Vcftools: Filtering By Multiple Regions (--Positions Flag?)

Question

[Matt W]

Is there a way to filter multiple positions from my VCF files? I am trying to use vcftools which basically gives me two different options.

1. --chr $(chrom) --from-bp $(start) --to-bp $(stop)

   The problem with this approach is I need multiple regions. So do I just reuse these flags multiple times? Specifically, there are 2192 regions I would like to extract.

2. --positions pos.txt

   According to the docs, the input file requires a "chromosome and position", but I need multiple regions. This would work if I could specify regions.

Am I misinterpreting how to use these flags? Or is there an easier way to extract multiple regions from VCF files?

Thanks!

Answer 1

[Pierre Lindenbaum]

use bedtools instead of vcftools : see http://bedtools.readthedocs.org/en/latest/content/tools/intersect.html with VCF/BED

[Tysler]

Code works for me :) bedtools v2.25.0

bedtools intersect -a myfile.vcf.gz -b myref.bed -header > output.vcf

[nhaituan]

Thank you very much, it works for me.

[Matt W]

I don't actually have a second input file. I only have a list of regions that I would like to extract. Does bedtools support an input that isn't BED/GFF/VCF?

[Pierre Lindenbaum]

" I only have a list of regions": means you have a BED (chrom/chromStart/chromEnd) https://genome.ucsc.edu/FAQ/FAQformat.html#format1

[Matt W]

Ah, silly question. Thanks for the reply. I should have read the docs before making an assumption about the format. Thanks!

Answer 2

[Erik Garrison]

vcfintersect in vcflib will do this.

vcfintersect -b regions.bed variants.vcf

You can also use another VCF file, but you'll need a reference (it checks the haplotypes to be sure that alleles are the same even if they are aligned differently).

vcfintersect -f ref.fa -i known.vcf new.vcf >results.vcf

Note that intersecting variants will remove alleles which don't overlap even if they are at the same position as variants which do. The records are all adjusted to reflect the fact that an allele has been removed to maintain semantic consistency in the file. Specifically, all Number=A and Number=G fields in INFO and in the sample fields are adjusted.

Answer 3

[Adam]

You could use the --bed option in vcftools (or use bedtools as Pierre suggests).