Phasing without a reference panel

Question

[chadapohn.chaosrikul]

Hi all,

I'm trying to make my own cohort from genotypes like HapMap project. I would like to know if there are any phasing tools which contain options for phasing genotypes when there's no reference panel?

Thank you.

Answer 1

[Wietje]

Beagle 4.0 offers the option to impute and phase without a reference panel

You should be able to run it with: java -jar ./beagle gt=[your.file.vcf] phase-its=5 nthreads=1 out=[your.output]

You can speed up the process with more threads and improve phasing accuracy by augmenting the phase-its number.

Documentation can be found here: https://faculty.washington.edu/browning/beagle/beagle.29Sep14.pdf

[Wietje]

And you might want to take a look at SHAPEIT as well (http://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html)...

[chadapohn.chaosrikul]

Thank you for your help! I'd like to know if I used SHAPEIT without a reference (by not specify --input-ref option) Does it use the default reference e.g 1000GP or use MCMC algorithm to estimate haplotypes instead?

[Wietje]

As far as I know, you have to invoke the reference panel option with --input-ref to make use of such a panel - otherwise SHAPEIT will phase without this kind of background information. If you want to use the 1000GP data for your phasing step, you'd have to download it as described on the SHAPEIT page (also check these instructions for further parameters and input/output options:

https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html#prephasing

[chadapohn.chaosrikul]

I've already tried both Beagle and SHAPEIT. Thank you for your suggestions.