- The Variant Call Format (VCF) and its binary counterpart, the Binary Call Format (BCF), are widely used in bioinformatics to represent, store, exchange, and analyse genetic variation identified through DNA sequencing. They are essential in genome analysis, population genetics, medical genetics, cancer genomics, evolutionary biology, and precision medicine. These formats describe genetic variants such as single-nucleotide variants (SNVs), insertions, deletions, and larger structural variants, together with information about their genomic positions, alleles, quality scores, supporting evidence, and sample genotypes. VCF is a text-based format that can be inspected with ordinary text editors, whereas BCF stores similar information in a compact binary representation that is generally faster and more efficient for computational processing.
- VCF was developed to provide a standardized way of recording variation relative to a reference genome. A VCF file typically contains a metadata section followed by a tab-delimited table in which each row represents a variant record. The metadata section begins with lines marked by ##, which describe the file format version, reference genome, annotation fields, filters, and other information. A header line beginning with #CHROM defines the columns in the variant table. When sample-level genotypes are included, additional columns identify the samples represented in the file.
- A simplified VCF file may look like this:
##fileformat=VCFv4.3
##reference=GRCh38
##INFO=<ID=DP,Number=1,Type=Integer,Description=”Total read depth”>
##FORMAT=<ID=GT,Number=1,Type=String,Description=”Genotype”>
##FORMAT=<ID=DP,Number=1,Type=Integer,Description=”Read depth”>
##FORMAT=<ID=GQ,Number=1,Type=Integer,Description=”Genotype quality”>
#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT Sample1
1 10001 rs123 A G 60 PASS DP=40 GT:DP:GQ 0/1:40:55
1 10500 . C T 45 PASS DP=32 GT:DP:GQ 1/1:32:42
- The first column, CHROM, identifies the chromosome or reference sequence containing the variant. The POS column gives its genomic coordinate, generally using one-based positioning. The ID column contains a known variant identifier, such as an rsID, when available; a period indicates that no identifier has been provided. The REF column records the reference allele, while ALT contains one or more alternative alleles observed in the sample or dataset. The QUAL column provides a Phred-scaled quality score for the variant record, and FILTER indicates whether the record passed the applied filters. The INFO column contains additional site-level annotations, while FORMAT defines the types of sample-level information provided in the following columns.
- The CHROM, POS, ID, REF, ALT, QUAL, FILTER, and INFO columns are the eight mandatory VCF columns. The FORMAT and sample columns are used when genotype or other sample-specific information is included. A VCF file containing only site-level variant records may therefore have eight columns, whereas a multisample VCF can contain many additional sample columns.
- The REF and ALT fields are particularly important because they define the variant in relation to the reference genome. For example, a record with REF=A and ALT=G indicates that the reference allele is A and an alternative allele G has been observed. A variant record with REF=AT and ALT=A represents a deletion of T relative to the reference sequence. Insertions and deletions are represented using a convention that includes an anchoring base, allowing the variant to be described consistently in the coordinate system. Complex variants may contain multiple nucleotide changes within a single record.
- VCF supports multiple alternative alleles at one genomic position. For example, a record may contain ALT=G,T, indicating that two alternative alleles have been observed relative to the same reference allele. Genotypes then refer to allele indices: 0 represents the reference allele, 1 represents the first alternative allele, and 2 represents the second alternative allele. A genotype such as 1/2 therefore indicates that the sample carries the first and second alternative alleles at that position. Correct interpretation of multiallelic records is important when performing genotype analysis, variant filtering, and downstream statistical studies.
- The QUAL field represents the Phred-scaled confidence in the variant record, based on the variant caller’s model. It is not the same as genotype quality, which is generally represented by the GQ field within the sample-level FORMAT information. A higher QUAL score indicates stronger evidence for the variant under the caller’s model, but it does not guarantee that the variant is biologically meaningful or free from technical artifacts. Similarly, a high GQ indicates greater confidence in a particular sample’s genotype call, not necessarily greater confidence in the functional effect of the variant.
- The FILTER column records the result of variant-level filtering. A value of PASS indicates that the record passed the filters represented in the file. Other values may identify filters that were not passed, such as low quality, insufficient read depth, strand bias, or poor mapping quality. A period indicates that no filter status has been provided. The filter names and their meanings should be defined in the metadata header, allowing downstream users to understand how the variants were evaluated.
- The INFO field contains site-level information that applies to the variant record. Common annotations include DP, representing total read depth; AF, representing an allele-frequency estimate; AC, representing the number of observed alternative alleles; AN, representing the total number of alleles considered; and MQ, representing mapping-quality information. Other fields may describe strand bias, read-position bias, variant type, population frequency, or functional annotation. The exact meaning of each field is defined by its corresponding metadata declaration in the VCF header. Researchers should not assume that similarly named fields are calculated identically by every variant caller or annotation tool.
- The FORMAT field defines the structure of the sample-specific data stored in subsequent columns. The GT field represents the genotype, while other common fields include DP for read depth, AD for allelic depths, GQ for genotype quality, and PL for normalized Phred-scaled genotype likelihoods. For a diploid organism, 0/0 commonly represents a homozygous reference genotype, 0/1 a heterozygous genotype, and 1/1 a homozygous alternative genotype. A vertical bar, as in 0|1, indicates a phased genotype, whereas a forward slash, as in 0/1, indicates an unphased genotype. Phasing provides information about which alleles occur together on the same chromosome, which can be important for haplotype analysis and disease interpretation.
- VCF also supports missing genotype information, commonly represented by a period. For example, ./. indicates a missing diploid genotype. Missingness can arise from insufficient sequencing coverage, poor mapping, ambiguous evidence, or quality-control filtering. In organisms with different ploidy levels, genotype representations must be interpreted according to the biological context and the sample’s chromosome complement.
- VCF metadata is essential for understanding the contents of a file. The ##fileformat line identifies the VCF specification version, while ##reference may identify the reference genome used during variant calling. Metadata declarations for INFO, FORMAT, FILTER, and other fields describe their identifiers, data types, expected numbers of values, and meanings. The header may also contain information about contig names and lengths, sample descriptions, software versions, and annotation sources. Because VCF records are interpreted relative to a reference sequence, the reference assembly and contig naming conventions must be documented and consistent throughout the analysis.
- The VCF format is used for several types of genetic variation. Small variants include SNVs and short insertions and deletions. Structural variants may include deletions, duplications, inversions, translocations, and complex rearrangements. Symbolic alternative alleles, such as <DEL> or <DUP>, can be used to represent larger events that are not conveniently described as short nucleotide strings. Structural variant records may include additional information such as confidence intervals, event identifiers, affected lengths, and breakend descriptions. The representation of complex structural variation can be more complicated than that of simple SNVs, and the conventions used by callers should be checked carefully.
- VCF is also used to store variant calls from single-sample and multisample analyses. A single-sample VCF contains variant information for one individual, while a multisample VCF contains genotype data for multiple individuals at shared or jointly represented variant sites. Multisample files are useful in population genetics, genome-wide association studies, family-based analysis, genetic diversity assessment, and cohort-level disease research. They allow researchers to compare allele frequencies, genotype distributions, missingness, and patterns of genetic variation across samples.
- The Binary Call Format (BCF) is a binary representation designed to store variant information corresponding to VCF. It retains the main conceptual structure of VCF, including genomic coordinates, alleles, quality information, filters, annotations, and sample genotypes. Unlike VCF, however, BCF encodes the data in binary form rather than storing it as readable tab-delimited text. This reduces the need to repeatedly parse text and can improve computational efficiency when working with large datasets.
- BCF is particularly useful when processing large multisample variant datasets or performing repeated operations such as filtering, sorting, merging, querying, and genotype extraction. Binary encoding can reduce file size and improve processing speed, although the actual performance depends on the dataset, compression, storage system, and software implementation. BCF is commonly handled using BCFtools and HTSlib, which provide tools and libraries for reading, writing, and manipulating both VCF and BCF files.
- VCF and BCF are closely related, and many tools can convert between them. BCFtools is one of the most widely used command-line utilities for variant-file manipulation. The view command can convert a VCF file into compressed VCF or BCF, and it can also convert BCF back into readable VCF. For example, the following commands demonstrate common conversions:
# Convert an uncompressed VCF file to compressed VCF
bcftools view -Oz -o variants.vcf.gz variants.vcf
# Convert a VCF file to BCF
bcftools view -Ob -o variants.bcf variants.vcf
# Convert BCF back to readable VCF
bcftools view -Ov -o variants_from_bcf.vcf variants.bcf
# Display a BCF file as VCF on the terminal
bcftools view variants.bcf
- Compressed VCF files commonly use the .vcf.gz extension and are generally compressed using BGZF, a block-based form of gzip compression that supports efficient access to genomic regions when combined with an appropriate index. Such files are not identical to ordinary gzip-compressed text files for random-access purposes. A compressed VCF can be indexed using tools such as tabix or bcftools index, provided that it is correctly sorted and uses a supported compression format. BCF files can also be indexed for rapid region-based retrieval.
- Indexing is important when working with large variant datasets because it allows software to retrieve records from selected genomic regions without reading the entire file sequentially. For example:
# Index a BGZF-compressed VCF file
tabix -p vcf variants.vcf.gz
# Index a BCF file
bcftools index variants.bcf
# Retrieve variants from a chromosome region
bcftools view variants.vcf.gz chr1:100000-200000
# Retrieve variants from the same region in a BCF file
bcftools view variants.bcf chr1:100000-200000
- The region syntax uses genomic coordinates and contig names that must match those in the file. Some datasets use chromosome names such as 1, while others use names such as chr1. Such naming differences can prevent region queries and downstream comparisons from working correctly. Indexes should be regenerated whenever the underlying data file changes.
- BCFtools provides many additional operations for manipulating variant files. The query command extracts selected fields, while filter applies expressions to retain or exclude records. The norm command can left-align and normalize variants, split multiallelic records, and check reference alleles when supplied with the appropriate reference genome. The merge command combines compatible files, and concat joins files containing different genomic intervals or related sets of records. These operations are useful for preparing data for population-level analysis, annotation, and variant interpretation.
- The following examples illustrate common VCF and BCF processing tasks:
# Display chromosome, position, reference and alternative alleles
bcftools query -f ‘%CHROM\t%POS\t%REF\t%ALT\n’ variants.vcf.gz
# Retain variants that passed the filters
bcftools view -f PASS -Oz -o variants_pass.vcf.gz variants.vcf.gz
# Retain variants with a QUAL score of at least 30
bcftools filter -i ‘QUAL>=30’ -Oz -o high_quality.vcf.gz variants.vcf.gz
# Normalize variants against a reference genome
bcftools norm -f reference.fa -Oz -o normalized.vcf.gz variants.vcf.gz
# Split multiallelic records into separate records
bcftools norm -m -any -Oz -o split_variants.vcf.gz variants.vcf.gz
# Create an index for the normalized file
bcftools index normalized.vcf.gz
- Filtering expressions must be selected according to the study design, sequencing technology, variant caller, and biological question. A universal QUAL threshold is not appropriate for every dataset. Researchers should also distinguish between site-level filtering and sample-level genotype filtering. A variant site may be reliable overall while the genotype of one particular sample is uncertain. Conversely, a sample genotype may appear well supported even when the site itself has problematic mapping or systematic sequencing artifacts.
- VCF and BCF files are central to variant annotation and functional interpretation. After variant calling, records may be annotated with information about affected genes, transcripts, coding consequences, protein changes, regulatory regions, conservation, and population frequencies. Tools such as Ensembl Variant Effect Predictor (VEP), SnpEff, and ANNOVAR can add functional annotations. Variant records may also be compared with resources such as dbSNP, ClinVar, gnomAD, ClinGen, and disease-specific databases. Annotation depends on the reference genome, transcript set, database release, and software version, so these details should be retained for reproducibility.
- VCF and BCF are also used in genome-wide association studies and population-genetic analyses. Variant genotypes can be converted into formats used by programs such as PLINK and PLINK 2 for calculating allele frequencies, linkage disequilibrium, population structure, relatedness, and association statistics. Multisample variant files can support quality control, sample comparison, genotype imputation preparation, and haplotype analysis. However, VCF and BCF are general-purpose variant formats rather than complete analysis-ready representations for every statistical method. Some applications require additional transformations, metadata, or specialized formats.
- Several software packages support VCF and BCF. BCFtools and HTSlib are fundamental tools for reading, writing, indexing, filtering, and querying these formats. SAMtools supports related sequence and alignment workflows. GATK generates and processes VCF-based variant data in many germline and somatic workflows. FreeBayes, DeepVariant, and other variant callers produce VCF outputs, while annotation tools such as VEP and SnpEff add functional information. PLINK and PLINK 2 support downstream population and association analyses, often after conversion from VCF. The choice of software depends on the type of variant, sequencing technology, organism, and research objective.
- Despite their wide use, VCF and BCF have limitations. VCF is human-readable, but large multisample files can become very large and slow to parse. BCF is more computationally efficient for many operations, but it is not directly readable in a text editor and requires compatible software. Both formats depend on accurate reference genome information and consistent contig naming. Incorrect reference assemblies, coordinate mismatches, inconsistent allele representation, and incomplete metadata can lead to incorrect comparisons or downstream interpretation.
- Variant normalization is particularly important when comparing variant datasets. The same biological variant may be represented in more than one way, especially in repetitive regions or when dealing with insertions and deletions. Left alignment and normalization help produce consistent representations, making it easier to compare records, merge datasets, and identify matching variants. Multiallelic representation and complex variants also require careful handling. Normalization should be performed with the correct reference genome, and the resulting records should be checked to ensure that no biologically relevant information has been lost.
- VCF and BCF files must also be interpreted with attention to sample identity, ploidy, missing data, and genotype conventions. Sex chromosomes, mitochondrial genomes, polyploid organisms, pooled samples, and mixed-ploidy datasets may require special handling. A genotype such as 0/1 should not automatically be interpreted as a standard diploid heterozygote without considering the organism and genomic region. Similarly, allele frequencies calculated from a cohort may be affected by missing genotypes, population structure, relatedness, and sample selection.
- For reproducible research, the original VCF or BCF file should be preserved alongside the reference genome assembly, variant caller, software versions, command-line parameters, sample metadata, filtering criteria, and annotation resources. The header should be retained and inspected because it defines the meaning of many fields. Researchers should document whether variants were normalized, filtered, split into biallelic records, or annotated, and should keep track of any transformations that may alter the original representation. File integrity can be checked using checksums, while validation tools can help detect malformed records, inconsistent headers, and reference mismatches.
- VCF and BCF remain fundamental formats in modern bioinformatics because they provide standardized ways to represent genetic variation and sample genotypes. VCF offers a transparent, text-based structure that is convenient for inspection, exchange, and documentation, while BCF provides a compact binary representation suited to efficient computational workflows. Understanding their structure, metadata, genotype conventions, indexing, normalization, and software support is essential for accurate variant analysis, reliable data exchange, and reproducible genomic research.