I think Eagle or haploSep will do the job of reconstructing haplotype data, but I'm not sure what the 'standard' or best practice tool to use is.

I have a VCF/PLINK format file of ~8,000 individuals with ~150,000 SNPs each (from across a population with several sub-populations).

I want to use the resulting data to place new samples into one or more populations (e.g. ancestry analysis).

Before I go trying to make one of these tools work / look at accuracy, etc., am I doing it right in the first place?

  • $\begingroup$ I've answered, but you should provide some more information to get a proper answer. In particular, i) the number of individuals, ii) the number of SNPS, iii) what you plan on doing with the haplotypes once you have obtained them. All of these things matter to a different degree. Cheers. $\endgroup$
    – user438383
    Nov 5, 2021 at 16:00
  • $\begingroup$ I've tried to add more detail, but he downstream analysis isn't fixed yet... Each 'sub-population' is inbreeding, and I want to assign an individual to one or a mix of each sub-population... Perhaps I should just be using some other analysis? $\endgroup$
    – Dan Bolser
    Nov 6, 2021 at 16:25
  • 1
    $\begingroup$ That sounds reasonable, but unless you are using a method which requires phased data (e.g. ChromoPainter/fineSTRUCTURE), you don't need to infer the haplotypes. You could use something like ADMIXTURE for that (but you get less power than when using haplotype-based methods). $\endgroup$
    – user438383
    Nov 8, 2021 at 11:57

1 Answer 1


HaploSep uses allele frequency data - I presume you have genotype data.

In that case, I would suggest looking at either Beagle5.2, shapeit4 and Eagle2. I believe they more or less give the same accuracy (Eagle2 is slightly less accurate than the other two) and Beagle 5.2 is faster than the rest. Depending on what you plan on doing with the haplotypes, I don't think that the choice of phasing algorithm will affect downstream results a great deal.


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.