Lift-over on a VCF

Question

[win]

Hi all,

I have a VCF file that was generated using HG18 and I want to convert it to VCF but that has HG19 data.

I am aware that there is the lift-over tool from UCSC but how do I convert a VCF file from HG18 to a VCF file to HG19, since it seems the lift-over tools produces a BED file instead.

Thanks in advance.

Answer 1

[Ashutosh Pandey]

You may use GATK "LiftoverVariants". It allows you to use VCF file as an input and outputs a VCF file corresponding to the new assembly. Link: https://www.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_gatk_tools_walkers_variantutils_LiftoverVariants.php

[A. Domingues]

It seems that LiftoverVariants is no longer provided with GATK 3.5:

java -jar GATK/3.5/GenomeAnalysisTK.jar -T LiftoverVariants -h
##### ERROR MESSAGE: Invalid command line: Malformed walker argument: Could not find walker with name: LiftoverVariants

The link in the answer is also dead supporting the deprecation of the tool, and GATK seems to be recommending Picard LiftoverVcf.

[win]

Do we get the chain file from UCSC?

[Sean Davis]

Yes, this would be a UCSC chain file.

Answer 2

[Tommy Carstensen]

Picard LiftoverVcf, which is supported by the Broad: http://broadinstitute.github.io/picard/command-line-overview.html#LiftoverVcf

[rodd]

This worked like a charm! :) (P.S.: chromosomes in input vcf must be numerical, not preceded by "chr", and to avoid running into memory problems, use the -Xxm15G flag for the java command [if you have up to this amount of memory available]).

Answer 3

[Sean Davis]

NCBI offers a remapping service (equivalent of liftover) that works with multiple formats, including VCF:

http://www.ncbi.nlm.nih.gov/genome/tools/remap

Answer 4

[yhoogstrate]

CrossMap can do the trick as well

• http://crossmap.sourceforge.net/

Answer 5

[liqigang]

A simple script is here, https://github.com/liqg/vcf-liftover. Using it is very simple, like this: zcat test.vcf.gz | ./vcf-liftover GRCh37_to_GRCh38.chain.gz

The converted VCF will output to stdout,you can save the unmapped variants by add a file path to store them.

[rodd]

This hasn't worked for me: the resulting VCF was completely unstructured (but I didn't have time to investigate why this happened). Picard (as per one of the responses above) was very straightforward.

Answer 6

[brianpenghe]

You can use ENSEMBL: http://www.ensembl.org/Homo_sapiens/Tools/AssemblyConverter?db=core