CopyFrom(self.getExperiment())
self.serializeAttributes(readGroup)
return readGroup
def toProtocolElement(self):
"""
Returns the GA4GH protocol representation of this ReadGroup.
"""
# TODO this is very incomplete, but we don't have the
# implementation to fill out the rest of the fields currently
readGroup = protocol.ReadGroup()
readGroup.id = self.getId()
readGroup.created = self._creationTime
readGroup.updated = self._updateTime
dataset = self.getParentContainer().getParentContainer()
readGroup.dataset_id = dataset.getId()
readGroup.name = self.getLocalId()
readGroup.predicted_insert_size = pb.int(self.getPredictedInsertSize())
referenceSet = self._parentContainer.getReferenceSet()
readGroup.sample_name = pb.string(self.getSampleName())
readGroup.biosample_id = pb.string(self.getBiosampleId())
if referenceSet is not None:
readGroup.reference_set_id = referenceSet.getId()
readGroup.stats.CopyFrom(self.getStats())
readGroup.programs.extend(self.getPrograms())
readGroup.description = pb.string(self.getDescription())
readGroup.experiment.CopyFrom(self.getExperiment())
self.serializeAttributes(readGroup)
return readGroup
def toProtocolElement(self):
dataset = protocol.Dataset()
dataset.id = self.getId()
dataset.name = pb.string(self.getLocalId())
dataset.description = pb.string(self.getDescription())
for key in self.getInfo():
dataset.info[key].values.extend(_encodeValue(self._info[key]))
return dataset
def toProtocolElement(self):
"""
Returns the GA4GH protocol representation of this ReferenceSet.
"""
ret = protocol.ReferenceSet()
ret.assembly_id = pb.string(self.getAssemblyId())
ret.description = pb.string(self.getDescription())
ret.id = self.getId()
ret.is_derived = self.getIsDerived()
ret.md5checksum = self.getMd5Checksum()
ret.ncbi_taxon_id = pb.int(self.getNcbiTaxonId())
ret.source_accessions.extend(self.getSourceAccessions())
ret.source_uri = pb.string(self.getSourceUri())
ret.name = self.getLocalId()
return ret
def testToProtocolElement(self):
dataset = datasets.Dataset('dataset1')
term = protocol.OntologyTerm()
term.term = "male genotypic sex"
term.id = "PATO:0020001"
term.source_name = "PATO"
term.source_version = pb.string("2015-11-18")
# Write out a valid input
print(protocol.toJsonDict(term))
validIndividual = protocol.Individual(
name="test",
created="2016-05-19T21:00:19Z",
updated="2016-05-19T21:00:19Z",
sex=term)
validIndividual.info['test'].values.add().string_value = 'test-info'
# pass through protocol creation
individual = bioMetadata.Individual(
dataset, "test")
individual.populateFromJson(protocol.toJson(validIndividual))
gaIndividual = individual.toProtocolElement()
# Verify elements exist
self.assertEqual(gaIndividual.created, validIndividual.created)
self.assertEqual(gaIndividual.updated, validIndividual.updated)
# Invalid input
invalidIndividual = '{"bad:", "json"}'
individual = bioMetadata.Individual(dataset, "test")
# Should fail
self.assertRaises(
exceptions.InvalidJsonException,
individual.populateFromJson,
invalidIndividual)
def search_reads(self,
read_group_ids,
reference_id=None,
start=None,
end=None):
"""
Returns an iterator over the Reads fulfilling the specified
conditions from the specified read_group_ids.
:param str read_group_ids: The IDs of the
:class:`ga4gh.protocol.ReadGroup` of interest.
:param str reference_id: The name of the
:class:`ga4gh.protocol.Reference` we wish to return reads
mapped to.
:param int start: The start position (0-based) of this query. If a
reference is specified, this defaults to 0. Genomic positions are
non-negative integers less than reference length. Requests spanning
the join of circular genomes are represented as two requests one on
each side of the join (position 0).
:param int end: The end position (0-based, exclusive) of this query.
If a reference is specified, this defaults to the reference's
length.
:return: An iterator over the
:class:`ga4gh.protocol.ReadAlignment` objects defined by
the query parameters.
:rtype: iter
"""
request = protocol.SearchReadsRequest()
request.read_group_ids.extend(read_group_ids)
request.reference_id = pb.string(reference_id)
request.start = pb.int(start)
request.end = pb.int(end)
request.page_size = pb.int(self._page_size)
return self._run_search_request(request, "reads",
protocol.SearchReadsResponse)
def getSerializedResponse(self):
"""
Returns a string version of the SearchResponse that has
been built by this SearchResponseBuilder.
"""
self._protoObject.next_page_token = pb.string(self._nextPageToken)
s = protocol.serialize(self._protoObject, self._return_mimetype)
return s
def getSerializedResponse(self):
"""
Returns a string version of the SearchResponse that has
been built by this SearchResponseBuilder.
"""
self._protoObject.next_page_token = pb.string(self._nextPageToken)
s = protocol.toJson(self._protoObject)
return s
def search_read_group_sets(self, dataset_id, name=None, biosample_id=None):
"""
Returns an iterator over the ReadGroupSets fulfilling the specified
conditions from the specified Dataset.
:param str name: Only ReadGroupSets matching the specified name
will be returned.
:param str biosample_id: Only ReadGroups matching the specified
biosample will be included in the response.
:return: An iterator over the :class:`ga4gh.protocol.ReadGroupSet`
objects defined by the query parameters.
:rtype: iter
"""
request = protocol.SearchReadGroupSetsRequest()
request.dataset_id = dataset_id
request.name = pb.string(name)
request.biosample_id = pb.string(biosample_id)
request.page_size = pb.int(self._page_size)
return self._run_search_request(request, "readgroupsets",
protocol.SearchReadGroupSetsResponse)
def toProtocolElement(self, tier=0):
"""
Returns the GA4GH protocol representation of this ReferenceSet.
"""
ret = protocol.ReferenceSet()
ret.assembly_id = pb.string(self.getAssemblyId())
ret.description = pb.string(self.getDescription())
ret.id = self.getId()
ret.is_derived = self.getIsDerived()
ret.md5checksum = self.getMd5Checksum()
if self.getSpecies():
term = protocol.fromJson(json.dumps(self.getSpecies()),
protocol.OntologyTerm)
ret.species.term_id = term.term_id
ret.species.term = term.term
ret.source_accessions.extend(self.getSourceAccessions())
ret.source_uri = pb.string(self.getSourceUri())
ret.name = self.getLocalId()
self.serializeAttributes(ret)
return ret
def toProtocolElement(self):
"""
Returns the GA4GH protocol representation of this ReferenceSet.
"""
ret = protocol.ReferenceSet()
ret.assembly_id = pb.string(self.getAssemblyId())
ret.description = pb.string(self.getDescription())
ret.id = self.getId()
ret.is_derived = self.getIsDerived()
ret.md5checksum = self.getMd5Checksum()
if self.getSpecies():
term = protocol.fromJson(
json.dumps(self.getSpecies()), protocol.OntologyTerm)
ret.species.term_id = term.term_id
ret.species.term = term.term
ret.source_accessions.extend(self.getSourceAccessions())
ret.source_uri = pb.string(self.getSourceUri())
ret.name = self.getLocalId()
self.serializeAttributes(ret)
return ret
def search_bio_samples(self, dataset_id, name=None, individual_id=None):
"""
Returns an iterator over the BioSamples fulfilling the specified
conditions.
:param str dataset_id: The dataset to search within.
:param str name: Only BioSamples matching the specified name will
be returned.
:param str individual_id: Only BioSamples matching matching this
id will be returned.
:return: An iterator over the :class:`ga4gh.protocol.BioSample`
objects defined by the query parameters.
"""
request = protocol.SearchBioSamplesRequest()
request.dataset_id = dataset_id
request.name = pb.string(name)
request.individual_id = pb.string(individual_id)
request.page_size = pb.int(self._page_size)
return self._run_search_request(request, "biosamples",
protocol.SearchBioSamplesResponse)
def search_call_sets(self, variant_set_id, name=None, bio_sample_id=None):
"""
Returns an iterator over the CallSets fulfilling the specified
conditions from the specified VariantSet.
:param str variant_set_id: Find callsets belonging to the
provided variant set.
:param str name: Only CallSets matching the specified name will
be returned.
:param str bio_sample_id: Only CallSets matching this id will
be returned.
:return: An iterator over the :class:`ga4gh.protocol.CallSet`
objects defined by the query parameters.
"""
request = protocol.SearchCallSetsRequest()
request.variant_set_id = variant_set_id
request.name = pb.string(name)
request.bio_sample_id = pb.string(bio_sample_id)
request.page_size = pb.int(self._page_size)
return self._run_search_request(request, "callsets",
protocol.SearchCallSetsResponse)
def search_variants(self,
variant_set_id,
start=None,
end=None,
reference_name=None,
call_set_ids=None):
"""
Returns an iterator over the Variants fulfilling the specified
conditions from the specified VariantSet.
:param str variant_set_id: The ID of the
:class:`ga4gh.protocol.VariantSet` of interest.
:param int start: Required. The beginning of the window (0-based,
inclusive) for which overlapping variants should be returned.
Genomic positions are non-negative integers less than reference
length. Requests spanning the join of circular genomes are
represented as two requests one on each side of the join
(position 0).
:param int end: Required. The end of the window (0-based, exclusive)
for which overlapping variants should be returned.
:param str reference_name: The name of the
:class:`ga4gh.protocol.Reference` we wish to return variants from.
:param list call_set_ids: Only return variant calls which belong to
call sets with these IDs. If an empty array, returns variants
without any call objects. If null, returns all variant calls.
:return: An iterator over the :class:`ga4gh.protocol.Variant` objects
defined by the query parameters.
:rtype: iter
"""
request = protocol.SearchVariantsRequest()
request.reference_name = pb.string(reference_name)
request.start = pb.int(start)
request.end = pb.int(end)
request.variant_set_id = variant_set_id
request.call_set_ids.extend(pb.string(call_set_ids))
request.page_size = pb.int(self._page_size)
return self._run_search_request(request, "variants",
protocol.SearchVariantsResponse)
def toProtocolElement(self):
"""
Returns the representation of this FeatureSet as the corresponding
ProtocolElement.
"""
gaFeatureSet = protocol.FeatureSet()
gaFeatureSet.id = self.getId()
gaFeatureSet.dataset_id = self.getParentContainer().getId()
gaFeatureSet.reference_set_id = pb.string(self._referenceSet.getId())
gaFeatureSet.name = self._name
gaFeatureSet.source_uri = self._sourceUri
for key in self._info:
gaFeatureSet.info[key].values.extend(self._info[key])
return gaFeatureSet
def search_references(self,
reference_set_id,
accession=None,
md5checksum=None):
"""
Returns an iterator over the References fulfilling the specified
conditions from the specified Dataset.
:param str reference_set_id: The ReferenceSet to search.
:param str accession: If not None, return the references for which the
`accession` matches this string (case-sensitive, exact match).
:param str md5checksum: If not None, return the references for which
the `md5checksum` matches this string (case-sensitive, exact
match).
:return: An iterator over the :class:`ga4gh.protocol.Reference`
objects defined by the query parameters.
"""
request = protocol.SearchReferencesRequest()
request.reference_set_id = reference_set_id
request.accession = pb.string(accession)
request.md5checksum = pb.string(md5checksum)
request.page_size = pb.int(self._page_size)
return self._run_search_request(request, "references",
protocol.SearchReferencesResponse)
def toProtocolElement(self):
"""
Returns the representation of this FeatureSet as the corresponding
ProtocolElement.
"""
gaFeatureSet = protocol.FeatureSet()
gaFeatureSet.id = self.getId()
gaFeatureSet.dataset_id = self.getParentContainer().getId()
gaFeatureSet.reference_set_id = pb.string(self._referenceSet.getId())
gaFeatureSet.name = self._name
gaFeatureSet.source_uri = self._sourceUri
attributes = self.getAttributes()
for key in attributes:
gaFeatureSet.attributes.attr[key] \
.values.extend(protocol.encodeValue(attributes[key]))
return gaFeatureSet
def toProtocolElement(self):
"""
Returns the representation of this FeatureSet as the corresponding
ProtocolElement.
"""
gaFeatureSet = protocol.FeatureSet()
gaFeatureSet.id = self.getId()
gaFeatureSet.dataset_id = self.getParentContainer().getId()
gaFeatureSet.reference_set_id = pb.string(self._referenceSet.getId())
gaFeatureSet.name = self._name
gaFeatureSet.source_uri = self._sourceUri
attributes = self.getAttributes()
for key in attributes:
gaFeatureSet.attributes.attr[key] \
.values.extend(protocol.encodeValue(attributes[key]))
return gaFeatureSet
def getExperiment(self):
"""
Returns the GA4GH protocol representation of this read group's
Experiment.
"""
experiment = protocol.Experiment()
experiment.id = self.getExperimentId()
experiment.instrument_model = pb.string(self.getInstrumentModel())
experiment.sequencing_center = pb.string(self.getSequencingCenter())
experiment.description = pb.string(self.getExperimentDescription())
experiment.library = pb.string(self.getLibrary())
experiment.platform_unit = pb.string(self.getPlatformUnit())
experiment.message_create_time = self._iso8601
experiment.message_update_time = self._iso8601
experiment.run_time = pb.string(self.getRunTime())
return experiment
def getExperiment(self):
"""
Returns the GA4GH protocol representation of this read group's
Experiment.
"""
experiment = protocol.Experiment()
experiment.id = self.getExperimentId()
experiment.instrument_model = pb.string(self.getInstrumentModel())
experiment.sequencing_center = pb.string(self.getSequencingCenter())
experiment.description = pb.string(self.getExperimentDescription())
experiment.library = pb.string(self.getLibrary())
experiment.platform_unit = pb.string(self.getPlatformUnit())
experiment.message_create_time = self._iso8601
experiment.message_update_time = self._iso8601
experiment.run_time = pb.string(self.getRunTime())
return experiment
def getGaTermByName(self, name):
"""
Returns a GA4GH OntologyTerm object by name.
:param name: name of the ontology term, ex. "gene".
:return: GA4GH OntologyTerm object.
"""
# TODO what is the correct value when we have no mapping??
termIds = self.getTermIds(name)
if len(termIds) == 0:
termId = ""
# TODO add logging for missed term translation.
else:
# TODO what is the correct behaviour here when we have multiple
# IDs matching a given name?
termId = termIds[0]
term = protocol.OntologyTerm()
term.term = name
term.id = termId
term.source_name = self._sourceName
term.source_version = pb.string(self._sourceVersion)
return term
def testString(self):
self.assertEqual(pb.DEFAULT_STRING, pb.string(None))
self.assertEqual('A', pb.string('A'))
def testString(self):
self.assertEqual(pb.DEFAULT_STRING, pb.string(None))
self.assertEqual('A', pb.string('A'))