TP considered FN about truvari HOT 5 CLOSED

I would need to see more context but something could be happening is that this run with --pick single. The --pick parameter controls how many matches a call is allowed to participate in (details).

Are there other calls in this region which were marked as true positive? If so consider using --pick multi

Otherwise I'm not sure. The base vcf entry you've provided shows the call was annotated:

PctSizeSimilarity=1;PctRecOverlap=0.9565;SizeDiff=0;StartDistance=-68;EndDistance=-68;GTMatch=0;TruScore=65;MatchId=94.0.0

So you could look through the tp-comp/fp VCFs for MatchId=94.0.* To see what it was compared to. (MatchId details)

Also, If you turn on --debug the logging is pretty verbose about exactly why any match decision is made.

If you're still stuck, if you could send me the input base/comp VCFs in the region, say chr12: 71959796-71965796, I could help investigate it.

from truvari.

Thank you @ACEnglish! I'll start with your suggestions and I'll let you know about the outcomes.

from truvari.

Here it is the matched variant in the fp.vcf.gz, that is actually the one I was expecting to match:

chr12 71962796 DRAGEN:LOSS:chr12:71962797-71964360 N <DEL> 150 PASS SVLEN=-1564;SVTYPE=CNV;END=71964360;REFLEN=1564;OrigCnvPos=71962186;OrigCnvEnd=71964742;SVCLAIM=DJ;MatchSv=DRAGEN:DEL:183383:0:1:0:0:0;PctSeqSimilarity=0;PctSizeSimilarity=1;PctRecOverlap=0.9565;SizeDiff=0;StartDistance=-68;EndDistance=-68;GTMatch=0;TruScore=65;MatchId=94.0.0 GT:SM:CN:BC:GC:CT:AC:PE 1/1:0.263257:1:2:0.375978:0.507825:0.483959:39,42

They are also annotated with a high Overlapping score (PctRecOverlap=0.9565) so I do not see a reason to exclude them. Moreover, in the region there in no other variant, so I presume the --pick multiple won't help.

from truvari.

Ah! I see what's happening, now. So there's another 'threshold' that wasn't accounted for typeignore. It is off (false) by default. These variants don't have matching types, so they're not passing the thresholds.

Now, you and I can see that they do have matching types (DEL), but if you look at how truvari determines svtype you'll also see that the SVTYPE=CNV in the comp vcf doesn't match the SVTYPE=DEL in the base.

I know that vcf v4.4 had a lot of changes recently and in it SVTYPE was deprecated in favor of symbolic alts. This was not my favorite move because it's obviously a breaking change.

Regardless, the quickest way to get these variants to match is for you to run bcftools annotate -x INFO/SVTYPE on the comp vcf.

from truvari.

Ok, great, thanks for the help!!
Now it makes sense!

Have a nice day,
Valentina

from truvari.