I've been trying without success to phase the variants of about 30 samples in a 5Mb non-recombining region. I have a phased genome assembly of the region but so far I have not been successful in using that to inform the phasing of the samples which have been sequenced to high coverage with short reads. Non-reference panel phasing using Beagle 5.2 and Shapeit4 have produced haplotypes that are composed of many small phased blocks which switch phase between them, far to many to manually correct.

I have created a "reference panel" of SNPs of known phase for the region and want to use it to guide the rest of the phasing of the non-"reference panel" SNPs but every method I have found only uses the intersection of the reference panel and the samples for analysis. Is there some technique that I can use to phase the genotypes in the samples? I believe this is a tractable problem since recombination is suppressed in this region so haplotype blocks should be large and contiguous.

  • $\begingroup$ I'm unsure what you mean by "but every method I have found only uses the intersection of the reference panel and the samples for analysis". Do you mean the intersection of SNPs or individuals? In either case, I don't see why that would be an issue. Also, phasing with 30 individuals without a reference panel will give very poor results at best. $\endgroup$
    – user438383
    Nov 3, 2021 at 11:41
  • $\begingroup$ Thanks for the reply. My goal was to use the reference panel as a guide to phase the genotypes between the reference panel SNPs. A sort of "reverse imputation". Even though the sample size is small my thought was that since the region in question does not recombine (it is a homozygous lethal inversion) that the nearly complete linkage would make this a workable problem. $\endgroup$
    – Jason H
    Nov 3, 2021 at 12:23

1 Answer 1


The fact that the reference assembly is phased doesn't do anything to help actually phase the 30 other samples. There is no way of knowing what the combination of variants on single haplotypes are in the 30 samples from the one reference genome + 30-sample VCF alone. With this info, it is possible to come up with a measure of "HaplotypeA from sample 1 is more similar to HaplotypeII than HaplotypeI from my phased reference", though.

To actually phase the 30 other samples it will take long reads and/or proximity ligation data to phase the variants over a 5Mb region. Hi-C, Chicago, or DNAse Hi-C/Dovetail Omni-C reads will be necessary to correctly phase all of the variants over a region that is as large as 5Mbp.

I recommend using the tool HapCUT2 to use the reads to phase the variants already called for each sample. When mapping the reads to your genome, only map the reads to a fasta file that contains a single haploid copy of the genome (2 homologous chromosomes/haplotypes represented by 1 scaffold), not a fasta file that contains both haplotypes from the phased assembly (2 homologous chromosomes/haplotypes represented by 2 scaffolds).


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.