def processVariantData(vcf_data, bam_file, strainID = '', refName='', minQuality = 30, minCount = 1,minSize=0):
'''
@return [vcf]
@summary:
Process variant data link combine with related sequencing reference.
Also creates a report of all variant data.
'''
#vReport = Report() #Testing new report object
vReport = DataReport()
vcfDataResult = []
print "Opening bam file [%s]" % (bam_file)
samFile = pysam.Samfile(bam_file,"rb")
refname = samFile.getrname(0)
for vcf in vcf_data:
vcf["STRAIN"] = strainID
irefname = vcf["CHROM"]
start = int(vcf["POS"])
ref = vcf["REF"]
alt = vcf["ALT"]
quality = vcf["QUAL"]
end = start + len(ref)
vcf["END"] = end
alreads = samFile.fetch(irefname,start,end)
alreads = list(alreads)
rCount = len(alreads)
vcf["READS"] = rCount
vcf["READFILE"] = bam_file
if float(quality < minQuality) or rCount < minCount or len(alt) < minSize:
#print "vcf [%s] quality [%s] count [%s] size [%s] has failed to pass" % (vcf,quality,rCount,len(alt))
continue
else:
vcfDataResult.append (vcf)
if irefname == '':
irefname = refname
vID = "%s_%s" % (strainID,start)
vReport.add(vID,"strainID", strainID)
vReport.add(vID,"start", start)
vReport.add(vID,"end", end)
vReport.add(vID,"chrom",irefname)
vReport.add(vID,"count",rCount)
vReport.add(vID,"qual",quality)
vReport.add(vID,"ref",ref)
vReport.add(vID,"alt",alt)
vReport.add(vID,"bamFile",bam_file)
#vReport.add(vID,"samFile",samFile)
return (vcfDataResult, vReport)
def vcfCollectionReport(self,vcfCollection,vcf,reorder=True,useCount=True,fill_blank="NA"):
'''
@summary:
Build report object form variant calls groups by regions into "variance collection".
Report is made to be easily written to delimited matrix / spread sheet format.
'''
result = DataReport()
coverageReport = DataReport()
data = vcfCollection.items()
data.sort()
strainIDs = vcfCollection.keys()
#sort strain IDs
#if reorder:
if True:
print "reordering report"
columnNames = strainIDs
cmap = {}
regex = ".*_([0-9]+).*"
for cName in columnNames:
match = re.match(regex,cName)
if match != None:
key = match.group(1)
else:
key = cName
cmap[key] = cName
strainKeys = cmap.keys()
strainKeys.sort()
#for (strainID,vcfRegions) in vcfCollection.items():
for strainKey in strainKeys:
strainID = cmap[strainKey]
vcfRegions = vcfCollection[strainID]
print "Collecting [%s] regions" % (strainID)
for (loc,vcfRegion) in vcfRegions.items():
count = vcfRegion["Count"]
vcfData = vcfRegion['vcfData']
item = "[%s]:" % (count)
coverageReport.add(loc,strainID,count)
if useCount:
if result.get(loc,"Region_Count") == None:
#result.add(loc,"Region_Count",0)
rCount = 0
else:
rCount = result.get(loc,"Region_Count")
rCount = float(rCount)
if len(vcfData) != 0:
rCount += 1
#result.add(loc,"Region_Count",rCount)
for vcf in vcfData:
item = item + "(%s,%s,%s):" % (vcf["POS"],vcf["READS"],vcf["ALT"])
if len(vcfData) != 0:
result.add(loc,strainID,item)
else:
result.add(loc,strainID,fill_blank)
return (result,coverageReport)