def run(self, outputFile):
img = IMG()
print 'Identifying all IMG prokaryotic genomes with valid data.'
metadata = img.genomeMetadata()
genomeIds = img.genomeIdsByTaxonomy('prokaryotes', metadata)
genomeMissingData = img.genomesWithMissingData(genomeIds)
genomeIds -= genomeMissingData
print ' Identified %d valid genomes.' % (len(genomeIds))
print 'Calculating gene copy number for each genome.'
countTable = img.geneCountTable(genomeIds)
counts = []
for _, count in countTable['pfam00318'].iteritems():
counts.append(count)
print len(genomeIds)
print len(counts)
print mean(counts)
fout = open(outputFile, 'w')
fout.write(str(countTable))
fout.close()
print 'Gene count dictionary to: ' + outputFile
def run(self, outputFile):
img = IMG()
print('Identifying all IMG prokaryotic genomes with valid data.')
metadata = img.genomeMetadata()
genomeIds = img.genomeIdsByTaxonomy('prokaryotes', metadata)
genomeMissingData = img.genomesWithMissingData(genomeIds)
genomeIds -= genomeMissingData
print(' Identified %d valid genomes.' % (len(genomeIds)))
print('Calculating gene copy number for each genome.')
countTable = img.geneCountTable(genomeIds)
counts = []
for _, count in countTable['pfam00318'].iteritems():
counts.append(count)
print(len(genomeIds))
print(len(counts))
print(mean(counts))
fout = open(outputFile, 'w')
fout.write(str(countTable))
fout.close()
print('Gene count dictionary to: ' + outputFile)
class MarkerSetSelection(object):
def __init__(self):
self.simFile = './experiments/simulation.tuning.genus.summary.tsv'
self.looRank = 5
self.markerSetBuilder = MarkerSetBuilder()
self.img = IMG()
def __stabilityTest(self,
genomeIds,
ubiquityThreshold=0.97,
singleCopyThreshold=0.97,
stabilityThreshold=0.05):
"""Test stability of marker set for a group of genomes using LOO-testing."""
# quick escape for lineage that are clearly stable
if len(genomeIds) > 200:
return True
# calculate marker sets using a LOO-testing
looMarkerGenes = []
for genomeId in genomeIds:
looGenomeIds = genomeIds.difference([genomeId])
# calculate marker genes
geneCountTable = self.img.geneCountTable(looGenomeIds)
markerGenes = self.markerSetBuilder.markerGenes(
looGenomeIds, geneCountTable,
ubiquityThreshold * len(looGenomeIds),
singleCopyThreshold * len(looGenomeIds))
tigrToRemove = self.img.identifyRedundantTIGRFAMs(markerGenes)
markerGenes = markerGenes - tigrToRemove
looMarkerGenes.append(markerGenes)
# calculate change in marker set for all pairs
markerSetSize = []
diffMarkerSet = []
for i in range(0, len(looMarkerGenes)):
markerSetSize.append(len(looMarkerGenes[i]))
for j in range(i + 1, len(looMarkerGenes)):
symmDiff = looMarkerGenes[i].symmetric_difference(
looMarkerGenes[j])
diffMarkerSet.append(len(symmDiff))
print(len(genomeIds), mean(diffMarkerSet), mean(markerSetSize))
return (float(mean(diffMarkerSet)) /
mean(markerSetSize)) <= stabilityThreshold
def __patristicDist(self, tree, taxa1, taxa2):
mrca = tree.mrca(taxon_labels=[taxa1.taxon.label, taxa2.taxon.label])
if mrca.parent_node == None:
# MRCA is the root of the tree
return taxa1.distance_from_root() + taxa2.distance_from_root()
else:
dist = taxa1.edge_length
parentNode = taxa1.parent_node
while parentNode != mrca:
dist += parentNode.edge_length
parentNode = parentNode.parent_node
dist += taxa2.edge_length
parentNode = taxa2.parent_node
while parentNode != mrca:
dist += parentNode.edge_length
parentNode = parentNode.parent_node
return dist
def __distToNodePercentileTest(self, genomeNode, markerSetNode, leaves,
percentileTest):
distToBin = self.__distanceToAncestor(genomeNode, markerSetNode)
distToLeaves = []
for leaf in leaves:
distToLeaves.append(self.__distanceToAncestor(leaf, markerSetNode))
return distToBin < percentile(distToLeaves, percentileTest)
def __selectMarkerSetNode(self, tree, genomeId, metadata,
taxonToGenomeIds):
"""Determine lineage-specific marker set to use for assessing the giving genome."""
# read genomes removed from tree as a result of duplicate sequences
duplicateSeqs = self.markerSetBuilder.readDuplicateSeqs()
# determine location of genome in tree
node = tree.find_node_with_taxon_label('IMG_' + genomeId)
# ascend tree to root looking for suitable marker set
curNode = node.parent_node
while curNode != None:
uniqueId = curNode.label.split('|')[0]
genomeIds = set()
for leaf in curNode.leaf_nodes():
genomeIds.add(leaf.taxon.label.replace('IMG_', ''))
duplicateGenomes = duplicateSeqs.get(leaf.taxon.label, [])
for dup in duplicateGenomes:
genomeIds.add(dup.replace('IMG_', ''))
# remove genome (LOO-style analysis)
print('Full:', len(genomeIds))
genomeIds.difference_update([genomeId])
print('LOO:', len(genomeIds))
# remove all genomes from the same taxonomic group as the genome of interest
taxon = metadata[genomeId]['taxonomy'][self.looRank]
genomeIds.difference_update(taxonToGenomeIds[taxon])
print('Rank reduced:', len(genomeIds))
print(uniqueId)
if len(genomeIds) > 10 and self.__stabilityTest(genomeIds):
uidSelected = uniqueId
break
curNode = curNode.parent_node
if curNode == None:
# reach root so use universal marker set
uidSelected = uniqueId
return uidSelected
def __bestMarkerSet(self, genomeId, simResults):
"""Get stats for best marker set."""
curBest = 1000
bestUID = None
for uid, results in simResults[genomeId].items():
numDescendants, dComp, dCont = results
if (dComp + dCont) < curBest:
numDescendantsBest = numDescendants
dCompBest = dComp
dContBest = dCont
bestUID = uid
curBest = dComp + dCont
return bestUID, numDescendantsBest, dCompBest, dContBest
def __workerThread(self, tree, simResults, metadata, taxonToGenomeIds,
queueIn, queueOut):
"""Process each data item in parallel."""
while True:
testGenomeId = queueIn.get(block=True, timeout=None)
if testGenomeId == None:
break
uidSelected = self.__selectMarkerSetNode(tree, testGenomeId,
metadata,
taxonToGenomeIds)
numDescendantsSelected, dCompSelected, dContSelected = simResults[
testGenomeId][uidSelected]
# find best marker set
bestUID, numDescendantsBest, dCompBest, dContBest = self.__bestMarkerSet(
testGenomeId, simResults)
queueOut.put((testGenomeId, uidSelected, numDescendantsSelected,
dCompSelected, dContSelected, bestUID,
numDescendantsBest, dCompBest, dContBest))
def __writerThread(self, numTestGenomes, writerQueue):
"""Store or write results of worker threads in a single thread."""
fout = open('./experiments/markerSetSelection.tsv', 'w')
fout.write(
'Genome Id\tSelected UID\t# descendants\tSelected dComp\tSelected dCont\tBest UID\t# descendants\tBest dComp\tBest dCont\tdDescendants\tdComp\tdCont\n'
)
itemsToProcess = 0
dComps = []
dConts = []
dCompsPer = []
dContsPer = []
bestComp = []
bestCont = []
selectedComp = []
selectedCont = []
while True:
testGenomeId, uidSelected, numDescendantsSelected, dCompSelected, dContSelected, bestUID, numDescendantsBest, dCompBest, dContBest = writerQueue.get(
block=True, timeout=None)
if testGenomeId == None:
break
itemsToProcess += 1
statusStr = ' Finished processing %d of %d (%.2f%%) test genomes.' % (
itemsToProcess, numTestGenomes, float(itemsToProcess) * 100 /
(numTestGenomes))
sys.stdout.write('%s\r' % statusStr)
sys.stdout.flush()
dComp = abs(dCompSelected - dCompBest)
dCont = abs(dContSelected - dContBest)
dDescendants = abs(numDescendantsSelected - numDescendantsBest)
fout.write(
'%s\t%s\t%d\t%.4f\t%.4f\t%s\t%d\t%.4f\t%.4f\t%d\t%.4f\t%.4f\n'
% (testGenomeId, uidSelected, numDescendantsSelected,
dCompSelected, dContSelected, bestUID, numDescendantsBest,
dCompBest, dContBest, dDescendants, dComp, dCont))
dComps.append(dComp)
dConts.append(dCont)
dCompsPer.append(dComp * 100.0 / dCompBest)
dContsPer.append(dCont * 100.0 / max(dContBest, 0.01))
bestComp.append(dCompBest)
bestCont.append(dContBest)
selectedComp.append(dCompSelected)
selectedCont.append(dContSelected)
sys.stdout.write('\n')
fout.close()
print('')
print(' General results:')
print(' Best comp: %.2f +/- %.2f' % (mean(bestComp), std(bestComp)))
print(' Best cont: %.2f +/- %.2f' % (mean(bestCont), std(bestCont)))
print(' Selected comp: %.2f +/- %.2f' %
(mean(selectedComp), std(selectedComp)))
print(' Selected cont: %.2f +/- %.2f' %
(mean(selectedCont), std(selectedCont)))
print('')
print(' Delta comp: %.2f +/- %.2f' % (mean(dComps), std(dComps)))
print(' Delta cont: %.2f +/- %.2f' % (mean(dConts), std(dConts)))
print(' Delta comp per error: %.1f +/- %.1f' %
(mean(dCompsPer), std(dCompsPer)))
print(' Delta cont per error: %.1f +/- %.1f' %
(mean(dContsPer), std(dContsPer)))
def __distanceToAncestor(self, leaf, ancestor):
dist = 0
curNode = leaf
while curNode != ancestor:
dist += curNode.edge_length
curNode = curNode.parent_node
return dist
def __bestNodeProperties(self, genomeId, tree, bestUID):
# determine location of genome in tree
node = tree.find_node_with_taxon_label('IMG_' + genomeId)
# find node of best marker set
curNode = node.parent_node
nodesToBin = 0
distanceToBin = node.edge_length
distanceToLeaves = []
while curNode != None:
uniqueId = curNode.label.split('|')[0]
nodesToBin += 1
if uniqueId == bestUID:
for leaf in curNode.leaf_nodes():
if leaf != node:
dist = self.__distanceToAncestor(leaf, curNode)
distanceToLeaves.append(dist)
break
distanceToBin += curNode.edge_length
curNode = curNode.parent_node
return nodesToBin, distanceToBin, mean(distanceToLeaves)
def __propertiesOfBestMarkerSets(self, tree, simResults):
numDescendants = []
nodesToBin = []
distanceToBin = []
avgDistanceToLeaf = []
percDiffs = []
for genomeId in simResults:
bestUID, numDescendantsBest, _, _ = self.__bestMarkerSet(
genomeId, simResults)
nodesToBinBest, distanceToBinBest, avgDistanceToLeafBest = self.__bestNodeProperties(
genomeId, tree, bestUID)
numDescendants.append(numDescendantsBest)
nodesToBin.append(nodesToBinBest)
distanceToBin.append(distanceToBinBest)
avgDistanceToLeaf.append(avgDistanceToLeafBest)
percDiff = abs(distanceToBinBest -
avgDistanceToLeafBest) * 100 / distanceToBinBest
percDiffs.append(percDiff)
print(' # descendants: %.2f +/- %.2f' %
(mean(numDescendants), std(numDescendants)))
print(' # nodes to bin: %.2f +/- %.2f' %
(mean(nodesToBin), std(nodesToBin)))
print(' Distance to bin: %.2f +/- %.2f' %
(mean(distanceToBin), std(distanceToBin)))
distanceToBin = array(distanceToBin)
avgDistanceToLeaf = array(avgDistanceToLeaf)
print(' Distance to bin - average distance to leaf: %.2f +/- %.2f' %
(mean(abs(distanceToBin - avgDistanceToLeaf)),
std(abs(distanceToBin - avgDistanceToLeaf))))
print(
' Percent difference to average leaf distance: %.2f +/- %.2f' %
(mean(percDiffs), std(percDiffs)))
print('')
def run(self, numThreads):
# read reference tree
print('\n Reading reference genome tree.')
treeFile = os.path.join(os.path.dirname(sys.argv[0]), '..', 'data',
'genome_tree', 'genome_tree_prok.refpkg',
'genome_tree.final.tre')
tree = dendropy.Tree.get_from_path(treeFile,
schema='newick',
as_rooted=True,
preserve_underscores=True)
# get all genomes with a given taxon label
metadata = self.img.genomeMetadata()
taxonToGenomeIds = defaultdict(set)
for genomeId in metadata:
for t in metadata[genomeId]['taxonomy']:
taxonToGenomeIds[t].add(genomeId)
# read simulation results
print(' Reading simulation results.')
simResults = defaultdict(dict)
with open(self.simFile) as f:
f.readline()
for line in f:
lineSplit = line.split('\t')
simId = lineSplit[0] + '-' + lineSplit[1] + '-' + lineSplit[
2] + '-' + lineSplit[3]
uid = lineSplit[5].split('|')[0].strip()
numDescendants = int(lineSplit[6])
comp = float(lineSplit[21])
cont = float(lineSplit[23])
simResults[simId][uid] = [numDescendants, comp, cont]
#print ''
#print ' Properties of best marker sets:'
#self.__propertiesOfBestMarkerSets(tree, simResults)
print(' Evaluating %d test genomes.' % len(simResults))
workerQueue = mp.Queue()
writerQueue = mp.Queue()
for testGenomeId in simResults:
workerQueue.put(testGenomeId)
for _ in range(numThreads):
workerQueue.put(None)
workerProc = [
mp.Process(target=self.__workerThread,
args=(tree, simResults, metadata, taxonToGenomeIds,
workerQueue, writerQueue))
for _ in range(numThreads)
]
writeProc = mp.Process(target=self.__writerThread,
args=(len(simResults), writerQueue))
writeProc.start()
for p in workerProc:
p.start()
for p in workerProc:
p.join()
writerQueue.put((None, None, None, None, None, None, None, None, None))
writeProc.join()
class Simulation(object):
def __init__(self):
self.markerSetBuilder = MarkerSetBuilder()
self.img = IMG()
self.simContigLen = 10000
def __selectMarkerSet(self, tree, internalNode, metadata, ubiquityThreshold, singleCopyThreshold, queueOut):
"""Select marker set for parent edge of specified internal node."""
# get genomes descendant from each child of the specified internal node
leaves = []
for child in internalNode.child_nodes():
genomeIds = set()
for leaf in child.leaf_nodes():
genomeId = leaf.taxon.label.replace('IMG_', '')
genomeIds.add(genomeId)
duplicateGenomes = self.markerSetBuilder.duplicateSeqs.get(leaf.taxon.label, [])
for dup in duplicateGenomes:
dupId = dup.replace('IMG_', '')
genomeIds.add(dupId)
leaves.append(genomeIds)
# make sure each set of leaves contains at least a minimum number of genomes
orderedLeaves = sorted(leaves, key=len)
if len(orderedLeaves[0]) < 5:
queueOut.put(('NA', -1, -1, -1, -1, -1))
return
# calculate marker genes with all genomes in lineage with the fewest genomes removed
binMarkerGenes, _ = self.markerSetBuilder.buildBinMarkerSet(tree, internalNode, ubiquityThreshold, singleCopyThreshold, bMarkerSet = False, genomeIdsToRemove = orderedLeaves[0])
# evaluate accuracy of completeness and contamination estimations on different partial genomes from lineage with fewest genomes
testGenomeIds = random.sample(orderedLeaves[0], min(len(orderedLeaves[0]), 100))
deltaComp = defaultdict(list)
deltaCont = defaultdict(list)
for testGenomeId in testGenomeIds:
geneDistTable = self.img.geneDistTable([testGenomeId], binMarkerGenes.getMarkerGenes(), spacingBetweenContigs=0)
genomeSize = readFastaBases(os.path.join(self.img.genomeDir, testGenomeId, testGenomeId + '.fna'))
repsPerGenome = 100
for _ in xrange(0, repsPerGenome):
testComp = random.uniform(0.5, 1.0)
testCont = random.uniform(0, 0.2)
trueComp, trueCont, startPartialGenomeContigs = self.markerSetBuilder.sampleGenome(genomeSize, testComp, testCont, self.simContigLen)
for ms in binMarkerGenes.markerSetIter():
containedMarkerGenes = self.markerSetBuilder.containedMarkerGenes(ms.getMarkerGenes(), geneDistTable[testGenomeId], startPartialGenomeContigs, self.simContigLen)
completeness, contamination = ms.genomeCheck(containedMarkerGenes, bIndividualMarkers=True)
if completeness == 0.0:
print ms.getMarkerGenes()
print geneDistTable[testGenomeId]
print startPartialGenomeContigs
print genomeSize
print '*****************' + testGenomeId
sys.exit()
deltaComp[ms.lineageStr].append(completeness - trueComp)
deltaCont[ms.lineageStr].append(contamination - trueCont)
# determine lineage-specific marker set with best average performance
curBest = 1000
bestUID = None
dCompBest = 0
dContBest = 0
for lineageStr in deltaComp:
dComp, dCont = mean(abs(array(deltaComp[lineageStr]))), mean(abs(array(deltaCont[lineageStr])))
if (dComp + dCont) < curBest:
dCompBest = dComp
dContBest = dCont
dCompStdBest = std(abs(array(deltaComp[lineageStr])))
dContStdBest = std(abs(array(deltaCont[lineageStr])))
bestUID = lineageStr.split('|')[0]
curBest = dComp + dCont
queueOut.put((internalNode, bestUID, dCompBest, dCompStdBest, dContBest, dContStdBest))
def __workerThread(self, tree, metadata, ubiquityThreshold, singleCopyThreshold, queueIn, queueOut):
"""Process each data item in parallel."""
while True:
internalNode = queueIn.get(block=True, timeout=None)
if internalNode == None:
break
self.__selectMarkerSet(tree, internalNode, metadata, ubiquityThreshold, singleCopyThreshold, queueOut)
def __writerThread(self, numInternalNodes, writerQueue):
"""Store or write results of worker threads in a single thread."""
fout = open('/tmp/simInferBestMarkerSet.tsv', 'w')
fout.write('Internal node ID\tMarker set ID\tmean % delta comp\tstd % delta comp\tmean % delta cont\tstd % delta cont\n')
itemsProcessed = 0
while True:
internalNode, bestUID, dCompBest, dCompStdBest, dContBest, dContStdBest = writerQueue.get(block=True, timeout=None)
if internalNode == None:
break
itemsProcessed += 1
statusStr = ' Finished processing %d of %d (%.2f%%) internal branches.' % (itemsProcessed, numInternalNodes, float(itemsProcessed)*100/(numInternalNodes))
sys.stdout.write('%s\r' % statusStr)
sys.stdout.flush()
if internalNode != 'NA':
fout.write(internalNode.label + '\t%s\t%.2f\t%.2f\t%.2f\t%.2f\n' % (bestUID, dCompBest, dCompStdBest, dContBest, dContStdBest))
fout.close()
sys.stdout.write('\n')
def run(self, ubiquityThreshold, singleCopyThreshold, numThreads):
random.seed(0)
print '\n Calculating global gene count table.'
metadata = self.img.genomeMetadata()
self.markerSetBuilder.globalGeneCountTable = self.img.geneCountTable(metadata.keys())
print '\n Reading reference genome tree.'
treeFile = os.path.join(os.path.dirname(sys.argv[0]), '..', 'data', 'genome_tree', 'genome_tree_prok.refpkg', 'genome_tree.final.tre')
tree = dendropy.Tree.get_from_path(treeFile, schema='newick', as_rooted=True, preserve_underscores=True)
print ' Evaluating %d internal nodes.' % len(tree.internal_nodes())
workerQueue = mp.Queue()
writerQueue = mp.Queue()
for internalNode in tree.internal_nodes():
if internalNode.parent_node != None:
workerQueue.put(internalNode)
for _ in range(numThreads):
workerQueue.put(None)
metadata = self.img.genomeMetadata()
workerProc = [mp.Process(target = self.__workerThread, args = (tree, metadata, ubiquityThreshold, singleCopyThreshold, workerQueue, writerQueue)) for _ in range(numThreads)]
writeProc = mp.Process(target = self.__writerThread, args = (len(tree.internal_nodes())-1, writerQueue))
writeProc.start()
for p in workerProc:
p.start()
for p in workerProc:
p.join()
writerQueue.put((None, None, None, None, None, None))
writeProc.join()
class MarkerSetStability(object):
def __init__(self):
self.img = IMG()
self.markerset = MarkerSet()
def __processLineage(self, metadata, ubiquityThreshold,
singleCopyThreshold, minGenomes, queueIn, queueOut):
"""Assess stability of marker set for a specific named taxonomic group."""
while True:
lineage = queueIn.get(block=True, timeout=None)
if lineage == None:
break
genomeIds = self.img.genomeIdsByTaxonomy(lineage, metadata,
'trusted')
changeMarkerSetSize = {}
markerGenes = []
if len(genomeIds) >= minGenomes:
# calculate marker set for all genomes in lineage
geneCountTable = self.img.geneCountTable(genomeIds)
markerGenes = self.markerset.markerGenes(
genomeIds, geneCountTable,
ubiquityThreshold * len(genomeIds),
singleCopyThreshold * len(genomeIds))
tigrToRemove = self.img.identifyRedundantTIGRFAMs(markerGenes)
markerGenes = markerGenes - tigrToRemove
for selectPer in range(50, 101, 5):
numGenomesToSelect = int(
float(selectPer) / 100 * len(genomeIds))
perChange = []
for _ in range(0, 10):
# calculate marker set for subset of genomes
subsetGenomeIds = random.sample(
genomeIds, numGenomesToSelect)
geneCountTable = self.img.geneCountTable(
subsetGenomeIds)
subsetMarkerGenes = self.markerset.markerGenes(
subsetGenomeIds, geneCountTable,
ubiquityThreshold * numGenomesToSelect,
singleCopyThreshold * numGenomesToSelect)
tigrToRemove = self.img.identifyRedundantTIGRFAMs(
subsetMarkerGenes)
subsetMarkerGenes = subsetMarkerGenes - tigrToRemove
perChange.append(
float(
len(
markerGenes.symmetric_difference(
subsetMarkerGenes))) * 100.0 /
len(markerGenes))
changeMarkerSetSize[selectPer] = [
mean(perChange), std(perChange)
]
queueOut.put((lineage, len(genomeIds), len(markerGenes),
changeMarkerSetSize))
def __storeResults(self, outputFile, totalLineages, writerQueue):
"""Store results to file."""
fout = open(outputFile, 'w')
fout.write(
'Lineage\t# genomes\t# markers\tsubsample %\tmean % change\tstd % change\n'
)
numProcessedLineages = 0
while True:
lineage, numGenomes, numMarkerGenes, changeMarkerSetSize = writerQueue.get(
block=True, timeout=None)
if lineage == None:
break
numProcessedLineages += 1
statusStr = ' Finished processing %d of %d (%.2f%%) lineages.' % (
numProcessedLineages, totalLineages,
float(numProcessedLineages) * 100 / totalLineages)
sys.stdout.write('%s\r' % statusStr)
sys.stdout.flush()
for selectPer in sorted(changeMarkerSetSize.keys()):
fout.write('%s\t%d\t%d\t%d\t%f\t%f\n' %
(lineage, numGenomes, numMarkerGenes, selectPer,
changeMarkerSetSize[selectPer][0],
changeMarkerSetSize[selectPer][1]))
sys.stdout.write('\n')
fout.close()
def run(self, outputFile, ubiquityThreshold, singleCopyThreshold,
minGenomes, mostSpecificRank, numThreads):
"""Calculate stability of marker sets for named taxonomic groups."""
print(' Calculating stability of marker sets:')
random.seed(1)
# process each sequence in parallel
workerQueue = mp.Queue()
writerQueue = mp.Queue()
metadata = self.img.genomeMetadata()
lineages = self.img.lineagesByCriteria(metadata, minGenomes,
mostSpecificRank)
#lineages = ['Bacteria']
#lineages += ['Bacteria;Proteobacteria']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales;Enterobacteriaceae']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales;Enterobacteriaceae;Escherichia']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales;Enterobacteriaceae;Escherichia;coli']
#lineages = ['Archaea']
#lineages += ['Archaea;Euryarchaeota']
#lineages += ['Archaea;Euryarchaeota;Methanomicrobia']
#lineages += ['Archaea;Euryarchaeota;Methanomicrobia;Methanosarcinales']
#lineages += ['Archaea;Euryarchaeota;Methanomicrobia;Methanosarcinales;Methanosarcinaceae']
for lineage in lineages:
workerQueue.put(lineage)
for _ in range(numThreads):
workerQueue.put(None)
calcProc = [
mp.Process(target=self.__processLineage,
args=(metadata, ubiquityThreshold, singleCopyThreshold,
minGenomes, workerQueue, writerQueue))
for _ in range(numThreads)
]
writeProc = mp.Process(target=self.__storeResults,
args=(outputFile, len(lineages), writerQueue))
writeProc.start()
for p in calcProc:
p.start()
for p in calcProc:
p.join()
writerQueue.put((None, None, None, None))
writeProc.join()
def run(self, inputMetadataFile, outputMetadataFile, outputDir,
ubiquityThreshold, singleCopyThreshold, trustedCompleteness,
trustedContamination):
img = IMG()
markerSetBuilder = MarkerSetBuilder()
allOut = open(os.path.join(outputDir, 'genomes_all.tsv'), 'w')
allOut.write(
'Genome Id\tLineage\tGenome size (Mbps)\tScaffold count\tGene count\tCoding base count\tN50\tBiotic Relationship\tStatus\tCompleteness\tContamination\tMissing markers\tDuplicate markers\n'
)
trustedOut = open(os.path.join(outputDir, 'genomes_trusted.tsv'), 'w')
trustedOut.write(
'Genome Id\tLineage\tGenome size (Mbps)\tScaffold count\tGene count\tCoding base count\tN50\tBiotic Relationship\tStatus\tCompleteness\tContamination\tMissing markers\tDuplicate markers\n'
)
filteredOut = open(os.path.join(outputDir, 'genomes_filtered.tsv'),
'w')
filteredOut.write(
'Genome Id\tLineage\tGenome size (Mbps)\tScaffold count\tGene count\tCoding base count\tN50\tBiotic Relationship\tStatus\tCompleteness\tContamination\tMissing markers\tDuplicate markers\n'
)
metadataOut = open(outputMetadataFile, 'w')
# read input metadata file
metadata = img.genomeMetadataFromFile(inputMetadataFile)
finishedGenomes = defaultdict(set)
allGenomes = defaultdict(set)
metadataLine = {}
bHeader = True
for line in open(inputMetadataFile):
if bHeader:
metadataOut.write(line)
bHeader = False
continue
lineSplit = line.split('\t')
genomeId = lineSplit[0]
domain = lineSplit[1]
status = lineSplit[2]
if status == 'Finished':
finishedGenomes[domain].add(genomeId)
allGenomes[domain].add(genomeId)
metadataLine[genomeId] = line
allTrustedGenomeIds = set()
for lineage, allLineageGenomeIds in allGenomes.items():
print('[' + lineage + ']')
print(' Number of genomes: %d' % len(allLineageGenomeIds))
# tabulate genomes from each phylum
allPhylumCounts = {}
for genomeId in allLineageGenomeIds:
taxon = metadata[genomeId]['taxonomy'][1]
allPhylumCounts[taxon] = allPhylumCounts.get(taxon, 0) + 1
# identify marker genes for finished genomes
print(
'\nDetermining initial marker gene sets for genome filtering.')
markerSet = markerSetBuilder.buildMarkerSet(
finishedGenomes[lineage], ubiquityThreshold,
singleCopyThreshold)
print(
' Marker set consists of %s marker genes organized into %d sets.'
% (markerSet.numMarkers(), markerSet.numSets()))
fout = open(
os.path.join(outputDir,
'trusted_marker_sets_' + lineage + '.txt'), 'w')
fout.write(str(markerSet.markerSet))
fout.close()
# identifying trusted genomes (highly complete, low contamination genomes)
print('\nIdentifying highly complete, low contamination genomes.')
trustedGenomeIds = set()
filteredGenomes = set()
retainedStatus = {}
filteredStatus = {}
geneCountTable = img.geneCountTable(allLineageGenomeIds)
for genomeId in allLineageGenomeIds:
completeness, contamination, missingMarkers, duplicateMarkers = markerSetBuilder.genomeCheck(
markerSet.markerSet, genomeId, geneCountTable)
genomeStr = self.__genomeString(genomeId, metadata,
completeness, contamination,
missingMarkers,
duplicateMarkers)
if completeness >= trustedCompleteness and contamination <= trustedContamination:
trustedGenomeIds.add(genomeId)
allTrustedGenomeIds.add(genomeId)
retainedStatus[metadata[genomeId]
['status']] = retainedStatus.get(
metadata[genomeId]['status'], 0) + 1
trustedOut.write(genomeStr)
allOut.write(genomeStr)
metadataOut.write(metadataLine[genomeId])
else:
filteredGenomes.add(genomeId)
filteredStatus[metadata[genomeId]
['status']] = filteredStatus.get(
metadata[genomeId]['status'], 0) + 1
filteredOut.write(genomeStr)
allOut.write(genomeStr)
print(' Filtered genomes: %d (%.2f%%)' %
(len(filteredGenomes),
len(filteredGenomes) * 100.0 / len(allLineageGenomeIds)))
print(' ' + str(filteredStatus))
print(' \nTrusted genomes: %d (%.2f%%)' %
(len(trustedGenomeIds),
len(trustedGenomeIds) * 100.0 / len(allLineageGenomeIds)))
print(' ' + str(retainedStatus))
# determine status of retained genomes
print('\nTrusted genomes by phylum:')
trustedPhylumCounts = {}
for genomeId in trustedGenomeIds:
taxon = metadata[genomeId]['taxonomy'][1]
trustedPhylumCounts[taxon] = trustedPhylumCounts.get(taxon,
0) + 1
for phylum, count in allPhylumCounts.items():
print(' ' + phylum + ': %d of %d' %
(trustedPhylumCounts.get(phylum, 0), count))
print('')
allOut.close()
trustedOut.close()
filteredOut.close()
metadataOut.close()
# write out lineage statistics for genome distribution
allStats = {}
trustedStats = {}
for r in range(0, 6): # Domain to Genus
for genomeId, data in metadata.items():
taxaStr = ';'.join(data['taxonomy'][0:r + 1])
allStats[taxaStr] = allStats.get(taxaStr, 0) + 1
if genomeId in allTrustedGenomeIds:
trustedStats[taxaStr] = trustedStats.get(taxaStr, 0) + 1
sortedLineages = img.lineagesSorted(metadata)
fout = open(os.path.join(outputDir, 'lineage_stats.tsv'), 'w')
fout.write('Lineage\tGenomes with metadata\tTrusted genomes\n')
for lineage in sortedLineages:
fout.write(lineage + '\t' + str(allStats.get(lineage, 0)) + '\t' +
str(trustedStats.get(lineage, 0)) + '\n')
fout.close()
class MarkerSetStabilityTest(object):
def __init__(self):
self.img = IMG()
self.markerset = MarkerSet()
def __processLineage(self, metadata, ubiquityThreshold,
singleCopyThreshold, minGenomes, queueIn, queueOut):
"""Assess stability of marker set for a specific named taxonomic group."""
while True:
lineage = queueIn.get(block=True, timeout=None)
if lineage == None:
break
genomeIds = self.img.genomeIdsByTaxonomy(lineage, metadata,
'trusted')
markerGenes = []
perChange = []
numGenomesToSelect = int(0.9 * len(genomeIds))
if len(genomeIds) >= minGenomes:
# calculate marker set for all genomes in lineage
geneCountTable = self.img.geneCountTable(genomeIds)
markerGenes = self.markerset.markerGenes(
genomeIds, geneCountTable,
ubiquityThreshold * len(genomeIds),
singleCopyThreshold * len(genomeIds))
tigrToRemove = self.img.identifyRedundantTIGRFAMs(markerGenes)
markerGenes = markerGenes - tigrToRemove
for _ in range(0, 100):
# calculate marker set for subset of genomes
subsetGenomeIds = random.sample(genomeIds,
numGenomesToSelect)
geneCountTable = self.img.geneCountTable(subsetGenomeIds)
subsetMarkerGenes = self.markerset.markerGenes(
subsetGenomeIds, geneCountTable,
ubiquityThreshold * numGenomesToSelect,
singleCopyThreshold * numGenomesToSelect)
tigrToRemove = self.img.identifyRedundantTIGRFAMs(
subsetMarkerGenes)
subsetMarkerGenes = subsetMarkerGenes - tigrToRemove
perChange.append(
float(
len(
markerGenes.symmetric_difference(
subsetMarkerGenes))) * 100.0 /
len(markerGenes))
if perChange != []:
queueOut.put(
(lineage, len(genomeIds), len(markerGenes),
numGenomesToSelect, mean(perChange), std(perChange)))
else:
queueOut.put((lineage, len(genomeIds), len(markerGenes),
numGenomesToSelect, -1, -1))
def __storeResults(self, outputFile, totalLineages, writerQueue):
"""Store results to file."""
fout = open(outputFile, 'w')
fout.write(
'Lineage\t# genomes\t# markers\t# sampled genomes\tmean % change\tstd % change\n'
)
numProcessedLineages = 0
while True:
lineage, numGenomes, numMarkerGenes, numSampledGenomes, meanPerChange, stdPerChange = writerQueue.get(
block=True, timeout=None)
if lineage == None:
break
numProcessedLineages += 1
statusStr = ' Finished processing %d of %d (%.2f%%) lineages.' % (
numProcessedLineages, totalLineages,
float(numProcessedLineages) * 100 / totalLineages)
sys.stdout.write('%s\r' % statusStr)
sys.stdout.flush()
fout.write('%s\t%d\t%d\t%d\t%f\t%f\n' %
(lineage, numGenomes, numMarkerGenes, numSampledGenomes,
meanPerChange, stdPerChange))
sys.stdout.write('\n')
fout.close()
def run(self, outputFile, ubiquityThreshold, singleCopyThreshold,
minGenomes, mostSpecificRank, numThreads):
"""Calculate stability of marker sets for named taxonomic groups."""
print(' Testing stability of marker sets:')
random.seed(1)
# process each sequence in parallel
workerQueue = mp.Queue()
writerQueue = mp.Queue()
metadata = self.img.genomeMetadata()
lineages = self.img.lineagesByCriteria(metadata, minGenomes,
mostSpecificRank)
for lineage in lineages:
workerQueue.put(lineage)
for _ in range(numThreads):
workerQueue.put(None)
calcProc = [
mp.Process(target=self.__processLineage,
args=(metadata, ubiquityThreshold, singleCopyThreshold,
minGenomes, workerQueue, writerQueue))
for _ in range(numThreads)
]
writeProc = mp.Process(target=self.__storeResults,
args=(outputFile, len(lineages), writerQueue))
writeProc.start()
for p in calcProc:
p.start()
for p in calcProc:
p.join()
writerQueue.put((None, None, None, None, None, None))
writeProc.join()
class MarkerSetStability(object):
def __init__(self):
self.img = IMG()
self.markerset = MarkerSet()
def __processLineage(self, metadata, ubiquityThreshold, singleCopyThreshold, minGenomes, queueIn, queueOut):
"""Assess stability of marker set for a specific named taxonomic group."""
while True:
lineage = queueIn.get(block=True, timeout=None)
if lineage == None:
break
genomeIds = self.img.genomeIdsByTaxonomy(lineage, metadata, 'trusted')
changeMarkerSetSize = {}
markerGenes = []
if len(genomeIds) >= minGenomes:
# calculate marker set for all genomes in lineage
geneCountTable = self.img.geneCountTable(genomeIds)
markerGenes = self.markerset.markerGenes(genomeIds, geneCountTable, ubiquityThreshold*len(genomeIds), singleCopyThreshold*len(genomeIds))
tigrToRemove = self.img.identifyRedundantTIGRFAMs(markerGenes)
markerGenes = markerGenes - tigrToRemove
for selectPer in xrange(50, 101, 5):
numGenomesToSelect = int(float(selectPer)/100 * len(genomeIds))
perChange = []
for _ in xrange(0, 10):
# calculate marker set for subset of genomes
subsetGenomeIds = random.sample(genomeIds, numGenomesToSelect)
geneCountTable = self.img.geneCountTable(subsetGenomeIds)
subsetMarkerGenes = self.markerset.markerGenes(subsetGenomeIds, geneCountTable, ubiquityThreshold*numGenomesToSelect, singleCopyThreshold*numGenomesToSelect)
tigrToRemove = self.img.identifyRedundantTIGRFAMs(subsetMarkerGenes)
subsetMarkerGenes = subsetMarkerGenes - tigrToRemove
perChange.append(float(len(markerGenes.symmetric_difference(subsetMarkerGenes)))*100.0 / len(markerGenes))
changeMarkerSetSize[selectPer] = [mean(perChange), std(perChange)]
queueOut.put((lineage, len(genomeIds), len(markerGenes), changeMarkerSetSize))
def __storeResults(self, outputFile, totalLineages, writerQueue):
"""Store results to file."""
fout = open(outputFile, 'w')
fout.write('Lineage\t# genomes\t# markers\tsubsample %\tmean % change\tstd % change\n')
numProcessedLineages = 0
while True:
lineage, numGenomes, numMarkerGenes, changeMarkerSetSize = writerQueue.get(block=True, timeout=None)
if lineage == None:
break
numProcessedLineages += 1
statusStr = ' Finished processing %d of %d (%.2f%%) lineages.' % (numProcessedLineages, totalLineages, float(numProcessedLineages)*100/totalLineages)
sys.stdout.write('%s\r' % statusStr)
sys.stdout.flush()
for selectPer in sorted(changeMarkerSetSize.keys()):
fout.write('%s\t%d\t%d\t%d\t%f\t%f\n' % (lineage, numGenomes, numMarkerGenes, selectPer, changeMarkerSetSize[selectPer][0], changeMarkerSetSize[selectPer][1]))
sys.stdout.write('\n')
fout.close()
def run(self, outputFile, ubiquityThreshold, singleCopyThreshold, minGenomes, mostSpecificRank, numThreads):
"""Calculate stability of marker sets for named taxonomic groups."""
print ' Calculating stability of marker sets:'
random.seed(1)
# process each sequence in parallel
workerQueue = mp.Queue()
writerQueue = mp.Queue()
metadata = self.img.genomeMetadata()
lineages = self.img.lineagesByCriteria(metadata, minGenomes, mostSpecificRank)
#lineages = ['Bacteria']
#lineages += ['Bacteria;Proteobacteria']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales;Enterobacteriaceae']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales;Enterobacteriaceae;Escherichia']
#lineages += ['Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacteriales;Enterobacteriaceae;Escherichia;coli']
#lineages = ['Archaea']
#lineages += ['Archaea;Euryarchaeota']
#lineages += ['Archaea;Euryarchaeota;Methanomicrobia']
#lineages += ['Archaea;Euryarchaeota;Methanomicrobia;Methanosarcinales']
#lineages += ['Archaea;Euryarchaeota;Methanomicrobia;Methanosarcinales;Methanosarcinaceae']
for lineage in lineages:
workerQueue.put(lineage)
for _ in range(numThreads):
workerQueue.put(None)
calcProc = [mp.Process(target = self.__processLineage, args = (metadata, ubiquityThreshold, singleCopyThreshold, minGenomes, workerQueue, writerQueue)) for _ in range(numThreads)]
writeProc = mp.Process(target = self.__storeResults, args = (outputFile, len(lineages), writerQueue))
writeProc.start()
for p in calcProc:
p.start()
for p in calcProc:
p.join()
writerQueue.put((None, None, None, None))
writeProc.join()
def run(self, inputMetadataFile, outputMetadataFile, outputDir, ubiquityThreshold, singleCopyThreshold, trustedCompleteness, trustedContamination):
img = IMG()
markerSetBuilder = MarkerSetBuilder()
allOut = open(os.path.join(outputDir, 'genomes_all.tsv'), 'w')
allOut.write('Genome Id\tLineage\tGenome size (Mbps)\tScaffold count\tGene count\tCoding base count\tN50\tBiotic Relationship\tStatus\tCompleteness\tContamination\tMissing markers\tDuplicate markers\n')
trustedOut = open(os.path.join(outputDir, 'genomes_trusted.tsv'), 'w')
trustedOut.write('Genome Id\tLineage\tGenome size (Mbps)\tScaffold count\tGene count\tCoding base count\tN50\tBiotic Relationship\tStatus\tCompleteness\tContamination\tMissing markers\tDuplicate markers\n')
filteredOut = open(os.path.join(outputDir, 'genomes_filtered.tsv'), 'w')
filteredOut.write('Genome Id\tLineage\tGenome size (Mbps)\tScaffold count\tGene count\tCoding base count\tN50\tBiotic Relationship\tStatus\tCompleteness\tContamination\tMissing markers\tDuplicate markers\n')
metadataOut = open(outputMetadataFile, 'w')
# read input metadata file
metadata = img.genomeMetadataFromFile(inputMetadataFile)
finishedGenomes = defaultdict(set)
allGenomes = defaultdict(set)
metadataLine = {}
bHeader = True
for line in open(inputMetadataFile):
if bHeader:
metadataOut.write(line)
bHeader = False
continue
lineSplit = line.split('\t')
genomeId = lineSplit[0]
domain = lineSplit[1]
status = lineSplit[2]
if status == 'Finished':
finishedGenomes[domain].add(genomeId)
allGenomes[domain].add(genomeId)
metadataLine[genomeId] = line
allTrustedGenomeIds = set()
for lineage, allLineageGenomeIds in allGenomes.iteritems():
print '[' + lineage + ']'
print ' Number of genomes: %d' % len(allLineageGenomeIds)
# tabulate genomes from each phylum
allPhylumCounts = {}
for genomeId in allLineageGenomeIds:
taxon = metadata[genomeId]['taxonomy'][1]
allPhylumCounts[taxon] = allPhylumCounts.get(taxon, 0) + 1
# identify marker genes for finished genomes
print '\nDetermining initial marker gene sets for genome filtering.'
markerSet = markerSetBuilder.buildMarkerSet(finishedGenomes[lineage], ubiquityThreshold, singleCopyThreshold)
print ' Marker set consists of %s marker genes organized into %d sets.' % (markerSet.numMarkers(), markerSet.numSets())
fout = open(os.path.join(outputDir, 'trusted_marker_sets_' + lineage + '.txt'), 'w')
fout.write(str(markerSet.markerSet))
fout.close()
# identifying trusted genomes (highly complete, low contamination genomes)
print '\nIdentifying highly complete, low contamination genomes.'
trustedGenomeIds = set()
filteredGenomes = set()
retainedStatus = {}
filteredStatus = {}
geneCountTable = img.geneCountTable(allLineageGenomeIds)
for genomeId in allLineageGenomeIds:
completeness, contamination, missingMarkers, duplicateMarkers = markerSetBuilder.genomeCheck(markerSet.markerSet, genomeId, geneCountTable)
genomeStr = self.__genomeString(genomeId, metadata, completeness, contamination, missingMarkers, duplicateMarkers)
if completeness >= trustedCompleteness and contamination <= trustedContamination:
trustedGenomeIds.add(genomeId)
allTrustedGenomeIds.add(genomeId)
retainedStatus[metadata[genomeId]['status']] = retainedStatus.get(metadata[genomeId]['status'], 0) + 1
trustedOut.write(genomeStr)
allOut.write(genomeStr)
metadataOut.write(metadataLine[genomeId])
else:
filteredGenomes.add(genomeId)
filteredStatus[metadata[genomeId]['status']] = filteredStatus.get(metadata[genomeId]['status'], 0) + 1
filteredOut.write(genomeStr)
allOut.write(genomeStr)
print ' Filtered genomes: %d (%.2f%%)' % (len(filteredGenomes), len(filteredGenomes)*100.0 / len(allLineageGenomeIds))
print ' ' + str(filteredStatus)
print ' \nTrusted genomes: %d (%.2f%%)' % (len(trustedGenomeIds), len(trustedGenomeIds)*100.0 / len(allLineageGenomeIds))
print ' ' + str(retainedStatus)
# determine status of retained genomes
print '\nTrusted genomes by phylum:'
trustedPhylumCounts = {}
for genomeId in trustedGenomeIds:
taxon = metadata[genomeId]['taxonomy'][1]
trustedPhylumCounts[taxon] = trustedPhylumCounts.get(taxon, 0) + 1
for phylum, count in allPhylumCounts.iteritems():
print ' ' + phylum + ': %d of %d' % (trustedPhylumCounts.get(phylum, 0), count)
print ''
allOut.close()
trustedOut.close()
filteredOut.close()
metadataOut.close()
# write out lineage statistics for genome distribution
allStats = {}
trustedStats = {}
for r in xrange(0, 6): # Domain to Genus
for genomeId, data in metadata.iteritems():
taxaStr = ';'.join(data['taxonomy'][0:r+1])
allStats[taxaStr] = allStats.get(taxaStr, 0) + 1
if genomeId in allTrustedGenomeIds:
trustedStats[taxaStr] = trustedStats.get(taxaStr, 0) + 1
sortedLineages = img.lineagesSorted(metadata)
fout = open(os.path.join(outputDir, 'lineage_stats.tsv'), 'w')
fout.write('Lineage\tGenomes with metadata\tTrusted genomes\n')
for lineage in sortedLineages:
fout.write(lineage + '\t' + str(allStats.get(lineage, 0))+ '\t' + str(trustedStats.get(lineage, 0))+ '\n')
fout.close()
class MarkerSetStabilityTest(object):
def __init__(self):
self.img = IMG()
self.markerset = MarkerSet()
def __processLineage(self, metadata, ubiquityThreshold, singleCopyThreshold, minGenomes, queueIn, queueOut):
"""Assess stability of marker set for a specific named taxonomic group."""
while True:
lineage = queueIn.get(block=True, timeout=None)
if lineage == None:
break
genomeIds = self.img.genomeIdsByTaxonomy(lineage, metadata, 'trusted')
markerGenes = []
perChange = []
numGenomesToSelect = int(0.9*len(genomeIds))
if len(genomeIds) >= minGenomes:
# calculate marker set for all genomes in lineage
geneCountTable = self.img.geneCountTable(genomeIds)
markerGenes = self.markerset.markerGenes(genomeIds, geneCountTable, ubiquityThreshold*len(genomeIds), singleCopyThreshold*len(genomeIds))
tigrToRemove = self.img.identifyRedundantTIGRFAMs(markerGenes)
markerGenes = markerGenes - tigrToRemove
for _ in xrange(0, 100):
# calculate marker set for subset of genomes
subsetGenomeIds = random.sample(genomeIds, numGenomesToSelect)
geneCountTable = self.img.geneCountTable(subsetGenomeIds)
subsetMarkerGenes = self.markerset.markerGenes(subsetGenomeIds, geneCountTable, ubiquityThreshold*numGenomesToSelect, singleCopyThreshold*numGenomesToSelect)
tigrToRemove = self.img.identifyRedundantTIGRFAMs(subsetMarkerGenes)
subsetMarkerGenes = subsetMarkerGenes - tigrToRemove
perChange.append(float(len(markerGenes.symmetric_difference(subsetMarkerGenes)))*100.0 / len(markerGenes))
if perChange != []:
queueOut.put((lineage, len(genomeIds), len(markerGenes), numGenomesToSelect, mean(perChange), std(perChange)))
else:
queueOut.put((lineage, len(genomeIds), len(markerGenes), numGenomesToSelect, -1, -1))
def __storeResults(self, outputFile, totalLineages, writerQueue):
"""Store results to file."""
fout = open(outputFile, 'w')
fout.write('Lineage\t# genomes\t# markers\t# sampled genomes\tmean % change\tstd % change\n')
numProcessedLineages = 0
while True:
lineage, numGenomes, numMarkerGenes, numSampledGenomes, meanPerChange, stdPerChange = writerQueue.get(block=True, timeout=None)
if lineage == None:
break
numProcessedLineages += 1
statusStr = ' Finished processing %d of %d (%.2f%%) lineages.' % (numProcessedLineages, totalLineages, float(numProcessedLineages)*100/totalLineages)
sys.stdout.write('%s\r' % statusStr)
sys.stdout.flush()
fout.write('%s\t%d\t%d\t%d\t%f\t%f\n' % (lineage, numGenomes, numMarkerGenes, numSampledGenomes, meanPerChange, stdPerChange))
sys.stdout.write('\n')
fout.close()
def run(self, outputFile, ubiquityThreshold, singleCopyThreshold, minGenomes, mostSpecificRank, numThreads):
"""Calculate stability of marker sets for named taxonomic groups."""
print ' Testing stability of marker sets:'
random.seed(1)
# process each sequence in parallel
workerQueue = mp.Queue()
writerQueue = mp.Queue()
metadata = self.img.genomeMetadata()
lineages = self.img.lineagesByCriteria(metadata, minGenomes, mostSpecificRank)
for lineage in lineages:
workerQueue.put(lineage)
for _ in range(numThreads):
workerQueue.put(None)
calcProc = [mp.Process(target = self.__processLineage, args = (metadata, ubiquityThreshold, singleCopyThreshold, minGenomes, workerQueue, writerQueue)) for _ in range(numThreads)]
writeProc = mp.Process(target = self.__storeResults, args = (outputFile, len(lineages), writerQueue))
writeProc.start()
for p in calcProc:
p.start()
for p in calcProc:
p.join()
writerQueue.put((None, None, None, None, None, None))
writeProc.join()
class MarkerSetSelection(object):
def __init__(self):
self.simFile = './experiments/simulation.tuning.genus.summary.tsv'
self.looRank = 5
self.markerSetBuilder = MarkerSetBuilder()
self.img = IMG()
def __stabilityTest(self, genomeIds, ubiquityThreshold = 0.97, singleCopyThreshold = 0.97, stabilityThreshold = 0.05):
"""Test stability of marker set for a group of genomes using LOO-testing."""
# quick escape for lineage that are clearly stable
if len(genomeIds) > 200:
return True
# calculate marker sets using a LOO-testing
looMarkerGenes = []
for genomeId in genomeIds:
looGenomeIds = genomeIds.difference([genomeId])
# calculate marker genes
geneCountTable = self.img.geneCountTable(looGenomeIds)
markerGenes = self.markerSetBuilder.markerGenes(looGenomeIds, geneCountTable, ubiquityThreshold*len(looGenomeIds), singleCopyThreshold*len(looGenomeIds))
tigrToRemove = self.img.identifyRedundantTIGRFAMs(markerGenes)
markerGenes = markerGenes - tigrToRemove
looMarkerGenes.append(markerGenes)
# calculate change in marker set for all pairs
markerSetSize = []
diffMarkerSet = []
for i in xrange(0, len(looMarkerGenes)):
markerSetSize.append(len(looMarkerGenes[i]))
for j in xrange(i+1, len(looMarkerGenes)):
symmDiff = looMarkerGenes[i].symmetric_difference(looMarkerGenes[j])
diffMarkerSet.append(len(symmDiff))
print len(genomeIds), mean(diffMarkerSet), mean(markerSetSize)
return (float(mean(diffMarkerSet)) / mean(markerSetSize)) <= stabilityThreshold
def __patristicDist(self, tree, taxa1, taxa2):
mrca = tree.mrca(taxon_labels=[taxa1.taxon.label, taxa2.taxon.label])
if mrca.parent_node == None:
# MRCA is the root of the tree
return taxa1.distance_from_root() + taxa2.distance_from_root()
else:
dist = taxa1.edge_length
parentNode = taxa1.parent_node
while parentNode != mrca:
dist += parentNode.edge_length
parentNode = parentNode.parent_node
dist += taxa2.edge_length
parentNode = taxa2.parent_node
while parentNode != mrca:
dist += parentNode.edge_length
parentNode = parentNode.parent_node
return dist
def __distToNodePercentileTest(self, genomeNode, markerSetNode, leaves, percentileTest):
distToBin = self.__distanceToAncestor(genomeNode, markerSetNode)
distToLeaves = []
for leaf in leaves:
distToLeaves.append(self.__distanceToAncestor(leaf, markerSetNode))
return distToBin < percentile(distToLeaves, percentileTest)
def __selectMarkerSetNode(self, tree, genomeId, metadata, taxonToGenomeIds):
"""Determine lineage-specific marker set to use for assessing the giving genome."""
# read genomes removed from tree as a result of duplicate sequences
duplicateSeqs = self.markerSetBuilder.readDuplicateSeqs()
# determine location of genome in tree
node = tree.find_node_with_taxon_label('IMG_' + genomeId)
# ascend tree to root looking for suitable marker set
curNode = node.parent_node
while curNode != None:
uniqueId = curNode.label.split('|')[0]
genomeIds = set()
for leaf in curNode.leaf_nodes():
genomeIds.add(leaf.taxon.label.replace('IMG_', ''))
duplicateGenomes = duplicateSeqs.get(leaf.taxon.label, [])
for dup in duplicateGenomes:
genomeIds.add(dup.replace('IMG_', ''))
# remove genome (LOO-style analysis)
print 'Full:', len(genomeIds)
genomeIds.difference_update([genomeId])
print 'LOO:', len(genomeIds)
# remove all genomes from the same taxonomic group as the genome of interest
taxon = metadata[genomeId]['taxonomy'][self.looRank]
genomeIds.difference_update(taxonToGenomeIds[taxon])
print 'Rank reduced:', len(genomeIds)
print uniqueId
if len(genomeIds) > 10 and self.__stabilityTest(genomeIds):
uidSelected = uniqueId
break
curNode = curNode.parent_node
if curNode == None:
# reach root so use universal marker set
uidSelected = uniqueId
return uidSelected
def __bestMarkerSet(self, genomeId, simResults):
"""Get stats for best marker set."""
curBest = 1000
bestUID = None
for uid, results in simResults[genomeId].iteritems():
numDescendants, dComp, dCont = results
if (dComp + dCont) < curBest:
numDescendantsBest = numDescendants
dCompBest = dComp
dContBest = dCont
bestUID = uid
curBest = dComp + dCont
return bestUID, numDescendantsBest, dCompBest, dContBest
def __workerThread(self, tree, simResults, metadata, taxonToGenomeIds, queueIn, queueOut):
"""Process each data item in parallel."""
while True:
testGenomeId = queueIn.get(block=True, timeout=None)
if testGenomeId == None:
break
uidSelected = self.__selectMarkerSetNode(tree, testGenomeId, metadata, taxonToGenomeIds)
numDescendantsSelected, dCompSelected, dContSelected = simResults[testGenomeId][uidSelected]
# find best marker set
bestUID, numDescendantsBest, dCompBest, dContBest = self.__bestMarkerSet(testGenomeId, simResults)
queueOut.put((testGenomeId, uidSelected, numDescendantsSelected, dCompSelected, dContSelected, bestUID, numDescendantsBest, dCompBest, dContBest))
def __writerThread(self, numTestGenomes, writerQueue):
"""Store or write results of worker threads in a single thread."""
fout = open('./experiments/markerSetSelection.tsv', 'w')
fout.write('Genome Id\tSelected UID\t# descendants\tSelected dComp\tSelected dCont\tBest UID\t# descendants\tBest dComp\tBest dCont\tdDescendants\tdComp\tdCont\n')
itemsToProcess = 0
dComps = []
dConts = []
dCompsPer = []
dContsPer = []
bestComp = []
bestCont = []
selectedComp = []
selectedCont = []
while True:
testGenomeId, uidSelected, numDescendantsSelected, dCompSelected, dContSelected, bestUID, numDescendantsBest, dCompBest, dContBest = writerQueue.get(block=True, timeout=None)
if testGenomeId == None:
break
itemsToProcess += 1
statusStr = ' Finished processing %d of %d (%.2f%%) test genomes.' % (itemsToProcess, numTestGenomes, float(itemsToProcess)*100/(numTestGenomes))
sys.stdout.write('%s\r' % statusStr)
sys.stdout.flush()
dComp = abs(dCompSelected - dCompBest)
dCont = abs(dContSelected - dContBest)
dDescendants = abs(numDescendantsSelected - numDescendantsBest)
fout.write('%s\t%s\t%d\t%.4f\t%.4f\t%s\t%d\t%.4f\t%.4f\t%d\t%.4f\t%.4f\n' % (testGenomeId, uidSelected, numDescendantsSelected, dCompSelected, dContSelected, bestUID, numDescendantsBest, dCompBest, dContBest, dDescendants, dComp, dCont))
dComps.append(dComp)
dConts.append(dCont)
dCompsPer.append(dComp*100.0 / dCompBest)
dContsPer.append(dCont*100.0 / max(dContBest, 0.01))
bestComp.append(dCompBest)
bestCont.append(dContBest)
selectedComp.append(dCompSelected)
selectedCont.append(dContSelected)
sys.stdout.write('\n')
fout.close()
print ''
print ' General results:'
print ' Best comp: %.2f +/- %.2f' % (mean(bestComp), std(bestComp))
print ' Best cont: %.2f +/- %.2f' % (mean(bestCont), std(bestCont))
print ' Selected comp: %.2f +/- %.2f' % (mean(selectedComp), std(selectedComp))
print ' Selected cont: %.2f +/- %.2f' % (mean(selectedCont), std(selectedCont))
print ''
print ' Delta comp: %.2f +/- %.2f' % (mean(dComps), std(dComps))
print ' Delta cont: %.2f +/- %.2f' % (mean(dConts), std(dConts))
print ' Delta comp per error: %.1f +/- %.1f' % (mean(dCompsPer), std(dCompsPer))
print ' Delta cont per error: %.1f +/- %.1f' % (mean(dContsPer), std(dContsPer))
def __distanceToAncestor(self, leaf, ancestor):
dist = 0
curNode = leaf
while curNode != ancestor:
dist += curNode.edge_length
curNode = curNode.parent_node
return dist
def __bestNodeProperties(self, genomeId, tree, bestUID):
# determine location of genome in tree
node = tree.find_node_with_taxon_label('IMG_' + genomeId)
# find node of best marker set
curNode = node.parent_node
nodesToBin = 0
distanceToBin = node.edge_length
distanceToLeaves = []
while curNode != None:
uniqueId = curNode.label.split('|')[0]
nodesToBin += 1
if uniqueId == bestUID:
for leaf in curNode.leaf_nodes():
if leaf != node:
dist = self.__distanceToAncestor(leaf, curNode)
distanceToLeaves.append(dist)
break
distanceToBin += curNode.edge_length
curNode = curNode.parent_node
return nodesToBin, distanceToBin, mean(distanceToLeaves)
def __propertiesOfBestMarkerSets(self, tree, simResults):
numDescendants = []
nodesToBin = []
distanceToBin = []
avgDistanceToLeaf = []
percDiffs = []
for genomeId in simResults:
bestUID, numDescendantsBest, _, _ = self.__bestMarkerSet(genomeId, simResults)
nodesToBinBest, distanceToBinBest, avgDistanceToLeafBest = self.__bestNodeProperties(genomeId, tree, bestUID)
numDescendants.append(numDescendantsBest)
nodesToBin.append(nodesToBinBest)
distanceToBin.append(distanceToBinBest)
avgDistanceToLeaf.append(avgDistanceToLeafBest)
percDiff = abs(distanceToBinBest - avgDistanceToLeafBest) * 100 / distanceToBinBest
percDiffs.append(percDiff)
print ' # descendants: %.2f +/- %.2f' % (mean(numDescendants), std(numDescendants))
print ' # nodes to bin: %.2f +/- %.2f' % (mean(nodesToBin), std(nodesToBin))
print ' Distance to bin: %.2f +/- %.2f' % (mean(distanceToBin), std(distanceToBin))
distanceToBin = array(distanceToBin)
avgDistanceToLeaf = array(avgDistanceToLeaf)
print ' Distance to bin - average distance to leaf: %.2f +/- %.2f' % (mean(abs(distanceToBin - avgDistanceToLeaf)), std(abs(distanceToBin - avgDistanceToLeaf)))
print ' Percent difference to average leaf distance: %.2f +/- %.2f' % (mean(percDiffs), std(percDiffs))
print ''
def run(self, numThreads):
# read reference tree
print '\n Reading reference genome tree.'
treeFile = os.path.join(os.path.dirname(sys.argv[0]), '..', 'data', 'genome_tree', 'genome_tree_prok.refpkg', 'genome_tree.final.tre')
tree = dendropy.Tree.get_from_path(treeFile, schema='newick', as_rooted=True, preserve_underscores=True)
# get all genomes with a given taxon label
metadata = self.img.genomeMetadata()
taxonToGenomeIds = defaultdict(set)
for genomeId in metadata:
for t in metadata[genomeId]['taxonomy']:
taxonToGenomeIds[t].add(genomeId)
# read simulation results
print ' Reading simulation results.'
simResults = defaultdict(dict)
with open(self.simFile) as f:
f.readline()
for line in f:
lineSplit = line.split('\t')
simId = lineSplit[0] + '-' + lineSplit[1] + '-' + lineSplit[2] + '-' + lineSplit[3]
uid = lineSplit[5].split('|')[0].strip()
numDescendants = int(lineSplit[6])
comp = float(lineSplit[21])
cont = float(lineSplit[23])
simResults[simId][uid] = [numDescendants, comp, cont]
#print ''
#print ' Properties of best marker sets:'
#self.__propertiesOfBestMarkerSets(tree, simResults)
print ' Evaluating %d test genomes.' % len(simResults)
workerQueue = mp.Queue()
writerQueue = mp.Queue()
for testGenomeId in simResults:
workerQueue.put(testGenomeId)
for _ in range(numThreads):
workerQueue.put(None)
workerProc = [mp.Process(target = self.__workerThread, args = (tree, simResults, metadata, taxonToGenomeIds, workerQueue, writerQueue)) for _ in range(numThreads)]
writeProc = mp.Process(target = self.__writerThread, args = (len(simResults), writerQueue))
writeProc.start()
for p in workerProc:
p.start()
for p in workerProc:
p.join()
writerQueue.put((None, None, None, None, None, None, None, None, None))
writeProc.join()
