def removeRedundantWithFile(folderName, mummerLink, inputFilename,
mummerTmpName, outputFileName):
thres = 10
os.system("sed -e 's/|//g' " + folderName + inputFilename + ".fasta > " +
folderName + inputFilename + "2.fasta")
os.system("cp " + folderName + inputFilename + "2.fasta " + folderName +
inputFilename + ".fasta")
if True:
alignerRobot.useMummerAlignBatch(mummerLink, folderName, [[
mummerTmpName, inputFilename + ".fasta", inputFilename + ".fasta",
""
]], houseKeeper.globalParallel)
# alignerRobot.useMummerAlign(mummerLink, folderName, "self", "contigs.fasta", "contigs.fasta")
# outputName, referenceName, queryName, specialName
dataList = alignerRobot.extractMumData(folderName, mummerTmpName + "Out")
dataList = alignerRobot.transformCoor(dataList)
lenDic = IORobot.obtainLength(folderName, inputFilename + '.fasta')
removeList = []
shortEmbedClusterDic = {}
for eachitem in lenDic:
shortEmbedClusterDic[eachitem] = clusterElem(eachitem)
for eachitem in dataList:
match1, match2, name1, name2 = eachitem[4], eachitem[5], eachitem[
7], eachitem[8]
if name1 != name2:
l1, l2 = lenDic[name1], lenDic[name2]
if abs(l1 - match1) < thres and abs(l2 - match2) > thres:
removeList.append(name1)
elif abs(l1 - match1) > thres and abs(l2 - match2) < thres:
removeList.append(name2)
elif abs(l1 - match1) < thres and abs(l2 - match2) < thres:
print "Both shortembedd", eachitem
union(shortEmbedClusterDic[name1], shortEmbedClusterDic[name2])
nameList = obtainComplement(lenDic, removeList)
returnList = []
for eachitem in nameList:
if find(shortEmbedClusterDic[eachitem]).id == eachitem:
returnList.append(eachitem)
print "len(nameList), len(returnList)", len(nameList), len(returnList)
IORobot.putListToFileO(folderName, inputFilename + ".fasta",
outputFileName, returnList)
def removeRedundantWithFile(folderName, mummerLink, inputFilename, mummerTmpName, outputFileName):
thres = 10
os.system("sed -e 's/|//g' " + folderName + inputFilename + ".fasta > " + folderName + inputFilename + "2.fasta")
os.system("cp " + folderName + inputFilename + "2.fasta " + folderName + inputFilename + ".fasta")
if True:
alignerRobot.useMummerAlignBatch(
mummerLink,
folderName,
[[mummerTmpName, inputFilename + ".fasta", inputFilename + ".fasta", ""]],
houseKeeper.globalParallel,
)
# alignerRobot.useMummerAlign(mummerLink, folderName, "self", "contigs.fasta", "contigs.fasta")
# outputName, referenceName, queryName, specialName
dataList = alignerRobot.extractMumData(folderName, mummerTmpName + "Out")
dataList = alignerRobot.transformCoor(dataList)
lenDic = IORobot.obtainLength(folderName, inputFilename + ".fasta")
removeList = []
shortEmbedClusterDic = {}
for eachitem in lenDic:
shortEmbedClusterDic[eachitem] = clusterElem(eachitem)
for eachitem in dataList:
match1, match2, name1, name2 = eachitem[4], eachitem[5], eachitem[7], eachitem[8]
if name1 != name2:
l1, l2 = lenDic[name1], lenDic[name2]
if abs(l1 - match1) < thres and abs(l2 - match2) > thres:
removeList.append(name1)
elif abs(l1 - match1) > thres and abs(l2 - match2) < thres:
removeList.append(name2)
elif abs(l1 - match1) < thres and abs(l2 - match2) < thres:
print "Both shortembedd", eachitem
union(shortEmbedClusterDic[name1], shortEmbedClusterDic[name2])
nameList = obtainComplement(lenDic, removeList)
returnList = []
for eachitem in nameList:
if find(shortEmbedClusterDic[eachitem]).id == eachitem:
returnList.append(eachitem)
print "len(nameList), len(returnList)", len(nameList), len(returnList)
IORobot.putListToFileO(folderName, inputFilename + ".fasta", outputFileName, returnList)
def removeEmbedded(folderName, mummerLink):
print "removeEmbedded"
thres = 10
os.system("sed -e 's/|//g' " + folderName + "contigs.fasta > " +
folderName + "contigs2.fasta")
os.system("cp " + folderName + "contigs2.fasta " + folderName +
"contigs.fasta")
if True:
print "removeEmbedded: Aligning contigs.fasta to contigs.fasta, outputs are self*.delta"
alignerRobot.useMummerAlignBatch(
mummerLink, folderName,
[["self", "contigs.fasta", "contigs.fasta", ""]],
houseKeeper.globalParallel)
# alignerRobot.useMummerAlign(mummerLink, folderName, "self", "contigs.fasta", "contigs.fasta")
# outputName, referenceName, queryName, specialName
print "removeEmbedded: Extracting MUMmer data from delta files to selfOut"
dataList = alignerRobot.extractMumData(folderName, "selfOut")
dataList = alignerRobot.transformCoor(dataList)
lenDic = IORobot.obtainLength(folderName, 'contigs.fasta')
removeList = []
for eachitem in dataList:
match1, match2, name1, name2 = eachitem[4], eachitem[5], eachitem[
7], eachitem[8]
if name1 != name2:
l1, l2 = lenDic[name1], lenDic[name2]
if abs(l1 - match1) < thres and abs(l2 - match2) > thres:
removeList.append(name1)
elif abs(l1 - match1) > thres and abs(l2 - match2) < thres:
removeList.append(name2)
elif abs(l1 - match1) < thres and abs(l2 - match2) < thres:
print "Both shortembedd", eachitem
nameList = []
for eachitem in lenDic:
nameList.append(eachitem)
print len(nameList)
for eachitem in removeList:
if eachitem in nameList:
nameList.remove(eachitem)
print len(nameList)
print "removeEmbedded: Outputting non-contained contigs to noEmbed.fasta"
IORobot.putListToFileO(folderName, "contigs.fasta", "noEmbed", nameList)
def removeEmbedded(folderName, mummerLink):
print "removeEmbedded"
thres = 10
command = r'''perl -pe 's/>[^\$]*$/">Seg" . ++$n ."\n"/ge' ''' + folderName + "raw_reads.fasta > " + folderName + houseKeeper.globalReadName
os.system(command)
command = r'''perl -pe 's/>[^\$]*$/">Seg" . ++$n ."\n"/ge' ''' + folderName + "contigs.fasta > " + folderName + houseKeeper.globalContigName
os.system(command)
if True:
alignerRobot.useMummerAlignBatch(mummerLink, folderName, [[
"self", houseKeeper.globalContigName, houseKeeper.globalContigName,
""
]], houseKeeper.globalParallel)
# alignerRobot.useMummerAlign(mummerLink, folderName, "self", "contigs.fasta", "contigs.fasta")
# outputName, referenceName, queryName, specialName
dataList = alignerRobot.extractMumData(folderName, "selfOut")
dataList = alignerRobot.transformCoor(dataList)
lenDic = IORobot.obtainLength(folderName, houseKeeper.globalContigName)
removeList = []
for eachitem in dataList:
match1, match2, name1, name2 = eachitem[4], eachitem[5], eachitem[
7], eachitem[8]
if name1 != name2:
l1, l2 = lenDic[name1], lenDic[name2]
if abs(l1 - match1) < thres and abs(l2 - match2) > thres:
removeList.append(name1)
elif abs(l1 - match1) > thres and abs(l2 - match2) < thres:
removeList.append(name2)
elif abs(l1 - match1) < thres and abs(l2 - match2) < thres:
print "Both shortembedd", eachitem
nameList = []
for eachitem in lenDic:
nameList.append(eachitem)
print len(nameList)
for eachitem in removeList:
if eachitem in nameList:
nameList.remove(eachitem)
print len(nameList)
IORobot.putListToFileO(folderName, houseKeeper.globalContigName, "noEmbed",
nameList)
def removeEmbedded(folderName , mummerLink):
print "removeEmbedded"
thres = 10
command= r'''perl -pe 's/>[^\$]*$/">Seg" . ++$n ."\n"/ge' ''' + folderName + "raw_reads.fasta > " + folderName + houseKeeper.globalReadName
os.system(command)
command= r'''perl -pe 's/>[^\$]*$/">Seg" . ++$n ."\n"/ge' ''' + folderName + "contigs.fasta > " + folderName + houseKeeper.globalContigName
os.system(command)
if True:
alignerRobot.useMummerAlignBatch(mummerLink, folderName, [["self", houseKeeper.globalContigName, houseKeeper.globalContigName, ""]], houseKeeper.globalParallel)
# alignerRobot.useMummerAlign(mummerLink, folderName, "self", "contigs.fasta", "contigs.fasta")
# outputName, referenceName, queryName, specialName
dataList = alignerRobot.extractMumData(folderName, "selfOut")
dataList = alignerRobot.transformCoor(dataList)
lenDic = IORobot.obtainLength(folderName, houseKeeper.globalContigName)
removeList = []
for eachitem in dataList:
match1, match2, name1, name2 = eachitem[4], eachitem[5], eachitem[7], eachitem[8]
if name1 != name2:
l1, l2 = lenDic[name1], lenDic[name2]
if abs(l1 - match1) < thres and abs(l2 - match2) > thres:
removeList.append(name1)
elif abs(l1 - match1) > thres and abs(l2 - match2) < thres:
removeList.append(name2)
elif abs(l1 - match1) < thres and abs(l2 - match2) < thres:
print "Both shortembedd", eachitem
nameList = []
for eachitem in lenDic:
nameList.append(eachitem)
print len(nameList)
for eachitem in removeList:
if eachitem in nameList:
nameList.remove(eachitem)
print len(nameList)
IORobot.putListToFileO(folderName, houseKeeper.globalContigName, "noEmbed", nameList)
def removeEmbedded(folderName, mummerLink):
print "removeEmbedded"
thres = 10
os.system("sed -e 's/|//g' " + folderName + "contigs.fasta > " + folderName + "contigs2.fasta")
os.system("cp " + folderName + "contigs2.fasta " + folderName + "contigs.fasta")
if True:
print "removeEmbedded: Aligning contigs.fasta to contigs.fasta, outputs are self*.delta"
alignerRobot.useMummerAlignBatch(
mummerLink, folderName, [["self", "contigs.fasta", "contigs.fasta", ""]], houseKeeper.globalParallel
)
# alignerRobot.useMummerAlign(mummerLink, folderName, "self", "contigs.fasta", "contigs.fasta")
# outputName, referenceName, queryName, specialName
print "removeEmbedded: Extracting MUMmer data from delta files to selfOut"
dataList = alignerRobot.extractMumData(folderName, "selfOut")
dataList = alignerRobot.transformCoor(dataList)
lenDic = IORobot.obtainLength(folderName, "contigs.fasta")
removeList = []
for eachitem in dataList:
match1, match2, name1, name2 = eachitem[4], eachitem[5], eachitem[7], eachitem[8]
if name1 != name2:
l1, l2 = lenDic[name1], lenDic[name2]
if abs(l1 - match1) < thres and abs(l2 - match2) > thres:
removeList.append(name1)
elif abs(l1 - match1) > thres and abs(l2 - match2) < thres:
removeList.append(name2)
elif abs(l1 - match1) < thres and abs(l2 - match2) < thres:
print "Both shortembedd", eachitem
nameList = []
for eachitem in lenDic:
nameList.append(eachitem)
print len(nameList)
for eachitem in removeList:
if eachitem in nameList:
nameList.remove(eachitem)
print len(nameList)
print "removeEmbedded: Outputting non-contained contigs to noEmbed.fasta"
IORobot.putListToFileO(folderName, "contigs.fasta", "noEmbed", nameList)
def observeOverlap(folderName):
dataList = alignerRobot.extractMumData(folderName, "selfOut")
dataList = alignerRobot.transformCoor(dataList)
lenDic = IORobot.obtainLength(folderName, 'contigs.fasta')
matchThres = 10000
nonMatchThres = 500
count = 0
newDataList = []
for eachitem in dataList:
name1, name2 = eachitem[-2], eachitem[-1]
matchLen1, matchLen2 = eachitem[4], eachitem[5]
start1, end1, start2, end2 = eachitem[0], eachitem[1], eachitem[
2], eachitem[3]
# if name1!= name2 and min(lenDic[name1] - end1, lenDic[name2] - end2 ) > nonMatchThres \
# and min(start1, start2) > nonMatchThres \
if name1!= name2 and ( min(lenDic[name1] - end1, lenDic[name2] - end2 ) > nonMatchThres \
or min(start1, start2) > nonMatchThres ) \
and matchLen1> matchThres:
print "eachitem ", eachitem, lenDic[name1], lenDic[name2]
count = count + 1
newDataList.append(eachitem)
print "Count: " + str(count)
blkDic = getBreakPointFromDataList(folderName, newDataList)
LCList = IORobot.loadContigsFromFile(folderName, "contigs.fasta")
contigList = []
for eachcontig in LCList:
#print eachcontig
if not eachcontig in blkDic:
contigList = contigList + [LCList[eachcontig]]
else:
contigList = contigList + tmpBreakAcBkPts(LCList[eachcontig],
blkDic[eachcontig])
print "len(contigList)", len(contigList)
IORobot.writeSegOut(contigList, folderName, "breakChains.fasta")
