error setting DHD in VCF when annotating

[chapmanb]

Brent;
Thanks for duphold, this is a brilliant idea to help improve SV filtering and I'm working to incorporate into bcbio right now.

When running validation cases I'm hitting an issue that I'm not sure how to fix/work around. It complains with:

error setting DHD in VCF

and exits with a 1 error code before producing any variants.

I put togther a small test case that has this issue. It's self-contained, except you need to edit REF to point at a GRCh37 fasta file (I couldn't figure out how to make it work with a reduced fasta file to include). I'm running with the v0.0.2 pre-built binary packaged up into conda.

https://s3.amazonaws.com/chapmanb/testcases/duphold_setting_DHD.tar.gz

The command is fairly vanilla:

duphold --threads 4 --vcf HCC2218BL-HCC2218-svs-smoove.genotyped.vcf.gz --bam HCC2218C-sort.bam --fasta $REF | bgzip -c > HCC2218BL-HCC2218-svs-smoove.genotyped-duphold.vcf.gz

and I get the same error whether using a bgzipped/tabix VCF or bcf input.

Do you have any clues about what I'm doing wrong? I can see the error message and check in the code (

duphold/src/duphold.nim

Line 204 in 5e5e9de

if variant.format.set("DHD", dhd) != Status.OK:

) but don't have a good clue why trying to set the calculated value would fail. Thanks much.

[brentp]

H Brad, thanks for finding this. This was a bad bug that only happens when run on > 1 sample since I was always testing on a single sample. I have made a fix locally.

I already have a change in smoove that incorporates duphold automatically (well, with an optional flag) so I'd hold off for a bit. I also stumbled on several improvements to duphold that make it even better. Those changes will be out this week.

Thanks again for the full test-case. It enabled the discovery of the silly bug quickly.

[chapmanb]

Brent;
Thanks so much for the quick fix. I'll definitely do more testing with this once you have a chance to roll out the new version this week. Thanks also for the work on incorporating into smoove, I'm using duphold for other callers (manta, CNVkit) but will swap over to your smoove support when that's all ready to go as well. Thanks again for duphold and all this work.

[brentp]

I have made a new release. Your testing is much appreciated:
https://github.com/brentp/duphold/releases/tag/v0.0.3

I think you had this for testing purposes, but note that:

duphold --threads 4 --vcf HCC2218BL-HCC2218-svs-smoove.genotyped.vcf.gz --bam HCC2218C-sort.bam \ 
    --fasta $REF -o HCC2218BL-HCC2218-svs-smoove.genotyped-duphold.vcf.gz

is preferable to:

duphold --threads 4 --vcf HCC2218BL-HCC2218-svs-smoove.genotyped.vcf.gz --bam HCC2218C-sort.bam \
    --fasta $REF | bgzip -c > HCC2218BL-HCC2218-svs-smoove.genotyped-duphold.vcf.gz

and that duphold DHD will be useful for manta, but not for cnvkit since it only looks right near the breakpoint. the fold-change stuff should work for both.