clarification of coverage in Sniffles2 VCF

[svedwards]

Thank you for the wonderful software and apologies if this question has been asked before but I couldn't find a satisfactory answer in other issues posted here. I'm simply wondering how to interpret the coverage data in the INFO field. Here is a line from my VCF:

CM083959.1 999150 Sniffles2.INS.1FS16 N TATATAACAATATATAACAATATATAACAATATATAGCAATATATAACATTATATAACAATATATAACAA 60 PASS PRECISE;SVTYPE=INS;SVLEN=70;END=999150;SUPPORT=14;COVERAGE=17,18,18,18,18;STRAND=+-;STDEV_LEN=0.000;STDEV_POS=0.000;SUPPORT_LONG=0;VAF=0.778 GT:GQ:DR:DV 0/1:0:4:14

I can clearly see the 14 reads supporting the 70 bp insertion in the IGV view. But I'm unsure on what the 17,18,18,18,18 coverage is (and so is ChatGPT). And as suggested by a previous issue, DV = SUPPORT, in this case = 14. And I can also see the 4 reads supporting the reference allele (DR). I must have had the RNAMES option turned off. Thanks for any help on the meaning of the four numbers in coverage.

-- Scott

[fritzsedlazeck]

Dear @svedwards
good question :) . We used the coverage as a feature we were interested in and thus maybe explains the lack of documentation. It should be the overall coverage left outside , left inside, middle, right inside and right outside of the SV event. For your insertion thats quite boring obviously but DEL , DUP or mapping artifacts could be highlighted with this. Note that this coverage is highly filtered based on MQ and other read properties. Also we just made an update on the way this is calculated.

Hope that helps
Fritz

[svedwards]

Thank you Fritz, this is very helpful. If I may, I have another related question. We are looking at trios (of birds!). We used this script to generate the joint vcf:

sniffles --input
/n/netscratch/edwards_lab/Lab/lbartel/firstround/sniffles2_joint/1873-10972.snf
/n/netscratch/edwards_lab/Lab/lbartel/firstround/sniffles2_joint/1713-90341.snf
/n/netscratch/edwards_lab/Lab/lbartel/firstround/sniffles2_joint/1873-49222.snf
--vcf /n/netscratch/edwards_lab/Lab/lbartel/firstround/sniffles2_joint/trio3_joint.vcf
--threads 16

We then filtered this VCF as follows to find de novo mutations in the chick:

[0]=Dad, [1]=Mom, [2]=Child

bcftools view -i 'QUAL>=60 && ((GT[2]=="0/1" || GT[2]=="1/1") && GT[1]=="0/0" && GT[0]=="0/0")'
-Ov -o $VCF_DIR/trio3_novo_filt.vcf $VCF_DIR/trio3_joint.vcf

But we are getting lots of putatively de novo chick mutations that also seem to be found in one of the parents. In the attached IGV screenshot, red = dad, green = mom and blue = chick, but some of the reads have the 94-bp insertion in the dad as well as the chick.

Could you suggest some additional sniffles2 options to produce a more reliable VCF? Again, likely a simple question but we appreciate the help!

[fritzsedlazeck]

Yeah we see something similar in our rare disease trios that points at that TR regions are sometimes not cleanly filterable with this approach... There are some technical challanges that makes this tricky (e.g. your region is probably due to some alignment properties).
We are using Kanpig now in our in house approach to clean this up further. https://github.com/ACEnglish/kanpig

I hope that helps
Fritz

[svedwards]

Great, thanks very much for the tip! We will continue to investigate. - Scott