Infer SV type from bedpe file

[andrewSharo]

I'm working to identify SVs in 10X linked-read data using methods alternative to longranger. What is the best way to call SVs from the adjacency info given in the bedpe file, or is there a tool that does this already? For example, it would be great if I could distinguish deletions from inversions. I understand that some SVs, such as inversions, create two novel adjacencies, while others, like deletions, create only one, which suggests I could write algorithm to find pairs of adjacencies. Maybe someone has already done this? Additionally, in the bedpe file, what does the orientation field indicate?
Thanks again for creating this method and making it easy to download and run!

[relyanow]

Hi Andrew,

The orientation field is actually key to distinguishing between SV types. An orientation is assigned to each breakpoint in the novel adjacency based on the orientation of the reads supporting the breakpoint, so a '+-' in the orientation field indicates that the first breakpoint has '+' orientation and the second breakpoint has '-' orientation. In the reference genome segments are aligned head to tail, so '+-' orientation is the reference orientation (which includes deletions). '--' and '++' indicate an novel adjacency created by a inversion and '-+' indicates a novel adjacency created by a tandem duplication.

Most methods that report structural variant calls do so using just this orientation information. I report only the orientation because the novel adjacency may be part of a complex rearrangement and I didn't want to make assumptions about how it was created. You could write an algorithm to pair up novel adjacencies, but if you want to reconstruct the structural variants, I think the best way is to use a method that performs local assembly (such as grocsvs or svaba). You could then filter SVs by those with novel adjacencies supported by both local assembly and NAIBR.

Hope that helps!

[andrewSharo]

Hi Rebecca,
Thanks for getting back to me so fast. The orientation field now makes sense to me, and I understand what you mean about having very complex rearrangements. I looked into groc-svs and svaba. I will give groc-svs a try and see how their calls compare to novel adjacencies identified by naibr. Please do let me know if you or anyone in your lab comes across a tool that will assembly SVs directly from the naibr bedpe output. Thanks again!