def annotation_scheduler():
""" This is run on scheduling_single_worker queue to avoid race conditions """
LOCK_EXPIRE = 60 * 5 # 5 minutes
lock_id = "annotation-scheduler-lock"
# cache.add fails if if the key already exists
acquire_lock = lambda: cache.add(lock_id, "true", LOCK_EXPIRE)
release_lock = lambda: cache.delete(lock_id)
try:
if acquire_lock():
try:
logging.info("Got the lock for annotation scheduler")
for genome_build in GenomeBuild.builds_with_annotation():
annotation_version = AnnotationVersion.latest(genome_build)
variant_annotation_version = annotation_version.variant_annotation_version
while True:
range_lock = _handle_variant_annotation_version(
variant_annotation_version)
if range_lock is None:
break
finally:
logging.info("Releasing lock")
release_lock()
else:
logging.info("Someone else has %s", lock_id)
except:
log_traceback()
def wait_for_node(node_id):
""" Used to build a dependency on a node that's already loading.
The danger here is that we'll end up taking up a celery worker, waiting for
the parent to become available... so we have to be careful
"""
EVENT_NAME = "wait_for_node"
#logging.info("wait_for_node: %s", node_id)
try:
TIME_BETWEEN_CHECKS = [5, 5, 10, 10, 30, 30, 60, MINUTE_SECS * 2]
total_time = 0
for sleep_time in TIME_BETWEEN_CHECKS:
# Any exception will exit this task which is ok
node = AnalysisNode.objects.get(pk=node_id)
#logging.info("node status: %s", node.status)
if NodeStatus.is_ready(node.status):
#logging.info("Node was ready")
return
try:
node_task = NodeTask.objects.get(node=node,
version=node.version)
if node_task.celery_task:
wait_for_task(node_task.celery_task)
return
except NodeTask.DoesNotExist:
pass
# QUEUED - there won't be a celery task yet
if node.status == NodeStatus.QUEUED:
details = f"Waiting on parent node {node_id} which is QUEUED"
details += f" - waiting for {sleep_time} secs, {total_time} so far!"
create_event(None,
EVENT_NAME,
details,
severity=LogLevel.WARNING)
else:
details = f"Waiting on parent node {node_id} status {node.status} no celery task!"
create_event(None,
EVENT_NAME,
details,
severity=LogLevel.ERROR)
return
logging.info("Sleeping for %d seconds", sleep_time)
sleep(sleep_time)
total_time += sleep_time
# Timeout
details = f"Waited on parent node {node_id} for {total_time} seconds. "
details += "Didn't become available, dying so we don't cause a deadlock!"
create_event(None, EVENT_NAME, details, severity=LogLevel.ERROR)
except:
log_traceback()
def delete_annotation_run(annotation_run_id):
try:
annotation_run = AnnotationRun.objects.get(pk=annotation_run_id)
annotation_run.status = AnnotationStatus.DELETING
annotation_run.save()
annotation_run.delete()
except:
log_traceback()
raise
def create_and_launch_analysis_tasks(analysis_id, run_async=True):
""" This is run in a single worker queue so that we avoid race conditions"""
try:
tasks = _get_analysis_update_tasks(analysis_id)
except:
log_traceback()
raise
for t in tasks:
execute_task(t, run_async=run_async)
def _verify_somalier_config() -> Optional[str]:
somalier_cfg = SomalierConfig()
somalier_bin = somalier_cfg.get_annotation("command")
somalier = None
try:
somalier_output = check_output([somalier_bin],
stderr=subprocess.STDOUT)
somalier = somalier_output.decode().split("\n", 1)[0]
except:
log_traceback()
return somalier
class NodeJSONViewMixin(View, ABC):
@abstractmethod
def _get_node(self, request, **kwargs) -> AnalysisNode:
pass
@abstractmethod
def _get_data(self, request, node: AnalysisNode, *args, **kwargs):
pass
def _get_redirect(self, request, node: AnalysisNode):
return None
def get_response(self, request, *args, **kwargs):
try:
node = self._get_node(request, **kwargs)
if ret := self._get_redirect(request, node):
return ret
data = self._get_data(request, *args, node=node, **kwargs)
status = HTTP_200_OK
except NonFatalNodeError as e:
log_traceback()
data = {"message": str(e), "non_fatal": True}
status = e.status
if isinstance(e, NodeNotFoundException):
# Is the node deleted, or perhaps it was just out of version?
if e.node_id:
if not AnalysisNode.objects.filter(pk=e.node_id).exists():
data["deleted_nodes"] = [e.node_id]
else:
# We could potentially redirect to the same page with new version
# but client will probably just ditch this and re-request latest anyway
pass
def annotation_versions(request):
anno_versions = {}
# Create VariantAnnotationVersion for build if not exists
for genome_build in GenomeBuild.builds_with_annotation():
try:
get_variant_annotation_version(genome_build)
except:
log_traceback()
qs = AnnotationVersion.objects.filter(
genome_build=genome_build).order_by("-annotation_date")
vep_command = get_vep_command("in.vcf", "out.vcf", genome_build,
genome_build.annotation_consortium)
vep_command = " ".join(vep_command).replace(" -", "\n")
anno_versions[genome_build.name] = (vep_command, qs)
context = {"annotation_versions": anno_versions}
return render(request, "annotation/annotation_versions.html", context)
def node_cache_task(node_id, version):
logging.info("node_cache_task: %s/%d", node_id, version)
try:
node = AnalysisNode.objects.get_subclass(pk=node_id, version=version)
except AnalysisNode.DoesNotExist: # @UndefinedVariable
# Node was deleted - this task is obsolete
return
node_cache = NodeCache.objects.get(node_version=node.node_version)
variant_collection = node_cache.variant_collection
if variant_collection.status != ProcessingStatus.CREATED:
return
if not (node.is_valid() and node.modifies_parents()):
logging.debug("Not doing anything for node %s", node.pk)
variant_collection.status = ProcessingStatus.SKIPPED
variant_collection.save()
return
variant_collection.status = ProcessingStatus.PROCESSING
variant_collection.save()
# Node could be READY, we're just generating cache afterwards (node doesn't need but children do)
if node.status != NodeStatus.READY:
node.set_node_task_and_status(node_cache_task.request.id,
NodeStatus.LOADING_CACHE)
try:
node.write_cache(variant_collection)
processing_status = ProcessingStatus.SUCCESS
status = NodeStatus.LOADING
except:
log_traceback()
processing_status = ProcessingStatus.ERROR
status = NodeStatus.ERROR
variant_collection.status = processing_status
variant_collection.save()
if node.status != NodeStatus.READY:
node.status = status
node.save()
if hl:
func, arg = hl
try:
match_type, records = func(arg)
if match_type == PhenotypeMatchTypes.HPO:
hpo_list.extend(records)
elif match_type == PhenotypeMatchTypes.OMIM:
omim_alias_list.extend(records)
elif match_type == PhenotypeMatchTypes.GENE:
gene_symbols.extend(records)
# logging.info("Got exact: %s => %s", text, records)
except Exception as e:
msg = f"Error: {e}, func: {func}, arg={arg}"
logging.error(msg)
log_traceback()
# raise ValueError(msg)
return hpo_list, omim_alias_list, gene_symbols
@classmethod
def _skip_word(cls, lower_text):
""" Return true to skip a word, throws SkipAllPhenotypeMatchException to skip all.
Only need to skip >MIN_LENGTH words as will do that later (after exact) """
# For multi-words where you want to skip components
SKIP_ALL = {
"library prep", "to cgf", "set up", "ad pattern", "recurrent eps",
"rest of"
}
if lower_text in SKIP_ALL:
def get_evidence_fields_from_preferred_transcript(
genome_build: GenomeBuild,
variant: Variant,
clingen_allele: ClinGenAllele,
hgvs_matcher: HGVSMatcher,
transcript_version: TranscriptVersion,
evidence_transcript_columns,
ekey_formatters,
annotation_version) -> AutopopulateData:
data = transcript_autopopulate(transcript_version)
data[SpecialEKeys.GENE_SYMBOL] = transcript_version.gene_version.gene_symbol_id
# Populate from TranscriptVersion data 1st (so can overwrite later)
if variant:
try:
vts = VariantTranscriptSelections(variant, genome_build, annotation_version)
transcript_data = vts.get_transcript_annotation(transcript_version)
for evidence_key, transcript_config in evidence_transcript_columns.items():
variant_column = transcript_config['col']
immutable = transcript_config['immutable']
# Getting out of dict directly, not joining via queryset
transcript_column = variant_column.replace("variantannotation__", "")
value = transcript_data.get(transcript_column)
if value is not None:
set_evidence(data, evidence_key, value, immutable, ekey_formatters)
gene_symbol = transcript_version.gene_version.gene_symbol
data[SpecialEKeys.INTERNAL_SAMPLES_20X_COVERAGE] = get_20x_gene_coverage(gene_symbol)
phastcons_dict = {
"phastcons_30_way_mammalian": "30 way mammalian",
"phastcons_46_way_mammalian": "46 way mammalian",
"phastcons_100_way_vertebrate": "100 way vertebrate",
}
data[SpecialEKeys.PHASTCONS] = get_set_fields_summary(transcript_data, phastcons_dict, phastcons_dict)
phylop_dict = {
"phylop_30_way_mammalian": "30 way mammalian",
"phylop_46_way_mammalian": "46 way mammalian",
"phylop_100_way_vertebrate": "100 way vertebrate",
}
data[SpecialEKeys.PHYLOP] = get_set_fields_summary(transcript_data, phylop_dict, phylop_dict)
if variant_annotation := vts.variant_annotation:
data[SpecialEKeys.SEARCH_TERMS] = variant_annotation.get_search_terms()
if settings.ANNOTATION_PUBMED_SEARCH_TERMS_ENABLED:
data[SpecialEKeys.PUBMED_SEARCH_TERMS] = variant_annotation.get_pubmed_search_terms()
except:
log_traceback()
try:
c_hgvs = hgvs_matcher.variant_to_c_hgvs_parts(variant, transcript_version.accession)
if c_hgvs:
data[SpecialEKeys.C_HGVS] = c_hgvs.full_c_hgvs
except Exception as e:
value_obj = {}
data.message = 'Could not parse HGVS value %s' % str(e)
data[SpecialEKeys.C_HGVS] = value_obj
if clingen_allele:
p_hgvs = clingen_allele.get_p_hgvs(transcript_version.accession, match_version=False)
else:
p_hgvs = None
data[SpecialEKeys.P_HGVS] = p_hgvs
gene_symbol_id = transcript_version.gene_version.gene_symbol_id
gnomad_oe_lof_summary = get_gnomad_oe_lof_summary(transcript_version)
if gnomad_oe_lof_summary:
data[SpecialEKeys.GNOMAD_OE_LOF] = gnomad_oe_lof_summary
gs_count = GeneSymbolPubMedCount.get_for_gene_symbol(gene_symbol_id)
data[SpecialEKeys.PUBMED_GENE_SEARCH_COUNT] = {"value": gs_count.count,
"note": f"Retrieved {gs_count.modified.date()}"}
return data
