Reading VCF Genotype Fields

Introduction

A VCF record can look like a long string of codes, but its structure is regular. The first columns describe a variant, while optional sample columns contain genotype information organized by keys listed in FORMAT. Reading those parts in the correct order prevents a common error: treating a value as if its meaning were fixed by its position in every VCF file.

VCF is a tab-delimited text format. Its header describes the file and names the columns; each data row describes one variant record. The precise definitions in a file's header matter, especially for FORMAT fields and fields whose values are represented by numbers. The examples below are hypothetical and use simplified records for teaching.

Read the Header Before the Records

VCF headers commonly begin with metadata lines that start with ##. These can specify the file format version and define INFO, FILTER, and FORMAT fields. A field declaration can include an ID, a Number, a Type, and a Description. The declaration for a field is the place to check its intended meaning and value structure; the short label alone may not provide every detail.

The final header line starts with #CHROM and names the columns. The first eight columns are fixed: CHROM, POS, ID, REF, ALT, QUAL, FILTER, and INFO. If sample data are included, a FORMAT column follows them, and the remaining columns give sample names. Each sample value in a data row corresponds to the sample name in the same column of the header.

A VCF with sample columns therefore has at least nine columns. The sample columns are not additional variant annotations: each contains values for one sample at that record. In a multi-sample file, keep the header's sample order in view while reading a row.

The Fixed Variant Columns

The fixed columns describe the location and alleles of the record, along with record-level annotations. CHROM identifies the chromosome or contig, and POS gives the position using the file's coordinate convention. ID contains an identifier when available, while a dot means that no value is supplied there. REF gives the reference allele, and ALT lists one or more alternate alleles. If more than one alternate allele is present, they are comma-separated.

QUAL is a record-level quality value; it is not the same field as an individual sample's genotype quality. FILTER records filtering status: it contains PASS if all applied filters passed, filter IDs for filters the record failed, or . if filters have not been applied. INFO contains record-level annotations, typically written as semicolon-separated entries. Some INFO entries are flags, while others have a value after an equals sign; consult their header declarations for definitions.

A dot is not a universal statement about why information is absent. Its meaning depends on the column or field where it appears. In particular, FILTER set to a dot says filters have not been applied; it does not mean the record failed a filter or that a filter result is merely unavailable.

GT: Allele Indices and Separators

GT is the genotype field. Its allele numbers refer to the alleles listed for that record: 0 refers to REF, 1 to the first ALT allele, 2 to the second ALT allele, and so on. A diploid biallelic GT of 0/1 therefore represents one REF allele and one ALT allele. A genotype may have a different number of allele entries when the record represents a different ploidy.

A slash separates unphased alleles, as in 0/1. A vertical bar separates phased alleles, as in 0|1. The separator communicates whether allele order is represented as phased; the two forms should not be treated as identical notation. A phased order by itself does not identify a parent of origin.

For a biallelic diploid record, an ALT-allele count can be read directly from GT by counting the number of allele indices equal to 1. For example, in 0/1 the count is \(0+1=1\), and in 1/1 it is \(1+1=2\). This simple count is specific to the biallelic example; with multiple ALT alleles, the index identifies which alternate allele is present.

DP, GQ, and AD

These FORMAT fields summarize different aspects of sample-level data. DP is the sample's read depth as represented by the caller. It describes read support at the site, not the genotype category itself. GQ is a phred-scaled measure of confidence in the called genotype. A larger GQ indicates stronger genotype-call confidence on that scale, but GQ should not be confused with the record-level QUAL value.

On the phred scale, an error probability \(e\) is represented by \(Q=-10\log_{10}(e)\). Equivalently, \(e=10^{-Q/10}\). For example, a GQ of 30 corresponds to \(10^{-30/10}=10^{-3}=0.001\), or one-thousandth, on that phred interpretation. Caller conventions and caps can affect reported values, so the field definition and caller documentation remain important.

AD gives allele depths, generally one value for REF followed by one value for each ALT allele. The header declaration commonly identifies this structure with Number=R, meaning one value for each allele including REF. In a biallelic record, AD such as 12,7 gives a REF-associated count of 12 and an ALT-associated count of 7. DP and AD are not interchangeable: DP is a depth field, while AD allocates depth by allele according to the caller's definition.

Missing Values and Field Order

The dot character is used to indicate a missing value in many VCF fields. In a diploid GT, ./. indicates that both allele entries are missing; 0/. indicates that one allele is specified and the other is missing. For phased notation, .|1 likewise has one missing allele and one specified allele. These are not complete genotype calls.

Other FORMAT values can be missing independently. For example, a sample can have a known GT but a dot for GQ, or a known DP but missing AD. Read the position of the dot against the FORMAT keys rather than assuming that the whole sample is missing. A sample field written as a single dot can indicate that sample data are missing for that record; always use the delimiters and keys to parse what is actually present.

FORMAT order can change from one record to another. A sample value with the same text in a different position can consequently mean something different. For example, GT:GQ:DP paired with 0/1:45:20 gives GQ 45 and DP 20, while GT:DP:GQ paired with 0/1:45:20 gives DP 45 and GQ 20. The keys, not a remembered customary order, determine the interpretation.

A Reliable Parsing Sequence

A brief, repeatable sequence helps prevent column shifts and mistaken interpretations. Use the record's own FORMAT keys for every sample, and consult header definitions when field structure or caller-specific wording is uncertain.

Common Mistakes

• Assuming FORMAT keys always have one fixed order. The wrong approach is to treat the first numeric value after GT as DP in every file. The correction is to match every sample value to the key in that record's FORMAT column.
• Reading a GT index without checking ALT order. The wrong approach is to treat index 1 as any alternate allele. It refers specifically to the first ALT listed; index 2 refers to the second ALT.
• Confusing QUAL and GQ. QUAL is in a fixed, record-level column; GQ is a FORMAT field for a sample. Their values describe different levels of the record.
• Treating every dot as the same status. A dot in FILTER means filters have not been applied. A dot in a sample's GQ means that GQ is missing, and a dot in ID means no ID value is supplied.
• Calling a partly missing genotype complete. A value such as 0/. has one specified allele and one missing allele; it does not provide a fully specified diploid genotype.