How To Filter Vcf By Coverage?

Question

[newDNASeqer]

I used GATK to make variant calling on an exome-seq data set that has 10 samples. I copied one line (for an SNP identified) from the VCF output as shown below. The genotype is indicated (for example 0/1 for heterozygous mutant). The reference and variant reads are shown immediately after the genotype 0/1 (for example).

My question is how do you filter the VCF by the coverage (ref reads + variant reads ?). What to do if some of the samples pass the filtration and other samples fail to pass?

0/1:2,10:12:28:256,0,28    0/1:13,18:31:99:427,0,315    0/1:6,9:15:99:246,0,155    0/1:8,8:16:99:176,0,187    0/0:8,0:8:24:0,24,259    0/1:5,6:11:99:144,0,136    0/1:5,5:10:99:103,0,110    0/1:4,6:10:99:161,0,103    0/0:10,0:10:30:0,30,277    0/1:16,7:23:99:192,0,529

[aadeokar]

Hi Pierre, Could you please inform me how i can filter my multi-sample vcf file (250 samples) for only variants with at five homozygous REF and five homozygous ALT call using VCFFilterJS. Thank you !

[Pierre Lindenbaum]

this is a new question, please open this as a new question: https://www.biostars.org/p/new/post/

[aadeokar]

Could you please inform me how i can filter my multi-sample vcf (250 samples) file for only variants with at least five homozygous REF and five homozygous ALT call using VCFFilterJS.

[Pierre Lindenbaum]

again, I'll give you the answer, but please, askthis a different question ( https://www.biostars.org/p/new/post/ ) (it's not How To Filter Vcf By Coverage? ). Ask your question as a NEW question. So everybody can contribute and follow the new thread. Thanks.

Answer 1

[Raony Guimarães]

You could use the parameter "--max-indv-meanDP" from vcftools.

--max-indv-meanDP <float> Calculate the mean coverage on a per-individual basis. Only individuals with coverage within the range specified by these options are included in subsequent analyses.

Source: http://manpages.ubuntu.com/manpages/trusty/man1/vcftools.1.html

[newDNASeqer]

thanks, but I found the latest version of vcftools does not support --max-indv-meanDP http://vcftools.sourceforge.net/man_latest.html

Answer 2

[Pierre Lindenbaum]

I wrote a tool VCFFilterJS to filter a VCF file with a javascript program/expression.

See https://github.com/lindenb/jvarkit/wiki/VCFFilterJS and A: GATK multi-sample VCF VariantFiltration for an example.