def LoadDataSet(vcfInfile, traningSet, qFaLen):
if len(traningSet) == 0: raise ValueError('[ERROR] No Training Data found')
if vcfInfile[-3:] == '.gz':
I = os.popen('gzip -dc %s' % vcfInfile)
else:
I = open(vcfInfile)
data, hInfo = [], VCF.VCFHeader()
while 1: # VCF format
lines = I.readlines(100000)
if not lines: break
for line in lines:
col = line.strip('\n').split()
if re.search(r'^#CHROM', line): col2sam = {i+9:sam for i,sam in enumerate(col[9:])}
# Record the header information
if re.search(r'^#', line):
hInfo.Record(line.strip('\n'))
continue
# Get inbreeding coefficient. If fail then continue.
# It's calculated like: 1.0 - hetCount/Expected_hetCount in VCF
#inbCoeff = re.search(r';F=([^;]+)', col[7])
inbCoeff = re.search(r';?InbCoeff=([^;]+)', col[7])
if not inbCoeff:
continue
#print >> sys.stderr, '[ERROR] No inbreeding coefficient "InbCoeff=..." in INFO field in vcf:\n%s\n' % vcfInfile
inbCoeff = float('%.2f' % float(inbCoeff.group(1)))
fmat = {k:i for i,k in enumerate(col[8].split(':'))} # Get Format
if 'QR' not in fmat: continue # Cause by INTERGAP. But We'd better delete this statment, because the error is cause by the USER
for tag in ['AA', 'QR', 'NR']:
if tag not in fmat: raise ValueError('[ERROR] The "Format" fields did not contian "%s" in VCF: %s\nAT: %s\n' %(tag, vcfInfile, line))
isBiallelic = True
if len(col[4].split(',')) > 1: isBiallelic = False
annotations = []
atleastOne = False
for i, sample in enumerate(col[9:]):
sampleId = col2sam[9+i]
if sample == './.': continue
field = sample.split(':')
if len(field[fmat['AA']].split(',')) != 4: continue
if len(field) < fmat['QR'] + 1: continue
qr = field[fmat['QR']].split(',')[-1]
if qr == '.': continue
atleastOne = True
qregion = np.array(qr.split('-'))
if len(qregion) > 3: qId = qregion[0] + '-' + qregion[1]
else : qId = qregion[0]
qSta = string.atoi(qregion[-2])
qEnd = string.atoi(qregion[-1])
if sampleId not in qFaLen : raise ValueError('[ERROR] The sample name $s(in vcf) is not in the name of Fa list.' % sampleId)
if qId not in qFaLen[sampleId]: raise ValueError('[ERROR]', qId, 'is not been found in fa file\n')
qSta = int(qSta * 100 / qFaLen[sampleId][qId] + 0.5)
qEnd = int(qEnd * 100 / qFaLen[sampleId][qId] + 0.5)
if qSta > 100: qSta = 100 # Bug!!! Should delete
if qEnd > 100: qEnd = 100 # Bug!!! Should delete
if qSta > 100 or qEnd > 100:
raise ValueError('[ERROR] Query size Overflow! sample: %s; scaffold: %s\n%s\n%s' % (sampleId, qId, sample, line))
leg = min(qSta, 100 - qEnd)
nn = string.atof(sample.split(':')[fmat['NR']])
n = int(1000 * nn + 0.5) / 10.0 # n ratio range: [0, 100]
alt = string.atoi(sample.split(':')[fmat['AA']].split(',')[1]) # Alternate perfect
bot = string.atoi(sample.split(':')[fmat['AA']].split(',')[3]) # Both imperfect
annotations.append([isBiallelic, inbCoeff, leg, n , alt, bot])
if not atleastOne: raise ValueError('[ERROR] All the samples don\'t contain this variant.', col)
datum = vd.VariantDatum()
datum.annotations = np.median(annotations, axis = 0)
pos = col[0] + ':' + col[1]
datum.variantOrder = pos
if pos in traningSet: datum.atTrainingSite = True
data.append(datum)
I.close()
return hInfo, np.array(data)
