Subset a multisample VCF file

Question

I need some help with GATK. I have a multisample vcf file containing 18125 samples. After performing PCA I have around 303 samples as outliers and I wanted to remove these samples from my multisample vcf file.

I tried doing it using the GATK SelectVariant tool. It does some stuff and generates a VCF file which is a bit smaller in size than the original file, but in the end the total count of samples remains the same. As I have no prior knowledge of GATK (working with it for the first time) I am just wondering if you guys can help me out in fixing this issue.

The command that I used was

gatk SelectVariants -V filtered.vcf -xl-sn remove.txt -O test.vcf 

remove.txt contains the sample names in each row (1 sample per row).

After running this command it generated a test.vcf file, for which I counted samples using:

bcftools query -l test.vcf | wc -l

However, I got exactly same number of samples as in the original VCF file filtered.vcf.

Answer 1

Based on the way you've called the command (using -xl-sn), I assume you are using GATK 4 and not GATK 3.

If this is the case, the documentation for the -xl-sn flag is available here: https://software.broadinstitute.org/gatk/documentation/tooldocs/current/org_broadinstitute_hellbender_tools_walkers_variantutils_SelectVariants.php#--exclude-sample-name.

By default, the -xl-sn SAMPLE means "exclude the sample called SAMPLE from my VCF". In your case, you're providing an entire list of samples, in a file, which is not the default behaviour. In order to tell GATK that you're providing a file and not a sample name, you need to ensure the file you provide has the extension .args.

From the documentation:

File names must use the extension ".args"