Is it possible to output compressed VCF files with --recode vcf?

[freeseek]

"plink --bfile binary_fileset --recode vcf" will convert the dataset to VCF format. This could be in principle handy to perform analyses with softwares like Beagle. Is it possible to output the file compressed?

[Christopher Chang]

This is not currently supported.  It would not be difficult to add a 'gz' modifier to some --recode operations, and I probably will do this for VCF soon, but I worry about where to draw the line.

[freeseek]

You have a good point. I feel bad asking. It would be nice if for outputs that generate only one file (like VCF, but this applies more in general to other kind of analyses) plink had a more general option to write the output file to /dev/stdout and the log file to /dev/stderr. This way there would be no need to encode all these extra features in plink. You could have it work with something like this:

plink --bfile binary_fileset --recode vcf --stdout 2>output.log | bgzip > output.gz

[Christopher Chang]

Yes, that's the clean Unix way of doing things.  For better or worse, though, PLINK is designed to generate more than one relevant output file at a time, so there is limited value in retrofitting it to be able to dump a single file to stdout.

So I will just target the common gzip cases.  VCF output is perhaps the most common case of all, so the February 6 development build has a basic "--recode vcf gz" implementation.  I will try to speed it up (on multicore systems, at least) over the weekend.

On Thursday, February 5, 2015 at 2:46:48 PM UTC-8, freeseek wrote:

[freeseek]

Though I forgot to stress the fact that VCF files should be bgzipped rather than gzipped, otherwise most programs will not handle them.

[Christopher Chang]

Hmm, that's true; this will probably be changed to "--recode vcf bgz" with the corresponding functionality.

[Christopher Chang]

"--recode vcf bgz" should be working in the February 9 development build.

[Florian Privé]

Thanks for this.