if profileReference == "begin":
lastwin = requiredWindows[-1]
elif profileReference == "end":
lastwin = requiredWindows[0]
elif profileReference == "stop3utr":
lastwin = maxNumWindows
else:
assert (False)
if (shufflesWithMissingWindows):
#cds.enqueueForProcessing(computationTag, shufflesWithMissingWindows, lastwin, windowStep)#
shuffleIdsToProcess = sorted(shufflesWithMissingWindows +
completelyMissingShuffles)
allSeqIds = cds.shuffledSeqIds(shuffleType)
def shuffleIdToSeqId(shuffleId):
if shuffleId == -1:
return cds.seqId()
else:
return allSeqIds[shuffleId]
requiredSeqIds = list(map(shuffleIdToSeqId, shuffleIdsToProcess))
queueItem = "%d/%s/%s/%s/%d/%d/%s/%d" % (
cds.getTaxId(), cds.getProtId(),
",".join(map(str, requiredSeqIds)), ",".join(
map(str, shufflesWithMissingWindows +
completelyMissingShuffles)), lastwin, windowStep,
profileReference, shuffleType)
def calculateMissingWindowsForSequence(
self,
taxId,
protId,
seqIds,
requestedShuffleIds,
firstWindow,
lastWindowStart,
windowStep,
reference="begin",
shuffleType=db.Sources.ShuffleCDSv2_python):
timerForPreFolding.start()
logging.warning("Parameters: %d %s %s %s %d %d %s %d" %
(taxId, protId, seqIds, requestedShuffleIds,
lastWindowStart, windowStep, reference, shuffleType))
f = self._logfile
assert (len(seqIds) > 0)
assert (len(seqIds) == len(requestedShuffleIds))
optimalSpeciesGrowthTemperature = None
if (self._seriesSourceNumber ==
db.Sources.RNAfoldEnergy_SlidingWindow40_v2_native_temp):
(numericalProp, _) = getSpeciesTemperatureInfo(taxId)
optimalSpeciesGrowthTemperature = numericalProp[0]
if optimalSpeciesGrowthTemperature is None:
raise Exception(
"No temperature value for taxid={}, can't calculate native-temperature folding profile..."
.format(taxId))
else:
optimalSpeciesGrowthTemperature = float(
optimalSpeciesGrowthTemperature)
assert (optimalSpeciesGrowthTemperature >= -30.0
and optimalSpeciesGrowthTemperature <= 150.0)
if (reference != "begin" and reference != "end"):
timerForPreFolding.stop()
e = "Specificed profile reference '%s' is not supported! (" % reference
logging.error(e)
raise Exception(e)
# We will process all listed shuffle-ids for the following protein record
cds = CDSHelper(taxId, protId)
if (cds.length() < self._windowWidth):
e = "Refusing to process item %s because the sequence length (%d nt) is less than the window size (%d nt)\n" % (
itemToProcess, cds.length(), self._windowWidth)
f.write(e)
logging.error(e)
timerForPreFolding.stop()
raise Exception(e)
# Create a list of the windows we need to calculate for this CDS
if reference == "begin":
requestedWindowStarts = frozenset(
range(
0,
min(lastWindowStart + 1,
cds.length() - self._windowWidth - 1), windowStep))
if (len(requestedWindowStarts) == 0):
e = "No windows exist for calculation taxid=%d, protId=%s, CDS-length=%d, lastWindowStart=%d, windowStep=%d, windowWidth=%d - Skipping...\n" % (
taxId, protId, cds.length(), lastWindowStart, windowStep,
self._windowWidth)
f.write(e)
logging.error(e)
timerForPreFolding.stop()
raise Exception(e)
elif reference == "end":
lastPossibleWindowStart = cds.length(
) - self._windowWidth #+ 1 # disregard lastWindowStart when reference=="end"
#lastWindowCodonStart = (lastPossibleWindowStart-3)-(lastPossibleWindowStart-3)%3
#lastPossibleWindowStart = seqLength - windowWidth # + 1 # disregard lastWindowStart when reference=="end"
requestedWindowStarts = frozenset(
filter(
lambda x: x >= lastWindowStart,
range(lastPossibleWindowStart % windowStep,
lastPossibleWindowStart + 1, windowStep)))
#requestedWindowStarts = frozenset(range(lastWindowCodonStart % windowStep, lastWindowCodonStart, windowStep))
#pass
else:
assert (False)
# First, read available results (for all shuffle-ids) in JSON format
# Array is indexed by shuffle-id, so results not requested will be represented by None (as will requested items that have no results yet).
logging.info("DEBUG: requestedShuffleIds (%d items): %s\n" %
(len(requestedShuffleIds), requestedShuffleIds))
existingResults = cds.getCalculationResult2(self._seriesSourceNumber,
requestedShuffleIds,
True,
shuffleType=shuffleType)
#assert(len(existingResults) >= len(requestedShuffleIds)) # The returned array must be at least as large as the requested ids list
assert (len(existingResults) == len(requestedShuffleIds))
logging.info("requestedShuffleIds: %s" % requestedShuffleIds)
logging.info("existingResults.keys(): %s" % existingResults.keys())
assert (frozenset(requestedShuffleIds) == frozenset(
existingResults.keys()))
#existingResults = [None] * (max(requestedShuffleIds)+1)
logging.info("DEBUG: existingResults (%d items): %s\n" %
(len(existingResults), existingResults))
# Check for which of the requested shuffle-ids there are values missing
shuffleIdsToProcess = {}
for shuffleId, r in existingResults.items():
if r is None:
# There are no existing results for shuffled-id n. If it was requested, it should be calculated now (including all windows)
if shuffleId in requestedShuffleIds:
shuffleIdsToProcess[shuffleId] = list(
requestedWindowStarts)
timerForPreFolding.stop()
# ------------------------------------------------------------------------------------
continue # TODO - verify this line; should we abort this sequence by throwing????
# ------------------------------------------------------------------------------------
logging.info("/// shuffleId r = %d %s" % (shuffleId, r))
logging.info("r[MFE-profile] %s" % r["MFE-profile"])
# Check the existing results for this shuffle
alreadyProcessedWindowStarts = frozenset([
i for i, x in enumerate(r["MFE-profile"]) if x is not None
]) # Get the indices (=window starts) of all non-None values
missingWindows = requestedWindowStarts - alreadyProcessedWindowStarts # Are there any requested windows that are not already computed?
if (missingWindows):
shuffleIdsToProcess[shuffleId] = missingWindows
if (not shuffleIdsToProcess):
e = "All requested shuffle-ids in (taxId: %d, protId: %s, seqs: %s) seem to have already been processed. Skipping...\n" % (
taxId, protId, str(list(zip(seqIds, requestedShuffleIds))))
logging.warning(e)
timerForPreFolding.stop()
return
logging.info("DEBUG: shuffleIdsToProcess (%d items): %s\n" %
(len(shuffleIdsToProcess), shuffleIdsToProcess))
logging.info("DEBUG: Before (%d items): %s\n" %
(len(existingResults), existingResults))
# Initialize new results records
for shuffleId in shuffleIdsToProcess.keys():
if existingResults[shuffleId] is None:
logging.info(seqIds)
logging.info(requestedShuffleIds)
logging.info(shuffleId)
thisSeqId = seqIds[requestedShuffleIds.index(shuffleId)]
existingResults[shuffleId] = {
"id":
"%s/%s/%d/%d" % (taxId, protId, thisSeqId, shuffleId),
"seq-crc": None,
"MFE-profile": [],
"MeanMFE": None,
"v": 2,
"shuffle-type": shuffleType
}
logging.info("DEBUG: existingResults (%d items): %s\n" %
(len(existingResults), existingResults))
timerForPreFolding.stop()
# Load the sequences of all shuffle-ids we need to work on
# TODO - combine loading of multiple sequences into one DB operation
for shuffleId, record in existingResults.items():
if record is None:
logging.info(
"DEBUG: skipping empty results record for shuffleId={}".
format(shuffleId))
continue
timerForPreFolding.start()
seq = None
annotatedSeqId = None
# Get the sequence for this entry
if (shuffleId < 0):
seq = cds.sequence()
annotatedSeqId = cds.seqId()
else:
seq = cds.getShuffledSeq(shuffleId, shuffleType)
annotatedSeqId = cds.getShuffledSeqId(shuffleId, shuffleType)
if (seq is None or (not seq is None and len(seq) == 0)):
seq2 = cds.getShuffledSeq2(annotatedSeqId)
seq3 = cds._fetchSequence(annotatedSeqId)
seq4 = cds._cache.get("%d:seq" % annotatedSeqId)
if not seq4 is None:
del cds._cache["%d:seq" % annotatedSeqId]
seq5 = cds.getShuffledSeq2(annotatedSeqId)
e = "Got empty sequence for shuffleId=%d, seqId=%d, taxId=%d, protId=%s, numShuffled=%d, ids[%d:%d]=%s, len(seq2)=%d, len(seq3)=%d, len(seq4)=%d, len(seq5)=%d" % (
shuffleId, annotatedSeqId, taxId, protId,
len(cds.shuffledSeqIds()), shuffleId - 2, shuffleId + 2,
cds.shuffledSeqIds()[shuffleId - 2:shuffleId + 2],
len(seq2) if not seq2 is None else -1, len(seq3)
if not seq3 is None else -1, len(seq4) if not seq4 is None
else -1, len(seq5) if not seq5 is None else -1)
logging.error(e)
timerForPreFolding.stop()
raise Exception(e)
#
# Disabled - calculation needn't include the native sequence...
#
#if( annotatedSeqId not in seqIds ):
# e = "Error: SeqId specified in queue item %s does not match annotated seq-id %d\n" % (itemToProcess, annotatedSeqId)
# f.write(e)
# f.write("Current shuffle-id: %d\n" % shuffleId)
# f.write("Ids in existing results:\n")
# for shuffleId, record in enumerate(existingResults):
# f.write(" %d) %s\n" % (shuffleId, record['id']))
# f.write("Debug info:\n")
# f.write("\n".join(cds.getDebugInfo()))
# f.write("\n")
# f.write("Skipping...\n")
# print("Skipping...")
# raise Exception(e)
expectedSeqLength = cds.length()
if (not expectedSeqLength is None):
if (expectedSeqLength != len(seq)):
e = "Warning: taxid=%d, protid=%s, seqid=%d - unexpected length %d (expected: %d)\n" % (
taxId, protId, annotatedSeqId, len(seq),
expectedSeqLength)
f.write(e)
logging.error(e)
timerForPreFolding.stop()
raise Exception(e)
if (len(seq) < self._windowWidth):
# Sequence is shorter than required window; skip
e = "Warning: skipping sequence because it is shorter than the requested window...\n"
f.write(e)
logging.error(e)
timerForPreFolding.stop()
raise Exception(e)
logging.info(
"DEBUG: Processing item taxId=%d, protId=%s, shuffle=%d (length=%d, %d windows)...\n"
% (taxId, protId, shuffleId, len(seq),
len(requestedWindowStarts)))
# TODO - Remove any old value stored in this key?
# Skip this for now
# This will be made redundant by completing the "updating" implementation
#
#if( cds.isCalculationDone( seriesSourceNumber, shuffleId )):
# # Sufficient data seems to exist. Skip...
# f.write("Item %s appears to be already completed, skipping..." % itemToProcess)
# continue
logging.info(seq[:50])
#f.write("\n")
MFEprofile = record["MFE-profile"]
#f.write("Profile: %s\n" % MFEprofile)
# Make sure the profile array contains enough entries for all new windows (and possibly, if windows are non-contiguous, entries between them that we are not going to compute right now)
if (len(MFEprofile) < max(requestedWindowStarts)):
entriesToAdd = max(requestedWindowStarts) - len(MFEprofile) + 1
MFEprofile.extend([None] * entriesToAdd)
assert (len(MFEprofile) >= max(requestedWindowStarts))
stats = RunningStats()
stats.extend([x for x in MFEprofile if x is not None])
timerForPreFolding.stop()
timerForFolding.start()
for start in requestedWindowStarts:
fragment = seq[start:(start + self._windowWidth)]
assert (len(fragment) == self._windowWidth)
if self._seriesSourceNumber == db.Sources.RNAfoldEnergy_SlidingWindow40_v2:
# Calculate the RNA folding energy. This is the computation-heavy part.
#strct, energy = RNA.fold(fragment)
energy = RNAfold_direct(fragment)
assert (energy <= 0.0)
elif self._seriesSourceNumber == db.Sources.RNAfoldEnergy_SlidingWindow40_v2_native_temp:
# Calculate the RNA folding energy. This is the computation-heavy part.
#strct, energy = RNA.fold(fragment)
energy = RNAfold_direct(fragment,
explicitCalculationTemperature=
optimalSpeciesGrowthTemperature)
assert (energy <= 0.0)
elif self._seriesSourceNumber == db.Sources.TEST_StepFunction_BeginReferenced:
if shuffleId < 0:
energy = 0
else:
energy = start % 50 - 20
elif self._seriesSourceNumber == db.Sources.TEST_StepFunction_EndReferenced:
if shuffleId < 0:
energy = 0
else:
energy = (expectedSeqLength - self._windowWidth -
start) % 50 - 20
else:
logging.error(
"Received unknown seriesSourceNumber {}".format(
self._seriesSourceNumber))
assert (False)
# Store the calculation result
#print("%d:%s --> %f" % (taxId, protId, energy))
stats.push(energy)
MFEprofile[start] = energy
print(
"/////////////////// shuffleId={} (len={}) //////////////////////////"
.format(shuffleId, expectedSeqLength))
prettyPrintProfile(MFEprofile)
timerForFolding.stop()
timerForPostFolding.start()
# Format
crc = calcCrc(seq)
#result = """{"id":"%s","seq-crc":%d,"MFE-profile":[%s],"MeanMFE":%.6g,v:2}""" % (itemToProcess, crc, ",".join(map(lambda x: "%.3g" % x, MFEprofile)), stats.mean())
record["seq-crc"] = crc
record["MFE-profile"] = [
round4(x) for x in MFEprofile
] # Round items down to save space (these are not exact numbers anyway)
record["MeanMFE"] = stats.mean()
result = json.dumps(record)
f.write(result)
f.write("\n")
if (not self._debugDoneWriteResults):
cds.saveCalculationResult2(self._seriesSourceNumber, result,
annotatedSeqId, False)
timerForPostFolding.stop()
timerForPostFolding.start()
if (not self._debugDoneWriteResults):
cds.commitChanges()
timerForPostFolding.stop()
skipped += 1
continue
else:
print(
"Warning: Could not find CDS length entry for taxid=%d, protid=%s"
% (taxIdForProcessing, protId))
skipped += 1
continue
#requiredNumWindows = seqLength - windowWidth + 1
cdsSeqId = cds.seqId()
match[cdsSeqId] = (protId, -1)
shuffledIds = cds.shuffledSeqIds()
computedShufflesCount = 0
missingShuffles = []
for n in range(50):
match[shuffledIds[n]] = (protId, n)
if shuffledIds[n] in computed:
computedShufflesCount += 1
else:
missingShuffles.append(n)
if (computedShufflesCount < 50):
print("%s - found only %d computed results" %
(protId, computedShufflesCount))
# Skip sequences that are too short
if(seqLength < numWindows + windowWidth + 1 ):
skipped += 1
continue
else:
print("Warning: Could not find CDS length entry for taxid=%d, protid=%s" % (taxIdForProcessing, protId) )
skipped += 1
continue
#requiredNumWindows = seqLength - windowWidth + 1
cdsSeqId = cds.seqId()
#match[cdsSeqId] = (protId,-1)
shuffledIds = cds.shuffledSeqIds()
computedShufflesCount = 0
for n in range(min(numShuffledGroups, len(shuffledIds))):
#match[shuffledIds[n]] = (protId,n)
if shuffledIds[n] in computed:
computedShufflesCount += 1
if( computedShufflesCount<requiredNumShuffledGroups or (not cdsSeqId in computed) ):
#print("%s - found only %d groups, skipping" % (protId, computedShufflesCount))
skipped += 1
continue
# TODO - do something here...
#cdsResults = cds.getCalculationResult( seriesSourceNumber, -1 )
numUniqueShuffles = Counter()
for taxId in species:
proteinsDone = 0
#nativeColumns = [[] for x in range(maxCodons)]
#shuffledColumns = [[[] for x in range(maxCodons)] for y in range(maxShuffles)]
allNativeSeqs = {}
allShuffledSeqs = {}
for protId in SpeciesCDSSource(taxId):
cds = CDSHelper(taxId, protId)
warnings.update(("total-cds", ))
allIds = cds.shuffledSeqIds(shuffleType=shuffleType)[:maxShuffles]
nativeSeq = cds.sequence()
if (len(nativeSeq) % 3 != 0):
warnings.update(("has-broken-codons", ))
continue
nativeCodons = Counter(splitCodons(nativeSeq))
hasMismatchedCodons = False
allNativeSeqs[protId] = nativeSeq
hashesForShuffles = set()
#for i, c in enumerate(splitCodons(nativeSeq)[:maxCodons]):
# nativeColumns[i].append(c)
def readSeriesResultsForSpecies(seriesSourceNumber,
species,
minShuffledGroups=20,
maxShuffledGroups=20,
shuffleType=db.Sources.ShuffleCDSv2_python,
cdsFilter=None,
returnCDS=True):
if isinstance(
species, Iterable
): # usually, species will be a sequence of numeric taxid values
if isinstance(species, basestring):
raise Exception("species cannot be string")
# all set - proceed...
else:
species = (species, ) # assume we got a single (numeric) taxid value
assert (minShuffledGroups <= maxShuffledGroups)
for taxIdForProcessing in species:
print("Procesing %d sequences for tax-id %d (%s)..." %
(countSpeciesCDS(taxIdForProcessing), taxIdForProcessing,
getSpeciesName(taxIdForProcessing)))
computed = getAllComputedSeqsForSpecies(seriesSourceNumber,
taxIdForProcessing,
maxShuffledGroups,
shuffleType=shuffleType)
computedIds = frozenset(computed.keys())
print("Collecting data from %d computation results..." % len(computed))
skipped = 0
selected = 0
alreadyCompleted = 0
# Iterate over all CDS entries for this species
for protId in SpeciesCDSSource(taxIdForProcessing):
cds = CDSHelper(taxIdForProcessing, protId)
if (not cdsFilter is None) and (not cdsFilter(cds)):
continue
cdsSeqId = cds.seqId()
shuffledIds = cds.shuffledSeqIds(shuffleType=shuffleType)
# How many shuffles (for this cds) exist in the data we found?
computedShufflesCount = len(
computedIds.intersection(frozenset(shuffledIds)))
if (computedShufflesCount < minShuffledGroups
or (not cdsSeqId in computedIds)):
#print("%s - found only %d groups, skipping" % (protId, computedShufflesCount))
skipped += 1
continue
# Get the computed results for this CDS
seqIds = [cds.seqId()]
seqIds.extend(cds.shuffledSeqIds(shuffleType=shuffleType))
if (len(seqIds) > maxShuffledGroups + 1):
seqIds = seqIds[:maxShuffledGroups + 1]
results = [computed.get(x) for x in seqIds]
if (results is None or len([() for x in results if not x is None])
< minShuffledGroups):
print("Not enough results found for %s" % protId)
skipped += 1
continue
# Decode the results
results = list(
map(
lambda x: decodeJsonSeriesRecord(decompressSeriesRecord(x))
if not x is None else None, results))
if (returnCDS):
yield {
"taxid": taxIdForProcessing,
"content": results,
"cds": cds
}
else:
yield {"taxid": taxIdForProcessing, "content": results}
del results
del cds
selected += 1
if (rl()):
print("# %s - %d records included, %d records skipped" %
(datetime.now().isoformat(), selected, skipped))
if (len(cds.sequence()) != cds.length()):
print(
"WARNING: incorrect sequence length detected for record (taxid=%d, protId=%s); real-length=%d, recorded-length=%d."
% (taxId, protId, len(cds.sequence()), cds.length()))
warningsCount += 1
recomputedCrc = getCrc(cds.sequence())
annotatedCrc = cds.crc()
assert (recomputedCrc == annotatedCrc)
print(cds.sequence()[:15])
seq1trans = Seq(cds.sequence(), generic_dna).translate()
crc1 = getCrc(seq1trans)
shuffles = cds.shuffledSeqIds(shuffleType=db.Sources.ShuffleCDSv2_python)
unique = len(frozenset(shuffles))
if (unique != len(shuffles)):
print(
"WARNING: duplicate shuffles found in record (taxid=%d, protId=%s); count=%d, unique=%d."
% (taxId, protId, len(shuffles), unique))
warningsCount += 1
statsShuffles.push(len(shuffles))
shuffledCrcs = set()
shuffledTransCrcs = set()
for shuffledSeqId in shuffles:
shufSeq = cds.getShuffledSeq2(shuffledSeqId)