0
$\begingroup$

I am new to low-pass whole genome sequencing and have the basic idea of phasing and imputation. I have .vcf file after calling haplotypecaller tools from GATK. After searching the phasing and imputation tools I've encountered some tools like Beagle, Minimac, Shapeit, Glimpse, Eagle. The problem is that I didn't find enough hands-on material on how to process a .vcf from unphased to phased and then to an imputed one.

Can anyone suggests any hands-on tutorial regarding phasing and imputation?

$\endgroup$
5
  • $\begingroup$ What is your sample size, organism, and average coverage per sample? $\endgroup$
    – winni2k
    Jan 17, 2022 at 10:06
  • $\begingroup$ total sequences read 4093666, Human cell, average coverage (0.1x-6x). $\endgroup$ Jan 17, 2022 at 16:12
  • $\begingroup$ ok, here is the tricky bit: are you trying to impute germline or somatic mutations? I have tried beagle with somatic mutations, and it simply doesn't (and shouldn't) work. If you are using germline mutations, what is your reference panel size? $\endgroup$
    – winni2k
    Jan 19, 2022 at 14:47
  • $\begingroup$ I am trying to impute germline. I'm trying to impute it with 1000kg phase3 reference panel. $\endgroup$ Jan 21, 2022 at 6:25
  • $\begingroup$ 1000 genome phase3 reference panel. $\endgroup$ Jan 21, 2022 at 6:41

1 Answer 1

3
$\begingroup$

This is all relevant for data 0.5x-1x coverage.

Assuming you have genotype likelihoods data, if you want to phase low-coverage data, the most suitable options are GLIMPSE, Beagle4 and QUILT. Of the three, GLIMSPE and QUILT are the most recent and similar in performance. shapeit, eagle and minimac aren't suitable for low-coverage data.

I have found GLIMPSE to be fast and memory efficient, but it requires several different steps; breaking the genome into chunks, imputing, reforming the chunks and then sampling from the haplotypes to phase the data. It has a fairly good tutorial here.

QUILT has fewer stages and faster, but is unlikely to make much of a difference unless you are working on very large datasets. It has a tutorial here.

Which one you use probably doesn't matter a whole lot, but I would probably suggest using QUILT as it has fewer stages and is slightly more recent.

$\endgroup$
12
  • $\begingroup$ I am interested in using Beagle. But it doesn't have good documentation and tutorial. Can you suggest any? $\endgroup$ Dec 27, 2021 at 17:09
  • $\begingroup$ Only beagle4 is suitable for low coverage data and it’s pretty old and doesn’t scale well, so I’d advise against using it unless you have a good reason to? $\endgroup$
    – user438383
    Dec 27, 2021 at 17:12
  • $\begingroup$ Also, I guess Shapit is only for phasing. I found some tutorial on Shapit+impute2 mathgen.stats.ox.ac.uk/impute/impute_v2.html#examples. What do you suggest about it? $\endgroup$ Dec 27, 2021 at 17:18
  • $\begingroup$ Glimpse or Quilt which one is much older? $\endgroup$ Dec 27, 2021 at 17:22
  • $\begingroup$ Also, there is a new Beagle5.2 version. Will it be suitable for low pass sequencing? $\endgroup$ Dec 27, 2021 at 17:24

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.