# Exclude some sequences from the calculation
# ------------------------------------------------------------------------------------------
# Skip sequences with partial CDS annotations
#if(r.exists("CDS:taxid:%d:protid:%s:partial" % (taxIdForProcessing, protId))):
# skipped += 1
# continue
#if( not r.exists(nativeCdsSeqIdKey % (taxIdForProcessing, protId)) ):
# skipped +=1
# continue
cds = CDSHelper(taxIdForProcessing, protId)
seqLength = cds.length()
stopCodonPos = cds.CDSlength()
if seqLength is None:
print(
"Warning: Could not find CDS length entry for taxid=%d, protid=%s"
% (taxIdForProcessing, protId))
skipped += 1
stats['skipped-cds-length-missing'] += 1
continue
# Skip sequences with length <40nt (window width)
if (seqLength < windowWidth + 1):
print("short seq")
stats['skipped-short-seq'] += 1
skipped += 1
continue
def calculateMissingWindowsForSequence(self, taxId, protId, seqIds, requestedShuffleIds, firstWindow, lastWindowStart, windowStep, reference="begin", shuffleType=db.Sources.ShuffleCDSv2_python, debug=False):
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))
# ------------------------------------------------------------------------
# Obtain species-dependent properties needed for some calculations
# ----------------
# Optimal Temp
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)
# ----------------
# Genomic translation table
genomicTranslationTable = None
if( self._seriesSourceNumber in (db.Sources.StopCodon_content_SlidingWindow30, db.Sources.StopCodon_content_SlidingWindow40, db.Sources.StopCodon_content_SlidingWindow50 )):
genomicTranslationTable = getSpeciesTranslationTable(taxId)
assert(genomicTranslationTable>0 and genomicTranslationTable<=31)
if( reference != "begin" and reference != "end" and reference != "stop3utr"):
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
if( reference == "begin" or reference == "end" ):
regionOfInterest = RegionsOfInterset.CDSonly
elif reference == "stop3utr":
regionOfInterest = RegionsOfInterset.CDSand3UTR
else:
assert(False)
cds = CDSHelper( taxId, protId, regionOfInterest=regionOfInterest )
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(list(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([x for x in range(lastPossibleWindowStart % windowStep, lastPossibleWindowStart+1, windowStep) if x>=lastWindowStart])
elif reference == "stop3utr":
seqLength = cds.length()
stopCodonPos = cds.CDSlength()
isRequired = [1 if abs(pos-stopCodonPos)<((lastWindowStart//2)*windowStep) else 0 for pos in range(0, seqLength - self._windowWidth, windowStep)]
requestedWindowStarts = frozenset( compress( range(seqLength), isRequired ) )
#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" % list(existingResults.keys()))
assert(frozenset(requestedShuffleIds)==frozenset(list(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 list(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 list(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 list(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 in (db.Sources.RNAfoldEnergy_SlidingWindow30_v2, db.Sources.RNAfoldEnergy_SlidingWindow40_v2, db.Sources.RNAfoldEnergy_SlidingWindow50_v2):
# Calculate the RNA folding energy. This is the computation-heavy part.
#strct, energy = RNA.fold(fragment)
result = RNAfold_direct(fragment)
assert(result <= 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)
result = RNAfold_direct(fragment, explicitCalculationTemperature = optimalSpeciesGrowthTemperature)
assert(result <= 0.0)
elif self._seriesSourceNumber == db.Sources.GC_content_SlidingWindow40:
result = calcWindowGCContent( fragment )
assert( isnan(result) or (result >= 0.0 and result <= 1.0) )
elif self._seriesSourceNumber == db.Sources.Purine_content_SlidingWindow40:
result = calcWindowPurineContent( fragment )
assert( isnan(result) or (result >= 0.0 and result <= 1.0) )
elif self._seriesSourceNumber in (db.Sources.StopCodon_content_SlidingWindow30, db.Sources.StopCodon_content_SlidingWindow40, db.Sources.StopCodon_content_SlidingWindow50):
result = calcWindowStopCodonContent( fragment, translationTable=genomicTranslationTable, phase=start%3 )
assert( result >= 0.0 and result <= 1.0 )
elif self._seriesSourceNumber == db.Sources.TEST_StepFunction_BeginReferenced:
if shuffleId < 0:
result = 0
else:
result = start%50 - 20
elif self._seriesSourceNumber == db.Sources.TEST_StepFunction_EndReferenced:
if shuffleId < 0:
result = 0
else:
result = (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(result)
MFEprofile[start] = result
#print("/////////////////// shuffleId={} (len={}) //////////////////////////".format(shuffleId, expectedSeqLength))
if debug:
prettyPrintProfile(MFEprofile)
timerForFolding.stop()
timerForPostFolding.start()
# Format
crc = getCrc(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()
if reference == "stop3utr":
record["stop-codon-pos"] = cds.CDSlength()
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()
newValues = sum([1 for x in profile1 if not x is None])
windowsAddedToProfiles.update( (newValues,) )
rawRecordId = updateRecord[1]['id']
try:
recordId = splitLongSequenceIdentifier(rawRecordId)
except Exception as e:
err[ErrorTypes.UpdateRecordFormatError] += 1
badUpdateRecords.add(updateRecord[2])
print(e)
continue
cds = CDSHelper(recordId[0], recordId[1] )
cdsLength = cds.CDSlength()
newPositions = [x[0] for x in enumerate(profile1) if not x[1] is None]
for pos in newPositions:
windowsAddedToProfiles_DistanceFromStart.update( (pos,) )
windowsAddedToProfiles_DistanceFromEnd.update( (cdsLength-pos,) )
windowsAddedToProfiles_FrameRelativeToStart.update( (pos%10,) )
windowsAddedToProfiles_FrameRelativeToEnd.update( ((cdsLength-pos)%10,) )
if(rl()):
print(total, err, recordsByTaxId)
# DEBUG ONLY #### DEBUG ONLY #### DEBUG ONLY #### DEBUG ONLY #### DEBUG ONLY #### DEBUG ONLY #### DEBUG ONLY #
#if( total > 90000):
def randomize(self, nucleotideSeq: str, protId: str) -> (int, float, str):
#print("-----------"*5)
cds = CDSHelper(self.taxId, protId)
# Get metadata from genome model
#gm = cds.getGenomeModel()
#found = gm.findFeatureById( protId )
#if found is None:
# raise Exception("Failed to find feature matching protein-id={} in genome model".format(protId))
#(moleculeId, currFeature) = found
#if gm.moleculeModels[moleculeId].find3PrimeFlankingRegion( currFeature, debug=True ) is None:
# pass
#print((moleculeId, feature))
cdsLengthNt = cds.CDSlength()
assert (cdsLengthNt % 3 == 0)
flankingRegionLengthNt = cds.flankingRegion3UtrLength()
nextCDSOppositeStrand = cds.nextCDSOnOppositeStrand()
# Case 1 (no overlap):
# +--------intergenic--------+
# | |
# +-------CDS1--------+ +------------CDS2-----------+
# | | | |
# +===================+--------------------------+===========================+
# | | | |
# +===================+--------------------------+===========================+
# |<---cdsLengthNt--->|<-flankingRegionLengthNt->| |
# | (>= 0) |
# |<---------------------------cds.totalLength()---------------------------->|
# Case 2 (overlap):
# +--------------------------CDS2-------------------------+
# | |
# +---------------------CDS1----------------------+ |
# | | | |
# +===================+===========================+===========================+
# | | | |
# +===================+===========================+===========================+
# | |<-flankingRegionLengthNt-->| |
# | (<= 0) | |
# |<----------------cdsLengthNt------------------>| |
# |<----------------------------cds.totalLength()---------------------------->|
if flankingRegionLengthNt < 0 and -flankingRegionLengthNt > cdsLengthNt:
#flankingRegionLengthNt = -cdsLengthNt
raise Exception("Next CDS is fully overlapping...")
#-----------------------------------------------------------------------------
# Randomize the "main" CDS
#-----------------------------------------------------------------------------
# First, determine which region to randomize...
if (not self.constantOverlaps) or (
flankingRegionLengthNt >=
0): # no overlap, or overlap should be randomized
CDSseq = nucleotideSeq[:cdsLengthNt]
assert (len(CDSseq) == cdsLengthNt)
else: # constant overlaps requested and this CDS is overlapping the next. Remove the overlap from the CDS (it will not be randomized):
lastNucBeforeOverlap = cdsLengthNt + flankingRegionLengthNt
assert (lastNucBeforeOverlap < cdsLengthNt)
lastNucToRandomize = lastNucBeforeOverlap - (lastNucBeforeOverlap %
3)
CDSseq = nucleotideSeq[:lastNucToRandomize]
assert (len(CDSseq) % 3 == 0)
# Then, do the randomization...
(CDSpermCount, CDSidentity,
randomizedCDS) = self.cdsRand.randomizeAmbiguousSequence(CDSseq)
# Finally, add the non-randomized part of the CDS (if any)
if (not self.constantOverlaps) or (
flankingRegionLengthNt >=
0): # no overlap, or overlap should be randomized
pass
else: # constant overlaps requested and this CDS is overlapping the next.
randomizedCDS = randomizedCDS + nucleotideSeq[
lastNucToRandomize:cdsLengthNt]
assert (len(randomizedCDS) % 3 == 0)
assert (
len(randomizedCDS) == cdsLengthNt
) # the length of the resulting sequence matches the original CDS sequence
#-----------------------------------------------------------------------------
# Randomize the 3'UTR
#-----------------------------------------------------------------------------
if flankingRegionLengthNt > 0:
_3UTRseq = nucleotideSeq[cdsLengthNt:cdsLengthNt +
flankingRegionLengthNt]
assert (len(_3UTRseq) == flankingRegionLengthNt)
(UTRpermCount, UTRidentity,
randomizedUTR) = self.utrRand.randomizeAmbiguousSequence(_3UTRseq)
else:
_3UTRseq = ""
UTRpermCount = 1
UTRidentity = 1.0
randomizedUTR = ""
#-----------------------------------------------------------------------------
# Randomize the downstream CDS
#-----------------------------------------------------------------------------
nextCDSseq = nucleotideSeq[
cdsLengthNt +
flankingRegionLengthNt:] # Should work for positive and negative length UTRs
assert (len(nextCDSseq) % 3 == 0)
#nextCDSseq = nextCDSseq[(len(nextCDSseq)%3):] # remove partial codons from the start (caused due to the overlap; we can only randomize each codon as part of one CDS, although in the overlap region codons belong to two CDSs...)
if nextCDSOppositeStrand:
nextCDSseq = str(Seq(nextCDSseq, generic_dna).reverse_complement())
assert (len(nextCDSseq) % 3 == 0)
(nextCDSpermCount, nextCDSidentity, randomizedNextCDS
) = self.cdsRand.randomizeAmbiguousSequence(nextCDSseq)
if nextCDSOppositeStrand: # if the next CDS is on the opposite strand, revcomp it back to its original frame
randomizedNextCDS = str(
Seq(randomizedNextCDS, generic_dna).reverse_complement())
if flankingRegionLengthNt < 0:
randomizedNextCDS = randomizedNextCDS[-flankingRegionLengthNt:]
totalPerms = CDSpermCount * UTRpermCount * nextCDSpermCount
totalIdentity = ((CDSidentity * len(CDSseq)) +
(UTRidentity * len(_3UTRseq)) +
(nextCDSidentity * len(nextCDSseq))) / (
len(CDSseq) + len(_3UTRseq) + len(nextCDSseq))
return (totalPerms, totalIdentity,
randomizedCDS + randomizedUTR + randomizedNextCDS)