def filterAlignmentByReferenceIDs(cls, inputFname, outputFname=None, refNameSet = set(['Contig0', 'Contig1', 'Contig2']), \
readGroup="", platform='LS454'):
"""
2011-7-8
"""
import os, sys
import pysam, copy
samfile = BamFile(inputFname, 'rb')
header = copy.deepcopy(samfile.header)
if readGroup: #add read group if it's missing
if "RG" not in header:
header['RG'] = []
header['RG'].append({'ID':readGroup, 'PL':platform, 'LB':platform, 'SM':readGroup})
# remove SQ entries that are not in refNameSet, from the header
newSQList = []
for SQ_entry in header["SQ"]:
if SQ_entry['SN'] in refNameSet:
newSQList.append(SQ_entry)
header["SQ"] = newSQList
refName2bamOutputF = {}
bamOutputF = pysam.Samfile(outputFname, 'wb', header=header) # template=samfile)
sys.stderr.write("Retain reads from %s only from these references %s ...\n"%(inputFname, refNameSet))
processor = cls.FilterAlignmentByReferenceIDs(refNameSet=refNameSet, readGroup=readGroup, bamOutputF=bamOutputF)
samfile.traverseBamByRead(processor=processor)
bamOutputF.close()
def filterAlignmentByReferenceIDs(cls, inputFname, outputDir=None, refNameSet = set(['Contig0', 'Contig1', 'Contig2']), \
readGroup="", platform='LS454'):
"""
2011-7-8
"""
import os, sys
import pysam, copy
samfile = BamFile(inputFname, 'rb')
header = copy.deepcopy(samfile.header)
if readGroup:
if "RG" not in header:
header['RG'] = []
header['RG'].append({'ID':readGroup, 'PL':platform, 'LB':platform, 'SM':readGroup})
refName2bamOutputF = {}
for refName in refNameSet:
inputFileBaseNamePrefix = os.path.splitext(os.path.basename(inputFname))[0]
outputFname = os.path.join(outputDir, '%s_%s.bam'%(inputFileBaseNamePrefix,refName))
bamOutputF = pysam.Samfile(outputFname, 'wb', header=header) # template=samfile)
refName2bamOutputF[refName] = bamOutputF
sys.stderr.write("Retain reads from %s only from these references %s ...\n"%(inputFname, refNameSet))
processor = cls.FilterAlignmentByReferenceIDs(refName2bamOutputF=refName2bamOutputF, refNameSet=refNameSet, readGroup=readGroup)
samfile.traverseBamByRead(processor=processor)
for refName, bamOutputF in refName2bamOutputF.iteritems():
bamOutputF.close()
