VariantAnnotation (version 1.18.5)

genotypeToSnpMatrix: Convert genotype calls from a VCF file to a SnpMatrix object

Description

Convert an array of genotype calls from the "GT", "GP", "GL" or "PL" FORMAT field of a VCF file to a SnpMatrix.

Usage

"genotypeToSnpMatrix"(x, uncertain=FALSE, ...)
"genotypeToSnpMatrix"(x, ref, alt, ...)

Arguments

x

A CollapsedVCF object or a array of genotype data from the "GT", "GP", "GL" or "PL" FORMAT field of a VCF file. This array is created with a call to readVcf and can be accessed with geno().

uncertain

A logical indicating whether the genotypes to convert should come from the "GT" field (uncertain=FALSE) or the "GP", "GL" or "PL" field (uncertain=TRUE).

ref

A DNAStringSet of reference alleles.

alt

A DNAStringSetList of alternate alleles.

...

Additional arguments, passed to methods.

Value

A list with the following elements,

genotypes

The output genotype data as an object of class "SnpMatrix". The columns are snps and the rows are the samples. See ?SnpMatrix details of the class structure.

map

A DataFrame giving the snp names and alleles at each locus. The ignore column indicates which variants were set to NA (see NA criteria in 'details' section).

Details

genotypeToSnpMatrix converts an array of genotype calls from the "GT", "GP", "GL" or "PL" FORMAT field of a VCF file into a SnpMatrix. The following caveats apply,

• no distinction is made between phased and unphased genotypes
• variants with >1 ALT allele are set to NA
• only single nucleotide variants are included; others are set to NA
• only diploid calls are included; others are set to NA

In VCF files, 0 represents the reference allele and integers greater than 0 represent the alternate alleles (i.e., 2, 3, 4 would indicate the 2nd, 3rd or 4th allele in the ALT field for a particular variant). This function only supports variants with a single alternate allele and therefore the alternate values will always be 1. Genotypes are stored in the SnpMatrix as 0, 1, 2 or 3 where 0 = missing, 1 = "0/0", 2 = "0/1" or "1/0" and 3 = "1/1". In SnpMatrix terminology, "A" is the reference allele and "B" is the risk allele. Equivalent statements to those made with 0 and 1 allele values would be 0 = missing, 1 = "A/A", 2 = "A/B" or "B/A" and 3 = "B/B". The genotype fields are defined as follows:

• GT : genotype, encoded as allele values separated by either of "/" or "|". The allele values are 0 for the reference allele and 1 for the alternate allele.
• GL : genotype likelihoods comprised of comma separated floating point log10-scaled likelihoods for all possible genotypes. In the case of a reference allele A and a single alternate allele B, the likelihoods will be ordered "A/A", "A/B", "B/B".
• PL : the phred-scaled genotype likelihoods rounded to the closest integer. The ordering of values is the same as for the GL field.
• GP : the phred-scaled genotype posterior probabilities for all possible genotypes; intended to store imputed genotype probabilities. The ordering of values is the same as for the GL field.

If uncertain=TRUE, the posterior probabilities of the three genotypes ("A/A", "A/B", "B/B") are encoded (approximately) as byte values. This encoding allows uncertain genotypes to be used in snpStats functions, which in some cases may be more appropriate than using only the called genotypes. The byte encoding conserves memory by allowing the uncertain genotypes to be stored in a two-dimensional raw matrix. See the snpStats documentation for more details.

References

http://www.1000genomes.org/wiki/Analysis/Variant%20Call%20Format/vcf-variant-call-format-version-41

See Also

readVcf, VCF, SnpMatrix

Examples

  ## ----------------------------------------------------------------
  ## Non-probability based snp encoding using "GT"
  ## ----------------------------------------------------------------
  fl <- system.file("extdata", "ex2.vcf", package="VariantAnnotation") 
  vcf <- readVcf(fl, "hg19")

  ## This file has no "GL" or "GP" field so we use "GT".
  geno(vcf)

  ## Convert the "GT" FORMAT field to a SnpMatrix.
  mat <- genotypeToSnpMatrix(vcf)

  ## The result is a list of length 2.
  names(mat)

  ## Compare coding in the VCF file to the SnpMatrix.
  geno(vcf)$GT
  t(as(mat$genotype, "character"))

  ## The 'ignore' column in 'map' indicates which variants 
  ## were set to NA. Variant rs6040355 was ignored because 
  ## it has multiple alternate alleles, microsat1 is not a 
  ## snp, and chr20:1230237 has no alternate allele.
  mat$map

  ## ----------------------------------------------------------------
  ## Probability-based encoding using "GL", "PL" or "GP"
  ## ----------------------------------------------------------------
  ## Read a vcf file with a "GL" field.
  fl <- system.file("extdata", "gl_chr1.vcf", package="VariantAnnotation") 
  vcf <- readVcf(fl, "hg19")
  geno(vcf)

  ## Convert the "GL" FORMAT field to a SnpMatrix
  mat <- genotypeToSnpMatrix(vcf, uncertain=TRUE)

  ## Only 3 of the 9 variants passed the filters.  The
  ## other 6 variants had no alternate alleles.
  mat$map

  ## Compare genotype representations for a subset of
  ## samples in variant rs180734498.
  ## Original called genotype
  geno(vcf)$GT["rs180734498", 14:16]

  ## Original genotype likelihoods
  geno(vcf)$GL["rs180734498", 14:16]

  ## Posterior probability (computed inside genotypeToSnpMatrix)
  GLtoGP(geno(vcf)$GL["rs180734498", 14:16, drop=FALSE])[1,]

  ## SnpMatrix coding.
  t(as(mat$genotype, "character"))["rs180734498", 14:16]
  t(as(mat$genotype, "numeric"))["rs180734498", 14:16]

  ## For samples NA11829 and NA11830, one probability is significantly
  ## higher than the others, so SnpMatrix calls the genotype.  These
  ## calls match the original coding: "0|1" -> "A/B", "0|0" -> "A/A".
  ## Sample NA11831 was originally called as "0|1" but the probability
  ## of "0|0" is only a factor of 3 lower, so SnpMatrix calls it as
  ## "Uncertain" with an appropriate byte-level encoding.