Somatic SV breakpoint boundary issue caused by supplementary alignment boundaries

[aganezov]

Version: Severus 0.1.2
Data:

• Tumor: colo829
• Normal: colo829bl

command executed:

Severus-0.1.2/severus.py -t 16 --target-bam colo829.chr15_41M.bam --control-bam colo829bl.chr15_41M.bam --out-dir severus_unhaplotagged --vaf-thr 0.05 --single-bp

and

Severus-0.1.2/severus.py -t 16 --target-bam colo829.chr15_41M.tagged.bam --control-bam colo829bl.chr15_41M.tagged.bam --out-dir severus_haplotagged --vaf-thr 0.05 --single-bp

Resulting somatic SV DUP call can be viewed on the IGV screenshots:


NB: the truthset is represented via individual breakends, while Severus vcf contains the more contiguous DUP call. Both refer to the same underlying somatic variant.

Somatic DUP left boundary extends beyond the truth set boundary into the supplementary alignment outermost location. As compared to the truth set DUP boundary breakpoint.

Issue persists with both unhaplotagged and haplotagged input data.

[aysegokce]

Hi @aganezov,
There is an overlapping sequence between the primary and supplementary alignment, possibly due to homology between breakpoints, and the overlap size (~200bp) is below our threshold (500bp). Therefore, Severus directly outputs the positions from the alignments, whereas since it is PCR-validated, truthset has a more precise position. Even If it is above the threshold, Severus would keep the start position and change the end position, so it may not be an exact match.

The truthset vcf is a junction vcf, so they represent everything (except unmapped insertions) as breakends, but in the INFO column, they specify the SVTYPE which should be DUP for this SV.

Ayse