Length of the inserted/deleted sequence does not match SVLEN – why?

I just noticed that the length of the deleted (in REF column of the output VCF) or inserted (in ALT column) does not always match the SVLEN. I cannot explain this. What am I missing?

Or in other words – how is the SVLEN calculated if not from the length of the inserted/deleted sequence?

Here is an example

chr20	4699380	0	N	GGTGGTGGCAGGGCGGCCTCATGGTGGTGGCTGGGGGCAGCCTCAT	.	STRANDBIAS	PRECISE;SVMETHOD=Snifflesv1.0.12;CHR2=chr20;END=4699404;STD_quant_start=3.558142;STD_quant_stop=7.716681;Kurtosis_quant_start=1.043092;Kurtosis_quant_stop=1.528971;SVTYPE=INS;SUPTYPE=AL;SVLEN=21;STRANDS=+-;STRANDS2=37,69,37,69;RE=106;REF_strand=6462,6730;Strandbias_pval=0.00447258;AF=0.00803517	GT:DR:DV	0/0:13086:106

SVLEN is +21, but length of the inserted sequence (GGT…CAT) is 46 bp.

I had a look at the previous questions but didn’t find anything about this. Feel free to link me if I missed it.

[mschatz]

We looked at this pretty carefully in the Jasmine paper and found -d 50 to be pretty much ideal (Especially Figure 2e): https://www.biorxiv.org/content/10.1101/2021.05.27.445886v1

Cheers

Mike

On Wed, Jul 7, 2021 at 1:10 PM Fritz Sedlazeck @.***> wrote:

So Sniffles uses the positions of the two breakpoints (start and stop) to see if SV across two reads are the same or not … This is where the distance come in ie. the allowed space between the start of the SV across two or more reads and also stop for that manner.

Default is 1kbp but a ratio is applied based on the size of the predicted SV… so smaller SV less tolerance. So I would not go too low , but maybe -d 50 is a start.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/fritzsedlazeck/Sniffles/issues/267#issuecomment-875778223, or unsubscribe https://github.com/notifications/unsubscribe-auth/AABP343VIA3QNTGGSHNUBM3TWSC77ANCNFSM4762FL7A .

[francoiskroll]

Brilliant. Thank you very much both