A VCFv4.0 and 4.1 parser for Python.

Online version of PyVCF documentation is available at

The intent of this module is to mimic the csv module in the Python stdlib, as opposed to more flexible serialization formats like JSON or YAML. vcf will attempt to parse the content of each record based on the data types specified in the meta-information lines – specifically the ##INFO and ##FORMAT lines. If these lines are missing or incomplete, it will check against the reserved types mentioned in the spec. Failing that, it will just return strings.

There main interface is the class: Reader. It takes a file-like object and acts as a reader:

>>> import vcf
>>> vcf_reader = vcf.Reader(open('vcf/test/example-4.0.vcf', 'r'))
>>> for record in vcf_reader:
...     print record
Record(CHROM=20, POS=14370, REF=G, ALT=[A])
Record(CHROM=20, POS=17330, REF=T, ALT=[A])
Record(CHROM=20, POS=1110696, REF=A, ALT=[G, T])
Record(CHROM=20, POS=1230237, REF=T, ALT=[None])
Record(CHROM=20, POS=1234567, REF=GTCT, ALT=[G, GTACT])

This produces a great deal of information, but it is conveniently accessed. The attributes of a Record are the 8 fixed fields from the VCF spec:

* ``Record.CHROM``
* ``Record.POS``
* ``Record.ID``
* ``Record.REF``
* ``Record.ALT``
* ``Record.QUAL``
* ``Record.FILTER``
* ``Record.INFO``

plus attributes to handle genotype information:

  • Record.FORMAT
  • Record.samples
  • Record.genotype

samples and genotype, not being the title of any column, are left lowercase. The format of the fixed fields is from the spec. Comma-separated lists in the VCF are converted to lists. In particular, one-entry VCF lists are converted to one-entry Python lists (see, e.g., Record.ALT). Semicolon-delimited lists of key=value pairs are converted to Python dictionaries, with flags being given a True value. Integers and floats are handled exactly as you’d expect:

>>> vcf_reader = vcf.Reader(open('vcf/test/example-4.0.vcf', 'r'))
>>> record =
>>> print record.POS
>>> print record.ALT
>>> print record.INFO['AF']

There are a number of convienience methods and properties for each Record allowing you to examine properties of interest:

>>> print record.num_called, record.call_rate, record.num_unknown
3 1.0 0
>>> print record.num_hom_ref, record.num_het, record.num_hom_alt
1 1 1
>>> print record.nucl_diversity, record.aaf
0.6 0.5
>>> print record.get_hets()
[Call(sample=NA00002, CallData(GT=1|0, GQ=48, DP=8, HQ=[51, 51]))]
>>> print record.is_snp, record.is_indel, record.is_transition, record.is_deletion
True False True False
>>> print record.var_type, record.var_subtype
snp ts
>>> print record.is_monomorphic

record.FORMAT will be a string specifying the format of the genotype fields. In case the FORMAT column does not exist, record.FORMAT is None. Finally, record.samples is a list of dictionaries containing the parsed sample column and record.genotype is a way of looking up genotypes by sample name:

>>> record =
>>> for sample in record.samples:
...     print sample['GT']
>>> print record.genotype('NA00001')['GT']

The genotypes are represented by Call objects, which have three attributes: the corresponding Record site, the sample name in sample and a dictionary of call data in data:

>>> call = record.genotype('NA00001')
>>> print
Record(CHROM=20, POS=17330, REF=T, ALT=[A])
>>> print call.sample
>>> print
CallData(GT=0|0, GQ=49, DP=3, HQ=[58, 50])

Please note that as of release 0.4.0, attributes known to have single values (such as DP and GQ above) are returned as values. Other attributes are returned as lists (such as HQ above).

There are also a number of methods:

>>> print call.called, call.gt_type, call.gt_bases, call.phased
True 0 T|T True

Metadata regarding the VCF file itself can be investigated through the following attributes:

  • Reader.metadata
  • Reader.infos
  • Reader.filters
  • Reader.formats
  • Reader.samples

For example:

>>> vcf_reader.metadata['fileDate']
>>> vcf_reader.samples
['NA00001', 'NA00002', 'NA00003']
>>> vcf_reader.filters
OrderedDict([('q10', Filter(id='q10', desc='Quality below 10')), ('s50', Filter(id='s50', desc='Less than 50% of samples have data'))])
>>> vcf_reader.infos['AA'].desc
'Ancestral Allele'

ALT records are actually classes, so that you can interrogate them:

>>> reader = vcf.Reader(open('vcf/test/example-4.1-bnd.vcf'))
>>> _ =; row =
>>> print row
Record(CHROM=1, POS=2, REF=T, ALT=[T[2:3[])
>>> bnd = row.ALT[0]
>>> print bnd.withinMainAssembly, bnd.orientation, bnd.remoteOrientation, bnd.connectingSequence
True False True T

Random access is supported for files with tabix indexes. Simply call fetch for the region you are interested in:

>>> vcf_reader = vcf.Reader(filename='vcf/test/tb.vcf.gz')
>>> for record in vcf_reader.fetch('20', 1110696, 1230237):  
...     print record
Record(CHROM=20, POS=1110696, REF=A, ALT=[G, T])
Record(CHROM=20, POS=1230237, REF=T, ALT=[None])

Or extract a single row:

>>> print vcf_reader.fetch('20', 1110696)  
Record(CHROM=20, POS=1110696, REF=A, ALT=[G, T])

The Writer class provides a way of writing a VCF file. Currently, you must specify a template Reader which provides the metadata:

>>> vcf_reader = vcf.Reader(filename='vcf/test/tb.vcf.gz')
>>> vcf_writer = vcf.Writer(open('/dev/null', 'w'), vcf_reader)
>>> for record in vcf_reader:
...     vcf_writer.write_record(record)

An extensible script is available to filter vcf files in VCF filters declared by other packages will be available for use in this script. Please see Filtering VCF files for full description.

Project Versions

Previous topic

PyVCF - A Variant Call Format Parser for Python

Next topic


This Page