If I'm looking for a specific SNP in my SNP-Chip data and it isn't there, are there any tools that let me quickly impute that SNP from surrounding SNPs rather than running a lengthy 'whole chromosome' imputation job?

If so, roughly how many upstream and downstream SNPs are generally required to predict the haploblock 'accurately'?


1 Answer 1


I would take something like 1.5Mb either side of the SNP of interest (so a 3Mb chunk), subset that out from the target and reference panel and then perform imputation on that chunk alone. I think that size will give you more than enough LD, since LD decays by about 500kB in most human populations by then.

shapeit4 has an option to just impute a single region:

shapeit4 \
    --input unphased.bcf \
    --map chr20.b37.gmap.gz \
    --region chr21:1231213-1413143 \ ## change this bit
    --reference reference.bcf \
    --output phased.bcf
  • $\begingroup$ Thanks, that's clear enough. I'm just looking for a tool to make that easy for me ;-) $\endgroup$
    – Dan Bolser
    Commented Apr 27, 2022 at 15:48
  • 1
    $\begingroup$ @dan see edit, shapeit4 can impute just a particular region $\endgroup$
    – user438383
    Commented Apr 27, 2022 at 15:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.