def runComparisonOfBlastScriptVsNaiveBlast(self, blastMode):
"""We compare the output with a naive run of the blast program, to check the results are nearly
equivalent.
"""
encodeRegions = [ "ENm00" + str(i) for i in xrange(1,2) ] #, 2) ] #Could go to six
species = ("human", "mouse", "dog")
#Other species to try "rat", "monodelphis", "macaque", "chimp"
for encodeRegion in encodeRegions:
regionPath = os.path.join(self.encodePath, encodeRegion)
for i in xrange(len(species)):
species1 = species[i]
for species2 in species[i+1:]:
seqFile1 = os.path.join(regionPath, "%s.%s.fa" % (species1, encodeRegion))
seqFile2 = os.path.join(regionPath, "%s.%s.fa" % (species2, encodeRegion))
#Run the random
runNaiveBlast(seqFile1, seqFile2, self.tempOutputFile)
logger.info("Ran the naive blast okay")
#Run the blast
jobTreeDir = os.path.join(getTempDirectory(self.tempDir), "jobTree")
if blastMode == "allAgainstAll":
runCactusBlast([ seqFile1, seqFile2 ], self.tempOutputFile2, jobTreeDir,
chunkSize=500000, overlapSize=10000)
else:
runCactusBlast([ seqFile1 ], self.tempOutputFile2, jobTreeDir,
chunkSize=500000, overlapSize=10000, targetSequenceFiles=[ seqFile2 ])
runJobTreeStatusAndFailIfNotComplete(jobTreeDir)
system("rm -rf %s " % jobTreeDir)
logger.info("Ran cactus_blast okay")
logger.critical("Comparing cactus_blast and naive blast; using mode: %s" % blastMode)
compareResultsFile(self.tempOutputFile, self.tempOutputFile2)
def testProgressiveOutgroupsVsAllOutgroups(self):
"""Tests the difference in outgroup coverage on an ingroup when
running in "ingroups vs. outgroups" mode and "set against set"
mode.
"""
encodeRegion = "ENm001"
ingroup = "human"
outgroups = ["macaque", "rabbit", "dog"]
regionPath = os.path.join(self.encodePath, encodeRegion)
ingroupPath = os.path.join(regionPath, ingroup + "." + encodeRegion + ".fa")
outgroupPaths = map(lambda x: os.path.join(regionPath, x + "." + encodeRegion + ".fa"), outgroups)
# Run in "set against set" mode, aligning the entire ingroup
# vs each outgroup
runCactusBlast([ingroupPath], self.tempOutputFile, os.path.join(self.tempDir, "setVsSetJobTree"),
chunkSize=500000, overlapSize=10000,
targetSequenceFiles=outgroupPaths)
# Run in "ingroup vs outgroups" mode, aligning the ingroup vs
# the outgroups in order, trimming away sequence that's
# already been aligned.
system("cactus_blast.py --ingroups %s --outgroups %s --cigars %s --jobTree %s/outgroupJobTree" % (ingroupPath, ",".join(outgroupPaths), self.tempOutputFile2, self.tempDir))
# Get the coverage on the ingroup, in bases, from each run.
coverageSetVsSet = int(popenCatch("cactus_coverage %s %s | awk '{ total += $3 - $2} END { print total }'" % (ingroupPath, self.tempOutputFile)))
coverageIngroupVsOutgroups = int(popenCatch("cactus_coverage %s %s | awk '{ total += $3 - $2} END { print total }'" % (ingroupPath, self.tempOutputFile2)))
print "total coverage on human (set vs set mode, %d outgroups): %d" % (len(outgroups), coverageSetVsSet)
print "total coverage on human (ingroup vs outgroup mode, %d outgroups): %d" % (len(outgroups), coverageIngroupVsOutgroups)
# Make sure we're getting a reasonable fraction of the
# alignments when using the trimming strategy.
self.assertTrue(float(coverageIngroupVsOutgroups)/coverageSetVsSet >= 0.95)
# Get the coverage on the ingroup, in bases, from just the
# last outgroup. Obviously this should be much higher in set
# vs set mode than in ingroup vs outgroup mode.
coverageFromLastOutgroupSetVsSet = int(popenCatch("grep %s %s | cactus_coverage %s /dev/stdin | awk '{ total += $3 - $2} END { print total }'" % (outgroups[-1], self.tempOutputFile, ingroupPath)))
coverageFromLastOutgroupInVsOut = int(popenCatch("grep %s %s | cactus_coverage %s /dev/stdin | awk '{ total += $3 - $2} END { print total }'" % (outgroups[-1], self.tempOutputFile2, ingroupPath)))
print "total coverage on human from last outgroup in set (%s) (set vs set mode): %d" % (outgroups[-1], coverageFromLastOutgroupSetVsSet)
print "total coverage on human from last outgroup in set (%s) (ingroup vs outgroup mode): %d" % (outgroups[-1], coverageFromLastOutgroupInVsOut)
self.assertTrue(float(coverageFromLastOutgroupInVsOut)/coverageFromLastOutgroupSetVsSet <= 0.10)
def testBlastRandom(self):
"""Make some sequences, put them in a file, call blast with random parameters
and check it runs okay.
"""
tempSeqFile = os.path.join(self.tempDir, "tempSeq.fa")
self.tempFiles.append(tempSeqFile)
for test in xrange(self.testNo):
seqNo = random.choice(xrange(0, 10))
seq = getRandomSequence(8000)[1]
fileHandle = open(tempSeqFile, 'w')
for fastaHeader, seq in [ (str(i), mutateSequence(seq, 0.3*random.random())) for i in xrange(seqNo) ]:
if random.random() > 0.5:
seq = reverseComplement(seq)
fastaWrite(fileHandle, fastaHeader, seq)
fileHandle.close()
chunkSize = random.choice(xrange(500, 9000))
overlapSize = random.choice(xrange(2, 100))
jobTreeDir = os.path.join(getTempDirectory(self.tempDir), "jobTree")
runCactusBlast([ tempSeqFile ], self.tempOutputFile, jobTreeDir, chunkSize, overlapSize)
runJobTreeStatusAndFailIfNotComplete(jobTreeDir)
if getLogLevelString() == "DEBUG":
system("cat %s" % self.tempOutputFile)
system("rm -rf %s " % jobTreeDir)