Closed fellen31 closed 1 month ago
I think we may as well! We already know trgt works well; will be good to see what the VCFs of the other two look like! Do you have current cases with them on like the dev area or on stage even?
Currently working on annotating SVs, but here's the current output files:
/home/proj/development/rare-disease/felix/scout_test/cnv.vcf.gz
(missing annotation and FOUND_IN tag) and /home/proj/development/rare-disease/felix/scout_test/test_svs.vcf.gz
(missing annotation). And for the future, those will very likely want to live in the same VCF. More conceptually correct naturally, but also merging them has been the best sens/spec tradeoff. Especially with LRS I would imagine the SVs should catch quite the bulk of the CNV break ends. Most of the SV callers both consider read depth as well as call DEL and DUP.
I know want them in the same VCF. Will work on merging those.
Does this make sense to you even if they have been filtered differently? For example, I think there's no reason to include CNV-calls in centromeres, the coverage is just too uneven and noisy. But for SV-calls, there might be a translocation with one clear breakpoint, while the other one ends up in the centromere, and you would not want to filter out those calls?
My impression is that Sniffles2 has missed a few pretty clear DELs and DUPs...maybe Sniffles1 was better at this.
Good, yes, I think it potentially makes sense to remove low quality calls before merging. We have been a bit back and forth on whether to frequency filter on the separate files or merged. But for both of these, it would be best to reevaluate since your relative accuracy for each type is different compared to Illumina SR.
Right, Im not so familiar with Sniffles, but I know @J35P312 had a decided preference for v1.
Currently, the following callers implemented in Nallo are missing tags in scout:
hificnv
(CNV)severus
(SV)trgt
(Repeats)Although the callers used in the pipeline might be subject to change, would you like to add them?