bench --dup-to-ins not handling DUP as INS

[jcslimmer]

Version :

Truvari v5.0.0 - built singularity image by converting Dockerfile to singularity definition file. Also occurs when using a pip install of v5.0.0.

Describe the bug :

SVTYPE=DUP is not being treated as SVTYPE=INS when enabling --dup-to-ins with truvari bench

To Reproduce :

I'm using using the diploidSV.vcf output from Manta on a 30X WGS HG002 sample.

truvari bench
--includebed GRCh38_HG2-T2TQ100-V1.1_stvar.benchmark.bed
-b GRCh38_HG2-T2TQ100-V1.1_stvar.vcf.gz -
-f GCA_000001405.15_GRCh38_no_alt_analysis_set.fasta
-c HG002_diploidSV.vcf.gz
--pick multi
-S 50
-s 50
--dup-to-ins
--sizemax 100000
-p 0

Expected behavior :

SVTYPE = DUP converted to SVTYPE = INS for truvari bench with --dup-to-ins

If I preprocess the VCF with a script to convert SVTYPE=DUP to SVTYPE=INS and rerun truvari bench then this produces the expected matches.

Example Data :

chr1     2602115     .     G       GTTCTTAGAGTCAGAGGCCACTCAGCAATCTAGAGGCCACGTCAGGGACCAGCCTCCCTCCAGGTAGAAGTCAGGTTCGTC     30     .       SVTYPE=INS;SVLEN=80;LCR=0.991794;REMAP=tandem;PctSeqSimilarity=0;PctSizeSimilarity=0.925;PctRecOverlap=0.0123;SizeDiff=6;StartDistance=0;EndDistance=0;GTMatch=0;TruScore=31;MatchId=53.0,53.0
chr1     2602115     MantaDUP:TANDEM:157:0:1:0:0:0     G     <DUP:TANDEM>     412     PASS       END=2602189;SVTYPE=DUP;SVLEN=74;SVINSLEN=6;SVINSSEQ=TTCGTC;PctSeqSimilarity=0;PctSizeSimilarity=0.925;PctRecOverlap=0.0123;SizeDiff=6;StartDistance=0;EndDistance=0;GTMatch=0;TruScore=31;MatchId=53.0,53.0     GT:FT:GQ:PL:PR:SR     0/1:PASS:412:462,0,619:25,0:40,15

These are two example records from fn.vcf and fp.vcf

Additional context :

Changing the ALT field from DUP:TANDEM to DUP and leaving SVTYPE as DUP did not fix this.

[ACEnglish]

Hello,

You are correct. The parameter was not being respected. I just pushed a change to develop that will match your example variants.

Unfortunately, this might be the most poorly timed ticked I've ever gotten 😆 . I had just hit the button to cut the v5.1.1 release when the github inbox notification appeared on the next page. Funny thing is I was sitting on cutting that release for a few days because I had a feeling another ticket like this might come in.

But you can still install v5.1.2-dev directly from the repo and I'll get around to cutting another release next week.

Have a great day,
~/Adam

[jcslimmer]

Haha, yeah that was pretty unfortunate timing for this ticket 😆

I've installed the v5.1.2-dev version and it works now. Cheers.