def testExtractEntities(self):
"""Test case - extract entities"""
try:
obEx = ObjectExtractor(
self.__cfgOb,
databaseName="pdbx_core",
collectionName="pdbx_core_polymer_entity",
cacheFilePath=os.path.join(self.__workPath,
"entity-data-test-cache.json"),
useCache=False,
keyAttribute="entity",
uniqueAttributes=["rcsb_id"],
cacheKwargs=self.__testEntryCacheKwargs,
objectLimit=self.__objectLimitTest,
)
eCount = obEx.getCount()
logger.info("Entity count is %d", eCount)
self.assertGreaterEqual(eCount, self.__objectLimitTest)
objD = obEx.getObjects()
for _, obj in objD.items():
obEx.genPathList(obj, path=None)
#
pL = obEx.getPathList(filterList=False)
logger.debug("Path list (unfiltered) %r", pL)
#
pL = obEx.getPathList()
logger.debug("Path list %r", pL)
obEx.setPathList(pL)
if self.__verbose:
for ky, obj in objD.items():
obEx.genValueList(obj, path=None)
tD = obEx.getValues()
logger.info("Index object %r %s", ky,
pprint.pformat(tD, indent=3, width=120))
except Exception as e:
logger.exception("Failing with %s", str(e))
self.fail()
def testExtractSelectedEntityContent(self):
"""Test case - extract selected entity content
"reference_sequence_identifiers": [
{
"database_name": "UniProt",
"database_accession": "Q5SHN1",
"provenance_source": "SIFTS"
},
{
"database_name": "UniProt",
"database_accession": "Q5SHN1",
"provenance_source": "PDB"
}
]
"""
try:
obEx = ObjectExtractor(
self.__cfgOb,
databaseName="pdbx_core",
collectionName="pdbx_core_polymer_entity",
cacheFilePath=os.path.join(
self.__workPath,
"entity-selected-content-test-cache.json"),
useCache=False,
keyAttribute="entity",
uniqueAttributes=["rcsb_id"],
cacheKwargs=self.__testEntryCacheKwargs,
# objectLimit=self.__objectLimitTest,
objectLimit=None,
selectionQuery={
"entity_poly.rcsb_entity_polymer_type": "Protein"
},
selectionList=[
"rcsb_id",
"rcsb_polymer_entity_container_identifiers.reference_sequence_identifiers"
],
)
eCount = obEx.getCount()
logger.info("Entity count is %d", eCount)
#
#
if self.__objectLimitTest is not None:
self.assertGreaterEqual(eCount, self.__objectLimitTest)
objD = obEx.getObjects()
for _, obj in objD.items():
obEx.genPathList(obj, path=None)
#
pL = obEx.getPathList(filterList=False)
logger.debug("Path list (unfiltered) %r", pL)
#
pL = obEx.getPathList()
logger.debug("Path list %r", pL)
obEx.setPathList(pL)
if self.__verbose:
for ky, obj in objD.items():
obEx.genValueList(obj, path=None)
tD = obEx.getValues()
logger.info("Index object %r %s", ky,
pprint.pformat(tD, indent=3, width=120))
objD = obEx.getObjects()
# logger.info("objD.keys() %r", list(objD.keys()))
totCount = 0
difCount = 0
pdbUnpIdD = defaultdict(int)
siftsUnpIdD = defaultdict(int)
pdbDifUnpIdD = defaultdict(int)
for entityKey, eD in objD.items():
try:
siftsS = set()
pdbS = set()
for tD in eD["rcsb_polymer_entity_container_identifiers"][
"reference_sequence_identifiers"]:
if tD["database_name"] == "UniProt":
if tD["provenance_source"] == "SIFTS":
siftsS.add(tD["database_accession"])
siftsUnpIdD[tD["database_accession"]] += 1
elif tD["provenance_source"] == "PDB":
pdbS.add(tD["database_accession"])
pdbUnpIdD[tD["database_accession"]] += 1
else:
logger.debug(
"No UniProt for %r",
eD["rcsb_polymer_entity_container_identifiers"]
)
logger.debug("PDB assigned sequence length %d", len(pdbS))
logger.debug("SIFTS assigned sequence length %d",
len(siftsS))
if pdbS and siftsS:
totCount += 1
if pdbS != siftsS:
difCount += 1
for idV in pdbS:
pdbDifUnpIdD[idV] += 1
except Exception as e:
logger.warning("No identifiers for %s with %s", entityKey,
str(e))
logger.info("Total %d differences %d", totCount, difCount)
logger.info("Unique UniProt ids PDB %d SIFTS %d", len(pdbUnpIdD),
len(siftsUnpIdD))
logger.info("Unique UniProt differences %d ", len(pdbDifUnpIdD))
except Exception as e:
logger.exception("Failing with %s", str(e))
self.fail()