What is the current standard for imputing missing genotypes between two genotyping panels? I have two populations genotyped using two different panels (A & B), and I would like to impute all the genotypes in population B for those positions on used in panel A.

I've read the examples for impute2, and I think the closest thing to what I am looking for is this example, "Imputation with one unphased reference panel".

Simply put, I want to provide a list of SNPs, some variant file for population B, and haplotype information from 1,000 Genomes and get imputed genotypes for each SNP in the list. Is impute2 the state of the art for this?

  • $\begingroup$ It depends, if are we imputing whole genome, all chromosomes then impute2 is pretty solid. Unphased panel would give a better result, but slow performance. If we are imputing a region I think beagle is better. $\endgroup$
    – zx8754
    Jun 1, 2017 at 7:59
  • $\begingroup$ I don't want to impute the whole genome, just certain specified sites. $\endgroup$
    – Greg
    Jun 1, 2017 at 8:00

1 Answer 1


Given that you mention wanting to use 1000 Genomes as a reference panel for imputing genotypes into your two SNP chip panels, I am going to assume that you are working with human data.

In that case there are several options you can go with:

  • If your two panels are of European descent, then you are probably best off using the HRC reference panel together with a fast genotype imputation tool such as Beagle 4.1 to impute genotypes in each of your two SNP chip panels separately.
  • If your panels are not of European descent, then you will likely want to use the 1000 Genomes phase 3 reference panel with Beagle 4.1, Impute2, or Minimac3.

In either case, there are two phasing services available that will do much of the heavy lifting for you 1,2.

The second Wellcome Trust Case-Control Consortium paper performed a cross-imputation analysis as you describe. I don't see many studies using multiple SNP chip panels. You will need to take care in your analysis that you are not hit by batch effects from using two different SNP chip panels.

Also, none of these methods will work if the region you are imputing into has too few variants. I'm not sure what the minimum number of variants is, but if you are using a whole genome genotyping panel of at least 500k SNPs, then you should be ok if you impute a whole chromosome at a time.

  • $\begingroup$ Thanks! These seem like good options. And yes, I'm working with human data. I don't know exactly what ethnicity the participants are, and I suspect it is likely a diverse population, so 1000 Genomes makes the most sense to me. You mention Beagle a couple times, is there some reason you have a preference for it? $\endgroup$
    – Greg
    Jun 2, 2017 at 16:38
  • $\begingroup$ I don't have enough experience comparing the three programs Beagle 4.1, Impute2, and Minimac3 to really voice a preference. You should get good quality imputation with any of them, but test for your self, which is easy to do by imputing held out genotypes. The only caveat to this is that I think Impute2 will likely take a bit longer than the other programs when imputing from a reference panel the size of the HRC. $\endgroup$
    – winni2k
    Jun 2, 2017 at 18:45
  • 1
    $\begingroup$ Beagle worked great and was easy to use. I highly recommend it $\endgroup$
    – Greg
    Jun 6, 2017 at 18:36

Your Answer

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

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