Circos plot from SV data of WES samples

Question

[Hasib]

Hi. I've detected SVs (Structural variants) from my WES data using the BreakDancer tool. Now I want to do a circos plot using those SV data. How do I go about doing it? I'm absolutely new in these waters so I'd very much appreciate some help.

Answer 1

[Getting there]

You can try ClicO FS which is web-based if you're new and don't have a lot of samples to go through. They have a demo project you can try out first to see the kind of visualizations you can get.
http://103.47.253.210:3000/projects?demo=1

Circos which is very popular at the command line, also has an online version. http://mkweb.bcgsc.ca/tableviewer/

[Hasib]

Thank you for your reply. I will definitely check them out!

Answer 2

[d-cameron]

I'm absolutely new in these waters so I'd very much appreciate some help.

There are a number of issues that have the potentially to fundamentally invalidate your results if you're not careful.

detected SVs (Structural variants) from my WES

WES is only able to detect breakpoints that are inside/immediately adjacent to exons. You'll be able to see exonic rearrangements, have a small chance of detecting gene fusions (fusions are also always intronic and you'll only find them if the breakpoint is right next to an exon - otherwise it'll be off-target and won't have any coverage), but are otherwise blind to genomic rearrangements.

You should assume that you are unable to detect any intronic or intergenic SVs. I don't see how a circos diagram is going to be meaningful as the only SVs you're going to see are going to be tiny compared to the size of chromosomes.

If you're dealing with cancer data and want a high-level overview of rearrangements then you should be using a WES CNV calling tool. You'll at least be able to see which chromosomal regions have been amplified/lost.

using the BreakDancer tool

BreakDancer is possibly the worst choice of available short read SV callers you could make. Not only does it not incorporate split read evidence but, unlike other read-pair based methods that refuse to run if the fragment size is less than twice the read length (since there's practically no read pair signal left - the signal is all in the split reads) it's failure mode is catastrophic and will return approximately the expected number of SVs but will have a 98%+ false positive rate. BreakDancer was designed for 2x36bp sequencing data and is the tool that practically every SV caller compares against because it so easy to make even a bad tool look good by feeding BreakDancer with modern (e.g. 2x150bp) short sequencing data.

If you're going to run a breakpoint-based SV caller, you should use one that a) uses both read pair and split read support, and b) performs local assembly. Manta and GRIDSS are two such tools that perform well in benchmarks.

TLDR: don't use BreakDancer; expect close to no meaningful results from any SV (breakpoint) caller; use a WES CNV caller instead.

[Hasib]

Thank you for your reply. I actually wanted to use Manta but it doesn't support the Linux version I have on my work PC. I've used CNVkit to call CNA from my tumor BAM files. Can I use CNVkit to detect SVs as well?