vcf-question

[test1]

I am receiving imputation data in vcf format from Michigan server. I want to know if there is any functionality of plink program on vcf quality measures that can be correspond to mach-rsq value (or info score in impute2)?

could it be --vcf-min-gp or --vcf-min-gq or --vcf-filter? and if yes what is the recommended threshold?

thanks

[Christopher Chang]

plink 1.x does not have a good way to compute mach-rsq (technically you can export a .gen file and then use --dosage, but that's horribly inefficient), but this will be included in v2.0.

[test1]

alright.....thank you Chris...hopefully for next new year!

[Colm O'Dushlaine]

Yeah, I have a similar question. It's easy enough to parse the info file to select e.g. Rsq > 0.3. I want to play with hard genotype calls also, however. Does something like this make sense? :

plink --vcf chr1.dose.vcf.gz --vcf-min-gp 0.3 ...

Colm

[Christopher Chang]

I will try to enable this by next week.

[Colm O'Dushlaine]

Thanks Chris!

What I've been doing us using DosageConvertor (http://genome.sph.umich.edu/wiki/DosageConvertor) to convert to PLINK dosage format, then parsing the "chr1.info.gz" that accompanies the "chr1.dose.vcf.gz" file from the imputation server to choose what SNPs to keep in the analysis (using Rsq>0.3 for example). I think if PLINK could read in '--minimac3' format or something like that, and be able to use the Rsq in the info file at the same time as being able to run dosage-based association, that would be fantastic. Essentially, if we could bypass the conversion step we could speed up a lot of the analyses (and save space!). 

Best wishes,

Colm 

[Colm O'Dushlaine]

This solution also seems to work well: https://www.biostars.org/p/205856/