def parseForINDEL( indel_file ) :
print indel_file
fin = open( indel_file )
(path,ext) = indel_file.split('.',1)
fname = path.split('/')[-1]
fout = open( "%s/intermediate_data/sift/input/%s_sift_input.csv" \
% (globes.DATA_DIR, fname), 'wb' )
print fout
#fast-forward through header lines
patients = broad.getPatients( fin )
indexOf = broad.COLUMN_MAP
for dataline in fin :
splt = dataline.strip().split('\t')
col_keys = ['chrom','pos','mut','ref']
chrom,pos,mut,ref = [ splt[ indexOf[k] ] for k in col_keys ]
dinfo = broad.makeInfoDict( splt[ indexOf["info"] ] )
try :
strand = dinfo["refseq.transcriptStrand"]
except KeyError :
try :
strand = dinfo["refseq.transcriptStrand_1"]
except KeyError : #dont have it guess '+'
strand = '+'
##raise Exception("what the f**k: %s" % splt[ indexOf["info"] ] )
if strand == "+" : strand = 1
elif strand == "-" : strand = -1
else : raise Exception("Strand is not + or - ??")
isInsertion = len(ref) == 1 and len(mut) > 1
isDeletion = len(ref) > 1 and len(mut) == 1
if isInsertion :
start = int(pos)
end = start
allele = mut
elif isDeletion :
start = int(pos)
end = start + (len(ref)-len(mut))
allele = '-/'
else : assert False
fout.write( "%s,%d,%d,%d,%s\n" % (chrom,start,end,strand,allele) )
fout.close()
fin.close()
def integrator( self, target, splts ) :
if len(splts) != 1 : assert "len isn't right"
for splt in splts :
##TODO generalize this to make it vendor independent, call start column is feature of VCF, not broad???
calls = splt[ broad.CALL_START: ]
base_calls = []
for pat_ix,c in enumerate(calls) :
sc = broad.splitCall(c)
gt = broad.convertGT( sc )
if broad.isMutated( gt ) or broad.noInf( gt ) :
base_calls.append( BaseCall(sc,pat_ix) )
fields = {}
keys = broad.COLUMN_MAP.keys()
for k in keys :
if k == "chrom" :
fields[k] = globes.chromNum( splt[self.indexOf[k]] )
elif k == "info" :
##TODO generalize this to make it vendor independent
dinfo = broad.makeInfoDict( splt[ self.indexOf[k] ] )
fields["AF"] = dinfo["AF"]
elif k == "dbSNP" :
value = splt[self.indexOf[k]]
#when we getFields, 'dbsnp' will be missing and yield null
if value == '.' : pass
#right now just take the first rs number if multiple
elif value.startswith('rs') :
fields[k] = [int(t.strip()[2:]) for t in value.split(';')][0]
else :
print "malformed rs number?", splt
assert False
else :
fields[k] = splt[ self.indexOf[k] ]
#according to: http://www.broadinstitute.org/gsa/wiki/index.php/Understanding_the_Unified_Genotyper's_VCF_files
#ref and alt are always given for the forward strand
fields['strand'] = True
return Variant( fields, base_calls )
