Count SNPs per individual FAST!

Question

[QVINTVS_FABIVS_MAXIMVS]

Say I have a large (N samples > 2000) VCF or plink bed file.

What's the quickest way to calculate the number of alleles (unique alleles and # ALT genotypes) for each sample?

What are the options that can quickly digest a 1Tb VCF (broken by chrom)?

Plink is ridiculously fast for this, but I don't think it can perform a per-sample count of variants

[Pierre Lindenbaum]

in the title you want to count the number of snp per individual, in the body you want the number of allele for each sample. Please, show us the expected output.

[QVINTVS_FABIVS_MAXIMVS]

Either

IID    #UNIQUE_ALLELES

Or

IID   #UNIQUE_ALLELES    #ALT_ALLELES

so the first file would output ID1 1 for 1/1 genotype but the second file you have ID1 1 2

Answer 1

[chrchang523]

plink --score can be abused for this purpose.

• Create an input file assigning weight 1 to every alt allele to get #ALT_ALLELES.
• You can then repeat the --score computation after erasing all the heterozygous calls ("plink --set-hh-missing --chr-set -26 --make-bed"; it may be necessary to use "--output-chr 26 --make-bed" first to force numeric chromosome codes). You should be able to infer #UNIQUE_ALLELES once you have both --score output files.