def stageWorkflow(outputSequenceDir,
configFile,
inputSequences,
toil,
restart=False,
outputSequences=[],
maskAlpha=False):
#Replace any constants
configNode = ET.parse(configFile).getroot()
if not outputSequences:
outputSequences = CactusPreprocessor.getOutputSequenceFiles(
inputSequences, outputSequenceDir)
else:
assert len(outputSequences) == len(inputSequences)
if configNode.find("constants") != None:
ConfigWrapper(
configNode).substituteAllPredefinedConstantsWithLiterals()
if maskAlpha:
ConfigWrapper(configNode).setPreprocessorActive(
"lastzRepeatMask", False)
ConfigWrapper(configNode).setPreprocessorActive("dna-brnn", True)
if not restart:
inputSequenceIDs = [
toil.importFile(makeURL(seq)) for seq in inputSequences
]
outputSequenceIDs = toil.start(
CactusPreprocessor(inputSequenceIDs, configNode))
else:
outputSequenceIDs = toil.restart()
for seqID, path in zip(outputSequenceIDs, outputSequences):
toil.exportFile(seqID, makeURL(path))
def run(self, fileStore):
self.configNode = ET.parse(fileStore.readGlobalFile(self.project.getConfigID())).getroot()
self.configWrapper = ConfigWrapper(self.configNode)
self.configWrapper.substituteAllPredefinedConstantsWithLiterals()
# Log the stats for the un-preprocessed assemblies
for name, sequence in self.project.getInputSequenceIDMap().items():
self.addChildJobFn(logAssemblyStats, "Before preprocessing", name, sequence)
# Create jobs to create the output sequences
logger.info("Reading config file from: %s" % self.project.getConfigID())
configFile = fileStore.readGlobalFile(self.project.getConfigID())
configNode = ET.parse(configFile).getroot()
ConfigWrapper(configNode).substituteAllPredefinedConstantsWithLiterals() #This is necessary..
#Add the preprocessor child job. The output is a job promise value that will be
#converted into a list of the IDs of the preprocessed sequences in the follow on job.
preprocessorJob = self.addChild(CactusPreprocessor(self.project.getInputSequenceIDs(), configNode))
self.project.setOutputSequenceIDs([preprocessorJob.rv(i) for i in range(len(self.project.getInputSequenceIDs()))])
#Now build the progressive-down job
schedule = Schedule()
schedule.loadProject(self.project, fileStore=fileStore)
schedule.compute()
self.options.event = self.project.mcTree.getRootName()
leafNames = [ self.project.mcTree.getName(i) for i in self.project.mcTree.getLeaves() ]
fileStore.logToMaster("Leaf names = %s" % leafNames)
self.options.globalLeafEventSet = set(leafNames)
return self.addFollowOn(RunCactusPreprocessorThenProgressiveDown2(options=self.options, project=self.project, event=self.options.event, schedule=schedule, memory=self.configWrapper.getDefaultMemory())).rv()
def stageWorkflow(outputSequenceDir,
configFile,
inputSequences,
toil,
restart=False,
outputSequences=[],
maskAlpha=False,
clipAlpha=None):
#Replace any constants
configNode = ET.parse(configFile).getroot()
if not outputSequences:
outputSequences = CactusPreprocessor.getOutputSequenceFiles(
inputSequences, outputSequenceDir)
else:
assert len(outputSequences) == len(inputSequences)
# Make sure we have the dna-brnn model in the filestore if we need it
loadDnaBrnnModel(toil, ET.parse(configFile).getroot(), maskAlpha=maskAlpha)
if configNode.find("constants") != None:
ConfigWrapper(
configNode).substituteAllPredefinedConstantsWithLiterals()
if maskAlpha or clipAlpha:
ConfigWrapper(configNode).setPreprocessorActive(
"lastzRepeatMask", False)
ConfigWrapper(configNode).setPreprocessorActive("dna-brnn", True)
if clipAlpha:
for node in configNode.findall("preprocessor"):
if getOptionalAttrib(node, "preprocessJob") == 'dna-brnn':
node.attrib["action"] = "clip"
node.attrib["minLength"] = clipAlpha
node.attrib["mergeLength"] = clipAlpha
if not restart:
inputSequenceIDs = []
for seq in inputSequences:
logger.info("Importing {}".format(seq))
inputSequenceIDs.append(toil.importFile(makeURL(seq)))
unzip_job = Job.wrapJobFn(unzip_then_pp, configNode, inputSequences,
inputSequenceIDs)
outputSequenceIDs = toil.start(unzip_job)
else:
outputSequenceIDs = toil.restart()
for seqID, path in zip(outputSequenceIDs, outputSequences):
try:
iter(seqID)
# dna-brnn will output a couple of bed files. we scrape those out here
toil.exportFile(seqID[0], makeURL(path))
toil.exportFile(seqID[1], makeURL(path) + '.bed')
toil.exportFile(seqID[2], makeURL(path) + '.mask.bed')
except:
toil.exportFile(seqID, makeURL(path))
def processConfig(self):
# read in the default right out of cactus
if self.options.configFile is not None:
configPath = self.options.configFile
else:
dir = cactusRootPath()
configPath = os.path.join(dir, "cactus_progressive_config.xml")
configXml = ET.parse(configPath).getroot()
self.configWrapper = ConfigWrapper(configXml)
# here we can go through the options and apply some to the config
self.configWrapper.setBuildHal(True)
self.configWrapper.setBuildFasta(True)
if self.options.outputMaf is not None:
self.configWrapper.setBuildMaf(True)
self.configWrapper.setJoinMaf(True)
# pre-emptively turn down maxParallelSubtree for singleMachine
# mode if not enough threads are provided to support it. Probably
# need to do something for other ?combined? batch systems?
if self.options.batchSystem == 'singleMachine' and \
self.options.database == 'kyoto_tycoon':
if int(self.options.maxThreads) < \
self.configWrapper.getMaxParallelSubtrees() * 3:
self.configWrapper.setMaxParallelSubtrees(
max(1,
int(self.options.maxThreads) / 3))
# this is a little hack to effectively toggle back to the
# non-progressive version of cactus (as published in Gen. Res. 2011)
# from the high-level interface.
if self.options.legacy is True:
self.configWrapper.setSubtreeSize(sys.maxint)
def processConfig(self):
log.info("Using config from path %s." % self.configPath)
configXml = ET.parse(self.configPath).getroot()
self.configWrapper = ConfigWrapper(configXml)
# here we can go through the options and apply some to the config
self.configWrapper.setBuildHal(True)
self.configWrapper.setBuildFasta(True)
def createFileStructure(mcProj, expTemplate, configTemplate, options):
if not os.path.exists(options.path):
os.makedirs(options.path)
mcProj.writeXML(os.path.join(options.path,
"%s_project.xml" % options.name))
seqMap = expTemplate.seqMap
portOffset = 0
for name, expPath in mcProj.expMap.items():
path = os.path.join(options.path, name)
seqMap[name] = os.path.join(path, name + '.fa')
for name, expPath in mcProj.expMap.items():
path = os.path.join(options.path, name)
children = mcProj.entireTree.getChildNames(name)
exp = copy.deepcopy(expTemplate)
# Get outgroups
outgroups = []
if configTemplate.getOutgroupStrategy() != 'none' \
and name in mcProj.outgroup.ogMap:
for og, ogDist in mcProj.outgroup.ogMap[name]:
assert og in seqMap, "No sequence found for outgroup: %s" % og
outgroups += [og]
# Get subtree connecting children + outgroups
assert len(children) > 0
subtree = mcProj.entireTree.extractSpanningTree(children + outgroups)
exp.updateTree(subtree, seqMap, outgroups)
exp.setConfigPath(os.path.join(path, "%s_config.xml" % name))
if not os.path.exists(path):
os.makedirs(path)
exp.writeXML(expPath)
config = ConfigWrapper(copy.deepcopy(configTemplate.xmlRoot))
config.setReferenceName(name)
config.writeXML(exp.getConfigPath())
def runCreateMultiCactusProject(expFile,
projectFile,
fixNames=False,
outgroupNames=None,
root=None,
overwrite=False):
options = CreateMultiCactusProjectOptions(expFile,
projectFile,
fixNames=fixNames,
outgroupNames=outgroupNames,
root=root,
overwrite=overwrite)
expTemplate = ExperimentWrapper(ET.parse(options.expFile).getroot())
configPath = expTemplate.getConfigPath()
confTemplate = ConfigWrapper(ET.parse(configPath).getroot())
if options.fixNames:
cleanEventTree(expTemplate)
tree = expTemplate.getTree()
if options.outgroupNames is not None:
options.outgroupNames = set(options.outgroupNames)
projNames = set([tree.getName(x) for x in tree.getLeaves()])
for outgroupName in options.outgroupNames:
if outgroupName not in projNames:
raise RuntimeError("Specified outgroup %s not found in tree" %
outgroupName)
mcProj = createMCProject(tree, expTemplate, confTemplate, options)
#Replace the sequences with output sequences
expTemplate.updateTree(mcProj.mcTree, expTemplate.buildSequenceMap())
#Now do the file tree creation
createFileStructure(mcProj, expTemplate, confTemplate, options)
def run(self, fileStore):
self.configNode = ET.parse(fileStore.readGlobalFile(self.project.getConfigID())).getroot()
self.configWrapper = ConfigWrapper(self.configNode)
self.configWrapper.substituteAllPredefinedConstantsWithLiterals()
logger.info("Progressive Up: " + self.event)
# open up the experiment
# note that we copy the path into the options here
experimentFile = fileStore.readGlobalFile(self.project.expIDMap[self.event])
expXml = ET.parse(experimentFile).getroot()
experiment = ExperimentWrapper(expXml)
configPath = fileStore.readGlobalFile(experiment.getConfigID())
configXml = ET.parse(configPath).getroot()
seqIDMap = dict()
tree = experiment.getTree()
seqNames = []
for node in tree.postOrderTraversal():
if tree.isLeaf(node):
name = tree.getName(node)
seqIDMap[name] = self.project.outputSequenceIDMap[name]
seqNames.append(name)
logger.info("Sequences in progressive, %s: %s" % (self.event, seqNames))
experimentFile = fileStore.getLocalTempFile()
experiment.writeXML(experimentFile)
self.options.experimentFileID = fileStore.writeGlobalFile(experimentFile)
# take union of command line options and config options for hal and reference
if self.options.buildReference == False:
refNode = findRequiredNode(configXml, "reference")
self.options.buildReference = getOptionalAttrib(refNode, "buildReference", bool, False)
halNode = findRequiredNode(configXml, "hal")
if self.options.buildHal == False:
self.options.buildHal = getOptionalAttrib(halNode, "buildHal", bool, False)
if self.options.buildFasta == False:
self.options.buildFasta = getOptionalAttrib(halNode, "buildFasta", bool, False)
# get parameters that cactus_workflow stuff wants
configFile = fileStore.readGlobalFile(experiment.getConfigID())
configNode = ET.parse(configFile).getroot()
workFlowArgs = CactusWorkflowArguments(self.options, experimentFile=experimentFile, configNode=configNode, seqIDMap = seqIDMap)
# copy over the options so we don't trail them around
workFlowArgs.buildReference = self.options.buildReference
workFlowArgs.buildHal = self.options.buildHal
workFlowArgs.buildFasta = self.options.buildFasta
workFlowArgs.globalLeafEventSet = self.options.globalLeafEventSet
if self.options.intermediateResultsUrl is not None:
# Give the URL prefix a special name for this particular
# subproblem (by suffixing it with the name of the
# internal node in the guide tree)
workFlowArgs.intermediateResultsUrl = self.options.intermediateResultsUrl + '-' + self.event
# Use the trimming strategy to blast ingroups vs outgroups.
finalExpWrapper = self.addChild(CactusTrimmingBlastPhase(cactusWorkflowArguments=workFlowArgs, phaseName="trimBlast")).rv()
logger.info("Going to create alignments and define the cactus tree")
return finalExpWrapper
def run(self, fileStore):
self.configNode = ET.parse(
fileStore.readGlobalFile(self.project.getConfigID())).getroot()
self.configWrapper = ConfigWrapper(self.configNode)
self.configWrapper.substituteAllPredefinedConstantsWithLiterals()
# Save preprocessed sequences
if self.options.intermediateResultsUrl is not None:
preprocessedSequences = self.project.outputSequenceIDMap
for genome, seqID in list(preprocessedSequences.items()):
fileStore.exportFile(
seqID, self.options.intermediateResultsUrl +
'-preprocessed-' + genome)
# Log the stats for the preprocessed assemblies
for name, sequence in list(self.project.outputSequenceIDMap.items()):
self.addChildJobFn(logAssemblyStats, "After preprocessing", name,
sequence)
project = self.addChild(
ProgressiveDown(
options=self.options,
project=self.project,
event=self.event,
schedule=self.schedule,
memory=self.configWrapper.getDefaultMemory())).rv()
#Combine the smaller HAL files from each experiment
return self.addFollowOnJobFn(
exportHal,
project=project,
memory=self.configWrapper.getDefaultMemory(),
disk=self.configWrapper.getExportHalDisk(),
preemptable=False).rv()
def createFileStructure(mcProj, expTemplate, configTemplate, options):
if not os.path.exists(options.path):
os.makedirs(options.path)
mcProj.writeXML(os.path.join(options.path, "%s_project.xml" % options.name))
for name, expPath in list(mcProj.expMap.items()):
path = os.path.join(options.path, name)
children = mcProj.entireTree.getChildNames(name)
# Get outgroups
outgroups = []
if configTemplate.getOutgroupStrategy() != 'none' \
and name in mcProj.outgroup.ogMap:
# Outgroup name is the first element of the ogMap tuples
outgroups.extend(list(map(itemgetter(0), mcProj.outgroup.ogMap[name])))
subtree = mcProj.entireTree.extractSpanningTree(children + [name] + outgroups)
exp = ExperimentWrapper.createExperimentWrapper(NXNewick().writeString(subtree),
children + [name] + outgroups,
databaseConf=expTemplate.confElem)
exp.setRootGenome(name)
exp.setOutgroupGenomes(outgroups)
if not os.path.exists(path):
os.makedirs(path)
config = ConfigWrapper(copy.deepcopy(configTemplate.xmlRoot))
if expTemplate.getSequenceID(name):
exp.setRootReconstructed(False)
exp.setSequenceID(name, expTemplate.getSequenceID(name))
else:
exp.setRootReconstructed(True)
exp.writeXML(expPath)
def validateInput(workDir, outputHalFile, options):
try:
if workDir.find(' ') >= 0:
raise RuntimeError("Cactus does not support spaces in pathnames: %s"
% workDir)
if not os.path.isdir(workDir):
os.makedirs(workDir)
if not os.path.isdir(workDir) or not os.access(workDir, os.W_OK):
raise
except:
raise RuntimeError("Can't write to workDir: %s" % workDir)
try:
open(outputHalFile, "w")
except:
raise RuntimeError("Unable to write to hal: %s" % outputHalFile)
if options.database != "tokyo_cabinet" and\
options.database != "kyoto_tycoon":
raise RuntimeError("Invalid database type: %s" % options.database)
if options.outputMaf is not None:
try:
open(options.outputMaf, "w")
except:
raise RuntimeError("Unable to write to maf: %s" % options.outputMaf)
if options.configFile is not None:
try:
ConfigWrapper(ET.parse(options.configFile).getroot())
except:
raise RuntimeError("Unable to read config: %s" % options.configFile)
if options.database == 'kyoto_tycoon':
if options.ktType.lower() != 'memory' and\
options.ktType.lower() != 'snapshot' and\
options.ktType.lower() != 'disk':
raise RuntimeError("Invalid ktserver type specified: %s. Must be "
"memory, snapshot or disk" % options.ktType)
def cactusPrepare(options, project):
""" annotate a SeqFile with ancestral names as well as paths for output sequences."""
# read the input
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
# prepare output sequence directory
# todo: support remote (ie s3) output directory
try:
os.makedirs(options.outSeqDir)
except:
pass
if not os.path.isdir(options.outSeqDir):
raise RuntimeError('Unable to create output sequence directory \'{}\''.format(options.outSeqDir))
if not os.access(options.outSeqDir, os.W_OK):
logger.warning('Output sequence directory is not writeable: \'{}\''.format(options.outSeqDir))
# hack the configfile to skip preprocessing and write it to the output dir
if options.preprocessOnly:
config.removePreprocessors()
options.configFile = os.path.join(options.outSeqDir, 'config.xml')
config.writeXML(options.configFile)
# pass through the config file to the options
# todo (don't like second hard-code check of .xml path)
if options.configFile != os.path.join(cactusRootPath(), "cactus_progressive_config.xml"):
options.cactusOptions += ' --configFile {}'.format(options.configFile)
# get the ancestor names
tree = MultiCactusTree(seqFile.tree)
tree.nameUnlabeledInternalNodes(prefix = config.getDefaultInternalNodePrefix())
# make the output
outSeqFile = SeqFile()
outSeqFile.tree= tree
outSeqFile.pathMap = seqFile.pathMap
outSeqFile.outgroups = seqFile.outgroups
# update paths for preprocessed leaves or inferred ancestors
for node in outSeqFile.tree.breadthFirstTraversal():
name = outSeqFile.tree.getName(node)
leaf = outSeqFile.tree.isLeaf(node)
if leaf or (not leaf and name not in seqFile.pathMap and not options.preprocessOnly):
out_basename = seqFile.pathMap[name] if name in seqFile.pathMap else '{}.fa'.format(name)
outSeqFile.pathMap[name] = os.path.join(options.outSeqDir, os.path.basename(out_basename))
# write the output
with open(options.outSeqFile, 'w') as out_sf:
out_sf.write(str(outSeqFile))
# write the instructions
print(get_plan(options, project, outSeqFile))
def runCactusProgressive(options):
with Toil(options) as toil:
importSingularityImage()
#Run the workflow
if options.restart:
halID = toil.restart()
else:
options.cactusDir = getTempDirectory()
#Create the progressive cactus project
projWrapper = ProjectWrapper(options)
projWrapper.writeXml()
pjPath = os.path.join(
options.cactusDir, ProjectWrapper.alignmentDirName,
'%s_project.xml' % ProjectWrapper.alignmentDirName)
assert os.path.exists(pjPath)
project = MultiCactusProject()
if not os.path.isdir(options.cactusDir):
os.makedirs(options.cactusDir)
project.readXML(pjPath)
#import the sequences
for genome, seq in project.inputSequenceMap.items():
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles([
os.path.join(seq, subSeq) for subSeq in os.listdir(seq)
], tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
project.inputSequenceIDMap[genome] = toil.importFile(seq)
#import cactus config
if options.configFile:
cactusConfigID = toil.importFile(makeURL(options.configFile))
else:
cactusConfigID = toil.importFile(
makeURL(project.getConfigPath()))
project.setConfigID(cactusConfigID)
project.syncToFileStore(toil)
configNode = ET.parse(project.getConfigPath()).getroot()
configWrapper = ConfigWrapper(configNode)
configWrapper.substituteAllPredefinedConstantsWithLiterals()
project.writeXML(pjPath)
halID = toil.start(
RunCactusPreprocessorThenProgressiveDown(
options, project, memory=configWrapper.getDefaultMemory()))
toil.exportFile(halID, makeURL(options.outputHal))
def processConfig(self):
# read in the default right out of cactus
if self.options.configFile is not None:
configPath = self.options.configFile
else:
dir = cactusRootPath()
configPath = os.path.join(dir, "cactus_progressive_config.xml")
configXml = ET.parse(configPath).getroot()
self.configWrapper = ConfigWrapper(configXml)
# here we can go through the options and apply some to the config
self.configWrapper.setBuildHal(True)
self.configWrapper.setBuildFasta(True)
def stageWorkflow(outputSequenceDir, configFile, inputSequences, toil, restart=False):
#Replace any constants
configNode = ET.parse(configFile).getroot()
outputSequences = CactusPreprocessor.getOutputSequenceFiles(inputSequences, outputSequenceDir)
if configNode.find("constants") != None:
ConfigWrapper(configNode).substituteAllPredefinedConstantsWithLiterals()
if not restart:
inputSequenceIDs = [toil.importFile(makeURL(seq)) for seq in inputSequences]
outputSequenceIDs = toil.start(CactusPreprocessor(inputSequenceIDs, configNode))
else:
outputSequenceIDs = toil.restart()
for seqID, path in zip(outputSequenceIDs, outputSequences):
toil.exportFile(seqID, makeURL(path))
def setUp(self):
self.batchSystem = "singleMachine"
if getBatchSystem() != None:
self.batchSystem = getBatchSystem()
unittest.TestCase.setUp(self)
self.useOutgroup = False
self.doSelfAlignment = False
#Load the config file, turn on the checks.
configWrapper = ConfigWrapper(ET.parse(os.path.join(cactusRootPath(), "cactus_progressive_config.xml")).getroot())
configWrapper.turnAllModesOn()
configWrapper.turnOffHeaderChecks()
self.tempDir = getTempDirectory(os.getcwd())
self.configFile = os.path.join(self.tempDir, "tempConfig.xml")
configWrapper.writeXML(self.configFile)
def run(self, fileStore):
self.configNode = ET.parse(fileStore.readGlobalFile(self.project.getConfigID())).getroot()
self.configWrapper = ConfigWrapper(self.configNode)
self.configWrapper.substituteAllPredefinedConstantsWithLiterals()
if not self.schedule.isVirtual(self.event):
tmpExp = fileStore.getLocalTempFile()
self.eventExpWrapper.writeXML(tmpExp)
self.project.expIDMap[self.event] = fileStore.writeGlobalFile(tmpExp)
followOnEvent = self.schedule.followOn(self.event)
if followOnEvent is not None:
logger.info("Adding follow-on event %s" % followOnEvent)
return self.addFollowOn(ProgressiveDown(self.options, self.project, followOnEvent,
self.schedule, memory=self.configWrapper.getDefaultMemory())).rv()
return self.project
def run(self, fileStore):
self.configNode = ET.parse(fileStore.readGlobalFile(self.project.getConfigID())).getroot()
self.configWrapper = ConfigWrapper(self.configNode)
self.configWrapper.substituteAllPredefinedConstantsWithLiterals()
logger.info("Progressive Down: " + self.event)
depProjects = dict()
deps = self.schedule.deps(self.event)
fileStore.logToMaster("There are %i dependent projects" % len(deps))
for child in deps:
fileStore.logToMaster("Adding dependent project %s" % child)
depProjects[child] = self.addChild(ProgressiveDown(self.options,
self.project, child,
self.schedule)).rv()
return self.addFollowOn(ProgressiveNext(self.options, self.project, self.event,
self.schedule, depProjects, memory=self.configWrapper.getDefaultMemory())).rv()
def loadProject(self, mcProject, fileStore = None):
self.inGraph = NX.DiGraph()
globTree = mcProject.mcTree
self.maxParallelSubtrees = None
leafEvents = [globTree.getName(i) for i in globTree.getLeaves()]
expMap = None
if fileStore:
expMap = dict()
for name in mcProject.expIDMap:
expMap[name] = fileStore.readGlobalFile(mcProject.expIDMap[name])
else:
expMap = mcProject.expMap
for name, expPath in list(expMap.items()):
exp = ExperimentWrapper(ET.parse(expPath).getroot())
tree = exp.getTree()
self.inGraph.add_node(name)
# Go through the species tree and add the correct
# dependencies (i.e. to the outgroups and the ingroups,
# but not to the other nodes that are just there because
# they are needed to form the correct paths).
for node in tree.postOrderTraversal():
nodeName = tree.getName(node)
# we don't add edges for leaves (in the global tree)
# as they are input sequences and do not form dependencies
# (it would be clever to maybe do the same with existing
# references when --overwrite is not specified but for now
# we just do the leaves)
if nodeName not in leafEvents and tree.isLeaf(node):
self.inGraph.add_edge(name, nodeName)
if fileStore:
configFile = fileStore.readGlobalFile(exp.getConfigID())
else:
# hack from running from cactus-prepare
configFile = exp.getConfigPath()
configElem = ET.parse(configFile).getroot()
conf = ConfigWrapper(configElem)
# load max parellel subtrees from the node's config
if self.maxParallelSubtrees is None:
self.maxParallelSubtrees = conf.getMaxParallelSubtrees()
else:
assert self.maxParallelSubtrees == conf.getMaxParallelSubtrees()
assert NX.is_directed_acyclic_graph(self.inGraph)
def runCactusGraphMapJoin(options):
with Toil(options) as toil:
importSingularityImage(options)
#Run the workflow
if options.restart:
wf_output = toil.restart()
else:
options.cactusDir = getTempDirectory()
#load cactus config
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
config.substituteAllPredefinedConstantsWithLiterals()
# load up the vgs
vg_ids = []
for vg_path in options.vg:
logger.info("Importing {}".format(vg_path))
vg_ids.append(toil.importFile(makeURL(vg_path)))
# tack on the decoys
if options.decoyGraph:
logger.info("Importing decoys {}".format(options.decoyGraph))
vg_ids.append(toil.importFile(makeURL(options.decoyGraph)))
# we'll treat it like any other graph downstream, except clipping
# where we'll check first using the path name
options.vg.append(options.decoyGraph)
# load up the hals
hal_ids = []
for hal_path in options.hal:
logger.info("Importing {}".format(hal_path))
hal_ids.append(toil.importFile(makeURL(hal_path)))
# run the workflow
wf_output = toil.start(
Job.wrapJobFn(graphmap_join_workflow, options, config, vg_ids,
hal_ids))
#export the split data
export_join_data(toil, options, wf_output[0], wf_output[1],
wf_output[2])
def run(self, fileStore):
self.configNode = ET.parse(fileStore.readGlobalFile(self.project.getConfigID())).getroot()
self.configWrapper = ConfigWrapper(self.configNode)
self.configWrapper.substituteAllPredefinedConstantsWithLiterals()
fileStore.logToMaster("Project has %i dependencies" % len(self.depProjects))
for projName in self.depProjects:
depProject = self.depProjects[projName]
for expName in depProject.expIDMap:
expID = depProject.expIDMap[expName]
experiment = ExperimentWrapper(ET.parse(fileStore.readGlobalFile(expID)).getroot())
fileStore.logToMaster("Reference ID for experiment %s: %s" % (expName, experiment.getReferenceID()))
if experiment.getReferenceID():
self.project.expIDMap[expName] = expID
self.project.outputSequenceIDMap[expName] = experiment.getReferenceID()
eventExpWrapper = None
logger.info("Progressive Next: " + self.event)
if not self.schedule.isVirtual(self.event):
eventExpWrapper = self.addChild(ProgressiveUp(self.options, self.project, self.event, memory=self.configWrapper.getDefaultMemory())).rv()
return self.addFollowOn(ProgressiveOut(self.options, self.project, self.event, eventExpWrapper, self.schedule, memory=self.configWrapper.getDefaultMemory())).rv()
def cactusPrepare(options, project):
""" annotate a SeqFile with ancestral names as well as paths for output sequences."""
# read the input
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
if not options.wdl:
# prepare output sequence directory
# todo: support remote (ie s3) output directory
try:
os.makedirs(options.outDir)
except:
pass
if not os.path.isdir(options.outDir):
raise RuntimeError('Unable to create output sequence directory \'{}\''.format(options.outDir))
if not os.access(options.outDir, os.W_OK):
logger.warning('Output sequence directory is not writeable: \'{}\''.format(options.outDir))
if options.preprocessOnly or options.gpu:
if options.preprocessOnly:
# hack the configfile to skip preprocessing and write it to the output dir
config.removePreprocessors()
if options.gpu:
# hack the configfile to toggle on gpu lastz
cafNode = findRequiredNode(config.xmlRoot, "caf")
cafNode.attrib["gpuLastz"] = "true"
# realigning doesn't mix well with lastz so we make sure it's off
# https://github.com/ComparativeGenomicsToolkit/cactus/issues/271
cafNode.attrib["realign"] = "0"
options.configFile = os.path.join(options.outDir, 'config-prepared.xml')
sys.stderr.write("configuration saved in {}\n".format(options.configFile))
config.writeXML(options.configFile)
# pass through the config file to the options
# todo (don't like second hard-code check of .xml path)
if options.configFile != os.path.join(cactusRootPath(), "cactus_progressive_config.xml") and not options.wdl:
options.cactusOptions += ' --configFile {}'.format(options.configFile)
# get the ancestor names
tree = MultiCactusTree(seqFile.tree)
tree.nameUnlabeledInternalNodes(prefix = config.getDefaultInternalNodePrefix())
# make the output
outSeqFile = SeqFile()
outSeqFile.tree= tree
outSeqFile.pathMap = copy.deepcopy(seqFile.pathMap)
outSeqFile.outgroups = copy.deepcopy(seqFile.outgroups)
# update paths for preprocessed leaves or inferred ancestors
for node in outSeqFile.tree.breadthFirstTraversal():
name = outSeqFile.tree.getName(node)
leaf = outSeqFile.tree.isLeaf(node)
if leaf or (not leaf and name not in seqFile.pathMap and not options.preprocessOnly):
out_basename = seqFile.pathMap[name] if name in seqFile.pathMap else '{}.fa'.format(name)
outSeqFile.pathMap[name] = os.path.join(options.outDir, os.path.basename(out_basename))
if options.wdl:
# uniquify name in wdl to prevent collisions
outSeqFile.pathMap[name] += '.pp'
# write the output
if options.outSeqFile:
with open(options.outSeqFile, 'w') as out_sf:
out_sf.write(str(outSeqFile))
# write the instructions
print(get_plan(options, project, seqFile, outSeqFile))
def runCactusBlastOnly(options):
with Toil(options) as toil:
importSingularityImage(options)
#Run the workflow
if options.restart:
alignmentID = toil.restart()
else:
options.cactusDir = getTempDirectory()
# apply path overrides. this was necessary for wdl which doesn't take kindly to
# text files of local paths (ie seqfile). one way to fix would be to add support
# for s3 paths and force wdl to use it. a better way would be a more fundamental
# interface shift away from files of paths throughout all of cactus
if options.pathOverrides:
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
tree = MultiCactusTree(seqFile.tree)
tree.nameUnlabeledInternalNodes(
prefix=config.getDefaultInternalNodePrefix())
for name, override in zip(options.pathOverrideNames,
options.pathOverrides):
seqFile.pathMap[name] = override
override_seq = os.path.join(options.cactusDir,
'seqFile.override')
with open(override_seq, 'w') as out_sf:
out_sf.write(str(seqFile))
options.seqFile = override_seq
#to be consistent with all-in-one cactus, we make sure the project
#isn't limiting itself to the subtree (todo: parameterize so root can
#be passed through from prepare to blast/align)
proj_options = copy.deepcopy(options)
proj_options.root = None
#Create the progressive cactus project (as we do in runCactusProgressive)
projWrapper = ProjectWrapper(proj_options,
proj_options.configFile,
ignoreSeqPaths=options.root)
projWrapper.writeXml()
pjPath = os.path.join(
options.cactusDir, ProjectWrapper.alignmentDirName,
'%s_project.xml' % ProjectWrapper.alignmentDirName)
assert os.path.exists(pjPath)
project = MultiCactusProject()
if not os.path.isdir(options.cactusDir):
os.makedirs(options.cactusDir)
project.readXML(pjPath)
# open up the experiment (as we do in ProgressiveUp.run)
# note that we copy the path into the options here
experimentFile = project.expMap[options.root]
expXml = ET.parse(experimentFile).getroot()
logger.info("Experiment {}".format(ET.tostring(expXml)))
experiment = ExperimentWrapper(expXml)
configPath = experiment.getConfigPath()
configXml = ET.parse(configPath).getroot()
seqIDMap = dict()
tree = MultiCactusTree(experiment.getTree()).extractSubTree(
options.root)
leaves = tree.getChildNames(tree.getRootName())
outgroups = experiment.getOutgroupGenomes()
genome_set = set(leaves + outgroups)
logger.info("Genomes in blastonly, {}: {}".format(
options.root, list(genome_set)))
print(str(project.inputSequenceMap))
#import the sequences (that we need to align for the given event, ie leaves and outgroups)
for genome, seq in list(project.inputSequenceMap.items()):
if genome in genome_set:
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles([
os.path.join(seq, subSeq)
for subSeq in os.listdir(seq)
], tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
project.inputSequenceIDMap[genome] = toil.importFile(seq)
else:
# out-of-scope sequences will only cause trouble later on
del project.inputSequenceMap[genome]
#import cactus config
if options.configFile:
cactusConfigID = toil.importFile(makeURL(options.configFile))
else:
cactusConfigID = toil.importFile(
makeURL(project.getConfigPath()))
project.setConfigID(cactusConfigID)
project.syncToFileStore(toil)
configNode = ET.parse(project.getConfigPath()).getroot()
configWrapper = ConfigWrapper(configNode)
configWrapper.substituteAllPredefinedConstantsWithLiterals()
workFlowArgs = CactusWorkflowArguments(
options,
experimentFile=experimentFile,
configNode=configNode,
seqIDMap=project.inputSequenceIDMap)
outWorkFlowArgs = toil.start(
CactusTrimmingBlastPhase(standAlone=True,
cactusWorkflowArguments=workFlowArgs,
phaseName="trimBlast"))
# export the alignments
toil.exportFile(outWorkFlowArgs.alignmentsID,
makeURL(options.outputFile))
# optional secondary alignments
if outWorkFlowArgs.secondaryAlignmentsID:
toil.exportFile(outWorkFlowArgs.secondaryAlignmentsID,
makeURL(options.outputFile) + '.secondary')
# outgroup fragments and coverage are necessary for cactus-align, as the sequence names got changed in the above alignemnts
for i, outgroupFragmentID in enumerate(
outWorkFlowArgs.outgroupFragmentIDs):
toil.exportFile(
outgroupFragmentID,
makeURL(options.outputFile) + '.og_fragment_{}'.format(i))
# cactus-align can recompute coverage on the fly, but we save them because we have them
for i, ingroupCoverageID in enumerate(
outWorkFlowArgs.ingroupCoverageIDs):
toil.exportFile(
ingroupCoverageID,
makeURL(options.outputFile) + '.ig_coverage_{}'.format(i))
def main():
options = get_options()
with Toil(options) as workflow:
setupBinaries(options)
importSingularityImage(options)
## Preprocessing:
if (options.pathOverrides or options.pathOverrideNames):
if not options.pathOverrides or not options.pathOverrideNames or \
len(options.pathOverrideNames) != len(options.pathOverrides):
raise RuntimeError(
'same number of values must be passed to --pathOverrides and --pathOverrideNames'
)
# apply path overrides. this was necessary for wdl which doesn't take kindly to
# text files of local paths (ie seqfile). one way to fix would be to add support
# for s3 paths and force wdl to use it. a better way would be a more fundamental
# interface shift away from files of paths throughout all of cactus
if options.pathOverrides:
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
tree = MultiCactusTree(seqFile.tree)
tree.nameUnlabeledInternalNodes(
prefix=config.getDefaultInternalNodePrefix())
for name, override in zip(options.pathOverrideNames,
options.pathOverrides):
seqFile.pathMap[name] = override
override_seq = os.path.join(options.cactusDir, 'seqFile.override')
with open(override_seq, 'w') as out_sf:
out_sf.write(str(seqFile))
options.seqFile = override_seq
# Import asms; by default, prepends unique IDs in the technique used in cactus-blast.
asms = get_asms_from_seqfile(options.seqFile, workflow)
## Perform alignments:
if not workflow.options.restart:
alignments = workflow.start(
Job.wrapJobFn(run_cactus_reference_align, asms, options.refID,
options.debug_export, options.dipcall_bed_filter,
options.dipcall_vcf_filter))
else:
alignments = workflow.restart()
if options.debug_export:
# first, ensure the debug dir exists.
if not os.path.isdir(options.debug_export_dir):
os.mkdir(options.debug_export_dir)
print(alignments)
# Then return value is: (all_primary, all_secondary, ref_mappings, primary_mappings, secondary_mappings)
for asm, mapping_file in alignments[2].items():
workflow.exportFile(
mapping_file, 'file://' +
os.path.abspath("mappings_for_" + asm + ".paf"))
for asm, mapping_file in alignments[3].items():
workflow.exportFile(
mapping_file, 'file://' +
os.path.abspath("mappings_for_" + asm + ".cigar"))
for asm, mapping_file in alignments[4].items():
workflow.exportFile(
mapping_file, 'file://' +
os.path.abspath("mappings_for_" + asm + ".cigar.secondry"))
## Save alignments:
if options.dipcall_vcf_filter: # this is substantially less restrictive than the dipcall_bed_filter.
dipcall_filtered = workflow.start(
Job.wrapJobFn(apply_dipcall_vcf_filter, alignments[0]))
workflow.exportFile(dipcall_filtered, makeURL(options.outputFile))
workflow.exportFile(
alignments[1],
makeURL(options.outputFile + ".unfiltered.secondary"))
else:
workflow.exportFile(alignments[0], makeURL(options.outputFile))
workflow.exportFile(alignments[1],
makeURL(options.outputFile + ".secondary"))
def runCactusGraphMapSplit(options):
with Toil(options) as toil:
importSingularityImage(options)
#Run the workflow
if options.restart:
wf_output = toil.restart()
else:
options.cactusDir = getTempDirectory()
#load cactus config
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
config.substituteAllPredefinedConstantsWithLiterals()
# load up the contigs if any
ref_contigs = set(options.refContigs)
# todo: use import?
if options.refContigsFile:
with open(options.refContigsFile, 'r') as rc_file:
for line in rc_file:
if len(line.strip()):
ref_contigs.add(line.strip().split()[0])
if options.otherContig:
assert options.otherContig not in ref_contigs
# get the minigraph "virutal" assembly name
graph_event = getOptionalAttrib(findRequiredNode(
configNode, "graphmap"),
"assemblyName",
default="_MINIGRAPH_")
# load the seqfile
seqFile = SeqFile(options.seqFile)
#import the graph
gfa_id = toil.importFile(makeURL(options.minigraphGFA))
#import the paf
paf_id = toil.importFile(makeURL(options.graphmapPAF))
#import the sequences (that we need to align for the given event, ie leaves and outgroups)
seqIDMap = {}
leaves = set([
seqFile.tree.getName(node)
for node in seqFile.tree.getLeaves()
])
if graph_event not in leaves:
raise RuntimeError(
"Minigraph name {} not found in seqfile".format(
graph_event))
if options.reference and options.reference not in leaves:
raise RuntimeError(
"Name given with --reference {} not found in seqfile".
format(options.reference))
for genome, seq in seqFile.pathMap.items():
if genome in leaves:
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles([
os.path.join(seq, subSeq)
for subSeq in os.listdir(seq)
], tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
logger.info("Importing {}".format(seq))
seqIDMap[genome] = (seq, toil.importFile(seq))
# run the workflow
wf_output = toil.start(
Job.wrapJobFn(graphmap_split_workflow, options, config,
seqIDMap, gfa_id, options.minigraphGFA, paf_id,
options.graphmapPAF, ref_contigs,
options.otherContig))
#export the split data
export_split_data(toil, wf_output[0], wf_output[1], wf_output[2:],
options.outDir, config)
def make_align_job(options, toil):
options.cactusDir = getTempDirectory()
# apply path overrides. this was necessary for wdl which doesn't take kindly to
# text files of local paths (ie seqfile). one way to fix would be to add support
# for s3 paths and force wdl to use it. a better way would be a more fundamental
# interface shift away from files of paths throughout all of cactus
if options.pathOverrides:
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
tree = MultiCactusTree(seqFile.tree)
tree.nameUnlabeledInternalNodes(
prefix=config.getDefaultInternalNodePrefix())
for name, override in zip(options.pathOverrideNames,
options.pathOverrides):
seqFile.pathMap[name] = override
override_seq = os.path.join(options.cactusDir, 'seqFile.override')
with open(override_seq, 'w') as out_sf:
out_sf.write(str(seqFile))
options.seqFile = override_seq
if not options.root:
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
mcTree = MultiCactusTree(seqFile.tree)
mcTree.nameUnlabeledInternalNodes(
prefix=config.getDefaultInternalNodePrefix())
options.root = mcTree.getRootName()
if options.acyclic:
seqFile = SeqFile(options.seqFile)
tree = MultiCactusTree(seqFile.tree)
leaves = [tree.getName(leaf) for leaf in tree.getLeaves()]
if options.acyclic not in leaves:
raise RuntimeError(
"Genome specified with --acyclic, {}, not found in tree leaves"
.format(options.acyclic))
#to be consistent with all-in-one cactus, we make sure the project
#isn't limiting itself to the subtree (todo: parameterize so root can
#be passed through from prepare to blast/align)
proj_options = copy.deepcopy(options)
proj_options.root = None
#Create the progressive cactus project (as we do in runCactusProgressive)
projWrapper = ProjectWrapper(proj_options,
proj_options.configFile,
ignoreSeqPaths=options.root)
projWrapper.writeXml()
pjPath = os.path.join(options.cactusDir, ProjectWrapper.alignmentDirName,
'%s_project.xml' % ProjectWrapper.alignmentDirName)
assert os.path.exists(pjPath)
project = MultiCactusProject()
if not os.path.isdir(options.cactusDir):
os.makedirs(options.cactusDir)
project.readXML(pjPath)
# open up the experiment (as we do in ProgressiveUp.run)
# note that we copy the path into the options here
experimentFile = project.expMap[options.root]
expXml = ET.parse(experimentFile).getroot()
experiment = ExperimentWrapper(expXml)
configPath = experiment.getConfigPath()
configXml = ET.parse(configPath).getroot()
seqIDMap = dict()
tree = MultiCactusTree(experiment.getTree()).extractSubTree(options.root)
leaves = [tree.getName(leaf) for leaf in tree.getLeaves()]
outgroups = experiment.getOutgroupGenomes()
genome_set = set(leaves + outgroups)
# this is a hack to allow specifying all the input on the command line, rather than using suffix lookups
def get_input_path(suffix=''):
base_path = options.cigarsFile[0]
for input_path in options.cigarsFile:
if suffix and input_path.endswith(suffix):
return input_path
if os.path.basename(base_path).startswith(
os.path.basename(input_path)):
base_path = input_path
return base_path + suffix
# import the outgroups
outgroupIDs = []
outgroup_fragment_found = False
for i, outgroup in enumerate(outgroups):
try:
outgroupID = toil.importFile(
makeURL(get_input_path('.og_fragment_{}'.format(i))))
outgroupIDs.append(outgroupID)
experiment.setSequenceID(outgroup, outgroupID)
outgroup_fragment_found = True
assert not options.pangenome
except:
# we assume that input is not coming from cactus blast, so we'll treat output
# sequences normally and not go looking for fragments
outgroupIDs = []
break
#import the sequences (that we need to align for the given event, ie leaves and outgroups)
for genome, seq in list(project.inputSequenceMap.items()):
if genome in leaves or (not outgroup_fragment_found
and genome in outgroups):
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles(
[os.path.join(seq, subSeq) for subSeq in os.listdir(seq)],
tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
logger.info("Importing {}".format(seq))
experiment.setSequenceID(genome, toil.importFile(seq))
if not outgroup_fragment_found:
outgroupIDs = [
experiment.getSequenceID(outgroup) for outgroup in outgroups
]
# write back the experiment, as CactusWorkflowArguments wants a path
experiment.writeXML(experimentFile)
#import cactus config
if options.configFile:
cactusConfigID = toil.importFile(makeURL(options.configFile))
else:
cactusConfigID = toil.importFile(makeURL(project.getConfigPath()))
project.setConfigID(cactusConfigID)
project.syncToFileStore(toil)
configNode = ET.parse(project.getConfigPath()).getroot()
configWrapper = ConfigWrapper(configNode)
configWrapper.substituteAllPredefinedConstantsWithLiterals()
if options.singleCopySpecies:
findRequiredNode(
configWrapper.xmlRoot,
"caf").attrib["alignmentFilter"] = "singleCopyEvent:{}".format(
options.singleCopySpecies)
if options.barMaskFilter:
findRequiredNode(
configWrapper.xmlRoot,
"bar").attrib["partialOrderAlignmentMaskFilter"] = str(
options.barMaskFilter)
if options.pangenome:
# turn off the megablock filter as it ruins non-all-to-all alignments
findRequiredNode(configWrapper.xmlRoot,
"caf").attrib["minimumBlockHomologySupport"] = "0"
findRequiredNode(
configWrapper.xmlRoot,
"caf").attrib["minimumBlockDegreeToCheckSupport"] = "9999999999"
# turn off mapq filtering
findRequiredNode(configWrapper.xmlRoot,
"caf").attrib["runMapQFiltering"] = "0"
# more iterations here helps quite a bit to reduce underalignment
findRequiredNode(configWrapper.xmlRoot,
"caf").attrib["maxRecoverableChainsIterations"] = "50"
# turn down minimum block degree to get a fat ancestor
findRequiredNode(configWrapper.xmlRoot,
"bar").attrib["minimumBlockDegree"] = "1"
# turn on POA
findRequiredNode(configWrapper.xmlRoot,
"bar").attrib["partialOrderAlignment"] = "1"
# save it
if not options.batch:
pg_file = options.outHal + ".pg-conf.xml"
if pg_file.startswith('s3://'):
pg_temp_file = getTempFile()
else:
pg_temp_file = pg_file
configWrapper.writeXML(pg_temp_file)
if pg_file.startswith('s3://'):
write_s3(pg_temp_file,
pg_file,
region=get_aws_region(options.jobStore))
logger.info("pangenome configuration overrides saved in {}".format(
pg_file))
workFlowArgs = CactusWorkflowArguments(options,
experimentFile=experimentFile,
configNode=configNode,
seqIDMap=project.inputSequenceIDMap)
#import the files that cactus-blast made
workFlowArgs.alignmentsID = toil.importFile(makeURL(get_input_path()))
workFlowArgs.secondaryAlignmentsID = None
if not options.pafInput:
try:
workFlowArgs.secondaryAlignmentsID = toil.importFile(
makeURL(get_input_path('.secondary')))
except:
pass
workFlowArgs.outgroupFragmentIDs = outgroupIDs
workFlowArgs.ingroupCoverageIDs = []
if outgroup_fragment_found and len(outgroups) > 0:
for i in range(len(leaves)):
workFlowArgs.ingroupCoverageIDs.append(
toil.importFile(
makeURL(get_input_path('.ig_coverage_{}'.format(i)))))
align_job = Job.wrapJobFn(run_cactus_align,
configWrapper,
workFlowArgs,
project,
checkpointInfo=options.checkpointInfo,
doRenaming=options.nonCactusInput,
pafInput=options.pafInput,
pafSecondaries=options.usePafSecondaries,
doVG=options.outVG,
doGFA=options.outGFA,
delay=options.stagger,
eventNameAsID=options.eventNameAsID,
acyclicEvent=options.acyclic)
return align_job
def runCactusGraphMap(options):
with Toil(options) as toil:
importSingularityImage(options)
#Run the workflow
if options.restart:
alignmentID = toil.restart()
else:
options.cactusDir = getTempDirectory()
# apply path overrides. this was necessary for wdl which doesn't take kindly to
# text files of local paths (ie seqfile). one way to fix would be to add support
# for s3 paths and force wdl to use it. a better way would be a more fundamental
# interface shift away from files of paths throughout all of cactus
if options.pathOverrides:
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
tree = MultiCactusTree(seqFile.tree)
tree.nameUnlabeledInternalNodes(
prefix=config.getDefaultInternalNodePrefix())
for name, override in zip(options.pathOverrideNames,
options.pathOverrides):
seqFile.pathMap[name] = override
override_seq = os.path.join(options.cactusDir,
'seqFile.override')
with open(override_seq, 'w') as out_sf:
out_sf.write(str(seqFile))
options.seqFile = override_seq
#load cactus config
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
config.substituteAllPredefinedConstantsWithLiterals()
# get the minigraph "virutal" assembly name
graph_event = getOptionalAttrib(findRequiredNode(
configNode, "refgraph"),
"assemblyName",
default="__MINIGRAPH_SEQUENCES__")
# load the seqfile
seqFile = SeqFile(options.seqFile)
logger.info("Genomes for graphmap, {}".format(seqFile.pathMap))
if not options.outputFasta and graph_event not in seqFile.pathMap:
raise RuntimeError(
"{} assembly not found in seqfile so it must be specified with --outputFasta"
.format(graph_event))
#import the graph
gfa_id = toil.importFile(makeURL(options.minigraphGFA))
#import the sequences (that we need to align for the given event, ie leaves and outgroups)
seqIDMap = {}
for genome, seq in seqFile.pathMap.items():
if genome != graph_event:
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles([
os.path.join(seq, subSeq)
for subSeq in os.listdir(seq)
], tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
seqIDMap[genome] = toil.importFile(seq)
# run the workflow
paf_id, gfa_fa_id = toil.start(
Job.wrapJobFn(minigraph_workflow, options, config, seqIDMap,
gfa_id, graph_event))
#export the paf
toil.exportFile(paf_id, makeURL(options.outputPAF))
if gfa_fa_id:
toil.exportFile(gfa_fa_id, makeURL(options.outputFasta))
# update the input seqfile (in place!)
add_genome_to_seqfile(options.seqFile, makeURL(options.outputFasta),
graph_event)
def runCactusAfterBlastOnly(options):
with Toil(options) as toil:
importSingularityImage(options)
#Run the workflow
if options.restart:
alignmentID = toil.restart()
else:
options.cactusDir = getTempDirectory()
#Create the progressive cactus project (as we do in runCactusProgressive)
projWrapper = ProjectWrapper(options,
options.configFile,
ignoreSeqPaths=options.root)
projWrapper.writeXml()
pjPath = os.path.join(
options.cactusDir, ProjectWrapper.alignmentDirName,
'%s_project.xml' % ProjectWrapper.alignmentDirName)
assert os.path.exists(pjPath)
project = MultiCactusProject()
if not os.path.isdir(options.cactusDir):
os.makedirs(options.cactusDir)
project.readXML(pjPath)
# open up the experiment (as we do in ProgressiveUp.run)
# note that we copy the path into the options here
experimentFile = project.expMap[options.root]
expXml = ET.parse(experimentFile).getroot()
experiment = ExperimentWrapper(expXml)
configPath = experiment.getConfigPath()
configXml = ET.parse(configPath).getroot()
seqIDMap = dict()
tree = MultiCactusTree(experiment.getTree()).extractSubTree(
options.root)
leaves = [tree.getName(leaf) for leaf in tree.getLeaves()]
outgroups = experiment.getOutgroupGenomes()
genome_set = set(leaves + outgroups)
# import the outgroups
outgroupIDs = []
cactus_blast_input = not options.nonBlastInput
for i, outgroup in enumerate(outgroups):
try:
outgroupID = toil.importFile(
makeURL(options.blastOutput) +
'.og_fragment_{}'.format(i))
outgroupIDs.append(outgroupID)
experiment.setSequenceID(outgroup, outgroupID)
except:
if cactus_blast_input:
raise
# we assume that input is not coming from cactus blast, so we'll treat output
# sequences normally and not go looking for fragments
outgroupIDs = []
break
#import the sequences (that we need to align for the given event, ie leaves and outgroups)
for genome, seq in list(project.inputSequenceMap.items()):
if genome in leaves or (not cactus_blast_input
and genome in outgroups):
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles([
os.path.join(seq, subSeq)
for subSeq in os.listdir(seq)
], tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
experiment.setSequenceID(genome, toil.importFile(seq))
if not cactus_blast_input:
outgroupIDs = [
experiment.getSequenceID(outgroup)
for outgroup in outgroups
]
# write back the experiment, as CactusWorkflowArguments wants a path
experiment.writeXML(experimentFile)
#import cactus config
if options.configFile:
cactusConfigID = toil.importFile(makeURL(options.configFile))
else:
cactusConfigID = toil.importFile(
makeURL(project.getConfigPath()))
project.setConfigID(cactusConfigID)
project.syncToFileStore(toil)
configNode = ET.parse(project.getConfigPath()).getroot()
configWrapper = ConfigWrapper(configNode)
configWrapper.substituteAllPredefinedConstantsWithLiterals()
workFlowArgs = CactusWorkflowArguments(
options,
experimentFile=experimentFile,
configNode=configNode,
seqIDMap=project.inputSequenceIDMap)
#import the files that cactus-blast made
workFlowArgs.alignmentsID = toil.importFile(
makeURL(options.blastOutput))
try:
workFlowArgs.secondaryAlignmentsID = toil.importFile(
makeURL(options.blastOutput) + '.secondary')
except:
workFlowArgs.secondaryAlignmentsID = None
workFlowArgs.outgroupFragmentIDs = outgroupIDs
workFlowArgs.ingroupCoverageIDs = []
if cactus_blast_input and len(outgroups) > 0:
for i in range(len(leaves)):
workFlowArgs.ingroupCoverageIDs.append(
toil.importFile(
makeURL(options.blastOutput) +
'.ig_coverage_{}'.format(i)))
halID = toil.start(
Job.wrapJobFn(run_cactus_align, configWrapper, workFlowArgs,
project, cactus_blast_input))
# export the hal
toil.exportFile(halID, makeURL(options.outputHal))
def runCactusAfterBlastOnly(options):
with Toil(options) as toil:
importSingularityImage(options)
#Run the workflow
if options.restart:
halID = toil.restart()
else:
options.cactusDir = getTempDirectory()
# apply path overrides. this was necessary for wdl which doesn't take kindly to
# text files of local paths (ie seqfile). one way to fix would be to add support
# for s3 paths and force wdl to use it. a better way would be a more fundamental
# interface shift away from files of paths throughout all of cactus
if options.pathOverrides:
seqFile = SeqFile(options.seqFile)
configNode = ET.parse(options.configFile).getroot()
config = ConfigWrapper(configNode)
tree = MultiCactusTree(seqFile.tree)
tree.nameUnlabeledInternalNodes(
prefix=config.getDefaultInternalNodePrefix())
for name, override in zip(options.pathOverrideNames,
options.pathOverrides):
seqFile.pathMap[name] = override
override_seq = os.path.join(options.cactusDir,
'seqFile.override')
with open(override_seq, 'w') as out_sf:
out_sf.write(str(seqFile))
options.seqFile = override_seq
#to be consistent with all-in-one cactus, we make sure the project
#isn't limiting itself to the subtree (todo: parameterize so root can
#be passed through from prepare to blast/align)
proj_options = copy.deepcopy(options)
proj_options.root = None
#Create the progressive cactus project (as we do in runCactusProgressive)
projWrapper = ProjectWrapper(proj_options,
proj_options.configFile,
ignoreSeqPaths=options.root)
projWrapper.writeXml()
pjPath = os.path.join(
options.cactusDir, ProjectWrapper.alignmentDirName,
'%s_project.xml' % ProjectWrapper.alignmentDirName)
assert os.path.exists(pjPath)
project = MultiCactusProject()
if not os.path.isdir(options.cactusDir):
os.makedirs(options.cactusDir)
project.readXML(pjPath)
# open up the experiment (as we do in ProgressiveUp.run)
# note that we copy the path into the options here
experimentFile = project.expMap[options.root]
expXml = ET.parse(experimentFile).getroot()
experiment = ExperimentWrapper(expXml)
configPath = experiment.getConfigPath()
configXml = ET.parse(configPath).getroot()
seqIDMap = dict()
tree = MultiCactusTree(experiment.getTree()).extractSubTree(
options.root)
leaves = [tree.getName(leaf) for leaf in tree.getLeaves()]
outgroups = experiment.getOutgroupGenomes()
genome_set = set(leaves + outgroups)
# this is a hack to allow specifying all the input on the command line, rather than using suffix lookups
def get_input_path(suffix=''):
base_path = options.cigarsFile[0]
for input_path in options.cigarsFile:
if suffix and input_path.endswith(suffix):
return input_path
if os.path.basename(base_path).startswith(
os.path.basename(input_path)):
base_path = input_path
return base_path + suffix
# import the outgroups
outgroupIDs = []
outgroup_fragment_found = False
for i, outgroup in enumerate(outgroups):
try:
outgroupID = toil.importFile(
makeURL(get_input_path('.og_fragment_{}'.format(i))))
outgroupIDs.append(outgroupID)
experiment.setSequenceID(outgroup, outgroupID)
outgroup_fragment_found = True
assert not options.pangenome
except:
# we assume that input is not coming from cactus blast, so we'll treat output
# sequences normally and not go looking for fragments
outgroupIDs = []
break
#import the sequences (that we need to align for the given event, ie leaves and outgroups)
for genome, seq in list(project.inputSequenceMap.items()):
if genome in leaves or (not outgroup_fragment_found
and genome in outgroups):
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles([
os.path.join(seq, subSeq)
for subSeq in os.listdir(seq)
], tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
experiment.setSequenceID(genome, toil.importFile(seq))
if not outgroup_fragment_found:
outgroupIDs = [
experiment.getSequenceID(outgroup)
for outgroup in outgroups
]
# write back the experiment, as CactusWorkflowArguments wants a path
experiment.writeXML(experimentFile)
#import cactus config
if options.configFile:
cactusConfigID = toil.importFile(makeURL(options.configFile))
else:
cactusConfigID = toil.importFile(
makeURL(project.getConfigPath()))
project.setConfigID(cactusConfigID)
project.syncToFileStore(toil)
configNode = ET.parse(project.getConfigPath()).getroot()
configWrapper = ConfigWrapper(configNode)
configWrapper.substituteAllPredefinedConstantsWithLiterals()
if options.pangenome:
# turn off the megablock filter as it ruins non-all-to-all alignments
configWrapper.disableCafMegablockFilter()
# the recoverable chains parameter does not seem to play nicely with star-like alignments either
#configWrapper.disableRecoverableChains()
workFlowArgs = CactusWorkflowArguments(
options,
experimentFile=experimentFile,
configNode=configNode,
seqIDMap=project.inputSequenceIDMap)
#import the files that cactus-blast made
workFlowArgs.alignmentsID = toil.importFile(
makeURL(get_input_path()))
workFlowArgs.secondaryAlignmentsID = None
if not options.pafInput:
try:
workFlowArgs.secondaryAlignmentsID = toil.importFile(
makeURL(get_input_path('.secondary')))
except:
pass
workFlowArgs.outgroupFragmentIDs = outgroupIDs
workFlowArgs.ingroupCoverageIDs = []
if outgroup_fragment_found and len(outgroups) > 0:
for i in range(len(leaves)):
workFlowArgs.ingroupCoverageIDs.append(
toil.importFile(
makeURL(get_input_path(
'.ig_coverage_{}'.format(i)))))
halID = toil.start(
Job.wrapJobFn(run_cactus_align,
configWrapper,
workFlowArgs,
project,
doRenaming=options.nonCactusInput,
pafInput=options.pafInput))
# export the hal
toil.exportFile(halID, makeURL(options.outputHal))
def main():
parser = ArgumentParser()
Job.Runner.addToilOptions(parser)
addCactusWorkflowOptions(parser)
parser.add_argument("seqFile", help="Seq file")
parser.add_argument("outputHal", type=str, help="Output HAL file")
#Progressive Cactus Options
parser.add_argument("--database",
dest="database",
help="Database type: tokyo_cabinet or kyoto_tycoon"
" [default: %(default)s]",
default="kyoto_tycoon")
parser.add_argument("--configFile",
dest="configFile",
help="Specify cactus configuration file",
default=None)
parser.add_argument(
"--root",
dest="root",
help="Name of ancestral node (which"
" must appear in NEWICK tree in <seqfile>) to use as a "
"root for the alignment. Any genomes not below this node "
"in the tree may be used as outgroups but will never appear"
" in the output. If no root is specifed then the root"
" of the tree is used. ",
default=None)
parser.add_argument("--latest",
dest="latest",
action="store_true",
help="Use the latest, locally-built docker container "
"rather than pulling from quay.io")
parser.add_argument("--binariesMode",
choices=["docker", "local", "singularity"],
help="The way to run the Cactus binaries",
default=None)
options = parser.parse_args()
options.cactusDir = getTempDirectory()
setupBinaries(options)
setLoggingFromOptions(options)
# Mess with some toil options to create useful defaults.
# Caching generally slows down the cactus workflow, plus some
# methods like readGlobalFileStream don't support forced
# reads directly from the job store rather than from cache.
options.disableCaching = True
# Job chaining breaks service termination timing, causing unused
# databases to accumulate and waste memory for no reason.
options.disableChaining = True
# The default deadlockWait is currently 60 seconds. This can cause
# issues if the database processes take a while to actually begin
# after they're issued. Change it to at least an hour so that we
# don't preemptively declare a deadlock.
if options.deadlockWait is None or options.deadlockWait < 3600:
options.deadlockWait = 3600
if options.retryCount is None:
# If the user didn't specify a retryCount value, make it 5
# instead of Toil's default (1).
options.retryCount = 5
#Create the progressive cactus project
projWrapper = ProjectWrapper(options)
projWrapper.writeXml()
pjPath = os.path.join(options.cactusDir, ProjectWrapper.alignmentDirName,
'%s_project.xml' % ProjectWrapper.alignmentDirName)
assert os.path.exists(pjPath)
project = MultiCactusProject()
if not os.path.isdir(options.cactusDir):
os.makedirs(options.cactusDir)
with Toil(options) as toil:
importSingularityImage()
#Run the workflow
if options.restart:
halID = toil.restart()
else:
project.readXML(pjPath)
#import the sequences
seqIDs = []
for seq in project.getInputSequencePaths():
if os.path.isdir(seq):
tmpSeq = getTempFile()
catFiles([
os.path.join(seq, subSeq) for subSeq in os.listdir(seq)
], tmpSeq)
seq = tmpSeq
seq = makeURL(seq)
seqIDs.append(toil.importFile(seq))
project.setInputSequenceIDs(seqIDs)
#import cactus config
if options.configFile:
cactusConfigID = toil.importFile(makeURL(options.configFile))
else:
cactusConfigID = toil.importFile(
makeURL(project.getConfigPath()))
logger.info("Setting config id to: %s" % cactusConfigID)
project.setConfigID(cactusConfigID)
project.syncToFileStore(toil)
configNode = ET.parse(project.getConfigPath()).getroot()
configWrapper = ConfigWrapper(configNode)
configWrapper.substituteAllPredefinedConstantsWithLiterals()
project.writeXML(pjPath)
halID = toil.start(
RunCactusPreprocessorThenProgressiveDown(
options, project, memory=configWrapper.getDefaultMemory()))
toil.exportFile(halID, makeURL(options.outputHal))
