Sniffles
Sniffles copied to clipboard
fritzsedlazeck
difference between SUPPORT and DV
Hi,
I was looking at an SV event in IGV. The VCF record from sniffles2 has SUPPORT=1, but the DV was 2. SUPPORT means reads supporting the SVs. DV means variant supporting reads. What is the difference between the two? It was an insertion event and I could see only 1 read with the insertion event. Why was DV 2 in this case? Any explaination would be very helpful.
Regards, Prasun
Can you show the command that you ran? I am trying to understand if this is the non-germline or default mode. Thx Fritz
Hi,
This is the command and it was run in default mode:
sniffles --input snf_files_list.tsv --tandem-repeats $TANDEM_REPEAT_PATH/human_GRCh38_no_alt_analysis_set.trf.bed --minsvlen 50 --reference $REFERENCE --threads 16 --vcf output_multisample.vcf.gz
Regards, Prasun
Any update on this issue? I have the same question since we are using the DV field for variant filtering but doesn't seem to match the SUPPORT field. I ran it on the default mode.
sorry all for the slow reply. DV refers to the number of reads; Support should be the number of samples in a multi sample VCF from sniffles. Retrospectively I see how that can be confusing. Sorry Fritz
Don't worry, we all know what it is like to be busy. Thanks for answering!
It seems that I completely misunderstand something fundamentally. In my opinion DV = |RNAMES|. I don't see this in any of my SVs.
One example:
15 44633572 Sniffles2.INS.22CS6 N AAAAAGA...AAAAAAAAAAAAAAAA 60 PASS PRECISE;SVTYPE=INS;SVLEN=4256;END=44633572;SUPPORT=6;RNAMES=8818c927-5059-46ee-a14f-93f60b12f1e4,06ae71db-bd97-4956-a277-60aad89e8695,2a67c270-01c5-4020-8461-d6694d7c0a27,67b0b91c-e6d9-4c33-9fb6-1ab8d5c381e2,79e3f01b-4bf9-4af5-9e45-4bb0d87e64c0,93a83dd7-dc2a-4fa5-b9cc-4f0cd14f041e;COVERAGE=12,12,14,14,15;STRAND=+-;AF=0.929;PHASE=1,44606444,6,6,PASS,PASS;STDEV_LEN=14.660;STDEV_POS=0.000;SUPPORT_LONG=0 GT:GQ:DR:DV 1/1:26:1:13
(I shortened the sequence in the ALT field for readability)
We deal with only 1 sample, so no mutli-sample calling. There are 6 RNAMES, so 6 reads supporting my SV. I can visually confirm these via IGV. SUPPORT=6, so this likely refers to these 6 reads mentioned in RNAMES. I would (visually) agree with 6 out of 14, with is about 50%.
But then
DV=13 DR=1 AF=0.929 GT=1/1
The value of DV doesn't reflect the number of RNAMES at all and apparently my assumption is wrong. Would you help me? Maybe there are internal realignments in sniffles and there are more reads supporting and counted in DV?
In case it helps I also attach a screenshot of the region.
THANK YOU!
I've looked through your code and I've found the reason. You are rescaling the number of supports for long insertions https://github.com/fritzsedlazeck/Sniffles/blob/fd31ee3962d7c588752e1926176c79912691d54e/src/sniffles/postprocessing.py#L142-L148
in order to increase the sensitivity to detect them
https://github.com/fritzsedlazeck/Sniffles/blob/fd31ee3962d7c588752e1926176c79912691d54e/src/sniffles/postprocessing.py#L170
That seems to be valid. But you also rescale the support when genotyping the call
https://github.com/fritzsedlazeck/Sniffles/blob/fd31ee3962d7c588752e1926176c79912691d54e/src/sniffles/postprocessing.py#L347
and write this rescaled support count in its genotype at a position, that is later used as DV
https://github.com/fritzsedlazeck/Sniffles/blob/fd31ee3962d7c588752e1926176c79912691d54e/src/sniffles/postprocessing.py#L414
But this new rescaled support does not represent the number of supporting reads anymore and also the derived allele-frequency is wrong. This is a bug in my opinion. If you agree, I would write a pull request to change this behavior.
Thanks for digging in. There are two modi. One is default/gremline and the other is mosaic Sv calling. I think the scaling is on the mosaic. @hermannromanek and I can check this out to make sure. Thanks Fritz