VCF (variant call format) parser

npm install @gmod/vcf

import { TabixIndexedFile } from '@gmod/tabix'
import VCF, { type Variant } from '@gmod/vcf'

const tbiIndexed = new TabixIndexedFile({ path: '/path/to/my.vcf.gz' })

const headerText = await tbiIndexed.getHeader()
const parser = new VCF({ header: headerText })

const variants: Variant[] = []
await tbiIndexed.getLines('ctgA', 200, 300, line => {
  variants.push(parser.parseLine(line))
})

Reuse one parser for all lines — the header is parsed once per VCF.

strict (default true) makes parseLine throw on a line with no INFO column; the VCF spec requires at least a . there.

parseLine(line) returns a Variant:

{
  CHROM: 'contigA',
  POS: 3000,
  ID: ['rs17883296'],  // undefined if '.'
  REF: 'G',
  ALT: ['T', 'A'],     // undefined if '.'
  QUAL: 100,           // undefined if '.'
  FILTER: 'PASS',      // 'PASS' | string[] of filter names | undefined if '.'
  FORMAT: 'GT:DP',     // undefined if the file has no samples
  INFO: {
    NS: [3],
    DP: [14],
    AF: [0.5],
    DB: true,   // Type=Flag, a bare key, or KEY= with an empty value
    XYZ: ['5'], // unknown fields default to Number=1, Type=String
  },
}

INFO and FORMAT values are typed using header metadata. Values are arrays unless Type=Flag, in which case they are true. Fields defined in the VCF spec are typed even without a header entry. . inside a value becomes undefined, and %XX escapes are percent-decoded.

JSON.stringify(variant) serializes the columns above only — sample data is not included.

Sample data is not touched until one of these is called, so lines from a many-sample file are cheap to parse if you only need the columns above.

• variant.SAMPLES() — all FORMAT fields, keyed by sample name
• variant.GENOTYPES() — GT strings only, keyed by sample name
• variant.processGenotypes(callback) — GT positions only, no allocation

let homRef = 0
variant.processGenotypes((str, start, end, sampleIdx) => {
  if (
    end - start === 3 && // e.g. "0|0"
    str.charCodeAt(start) === 48 && // 48 = '0'
    str.charCodeAt(start + 2) === 48
  ) {
    homRef++
  }
})

For files with many samples:

• SAMPLES() re-parses on every call. Call it once and keep the result.
• processGenotypes ignores the callback's return value, so there is no early exit — it always visits every sample.
• GENOTYPES() returns a null-prototype object; use Object.keys / in, not hasOwnProperty.
• INFO is parsed eagerly for every line, unlike sample data. Files with large INFO columns pay that cost even if you only read CHROM/POS.
• A retained Variant holds a reference to its whole input line. Streaming is fine, but collecting variants from a many-sample file keeps every line in memory.

parser.getMetadata(...keys) returns header metadata, filtered by the keys provided:

parser.getMetadata('INFO', 'DP')
// { Number: 1, Type: 'Integer', Description: 'combined depth across samples' }

parser.getMetadata('INFO', 'DP', 'Number')
// 1

Call with no arguments to get all metadata. parser.samples lists sample names.

Without tabix, collect header lines until the first non-header line, then construct the parser:

import fs from 'node:fs'
import readline from 'node:readline'
import VCF from '@gmod/vcf'

const rl = readline.createInterface({
  input: fs.createReadStream('file.vcf'),
})

const header: string[] = []
let parser: VCF | undefined

for await (const line of rl) {
  if (line.startsWith('#')) {
    header.push(line)
  } else {
    parser ??= new VCF({ header: header.join('\n') })
    const variant = parser.parseLine(line)
    console.log(variant.CHROM, variant.POS)
  }
}

For .vcf.gz, pipe the read stream through createGunzip() from node:zlib.

parseBreakend(alt) parses a breakend ALT string, returning undefined for anything that isn't one:

import { parseBreakend } from '@gmod/vcf'

parseBreakend('C[2:321682[')
// { MateDirection: 'right', Replacement: 'C', MatePosition: '2:321682', Join: 'right' }

All four bracket forms from the VCF spec:

• Join — whether the replacement base appears before (right) or after (left) the mate position
• MateDirection — [ means the mate sequence extends rightward; ] means leftward

When the ALT starts or ends with ., the result has SingleBreakend: true and no MatePosition:

parseBreakend('C.')
// { Join: 'right', Replacement: 'C', SingleBreakend: true }

parseBreakend('.ACGT')
// { Join: 'left', Replacement: 'ACGT', SingleBreakend: true }

An ALT containing <...> uses the symbol as the mate contig:

parseBreakend('<DUP>ACGT')
// { Join: 'left', Replacement: 'ACGT', MateDirection: 'right', MatePosition: '<DUP>:1' }

parseBreakend('ACGT<DUP>')
// { Join: 'right', Replacement: 'ACGT', MateDirection: 'right', MatePosition: '<DUP>:1' }

default (the parser), Variant, parseBreakend, and the types Breakend, Samples, SampleData, SampleValue, InfoValue, MetaMap, MetaField, GenotypeCallback.

See CONTRIBUTING.md.