search

fritzsedlazeck / Sniffles

Structural variation caller using third generation sequencing
Other
560 stars 94 forks source link

false genotyping large deletion #505

Open DHmeduni opened 2 months ago

DHmeduni commented 2 months ago

Hi, I have a 62kb deletion that is very obviously heterzygous but sniffles is calling it as homozygous even the coverage field its's obvious its heterozygous..... anybody have an idea what to do, or is this a bug?

chr7 148238576 Sniffles2.DEL.4A5S6 N 60 PASS PRECISE;SVTYPE=DEL;SVLEN=-62659;END=148301235;SUPPORT=83;COVERAGE=154,75,57,79,157;STRAND=+-;AF=1.000;STDEV_LEN=0.000;STDEV_POS=0.000 GT:GQ:DR:DV 1/1:60:0:83

mletexier-cnrgh commented 2 months ago

Hi @DHmeduni & @fritzsedlazeck , I have the same phenomenon, which is repeated on several events, for example, a 39 Kb deletion, which is seen with the 1/1 genotype with sniffles (VAF=100%, (0, 25)) whereas visually in IGV, it is 0/1. Sequences in IGV are sorted according to haplotypes. I've attached the IGV screenshot. The parameters used are: --minsupport 5 --tandem-repeats hs38me.trf.bed --minsvlen 50 --min-alignment-length 2000 --mapq 20

FalseGenotype_DEL_IGV

Thanks for any help you can give Mélanie

fritzsedlazeck commented 2 months ago

Thanks we can reproduce this now also here.. @lfpaulin is working on that. Hope we have a fix on the next verison.

We found the issue , which is basically handeling the read counting wrong for large split read events.. which is a bug in the deletions but works on all other SV events. In short we are by far undercounting the ref supporting reads , which leads to a 1/1 GT.

Thanks Fritz

mletexier-cnrgh commented 2 months ago

Thank you very much for your prompt feedback, Fritz. Looking forward to the next version :)

Mélanie

tuannguyen8390 commented 3 weeks ago

Hi @fritzsedlazeck! Do you have any estimate for when the next release might be coming? We're really looking forward to seeing this fix as this having an impact on our recessive lethal tracking 🍄

fritzsedlazeck commented 3 weeks ago

We have a working prototype that goes through testing right now. It looks like we resolved it. We hope that we can roll out that new version next week. Thanks Fritz

tuannguyen8390 commented 3 weeks ago

Thanks Fritz, really looking forward to it :)

Tuan

tuannguyen8390 commented 3 weeks ago

Sorry for having a back-to-back message Fritz, but would the bug cause error in the existing snf files or can we just run the merge SNF with the new Sniffles version ?

fritzsedlazeck commented 3 weeks ago

So this is still something we are investigating. @lfpaulin can maybe expand a bit more... the SNF file holds all variants but they have set filters.. These filters can be overwritten with new behavior during the merge, which also applies the same filters.

lfpaulin commented 3 weeks ago

The SNF file is "primed" with a filter, however during the merge it can be override if there is support in other samples. Regarding the genotype we are still testing that feature. Theoretically, if the coverage support values are correct in the SNF then they should provide the correct GT, otherwise you may need to re-run them. We have one test case I will check tomorrow, and will let you know our findings

tuannguyen8390 commented 3 weeks ago

Thanks for the input @fritzsedlazeck & @lfpaulin. The files that we received from partners are from v2.2 so it poses some concern our end. We have a few known recessive SV that are not supposed to be 1/1 so it's a relatively straightforward test our end.