def testDynamicOutgroupsJustLeaves(self):
tree = '((((HUMAN:0.006969,CHIMP:0.009727)Anc7:0.025291,BABOON:0.044568)Anc6:0.11,(MOUSE:0.072818,RAT:0.081244)Anc5:0.260342)Anc4:0.023260,((DOG:0.07,CAT:0.07)Anc3:0.087381,(PIG:0.06,COW:0.06)Anc2:0.104728)Anc1:0.04)Anc0;'
mcTree = MultiCactusTree(NXNewick().parseString(tree, addImpliedRoots = False))
mcTree.computeSubtreeRoots()
og = DynamicOutgroup()
og.importTree(mcTree, self.blanchetteSeqMap)
og.compute(maxNumOutgroups=3, sequenceLossWeight=0.)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# and for all entries, the closest must be first.
assert all(map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))
# ordering is important!
assert og.ogMap['Anc1'][0][0] == 'HUMAN'
assert og.ogMap['Anc7'][0][0] == 'BABOON'
og = DynamicOutgroup()
og.importTree(mcTree, self.blanchetteSeqMap)
og.compute(maxNumOutgroups=3)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# we keep dynamic outgroups sorted by distance too
assert all(map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))
def fillInOutgroups(mcProj, outgroupNames, config, alignmentRootId):
"""
Determines the outgroups for a MultiCactusProject using the strategy from the config.
"""
mcProj.outgroup = None
if config.getOutgroupStrategy() == 'greedy':
# use the provided outgroup candidates, or use all outgroups
# as candidates if none are given
mcProj.outgroup = GreedyOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree, alignmentRootId)
mcProj.outgroup.greedy(
threshold=config.getOutgroupThreshold(),
candidateSet=outgroupNames,
candidateChildFrac=config.getOutgroupAncestorQualityFraction(),
maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() == 'greedyLeaves':
# use all leaves as outgroups, unless outgroup candidates are given
mcProj.outgroup = GreedyOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree, alignmentRootId)
mcProj.outgroup.greedy(threshold=config.getOutgroupThreshold(),
candidateSet=outgroupNames,
candidateChildFrac=2.0,
maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() == 'greedyPreference':
# prefer the provided outgroup candidates, if any, but use
# other nodes as "filler" if we can't find enough.
mcProj.outgroup = GreedyOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree, alignmentRootId)
mcProj.outgroup.greedy(
threshold=config.getOutgroupThreshold(),
candidateSet=outgroupNames,
candidateChildFrac=config.getOutgroupAncestorQualityFraction(),
maxNumOutgroups=config.getMaxNumOutgroups())
mcProj.outgroup.greedy(
threshold=config.getOutgroupThreshold(),
candidateSet=None,
candidateChildFrac=config.getOutgroupAncestorQualityFraction(),
maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() == 'dynamic':
# dynamic programming algorithm that exactly optimizes probability
# that base in target node aligns to at least one base in the
# outgroup set. Caveats are that it only returns leaves, and
# the model used for optimization is super naive. Still, it does
# some things better than greedy approaches such as properly account
# for phylogenetic redundancy, as well as try to factor assembly
# size/quality automatically.
mcProj.outgroup = DynamicOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree,
mcProj.inputSequenceMap,
alignmentRootId,
candidateSet=outgroupNames)
mcProj.outgroup.compute(maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() != 'none':
raise RuntimeError("Could not understand outgroup strategy %s" %
config.getOutgroupStrategy())
def testDynamicOutgroupsJustLeaves(self):
og = DynamicOutgroup()
og.importTree(self.borMcTree, self.blanchetteSeqMap)
og.compute(maxNumOutgroups=3, sequenceLossWeight=0.)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# and for all entries, the closest must be first.
assert all(map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))
# ordering is important!
assert og.ogMap['Anc1'][0][0] == 'HUMAN'
assert og.ogMap['Anc7'][0][0] == 'BABOON'
og = DynamicOutgroup()
og.importTree(self.borMcTree, self.blanchetteSeqMap)
og.compute(maxNumOutgroups=3)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# we keep dynamic outgroups sorted by distance too
assert all(map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))
def testDynamicOutgroupsOnRandomTrees(self):
for tree, seqMap in zip(self.mcTrees, self.dummySeqMaps):
degree = max([
len(tree.getChildren(x)) for x in tree.breadthFirstTraversal()
])
if degree < 8:
og = DynamicOutgroup()
og.edgeLen = 5
og.importTree(tree, seqMap)
og.compute(maxNumOutgroups=3)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# and for all entries, the closest must be first.
# (this will be true because all sequences are the same)
assert all(
map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))
def testDynamicOutgroupsOnRandomTrees(self):
for tree, seqMap in zip(self.mcTrees, self.dummySeqMaps):
degree = max([len(tree.getChildren(x)) for x in
tree.breadthFirstTraversal()])
if degree < 8:
og = DynamicOutgroup()
og.edgeLen = 5
og.importTree(tree, seqMap)
og.compute(maxNumOutgroups=3)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# and for all entries, the closest must be first.
# (this will be true because all sequences are the same)
assert all(map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))
def createMCProject(tree, experiment, config, options):
mcTree = MultiCactusTree(tree, config.getSubtreeSize())
mcTree.nameUnlabeledInternalNodes(config.getDefaultInternalNodePrefix())
mcTree.computeSubtreeRoots()
mcProj = MultiCactusProject()
mcProj.mcTree = mcTree
mcProj.inputSequences = experiment.getSequences()[:]
if config.getDoSelfAlignment():
mcTree.addSelfEdges()
for name in mcProj.mcTree.getSubtreeRootNames():
expPath = "%s/%s/%s_experiment.xml" % (options.path, name, name)
mcProj.expMap[name] = os.path.abspath(expPath)
alignmentRootId = mcProj.mcTree.getRootId()
if options.root is not None:
try:
alignmentRootId = mcProj.mcTree.getNodeId(options.root)
except:
raise RuntimeError("Specified root name %s not found in tree" %
options.root)
mcProj.outgroup = None
if config.getOutgroupStrategy() == 'greedy':
# use the provided outgroup candidates, or use all outgroups
# as candidates if none are given
mcProj.outgroup = GreedyOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree, alignmentRootId)
mcProj.outgroup.greedy(
threshold=config.getOutgroupThreshold(),
candidateSet=options.outgroupNames,
candidateChildFrac=config.getOutgroupAncestorQualityFraction(),
maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() == 'greedyLeaves':
# use all leaves as outgroups, unless outgroup candidates are given
mcProj.outgroup = GreedyOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree, alignmentRootId)
ogSet = options.outgroupNames
if ogSet is None:
ogSet = set(
[mcProj.mcTree.getName(x) for x in mcProj.mcTree.getLeaves()])
mcProj.outgroup.greedy(threshold=config.getOutgroupThreshold(),
candidateSet=ogSet,
candidateChildFrac=2.0,
maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() == 'greedyPreference':
# prefer the provided outgroup candidates, if any, but use
# other nodes as "filler" if we can't find enough.
mcProj.outgroup = GreedyOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree, alignmentRootId)
mcProj.outgroup.greedy(
threshold=config.getOutgroupThreshold(),
candidateSet=options.outgroupNames,
candidateChildFrac=config.getOutgroupAncestorQualityFraction(),
maxNumOutgroups=config.getMaxNumOutgroups())
mcProj.outgroup.greedy(
threshold=config.getOutgroupThreshold(),
candidateSet=None,
candidateChildFrac=config.getOutgroupAncestorQualityFraction(),
maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() == 'dynamic':
# dynamic programming algorithm that exactly optimizes probability
# that base in target node aligns to at least one base in the
# outgroup set. Caveats are that it only returns leaves, and
# the model used for optimization is super naive. Still, it does
# some things better than greedy approaches such as properly account
# for phylogenetic redundancy, as well as try to factor assembly
# size/quality automatically.
mcProj.outgroup = DynamicOutgroup()
mcProj.outgroup.importTree(mcProj.mcTree,
mcProj.getInputSequenceMap(),
alignmentRootId,
candidateSet=options.outgroupNames)
mcProj.outgroup.compute(maxNumOutgroups=config.getMaxNumOutgroups())
elif config.getOutgroupStrategy() != 'none':
raise RuntimeError("Could not understand outgroup strategy %s" %
config.getOutgroupStrategy())
# if necessary, we reroot the tree at the specified alignment root id. all leaf genomes
# that are no longer in the tree, but still used as outgroups, are moved into special fields
# so that we can remember to, say, get their paths for preprocessing.
specifyAlignmentRoot(mcProj, alignmentRootId)
return mcProj
def testDynamicOutgroupsJustLeaves(self):
og = DynamicOutgroup()
og.importTree(self.borMcTree, self.blanchetteSeqMap)
og.compute(maxNumOutgroups=3, sequenceLossWeight=0.)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# and for all entries, the closest must be first.
assert all(
map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))
# ordering is important!
assert og.ogMap['Anc1'][0][0] == 'HUMAN'
assert og.ogMap['Anc7'][0][0] == 'BABOON'
og = DynamicOutgroup()
og.importTree(self.borMcTree, self.blanchetteSeqMap)
og.compute(maxNumOutgroups=3)
# make sure all entries have <= 3 outgroups.
assert all(map(lambda x: len(x) <= 3, og.ogMap.values()))
# we keep dynamic outgroups sorted by distance too
assert all(
map(lambda x: x == sorted(x, key=itemgetter(1)),
og.ogMap.values()))