Loqusdb archive frequency not annotated for some BNDs

[dnil]

Describe the bug
Cf e.g. https://scout.scilifelab.se/cust002/F0073526/sv/variants/1e960ce3b036416873f3d0d1914d7bd8.

Intermediate size BND with slightly complicated ALT allele (bnd notation). Perhaps missed in loqusdb export or svdb markup?

[jemten]

Thanks for reporting!

The loqusdb export has this variant, which I assume is what we should match to

10  3209757 .   N   N]10:3209897]   .   .   Obs=1269;Frq=0.07007;SVTYPE=BND

We run svdb with an overlap setting of 0.6 and bnd disdance of 25000 so we should have annotated it. Maybe svdb is tripped by the alt field? Will look into it a bit more

[jemten]

OK so if I read the svdb code correct, the issue is that there are no overlap between these variants. The BNDs are close enough but svdb requires overlap as well? I assume that loqus matches sv variants differently.

https://github.com/J35P312/SVDB/blob/b1c3d1c104adb9226c3b2c8769e10b8f6d0cbc39/svdb/overlap_module.py#L17

[J35P312]

Hello!
I dont think that is the issue. BNDs should be tested using the precise overlap module, which only checks the distance and not the jaccard index.

Can you send me the tricky variant, Im curious, but I cant reach scout :p

[jemten]

I can share the vcf with you

[J35P312]

yes! I want the vcf!

[J35P312]

Srry! I had a look at the code as well, and I agree, since both breakpoints are at the same chromosome SVDB will also assess the jaccard Index, which will be zero for non-overlapping variants.

[dnil]

Perhaps not for this issue, but we probably want to double check INS as well - just had an issue with them in loqusdb. I imagine the interaction loqus-svdb might be a bit interesting there.