Skip to content


Repository files navigation


A C++ library for parsing and manipulating VCF files.

Github-CI AnacondaBadge DL BrewBadge GuixBadge DebianBadge C++0x Chat on Matrix

Vcflib and related tools are the workhorses in bioinformatics for processing the VCF variant calling format. See

A spectrum of free software tools for processing the VCF variant call format: vcflib, bio-vcf, cyvcf2, hts-nim and slivar. Garrison E, Kronenberg ZN, Dawson ET, Pedersen BS, Prins P (2022), PLoS Comput Biol 18(5): e1009123.

(bibtex reference at the end of this page)


Humpty Logo - CC0 license

January 2023: this is vcflib's first Humpty Dumpty release: vcfcreatemulti is the natural companion to vcfwave. Often variant callers are not perfect. vcfwave with its companion tool vcfcreatemulti can take an existing VCF file that contains multiple complex overlapping and even nested alleles and, unlike Humpty Dumpty, take them apart and put them together again. Thereby, hopefully, creating sane VCF output that is useful for analysis and getting rid of false positives.

We created these tools by including the state-of-the-art biWFA wavefront aligner. The tools are particularly useful for the output from structural variation callers and pangenome genotypers, such as used by the Human Pangenome Reference Consortium (HPRC) because of overlapping ALT segments.

See also

May 2022: the vcflib paper has been published on PLoS Computational Biology!

April 2022: vcflib has just gone pangenome!

By introducing the wavefront algorithm we can now realign long sequences and reduce call complexity (and FPs!) introduced by pangenome variant callers using the new vcfwave tool.

See also


The Variant Call Format (VCF) is a flat-file, tab-delimited textual format that describes reference-indexed variations between individuals. VCF provides a common interchange format for the description of variation in individuals and populations of samples, and has become the de facto standard reporting format for a wide array of genomic variant detectors.

vcflib provides methods to manipulate and interpret sequence variation described by VCF. It is both:

  • an API for parsing and operating on records of genomic variation as it can be described by the VCF format
  • a collection of command-line utilities for executing complex manipulations on VCF files

vclib is both a library (with an API) and a collection of useful tools. The API provides a quick and extremely permissive method to read and write VCF files. Extensions and applications of the library provided in the included utilities (*.cpp) comprise the vast bulk of the library's utility.

We have also added infrastructure to write Python bindings. See below.

Short index:


For latest updates see RELEASE NOTES.

Conda installs in user land without root access

conda install -c bioconda vcflib

Homebrew installs on Linux and Mac OSX

brew install brewsci/bio/vcflib

For Debian and Ubuntu

apt-get install libvcflib-tools libvcflib-dev

We develop against guix and vcflib is packaged as

guix package -i vcflib

See also the Guix shell below.


Users are encouraged to drive the utilities in the library in a streaming fashion, using Unix pipes to fully utilize resources on multi-core systems. Piping provides a convenient method to interface with other libraries (vcf-tools, BedTools, GATK, htslib, bio-vcf, bcftools, freebayes) which interface via VCF files, allowing the composition of an immense variety of processing functions. Examples can be found in the scripts, e.g. ./scripts.



filter command description
vcfuniq List unique genotypes. Similar to GNU uniq, but aimed at VCF records. vcfuniq removes records which have the same position, ref, and alt as the previous record on a sorted VCF file. Note that it does not adjust/combine genotypes in the output, but simply takes the first record. See also vcfcreatemulti for combining records.
vcfuniqalleles List unique alleles For each record, remove any duplicate alternate alleles that may have resulted from merging separate VCF files.
vcffilter VCF filter the specified vcf file using the set of filters


metrics command description
vcfhethomratio Generates the het/hom ratio for each individual in the file
vcfhetcount Calculate the heterozygosity rate: count the number of alternate alleles in heterozygous genotypes in all records in the vcf file
vcfdistance Adds a tag to each variant record which indicates the distance to the nearest variant. (defaults to BasesToClosestVariant if no custom tag name is given.
vcfentropy Annotate VCF records with the Shannon entropy of flanking sequence. Anotates the output VCF file with, for each record, EntropyLeft, EntropyRight, EntropyCenter, which are the entropies of the sequence of the given window size to the left, right, and center of the record. Also adds EntropyRef and EntropyAlt for each alt.
vcfcheck Validate integrity and identity of the VCF by verifying that the VCF record's REF matches a given reference file.


phenotype command description
permuteGPAT++ permuteGPAT++ is a method for adding empirical p-values to a GPAT++ score.


genotype command description
hapLrt HapLRT is a likelihood ratio test for haplotype lengths. The lengths are modeled with an exponential distribution. The sign denotes if the target has longer haplotypes (1) or the background (-1).
normalize-iHS normalizes iHS or XP-EHH scores.
abba-baba abba-baba calculates the tree pattern for four indviduals. This tool assumes reference is ancestral and ignores non abba-baba sites. The output is a boolian value: 1 = true , 0 = false for abba and baba. the tree argument should be specified from the most basal taxa to the most derived.


transformation command description
vcfintersect VCF set analysis
vcfleftalign Left-align indels and complex variants in the input using a pairwise ref/alt alignment followed by a heuristic, iterative left realignment process that shifts indel representations to their absolute leftmost (5') extent.
vcfglxgt Set genotypes using the maximum genotype likelihood for each sample.
vcfkeepsamples outputs each record in the vcf file, removing samples not listed on the command line
vcfallelicprimitives Note that this tool is considered legacy and will emit a warning! Use vcfwave instead.
vcfcommonsamples Generates each record in the first file, removing samples not present in the second
vcfcreatemulti Go through sorted VCF and if overlapping alleles are represented across multiple records, merge them into a single record. Currently only for indels.
vcfld Compute LD
vcfsamplenames List sample names
vcfsample2info Take annotations given in the per-sample fields and add the mean, median, min, or max to the site-level INFO.
vcflength Add length info field
vcfstreamsort Sorts the input (either stdin or file) using a streaming sort algorithm. Guarantees that the positional order is correct provided out-of-order variants are no more than 100 positions in the VCF file apart.
dumpContigsFromHeader Dump contigs from header
vcfgeno2haplo Convert genotype-based phased alleles within --window-size into haplotype alleles. Will break haplotype construction when encountering non-phased genotypes on input.
vcfinfo2qual Sets QUAL from info field tag keyed by [key]. The VCF file may be omitted and read from stdin. The average of the field is used if it contains multiple values.
vcfevenregions Generates a list of regions, e.g. chr20:10..30 using the variant density information provided in the VCF file to ensure that the regions have even numbers of variants. This can be use to reduce the variance in runtime when dividing variant detection or genotyping by genomic coordinates.
vcfbreakmulti If multiple alleles are specified in a single record, break the record into multiple lines, preserving allele-specific INFO fields.
vcfwave Realign reference and alternate alleles with WFA, parsing out the primitive alleles into multiple VCF records. New records have IDs that reference the source record ID. Genotypes are handled. Deletions generate haploid/missing genotypes at overlapping sites.
vcfafpath Display genotype paths
vcfremoveaberrantgenotypes strips samples which are homozygous but have observations implying heterozygosity. Remove samples for which the reported genotype (GT) and observation counts disagree (AO, RO).
vcf2tsv Converts VCF to per-allelle or per-genotype tab-delimited format, using null string to replace empty values in the table. Specifying -g will output one line per sample with genotype information. When there is more than one alt allele there will be multiple rows, one for each allele and, the info will match the 'A' index
vcfqual2info Puts QUAL into an info field tag keyed by [key].
vcfnumalt outputs a VCF stream where NUMALT has been generated for each record using sample genotypes
vcfprimers For each VCF record, extract the flanking sequences, and write them to stdout as FASTA records suitable for alignment.
vcf2fasta Generates sample_seq:N.fa for each sample, reference sequence, and chromosomal copy N in [0,1... ploidy]. Each sequence in the fasta file is named using the same pattern used for the file name, allowing them to be combined.
vcffixup Generates a VCF stream where AC and NS have been generated for each record using sample genotypes
vcfkeepinfo To decrease file size remove INFO fields not listed on the command line
vcfsamplediff Establish putative somatic variants using reported differences between germline and somatic samples. Tags each record where the listed sample genotypes differ with . The first sample is assumed to be germline, the second somatic. Each record is tagged with ={germline,somatic,loh} to specify the type of variant given the genotype difference between the two samples.
vcfflatten Removes multi-allelic sites by picking the most common alternate. Requires allele frequency specification 'AF' and use of 'G' and 'A' to specify the fields which vary according to the Allele or Genotype. VCF file may be specified on the command line or piped as stdin.
vcfglbound Adjust GLs so that the maximum GL is 0 by dividing all GLs for each sample by the max.
vcfecho Echo VCF to stdout (simple demo)
vcfgeno2alleles modifies the genotypes field to provide the literal alleles rather than indexes
vcf2dag Modify VCF to be able to build a directed acyclic graph (DAG)
vcfannotategenotypes Examine genotype correspondence. Annotate genotypes in the first file with genotypes in the second adding the genotype as another flag to each sample filed in the first file. annotation-tag is the name of the sample flag which is added to store the annotation. also adds a 'has_variant' flag for sites where the second file has a variant.
vcfremap For each alternate allele, attempt to realign against the reference with lowered gap open penalty. If realignment is possible, adjust the cigar and reference/alternate alleles. Observe how different alignment parameters, including context and entropy-dependent ones, influence variant classification and interpretation.
vcfremovesamples outputs each record in the vcf file, removing samples listed on the command line
vcfcleancomplex Removes reference-matching sequence from complex alleles and adjusts records to reflect positional change.
vcfindex Adds an index number to the INFO field (id=position)
vcfannotate Intersect the records in the VCF file with targets provided in a BED file. Intersections are done on the reference sequences in the VCF file. If no VCF filename is specified on the command line (last argument) the VCF read from stdin.
smoother smoothes is a method for window smoothing many of the GPAT++ formats.
vcfkeepgeno Reduce file size by removing FORMAT fields not listed on the command line from sample specifications in the output
vcfcombine Combine VCF files positionally, combining samples when sites and alleles are identical. Any number of VCF files may be combined. The INFO field and other columns are taken from one of the files which are combined when records in multiple files match. Alleles must have identical ordering to be combined into one record. If they do not, multiple records will be emitted.
vcfcat Concatenates VCF files
vcfinfosummarize Take annotations given in the per-sample fields and add the mean, median, min, or max to the site-level INFO.
vcfclassify Creates a new VCF where each variant is tagged by allele class: snp, ts/tv, indel, mnp
vcfoverlay Overlay records in the input vcf files with order as precedence.
vcfaddinfo Adds info fields from the second file which are not present in the first vcf file.
vcfnullgenofields Makes the FORMAT for each variant line the same (uses all the FORMAT fields described in the header). Fills out per-sample fields to match FORMAT. Expands GT values of '.' with number of alleles based on ploidy (eg: './doc/.' for dipolid).
vcfgenosamplenames Get samplenames


statistics command description
permuteSmooth permuteSmooth is a method for adding empirical p-values smoothed wcFst scores.
plotHaps plotHaps provides the formatted output that can be used with 'bin/plotHaplotypes.R'.
pVst pVst calculates vst, a measure of CNV stratification.
wcFst wcFst is Weir & Cockerham's Fst for two populations. Negative values are VALID, they are sites which can be treated as zero Fst. For more information see Evolution, Vol. 38 N. 6 Nov 1984. Specifically wcFst uses equations 1,2,3,4.
popStats General population genetic statistics for each SNP
vcfsitesummarize Summarize by site
vcfrandomsample Randomly sample sites from an input VCF file, which may be provided as stdin. Scale the sampling probability by the field specified in KEY. This may be used to provide uniform sampling across allele frequencies, for instance.
vcfroc Generates a pseudo-ROC curve using sensitivity and specificity estimated against a putative truth set. Thresholding is provided by successive QUAL cutoffs.
vcfstats Prints statistics about variants in the input VCF file.
bFst bFst is a Bayesian approach to Fst. Importantly bFst accounts for genotype uncertainty in the model using genotype likelihoods. For a more detailed description see: `A Bayesian approach to inferring population structure from dominant markers' by Holsinger et al. Molecular Ecology Vol 11, issue 7 2002. The likelihood function has been modified to use genotype likelihoods provided by variant callers. There are five free parameters estimated in the model: each subpopulation's allele frequency and Fis (fixation index, within each subpopulation), a free parameter for the total population's allele frequency, and Fst.
vcfcountalleles Count alleles
vcfparsealts Alternate allele parsing method. This method uses pairwise alignment of REF and ALTs to determine component allelic primitives for each alternate allele.
genotypeSummary Generates a table of genotype counts. Summarizes genotype counts for bi-allelic SNVs and indel
vcfgenosummarize Adds summary statistics to each record summarizing qualities reported in called genotypes. Uses: RO (reference observation count), QR (quality sum reference observations) AO (alternate observation count), QA (quality sum alternate observations)
vcfaltcount count the number of alternate alleles in all records in the vcf file
iHS iHS calculates the integrated haplotype score which measures the relative decay of extended haplotype homozygosity (EHH) for the reference and alternative alleles at a site (see: voight et al. 2006, Spiech & Hernandez 2014).
segmentFst segmentFst creates genomic segments (bed file) for regions with high wcFst
pFst pFst is a probabilistic approach for detecting differences in allele frequencies between two populations.
sequenceDiversity The sequenceDiversity program calculates two popular metrics of haplotype diversity: pi and extended haplotype homozygoisty (eHH). Pi is calculated using the Nei and Li 1979 formulation. eHH a convenient way to think about haplotype diversity. When eHH = 0 all haplotypes in the window are unique and when eHH = 1 all haplotypes in the window are identical.
vcfgenotypecompare adds statistics to the INFO field of the vcf file describing the amount of discrepancy between the genotypes (GT) in the vcf file and the genotypes reported in the . use this after vcfannotategenotypes to get correspondence statistics for two vcfs.
vcfrandom Generate a random VCF file
segmentIhs Creates genomic segments (bed file) for regions with high wcFst
vcfgenotypes Report the genotypes for each sample, for each variant in the VCF. Convert the numerical represenation of genotypes provided by the GT field to a human-readable genotype format.

See also


The vcflib source repository contains a number of additional scripts. Click on the link to see the source code.

script description
vcfclearinfo clear INFO field
vcfqualfilter quality filter
vcfnulldotslashdot rewrite null genotypes to ./.
vcfprintaltdiscrepancy.r show ALT discrepancies in a table
vcfremovenonATGC remove non-nucleotides in REF or ALT
plotSmoothed.R smooth plot of wcFst, pFst or abba-baba
vcf_strip_extra_headers strip headers
plotHapLrt.R plot results of pFst
vcfbiallelic remove anything that is not biallelic
vcfsort sort VCF using shell script
vcfnosnps remove SNPs
vcfmultiwayscripts more multiway comparisons annotates records in the first file with genotypes and sites from the second
plotPfst.R plot pFst
vcfregionreduce_and_cut reduce, gzip, and tabix
plotBfst.R plot results of pFst
vcfnobiallelicsnps remove biallelic SNPs
vcfindels show INDELS
vcfmultiway multiway comparison
vcfregionreduce reduce VCFs using a BED File, gzip them up and create tabix index runner
vcfclearid clear ID field
vcfcomplex remove all SNPs but keep SVs
vcffirstheader show first header
plotXPEHH.R plot XPEHH
vcfregionreduce_pipe reduce, gzip and tabix in a pipe plot ALT discrepancy runner runner
vcfnoindels remove INDELs
bgziptabix runs bgzip on the input and tabix indexes the result
plotHaplotypes.R plot results
vcfplotsitediscrepancy.r plot site discrepancy
vcfindelproximity show SNPs around an INDEL
bed2region convert VCF CHROM column in VCF file to region
vcfplotaltdiscrepancy.r plot ALT discrepancies
plot_roc.r plot ROC
vcfmultiallelic remove anything that is not multiallelic
vcfsnps show SNPs
vcfvarstats use fastahack to get stats
vcfregionreduce_uncompressed reduce, gzip and tabix
plotWCfst.R plot wcFst transform VCF to BED file
vcfjoincalls overlay files using QUAL and GT from a second VCF push VCF file into SQLite3 database using dbname


vcflib has rudimentary python bindings, but the are easy to build up on. See pyvcflib.


build from source

VCFLIB uses the cmake build system, after a recursive checkout of the sources make the files in the 'build' directory with:

git clone --recursive
cd vcflib
mkdir -p build && cd build
cmake --build .
cmake --install .

and to run the tests

ctest . --verbose

Executables are built into the ./build directory in the repository.

Note, if you have an existing repo update submodules with

git submodule update --init --recursive --progress
cd build
cmake --build . --target clean

The most interesting one, perhaps, is wfa2lib. Bundling is no longer the default, but it is a submodule that will install an optimized Set up compilation with

cmake --build .
ctest .
cmake --install .

Build dependencies can be viewed in the github-CI scripts (see badges above), as well as guix.scm used by us to create the build environment (for instructions see the header of guix.scm). Essentially:

  • cmake
  • C++ compiler
  • htslib
  • tabixpp
  • WFA2 (wfa2lib)
  • pybind11 (for testing)

For include files add

  • libhts-dev
  • libtabixpp-dev
  • libtabixpp0

And for some of the VCF executables

  • zig 0.10 (disable with cmake -DZIG=OFF)
  • python
  • perl

To use Zig compiled tools, such as vcfcreatemulti, make sure the zig compiler is in the PATH. The zig dependency is for the recently upgraded vcfcreatemulti that can be run after vcfwave to combine overlapping records. It is unfortunate that zig is not part of some large Linux distributions. More on zig can be found in the souce code README.

Using a recent C++ compiler

You may encounter something like version `GLIBCXX_3.4.32' not found (required by ...vcflib/build/ This is caused by the python pyvcflib not running against a python interpreter compiled with the same libstdc++ dependency. That is a bit nasty to solve. For now make sure to match C++ compilers for the python module.

Using a different htslib

Check out htslib in tabixpp (recursively) and

cmake -DHTSLIB_LOCAL:STRING=./htslib/ ..
cmake --build .

link library

The standard build creates build/vcflib.a. Take a hint from the cmake file that builds all the vcflib tools.

distro builds

Distros, such as Debian, should build with something like

cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo -DZIG=OFF ..

See the CMakeLists.txt header for more.

source code

See vcfecho.cpp for basic usage. Variant.h and Variant.cpp describe methods available in the API. vcflib is incorporated into several projects, such as freebayes, which may provide a point of reference for prospective developers. Note vcflib contains submodules (git repositories) comprising some dependencies. A full Guix development environment we use is defined here.

adding tests

vcflib uses different test systems. The most important one is the doctest because it doubles as documentation. For an example see which can be run from the command line with

cd test
python3 -m doctest -o NORMALIZE_WHITESPACE -o REPORT_UDIFF pytest/

We also added support for python bindings and unit tests. See for an example.


The developers are on the vcflib matrix channel. Please do not use the github issue tracker for support issues!


To contribute code to vcflib send a github pull request. We may ask you to add a working test case as described in 'adding tests'.


This software is distributed under the free software MIT LICENSE.


Citations are the bread and butter of Science. If you are using this software in your research and want to support our future work, please cite the following publication:

Please cite:

A spectrum of free software tools for processing the VCF variant call format: vcflib, bio-vcf, cyvcf2, hts-nim and slivar. Garrison E, Kronenberg ZN, Dawson ET, Pedersen BS, Prins P (2022), PLoS Comput Biol 18(5): e1009123.

Bibtex reference

    doi = {10.1371/journal.pcbi.1009123},
    author = {Garrison, Erik AND Kronenberg, Zev N. AND Dawson, Eric T. AND Pedersen, Brent S. AND Prins, Pjotr},
    journal = {PLOS Computational Biology},
    publisher = {Public Library of Science},
    title = {A spectrum of free software tools for processing the VCF variant call format: vcflib, bio-vcf, cyvcf2, hts-nim and slivar},
    year = {2022},
    month = {05},
    volume = {18},
    url = {},
    pages = {1-15}

Below the prepublished version of our paper

@article {Garrison2021.05.21.445151,
	author = {Garrison, Erik and Kronenberg, Zev N. and Dawson, Eric T. and Pedersen, Brent S. and Prins, Pjotr},
	title = {Vcflib and tools for processing the VCF variant call format},
	elocation-id = {2021.05.21.445151},
	year = {2021},
	doi = {10.1101/2021.05.21.445151},
	publisher = {Cold Spring Harbor Laboratory},
	URL = {},
	eprint = {},
	journal = {bioRxiv}