Picard LiftoverVCF Maps Variants to Incorrect Chromosomes When Processing Individual Chromosomes

Question

[jonas.andersson]

Hi!

I have tried processing individual chromosomes with picard LiftoverVcf from GRCh37 to GRCh38.

However, when I look at the results there are a few variants that has mapped to another chromosome.

This also happenes for allmost all chromosomes, with e.g. chr9 having 3 variants mapping to chr1 . Does anyone know why this is happening? Am I doing anything wrong? or is this completely normal behavior? I have made sure that the chromosome naming is the same in the vcf, chainfile and ref, they are all chr[1-22,XY]

The command used for liftovervcf:

 Java -Xmx128g -jar picard.jar LiftoverVcf \
-I chr2_MAF_0.05.vcf.gz  \
-O chr2_0.05_GRCh38.vcf.gz \
-CHAIN hg19ToHg38.over.chain.gz \
-REJECT chr2_rejected.vcf 
-R resources_broad_hg38_v0_Homo_sapiens_assembly38.fasta \
--TMP_DIR ./TMP \
--MAX_RECORDS_IN_RAM 10000

best Jonas

Answer 1

[Giulio Genovese]

The software you use for liftover does not matter. This is a property of the chain file which includes segments of the genome that map to different chromosomes across builds. In particular this is due to chain 1222507:

$ zgrep 1222507 hg19ToHg38.over.chain.gz
chain 1222507 chr2 243199373 + 91679897 91777075 chr1 248956422 - 104082337 104169381 306

which maps region

chr2:91679897-91777075

to region:

chr1:(248956422-104169381)-(248956422-104082337)

that is:

chr1:144787041-144874085

This specific mapping is due to misassembly of the 1q21.1 region in hg19 that was resolved in hg38 (see here)

[jonas.andersson]

Giulio Genovese Aha ok I see! now it makes sense:) thank you so much for clearing that out for me!:)