2 # Code to read, write and edit VCF files
4 # VCF lines are encoded as a dictionary with these keys (note: all lowercase):
9 # 'alt': list of strings
11 # 'filter': None (missing value), or list of keys (strings); empty list parsed as ["PASS"]
12 # 'info': dictionary of values (see below)
13 # 'format': list of keys (strings)
14 # sample keys: dictionary of values (see below)
16 # The sample keys are accessible through vcf.getsamples()
18 # A dictionary of values contains value keys (defined in ##INFO or ##FORMAT lines) which map
19 # to a list, containign integers, floats, strings, or characters. Missing values are replaced
20 # by a particular value, often -1 or .
22 # Genotypes are not stored as a string, but as a list of 1 or 3 elements (for haploid and diploid samples),
23 # the first (and last) the integer representing an allele, and the second the separation character.
24 # Note that there is just one genotype per sample, but for consistency the single element is stored in a list.
26 # Header lines other than ##INFO, ##FORMAT and ##FILTER are stored as (key, value) pairs and are accessible
29 # The VCF class can be instantiated with a 'regions' variable consisting of tuples (chrom,start,end) encoding
30 # 0-based half-open segments. Only variants with a position inside the segment will be parsed. A regions
31 # parser is available under parse_regions.
33 # When instantiated, a reference can be passed to the VCF class. This may be any class that supports a
34 # fetch(chrom, start, end) method.
38 # NOTE: the position that is returned to Python is 0-based, NOT 1-based as in the VCF file.
43 # only v4.0 writing is complete; alleles are not converted to v3.3 format
46 from collections import namedtuple, defaultdict
47 from operator import itemgetter
48 import sys, re, copy, bisect
52 gtsRegEx = re.compile("[|/\\\\]")
53 alleleRegEx = re.compile('^[ACGTN]+$')
55 # Utility function. Uses 0-based coordinates
56 def get_sequence(chrom, start, end, fa):
57 # obtain sequence from .fa file, without truncation
58 if end<=start: return ""
59 if not fa: return "N"*(end-start)
60 if start<0: return "N"*(-start) + get_sequence(chrom, 0, end, fa).upper()
61 sequence = fa.fetch(chrom, start, end).upper()
62 if len(sequence) < end-start: sequence += "N"*(end-start-len(sequence))
65 # Utility function. Parses a region string
66 def parse_regions( string ):
68 for r in string.split(','):
70 chrom, start, end = elts[0], 0, 3000000000
74 ielts = elts[1].split('-')
75 if len(ielts) != 2: ValueError("Don't understand region string '%s'" % r)
76 try: start, end = int(ielts[0])-1, int(ielts[1])
77 except: raise ValueError("Don't understand region string '%s'" % r)
79 raise ValueError("Don't understand region string '%s'" % r)
80 result.append( (chrom,start,end) )
84 FORMAT = namedtuple('FORMAT','id numbertype number type description missingvalue')
86 ###########################################################################################################
90 ###########################################################################################################
95 initialized from data and vcf meta
101 def __init__(self, data, vcf):
102 self.data, self.vcf = data, vcf
104 if len(data) != len(self.vcf._samples):
105 self.error(str(data),
106 self.BAD_NUMBER_OF_COLUMNS,
107 "expected %s for %s samples (%s), got %s" % \
108 (len(self.vcf._samples),
109 len(self.vcf._samples),
114 def contig( self ): return self.data[0]
121 def __get__( self ): return self.data[2]
125 # note: gerton substitutes reference if it can be fixed.
126 return self.data[3].upper()
130 # convert v3.3 to v4.0 alleles below
132 if alt == ".": alt = []
133 else: alt = alt.upper().split(',')
139 if qual == ".": qual = -1
141 try: qual = float(qual)
142 except: self.error(line,self.QUAL_NOT_NUMERICAL)
146 # postpone checking that filters exist. Encode missing filter or no filtering as empty list
147 if cols[6] == "." or cols[6] == "PASS" or cols[6] == "0": filter = []
148 else: filter = cols[6].split(';')
155 # dictionary of keys, and list of values
158 for blurp in col.split(';'):
159 elts = blurp.split('=')
160 if len(elts) == 1: v = None
161 elif len(elts) == 2: v = elts[1]
162 else: self.error(str(self.data),self.ERROR_INFO_STRING)
163 info[elts[0]] = self.parse_formatdata(elts[0], v, self.vcf._info, line)
168 return self.data[8].split(':')
170 def __getitem__(self, key):
172 # parse sample columns
173 values = self.data[self.vcf._sample2column[key]].split(':')
177 if len(values) > len(format):
178 self.error(line,self.BAD_NUMBER_OF_VALUES,"(found %s values in element %s; expected %s)" % (len(values),sample,len(format)))
181 for idx in range(len(format)):
182 expected = self.vcf.get_expected(format[idx], self.vcf._format, alt)
183 if idx < len(values): value = values[idx]
185 if expected == -1: value = "."
186 else: value = ",".join(["."]*expected)
188 result[format[idx]] = self.vcf.parse_formatdata(format[idx], value, self.vcf._format, line)
189 if expected != -1 and len(result[format[idx]]) != expected:
190 self.error(str(self.data),self.BAD_NUMBER_OF_PARAMETERS,
191 "id=%s, expected %s parameters, got %s" % (format[idx],expected,result[format[idx]]))
192 if len(result[format[idx]] ) < expected: result[format[idx]] += [result[format[idx]][-1]]*(expected-len(result[format[idx]]))
193 result[format[idx]] = result[format[idx]][:expected]
198 return str(self.data)
208 NT_PHASED_GENOTYPES = 5
210 _errors = { 0:"UNKNOWN_FORMAT_STRING:Unknown file format identifier",
211 1:"BADLY_FORMATTED_FORMAT_STRING:Formatting error in the format string",
212 2:"BADLY_FORMATTED_HEADING:Did not find 9 required headings (CHROM, POS, ..., FORMAT) %s",
213 3:"BAD_NUMBER_OF_COLUMNS:Wrong number of columns found (%s)",
214 4:"POS_NOT_NUMERICAL:Position column is not numerical",
215 5:"UNKNOWN_CHAR_IN_REF:Unknown character in reference field",
216 6:"V33_BAD_REF:Reference should be single-character in v3.3 VCF",
217 7:"V33_BAD_ALLELE:Cannot interpret allele for v3.3 VCF",
218 8:"POS_NOT_POSITIVE:Position field must be >0",
219 9:"QUAL_NOT_NUMERICAL:Quality field must be numerical, or '.'",
220 10:"ERROR_INFO_STRING:Error while parsing info field",
221 11:"ERROR_UNKNOWN_KEY:Unknown key (%s) found in formatted field (info; format; or filter)",
222 12:"ERROR_FORMAT_NOT_NUMERICAL:Expected integer or float in formatted field; got %s",
223 13:"ERROR_FORMAT_NOT_CHAR:Eexpected character in formatted field; got string",
224 14:"FILTER_NOT_DEFINED:Identifier (%s) in filter found which was not defined in header",
225 15:"FORMAT_NOT_DEFINED:Identifier (%s) in format found which was not defined in header",
226 16:"BAD_NUMBER_OF_VALUES:Found too many of values in sample column (%s)",
227 17:"BAD_NUMBER_OF_PARAMETERS:Found unexpected number of parameters (%s)",
228 18:"BAD_GENOTYPE:Cannot parse genotype (%s)",
229 19:"V40_BAD_ALLELE:Bad allele found for v4.0 VCF (%s)",
230 20:"MISSING_REF:Reference allele missing",
231 21:"V33_UNMATCHED_DELETION:Deleted sequence does not match reference (%s)",
232 22:"V40_MISSING_ANGLE_BRACKETS:Format definition is not deliminted by angular brackets",
233 23:"FORMAT_MISSING_QUOTES:Description field in format definition is not surrounded by quotes",
234 24:"V40_FORMAT_MUST_HAVE_NAMED_FIELDS:Fields in v4.0 VCF format definition must have named fields",
235 25:"HEADING_NOT_SEPARATED_BY_TABS:Heading line appears separated by spaces, not tabs",
236 26:"WRONG_REF:Wrong reference %s",
237 27:"ERROR_TRAILING_DATA:Numerical field ('%s') has semicolon-separated trailing data",
238 28:"BAD_CHR_TAG:Error calculating chr tag for %s",
239 29:"ZERO_LENGTH_ALLELE:Found zero-length allele",
240 30:"MISSING_INDEL_ALLELE_REF_BASE:Indel alleles must begin with single reference base"
243 # tag-value pairs; tags are not unique; does not include fileformat, INFO, FILTER or FORMAT fields
246 # version number; 33=v3.3; 40=v4.0
249 # info, filter and format data
254 # header; and required columns
255 _required = ["CHROM","POS","ID","REF","ALT","QUAL","FILTER","INFO","FORMAT"]
259 _ignored_errors = set([11]) # ERROR_UNKNOWN_KEY
260 _warn_errors = set([])
266 # regions to include; None includes everything
274 def __init__(self, _copy=None, reference=None, regions=None, lines=None, leftalign=False):
275 # make error identifiers accessible by name
276 for id in self._errors.keys(): self.__dict__[self._errors[id].split(':')[0]] = id
278 self._leftalign = _copy._leftalign
279 self._header = _copy._header[:]
280 self._version = _copy._version
281 self._info = copy.deepcopy(_copy._info)
282 self._filter = copy.deepcopy(_copy._filter)
283 self._format = copy.deepcopy(_copy._format)
284 self._samples = _copy._samples[:]
285 self._sample2column = copy.deepcopy(_copy._sample2column)
286 self._ignored_errors = copy.deepcopy(_copy._ignored_errors)
287 self._warn_errors = copy.deepcopy(_copy._warn_errors)
288 self._reference = _copy._reference
289 self._regions = _copy._regions
290 if reference: self._reference = reference
291 if regions: self._regions = regions
292 if leftalign: self._leftalign = leftalign
295 def error(self,line,error,opt=None):
296 if error in self._ignored_errors: return
297 errorlabel, errorstring = self._errors[error].split(':')
298 if opt: errorstring = errorstring % opt
299 errwarn = ["Error","Warning"][error in self._warn_errors]
300 sys.stderr.write("Line %s: '%s'\n%s %s: %s\n" % (self._lineno,line,errwarn,errorlabel,errorstring))
301 if error in self._warn_errors: return
302 raise ValueError(errorstring)
304 def parse_format(self,line,format,filter=False):
305 if self._version >= 40:
306 if not format.startswith('<'):
307 self.error(line,self.V40_MISSING_ANGLE_BRACKETS)
309 if not format.endswith('>'):
310 self.error(line,self.V40_MISSING_ANGLE_BRACKETS)
312 format = format[1:-1]
313 data = {'id':None,'number':None,'type':None,'descr':None}
315 while len(format.strip())>0:
316 elts = format.strip().split(',')
317 first, rest = elts[0], ','.join(elts[1:])
318 if first.find('=') == -1 or (first.find('"')>=0 and first.find('=') > first.find('"')):
319 if self._version >= 40: self.error(line,self.V40_FORMAT_MUST_HAVE_NAMED_FIELDS)
320 if idx == 4: self.error(line,self.BADLY_FORMATTED_FORMAT_STRING)
321 first = ["ID=","Number=","Type=","Description="][idx] + first
322 if first.startswith('ID='): data['id'] = first.split('=')[1]
323 elif first.startswith('Number='): data['number'] = first.split('=')[1]
324 elif first.startswith('Type='): data['type'] = first.split('=')[1]
325 elif first.startswith('Description='):
326 elts = format.split('"')
328 self.error(line,self.FORMAT_MISSING_QUOTES)
329 elts = first.split('=') + [rest]
330 data['descr'] = elts[1]
331 rest = '"'.join(elts[2:])
332 if rest.startswith(','): rest = rest[1:]
334 self.error(line,self.BADLY_FORMATTED_FORMAT_STRING)
337 if filter and idx==1: idx=3 # skip number and type fields for FILTER format strings
338 if not data['id']: self.error(line,self.BADLY_FORMATTED_FORMAT_STRING)
339 if not data['descr']:
340 self.error(line,self.BADLY_FORMATTED_FORMAT_STRING)
341 data['descr'] = '<none>'
342 if not data['type'] and not data['number']:
343 # fine, ##filter format
344 return FORMAT(data['id'],self.NT_NUMBER,0,"Flag",data['descr'],'.')
345 if not data['type'] in ["Integer","Float","Character","String","Flag"]:
346 self.error(line,self.BADLY_FORMATTED_FORMAT_STRING)
347 # I would like a missing-value field, but it isn't there
348 if data['type'] in ['Integer','Float']: data['missing'] = None # Do NOT use arbitrary int/float as missing value
349 else: data['missing'] = '.'
350 if not data['number']: self.error(line,self.BADLY_FORMATTED_FORMAT_STRING)
352 n = int(data['number'])
356 if data['number'] == '.': t = self.NT_UNKNOWN
357 elif data['number'] == '#alleles': t = self.NT_ALLELES
358 elif data['number'] == '#nonref_alleles': t = self.NT_NR_ALLELES
359 elif data['number'] == '#genotypes': t = self.NT_GENOTYPES
360 elif data['number'] == '#phased_genotypes': t = self.NT_PHASED_GENOTYPES
362 self.error(line,self.BADLY_FORMATTED_FORMAT_STRING)
363 return FORMAT(data['id'],t,n,data['type'],data['descr'],data['missing'])
366 def format_format( self, fmt, filter=False ):
367 values = [('ID',fmt.id)]
368 if fmt.number != None and not filter:
369 if fmt.numbertype == self.NT_UNKNOWN: nmb = "."
370 elif fmt.numbertype == self.NT_NUMBER: nmb = str(fmt.number)
371 elif fmt.numbertype == self.NT_ALLELES: nmb = "#alleles"
372 elif fmt.numbertype == self.NT_NR_ALLELES: nmb = "#nonref_alleles"
373 elif fmt.numbertype == self.NT_GENOTYPES: nmb = "#genotypes"
374 elif fmt.numbertype == self.NT_PHASED_GENOTYPES: nmb = "#phased_genotypes"
376 raise ValueError("Unknown number type encountered: %s" % fmt.numbertype)
377 values.append( ('Number',nmb) )
378 values.append( ('Type', fmt.type) )
379 values.append( ('Description', '"' + fmt.description + '"') )
380 if self._version == 33:
381 format = ",".join(v for k,v in values)
383 format = "<" + (",".join( "%s=%s" % (k,v) for (k,v) in values )) + ">"
386 def get_expected(self, format, formatdict, alt):
387 fmt = formatdict[format]
388 if fmt.numbertype == self.NT_UNKNOWN: return -1
389 if fmt.numbertype == self.NT_NUMBER: return fmt.number
390 if fmt.numbertype == self.NT_ALLELES: return len(alt)+1
391 if fmt.numbertype == self.NT_NR_ALLELES: return len(alt)
392 if fmt.numbertype == self.NT_GENOTYPES: return ((len(alt)+1)*(len(alt)+2)) // 2
393 if fmt.numbertype == self.NT_PHASED_GENOTYPES: return (len(alt)+1)*(len(alt)+1)
397 def _add_definition(self, formatdict, key, data, line ):
398 if key in formatdict: return
399 self.error(line,self.ERROR_UNKNOWN_KEY,key)
401 formatdict[key] = FORMAT(key,self.NT_NUMBER,0,"Flag","(Undefined tag)",".")
403 if data == []: data = [""] # unsure what type -- say string
404 if type(data[0]) == type(0.0):
405 formatdict[key] = FORMAT(key,self.NT_UNKNOWN,-1,"Float","(Undefined tag)",None)
407 if type(data[0]) == type(0):
408 formatdict[key] = FORMAT(key,self.NT_UNKNOWN,-1,"Integer","(Undefined tag)",None)
410 formatdict[key] = FORMAT(key,self.NT_UNKNOWN,-1,"String","(Undefined tag)",".")
413 # todo: trim trailing missing values
414 def format_formatdata( self, data, format, key=True, value=True, separator=":" ):
415 output, sdata = [], []
416 if type(data) == type([]): # for FORMAT field, make data with dummy values
418 for k in data: d[k] = []
420 # convert missing values; and silently add definitions if required
422 self._add_definition( format, k, data[k], "(output)" )
423 for idx,v in enumerate(data[k]):
424 if v == format[k].missingvalue: data[k][idx] = "."
425 # make sure GT comes first; and ensure fixed ordering; also convert GT data back to string
427 if k != 'GT': sdata.append( (k,data[k]) )
430 sdata = [('GT',map(self.convertGTback,data['GT']))] + sdata
434 if v != None: output.append( k+"="+','.join(map(str,v)) )
435 else: output.append( k )
436 elif key: output.append(k)
438 if v != None: output.append( ','.join(map(str,v)) )
439 else: output.append( "." ) # should not happen
440 # snip off trailing missing data
441 while len(output) > 1:
442 last = output[-1].replace(',','').replace('.','')
443 if len(last)>0: break
445 return separator.join(output)
448 def enter_default_format(self):
449 for f in [FORMAT('GT',self.NT_NUMBER,1,'String','Genotype','.'),
450 FORMAT('GQ',self.NT_NUMBER,1,'Integer','Genotype Quality',-1),
451 FORMAT('DP',self.NT_NUMBER,1,'Integer','Read depth at this position for this sample',-1),
452 FORMAT('HQ',self.NT_UNKNOWN,-1,'Integer','Haplotype Quality',-1), # unknown number, since may be haploid
453 FORMAT('FT',self.NT_NUMBER,1,'String','Sample Genotype Filter','.')]:
454 if f.id not in self._format:
455 self._format[f.id] = f
457 def parse_header( self, line ):
458 assert line.startswith('##')
459 elts = line[2:].split('=')
460 key = elts[0].strip()
461 value = '='.join(elts[1:]).strip()
462 if key == "fileformat":
463 if value == "VCFv3.3":
465 elif value == "VCFv4.0":
467 elif value == "VCFv4.1":
470 self.error(line,self.UNKNOWN_FORMAT_STRING)
472 f = self.parse_format(line, value)
473 self._info[ f.id ] = f
474 elif key == "FILTER":
475 f = self.parse_format(line, value, filter=True)
476 self._filter[ f.id ] = f
477 elif key == "FORMAT":
478 f = self.parse_format(line, value)
479 self._format[ f.id ] = f
481 # keep other keys in the header field
482 self._header.append( (key,value) )
485 def write_header( self, stream ):
486 stream.write("##fileformat=VCFv%s.%s\n" % (self._version // 10, self._version % 10))
487 for key,value in self._header: stream.write("##%s=%s\n" % (key,value))
488 for var,label in [(self._info,"INFO"),(self._filter,"FILTER"),(self._format,"FORMAT")]:
489 for f in var.itervalues(): stream.write("##%s=%s\n" % (label,self.format_format(f,filter=(label=="FILTER"))))
492 def parse_heading( self, line ):
493 assert line.startswith('#')
494 assert not line.startswith('##')
495 headings = line[1:].split('\t')
496 if len(headings)==1 and len(line[1:].split()) >= 9:
497 self.error(line,self.HEADING_NOT_SEPARATED_BY_TABS)
498 headings = line[1:].split()
500 for i,s in enumerate(self._required):
502 if len(headings)<=i or headings[i] != s:
504 if len(headings) <= i:
505 err = "(%sth entry not found)" % (i+1)
507 err = "(found %s, expected %s)" % (headings[i],s)
509 #self.error(line,self.BADLY_FORMATTED_HEADING,err)
511 # allow FORMAT column to be absent
512 if len(headings) == 8:
513 headings.append("FORMAT")
515 self.error(line,self.BADLY_FORMATTED_HEADING,err)
517 self._samples = headings[9:]
518 self._sample2column = dict( [(y,x) for x,y in enumerate( self._samples ) ] )
520 def write_heading( self, stream ):
521 stream.write("#" + "\t".join(self._required + self._samples) + "\n")
523 def convertGT(self, GTstring):
524 if GTstring == ".": return ["."]
526 gts = gtsRegEx.split(GTstring)
527 if len(gts) == 1: return [int(gts[0])]
528 if len(gts) != 2: raise ValueError()
529 if gts[0] == "." and gts[1] == ".": return [gts[0],GTstring[len(gts[0]):-len(gts[1])],gts[1]]
530 return [int(gts[0]),GTstring[len(gts[0]):-len(gts[1])],int(gts[1])]
532 self.error(self._line,self.BAD_GENOTYPE,GTstring)
536 def convertGTback(self, GTdata):
537 return ''.join(map(str,GTdata))
539 def parse_formatdata( self, key, value, formatdict, line ):
540 # To do: check that the right number of values is present
541 f = formatdict.get(key,None)
543 self._add_definition(formatdict, key, value, line )
546 if value is not None: self.error(line,self.ERROR_FLAG_HAS_VALUE)
548 values = value.split(',')
549 # deal with trailing data in some early VCF files
550 if f.type in ["Float","Integer"] and len(values)>0 and values[-1].find(';') > -1:
551 self.error(line,self.ERROR_TRAILING_DATA,values[-1])
552 values[-1] = values[-1].split(';')[0]
553 if f.type == "Integer":
554 for idx,v in enumerate(values):
556 if v == ".": values[idx] = f.missingvalue
557 else: values[idx] = int(v)
559 self.error(line,self.ERROR_FORMAT_NOT_NUMERICAL,values)
560 return [0] * len(values)
562 elif f.type == "String":
564 if f.id == "GT": values = map( self.convertGT, values )
566 elif f.type == "Character":
568 if len(v) != 1: self.error(line,self.ERROR_FORMAT_NOT_CHAR)
570 elif f.type == "Float":
571 for idx,v in enumerate(values):
572 if v == ".": values[idx] = f.missingvalue
573 try: return map(float,values)
575 self.error(line,self.ERROR_FORMAT_NOT_NUMERICAL,values)
576 return [0.0] * len(values)
579 self.error(line,self.ERROR_INFO_STRING)
582 def inregion(self, chrom, pos):
583 if not self._regions: return True
584 for r in self._regions:
585 if r[0] == chrom and r[1] <= pos < r[2]: return True
589 def parse_data( self, line, lineparse=False ):
590 cols = line.split('\t')
591 if len(cols) != len(self._samples)+9:
592 # gracefully deal with absent FORMAT column
593 if len(cols) == 8 and len(self._samples)==0:
597 self.BAD_NUMBER_OF_COLUMNS,
598 "expected %s for %s samples (%s), got %s" % (len(self._samples)+9, len(self._samples), self._samples, len(cols)))
602 # get 0-based position
603 try: pos = int(cols[1])-1
604 except: self.error(line,self.POS_NOT_NUMERICAL)
605 if pos < 0: self.error(line,self.POS_NOT_POSITIVE)
607 # implement filtering
608 if not self.inregion(chrom,pos): return None
610 # end of first-pass parse for sortedVCF
611 if lineparse: return chrom, pos, line
615 ref = cols[3].upper()
617 self.error(line,self.MISSING_REF)
618 if self._version == 33: ref = get_sequence(chrom,pos,pos+1,self._reference)
622 if c not in "ACGTN": self.error(line,self.UNKNOWN_CHAR_IN_REF)
623 if "N" in ref: ref = get_sequence(chrom,pos,pos+len(ref),self._reference)
625 # make sure reference is sane
627 left = max(0,pos-100)
628 faref_leftflank = get_sequence(chrom,left,pos+len(ref),self._reference)
629 faref = faref_leftflank[pos-left:]
630 if faref != ref: self.error(line,self.WRONG_REF,"(reference is %s, VCF says %s)" % (faref,ref))
633 # convert v3.3 to v4.0 alleles below
634 if cols[4] == ".": alt = []
635 else: alt = cols[4].upper().split(',')
637 if cols[5] == ".": qual = -1
639 try: qual = float(cols[5])
640 except: self.error(line,self.QUAL_NOT_NUMERICAL)
642 # postpone checking that filters exist. Encode missing filter or no filtering as empty list
643 if cols[6] == "." or cols[6] == "PASS" or cols[6] == "0": filter = []
644 else: filter = cols[6].split(';')
646 # dictionary of keys, and list of values
649 for blurp in cols[7].split(';'):
650 elts = blurp.split('=')
651 if len(elts) == 1: v = None
652 elif len(elts) == 2: v = elts[1]
653 else: self.error(line,self.ERROR_INFO_STRING)
654 info[elts[0]] = self.parse_formatdata(elts[0], v, self._info, line)
656 # Gracefully deal with absent FORMAT column
657 if cols[8] == "": format = []
658 else: format = cols[8].split(':')
660 # check: all filters are defined
662 if f not in self._filter: self.error(line,self.FILTER_NOT_DEFINED, f)
664 # check: format fields are defined
666 if f not in self._format: self.error(line,self.FORMAT_NOT_DEFINED, f)
668 # convert v3.3 alleles
669 if self._version == 33:
670 if len(ref) != 1: self.error(line,self.V33_BAD_REF)
672 have_deletions = False
674 if len(a) == 1: a = a + ref[1:] # SNP; add trailing reference
675 elif a.startswith('I'): a = ref[0] + a[1:] + ref[1:] # insertion just beyond pos; add first and trailing reference
676 elif a.startswith('D'): # allow D<seq> and D<num>
677 have_deletions = True
679 l = int(a[1:]) # throws ValueError if sequence
680 if len(ref) < l: # add to reference if necessary
681 addns = get_sequence(chrom,pos+len(ref),pos+l,self._reference)
683 for i,na in enumerate(newalts): newalts[i] = na+addns
684 a = ref[l:] # new deletion, deleting pos...pos+l
687 if len(ref) < len(s): # add Ns to reference if necessary
688 addns = get_sequence(chrom,pos+len(ref),pos+len(s),self._reference)
689 if not s.endswith(addns) and addns != 'N'*len(addns):
690 self.error(line,self.V33_UNMATCHED_DELETION,
691 "(deletion is %s, reference is %s)" % (a,get_sequence(chrom,pos,pos+len(s),self._reference)))
693 for i,na in enumerate(newalts): newalts[i] = na+addns
694 a = ref[len(s):] # new deletion, deleting from pos
696 self.error(line,self.V33_BAD_ALLELE)
699 # deletion alleles exist, add dummy 1st reference allele, and account for leading base
702 # Petr Danacek's: we can't have a leading nucleotide at (1-based) position 1
703 addn = get_sequence(chrom,pos+len(ref),pos+len(ref)+1,self._reference)
705 alt = [allele+addn for allele in alt]
707 addn = get_sequence(chrom,pos-1,pos,self._reference)
709 alt = [addn + allele for allele in alt]
712 # format v4.0 -- just check for nucleotides
714 if not alleleRegEx.match(allele):
715 self.error(line,self.V40_BAD_ALLELE,allele)
717 # check for leading nucleotide in indel calls
719 if len(allele) != len(ref):
720 if len(allele) == 0: self.error(line,self.ZERO_LENGTH_ALLELE)
721 if ref[0].upper() != allele[0].upper() and "N" not in (ref[0]+allele[0]).upper():
722 self.error(line,self.MISSING_INDEL_ALLELE_REF_BASE)
724 # trim trailing bases in alleles
725 for i in range(1,min(len(ref),min(map(len,alt)))):
726 if len(set(allele[-1].upper() for allele in alt)) > 1 or ref[-1].upper() != alt[0][-1].upper():
728 ref, alt = ref[:-1], [allele[:-1] for allele in alt]
730 # left-align alleles, if a reference is available
731 if self._leftalign and self._reference:
735 if len(allele) > len(ref):
736 longest, shortest = allele, ref
738 longest, shortest = ref, allele
739 if len(longest) == len(shortest) or longest[:len(shortest)].upper() != shortest.upper():
741 if longest[-1].upper() != longest[len(shortest)-1].upper():
746 alt = [allele[:-1] for allele in alt]
747 if min(len(allele) for allele in alt) == 0 or len(ref) == 0:
748 ref = faref_leftflank[pos-left-1] + ref
749 alt = [faref_leftflank[pos-left-1] + allele for allele in alt]
752 # parse sample columns
754 for sample in cols[9:]:
756 values = sample.split(':')
757 if len(values) > len(format):
758 self.error(line,self.BAD_NUMBER_OF_VALUES,"(found %s values in element %s; expected %s)" % (len(values),sample,len(format)))
759 for idx in range(len(format)):
760 expected = self.get_expected(format[idx], self._format, alt)
761 if idx < len(values): value = values[idx]
763 if expected == -1: value = "."
764 else: value = ",".join(["."]*expected)
765 dict[format[idx]] = self.parse_formatdata(format[idx], value, self._format, line)
766 if expected != -1 and len(dict[format[idx]]) != expected:
767 self.error(line,self.BAD_NUMBER_OF_PARAMETERS,
768 "id=%s, expected %s parameters, got %s" % (format[idx],expected,dict[format[idx]]))
769 if len(dict[format[idx]] ) < expected: dict[format[idx]] += [dict[format[idx]][-1]]*(expected-len(dict[format[idx]]))
770 dict[format[idx]] = dict[format[idx]][:expected]
771 samples.append( dict )
775 'pos':pos, # return 0-based position
783 for key,value in zip(self._samples,samples):
789 def write_data(self, stream, data):
790 required = ['chrom','pos','id','ref','alt','qual','filter','info','format'] + self._samples
792 if k not in data: raise ValueError("Required key %s not found in data" % str(k))
793 if data['alt'] == []: alt = "."
794 else: alt = ",".join(data['alt'])
795 if data['filter'] == None: filter = "."
796 elif data['filter'] == []:
797 if self._version == 33: filter = "0"
798 else: filter = "PASS"
799 else: filter = ';'.join(data['filter'])
800 if data['qual'] == -1: qual = "."
801 else: qual = str(data['qual'])
803 output = [data['chrom'],
804 str(data['pos']+1), # change to 1-based position
810 self.format_formatdata( data['info'], self._info, separator=";" ),
811 self.format_formatdata( data['format'], self._format, value=False ) ]
813 for s in self._samples:
814 output.append( self.format_formatdata( data[s], self._format, key=False ) )
816 stream.write( "\t".join(output) + "\n" )
818 def _parse_header(self, stream):
822 if line.startswith('##'):
823 self.parse_header( line.strip() )
824 elif line.startswith('#'):
825 self.parse_heading( line.strip() )
826 self.enter_default_format()
831 def _parse(self, line, stream):
832 if len(line.strip()) > 0:
833 d = self.parse_data( line.strip() )
837 if self._lines and self._lineno > self._lines: raise StopIteration
838 d = self.parse_data( line.strip() )
841 ######################################################################################################
845 ######################################################################################################
847 def getsamples(self):
848 """ List of samples in VCF file """
851 def setsamples(self,samples):
852 """ List of samples in VCF file """
853 self._samples = samples
856 """ List of header key-value pairs (strings) """
859 def setheader(self,header):
860 """ List of header key-value pairs (strings) """
861 self._header = header
864 """ Dictionary of ##INFO tags, as VCF.FORMAT values """
867 def setinfo(self,info):
868 """ Dictionary of ##INFO tags, as VCF.FORMAT values """
872 """ Dictionary of ##FORMAT tags, as VCF.FORMAT values """
875 def setformat(self,format):
876 """ Dictionary of ##FORMAT tags, as VCF.FORMAT values """
877 self._format = format
880 """ Dictionary of ##FILTER tags, as VCF.FORMAT values """
883 def setfilter(self,filter):
884 """ Dictionary of ##FILTER tags, as VCF.FORMAT values """
885 self._filter = filter
887 def setversion(self, version):
888 if version not in [33,40,41]: raise ValueError("Can only handle v3.3, v4.0 and v4.1 VCF files")
889 self._version = version
891 def setregions(self, regions):
892 self._regions = regions
894 def setreference(self, ref):
895 """ Provide a reference sequence; a Python class supporting a fetch(chromosome, start, end) method, e.g. PySam.FastaFile """
896 self._reference = ref
898 def ignoreerror(self, errorstring):
899 try: self._ignored_errors.add(self.__dict__[errorstring])
900 except KeyError: raise ValueError("Invalid error string: %s" % errorstring)
902 def warnerror(self, errorstring):
903 try: self._warn_errors.add(self.__dict__[errorstring])
904 except KeyError: raise ValueError("Invalid error string: %s" % errorstring)
906 def parse(self, stream):
907 """ Parse a stream of VCF-formatted lines. Initializes class instance and return generator """
908 last_line = self._parse_header(stream)
909 # now return a generator that does the actual work. In this way the pre-processing is done
910 # before the first piece of data is yielded
911 return self._parse(last_line, stream)
913 def write(self, stream, datagenerator):
914 """ Writes a VCF file to a stream, using a data generator (or list) """
915 self.write_header(stream)
916 self.write_heading(stream)
917 for data in datagenerator: self.write_data(stream,data)
919 def writeheader(self, stream):
920 """ Writes a VCF header """
921 self.write_header(stream)
922 self.write_heading(stream)
924 def compare_calls(self, pos1, ref1, alt1, pos2, ref2, alt2):
925 """ Utility function: compares two calls for equality """
926 # a variant should always be assigned to a unique position, one base before
927 # the leftmost position of the alignment gap. If this rule is implemented
928 # correctly, the two positions must be equal for the calls to be identical.
929 if pos1 != pos2: return False
930 # from both calls, trim rightmost bases when identical. Do this safely, i.e.
931 # only when the reference bases are not Ns
932 while len(ref1)>0 and len(alt1)>0 and ref1[-1] == alt1[-1]:
935 while len(ref2)>0 and len(alt2)>0 and ref2[-1] == alt2[-1]:
938 # now, the alternative alleles must be identical
941 ###########################################################################################################
942 ###########################################################################################################
943 ## API functions added by Andreas
944 ###########################################################################################################
946 def connect( self, filename ):
947 '''connect to tabix file.'''
948 self.tabixfile = pysam.Tabixfile( filename )
949 self._parse_header(self.tabixfile.header)
956 """ Parse a stream of VCF-formatted lines. Initializes class instance and return generator """
958 iter = self.tabixfile.fetch( reference, start, end, region, parser = pysam.asVCF() )
960 yield VCFRecord( x, self )
962 def validate( self, record ):
963 '''validate vcf record.
965 returns a validated record.
968 chrom, pos = record.chrom, record.pos
973 self.error(str(record),self.MISSING_REF)
974 if self._version == 33: ref = get_sequence(chrom,pos,pos+1,self._reference)
978 if c not in "ACGTN": self.error(str(record),self.UNKNOWN_CHAR_IN_REF)
979 if "N" in ref: ref = get_sequence(chrom,
984 # make sure reference is sane
986 left = max(0,self.pos-100)
987 faref_leftflank = get_sequence(chrom,left,self.pos+len(ref),self._reference)
988 faref = faref_leftflank[pos-left:]
989 if faref != ref: self.error(line,self.WRONG_REF,"(reference is %s, VCF says %s)" % (faref,ref))
992 # check: format fields are defined
993 for f in record.format:
994 if f not in self._format: self.error(str(record),self.FORMAT_NOT_DEFINED, f)
996 # check: all filters are defined
997 for f in record.filter:
998 if f not in self._filter: self.error(str(record),self.FILTER_NOT_DEFINED, f)
1000 # convert v3.3 alleles
1001 if self._version == 33:
1002 if len(ref) != 1: self.error(line,self.V33_BAD_REF)
1004 have_deletions = False
1006 if len(a) == 1: a = a + ref[1:] # SNP; add trailing reference
1007 elif a.startswith('I'): a = ref[0] + a[1:] + ref[1:] # insertion just beyond pos; add first and trailing reference
1008 elif a.startswith('D'): # allow D<seq> and D<num>
1009 have_deletions = True
1011 l = int(a[1:]) # throws ValueError if sequence
1012 if len(ref) < l: # add to reference if necessary
1013 addns = get_sequence(chrom,pos+len(ref),pos+l,self._reference)
1015 for i,na in enumerate(newalts): newalts[i] = na+addns
1016 a = ref[l:] # new deletion, deleting pos...pos+l
1019 if len(ref) < len(s): # add Ns to reference if necessary
1020 addns = get_sequence(chrom,pos+len(ref),pos+len(s),self._reference)
1021 if not s.endswith(addns) and addns != 'N'*len(addns):
1022 self.error(line,self.V33_UNMATCHED_DELETION,
1023 "(deletion is %s, reference is %s)" % (a,get_sequence(chrom,pos,pos+len(s),self._reference)))
1025 for i,na in enumerate(newalts): newalts[i] = na+addns
1026 a = ref[len(s):] # new deletion, deleting from pos
1028 self.error(line,self.V33_BAD_ALLELE)
1031 # deletion alleles exist, add dummy 1st reference allele, and account for leading base
1034 # Petr Danacek's: we can't have a leading nucleotide at (1-based) position 1
1035 addn = get_sequence(chrom,pos+len(ref),pos+len(ref)+1,self._reference)
1037 alt = [allele+addn for allele in alt]
1039 addn = get_sequence(chrom,pos-1,pos,self._reference)
1041 alt = [addn + allele for allele in alt]
1044 # format v4.0 -- just check for nucleotides
1046 if not alleleRegEx.match(allele):
1047 self.error(line,self.V40_BAD_ALLELE,allele)
1050 # check for leading nucleotide in indel calls
1052 if len(allele) != len(ref):
1053 if len(allele) == 0: self.error(line,self.ZERO_LENGTH_ALLELE)
1054 if ref[0].upper() != allele[0].upper() and "N" not in (ref[0]+allele[0]).upper():
1055 self.error(line,self.MISSING_INDEL_ALLELE_REF_BASE)
1057 # trim trailing bases in alleles
1058 for i in range(1,min(len(ref),min(map(len,alt)))):
1059 if len(set(allele[-1].upper() for allele in alt)) > 1 or ref[-1].upper() != alt[0][-1].upper():
1061 ref, alt = ref[:-1], [allele[:-1] for allele in alt]
1063 # left-align alleles, if a reference is available
1064 if self._leftalign and self._reference:
1068 if len(allele) > len(ref):
1069 longest, shortest = allele, ref
1071 longest, shortest = ref, allele
1072 if len(longest) == len(shortest) or longest[:len(shortest)].upper() != shortest.upper():
1074 if longest[-1].upper() != longest[len(shortest)-1].upper():
1079 alt = [allele[:-1] for allele in alt]
1080 if min(len(allele) for allele in alt) == 0 or len(ref) == 0:
1081 ref = faref_leftflank[pos-left-1] + ref
1082 alt = [faref_leftflank[pos-left-1] + allele for allele in alt]