converting genotypes to missing in different samples in different site

[mgrzy]

Hi,

I am trying to calculate imputation accuracy. For that purpose I want to mask some small fraction of know genotypes, impute and then compare imputed set with true values. For example, suppose I have 5 samples: A, B, C, D and E and I would like to mask (convert to missing) first site (variant) in sample A and B, then fifth site in D and E, and so on... I would like to keep genotypes for rest sample in given site, so if in site one A and B will be masked, I would like to keep genotypes for C,D,E.

Is there any clever way to do it with bcftools?

Best,
Marcin

[pd3]

Currently there is no such clever way. However, it would be straightforward to extend the +setGT plugin to support a new mode that randomly selects a proportion of genotypes. It could work like this

# set 10% of genotypes to missing
bcftools +setGT in.vcf -- -t r:0.1 -n .

[mgrzy]

Thank you for quick response. I will deal with it in other way in that case.
It will be great if it will be possible to add such option in future build of bcftools.

[pd3]

This is now supported