Tools for SV & CNV detection

Question

[alons]

Hi all,

We're performing NGS on cancer DNA samples with a custom panel of target regions in different genes.

From your experience & knowledge, which tools do you recommend using for our purpose:

1. Finding structural variants (SVs) such as inversions, large indels, translocations, tandem duplications, etc.
2. CNV (copy number variantion) discovery for only a single sample (many tools I've found were for multiple samples).

If I'm not mistaken I should look for tools that are suitable for exome data rather than whole genome because of our sequencing methods and custom panel.

Thank you very much in advance!

---

I realize this has been somewhat asked before but the posts I've found were at least 2-3 years old and didn't elaborate on cancer genomics or exome sequencing. It was also unclear to me which CNV discovery tools are suitable for single samples. I hope this isn't out of line.

[Tanvir Ahamed]

For CNV, you can try conumee package.

[alons]

Googled it, will check it out.

[alons]

Checked it. Seems like it's not suitable for NGS data.

Answer 1

[Eric T.]

Lumpy, Delly, Manta, and Pindel are good for detecting structural variation and will work on exome/targeted samples. For CNV detection, some recent review articles give an overview of what's available for your application:

• http://bib.oxfordjournals.org/cgi/doi/10.1093/bib/bbu004
• http://bmcgenomics.biomedcentral.com/articles/10.1186/1471-2164-15-732

These articles are not comprehensive. A couple of tools not mentioned there but worth checking out are CopywriteR (Bioconductor) and CNVkit (mine; Unix command line) -- these are relevant if you're using hybrid capture, but less so for targeted amplicon capture. In the latter case, try OncoCNV.

In general you can get a good list of tools that are worth using by looking at existing validated pipelines like bcbio-nextgen and speedseq. Even if you don't use these pipelines directly, you can learn from others' experience by reading the documentation they provide.

[alons]

Thank you very much for the detailed answer!
We're already looking to use Pindel & Delly, will definitely check out Lumpy, Manta and the others.
Your program, CNVkit, seems to be relevant actually because we're using hybrid capture.

[Chris Saunders]

Re: Manta, note default configuration is WGS -- enrichment/panel analysis requires `--exome` config option:

https://github.com/Illumina/manta/blob/v0.29.3/src/markdown/mantaUserGuide.md#exometargeted

[alons]

Thanks, we might try Manta next. Will keep that in mind!

Answer 2

[Chris Miller]

A collaborator made cc2 (https://github.com/abelhj/cc2) for copy number calling from panel sequencing (using a panel of normals run on the same panel). I can't offer specific advice on using it, but their results have always been really nice.

[alons]

Thanks! I will check it out.

Answer 3

[Leandro Lima]

I also recommend SVDetect.

[alons]

Thanks. I'm reading about it now. From your experience, in terms of speed, how does it compare to the other SV detection programs?

[Leandro Lima]

Alons, I haven't compared speed, but for me it works very well.

Maybe it would be good for you to take a look at this paper:

http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0059128

Regards,

Leandro

[Samarth Kulshrestha]

@Leandro Lima: I would like to discuss a problem i am facing during my analysis. I have BAM files (Mate-pair, ~3GB, after pre-processing step). I tried to run SVDetect on 64GB and 128GB machine but every time program crashes during "links2compare" process due to high memory usage.

Have you faced this memory related issue ever ? Any recommendation will be helpful to carry my work to next level.

P.S. I am using SVDetect_r0.8b with multiple threads.

Thanks Samarth

[Leandro Lima]

Hi Samarth.

How much memory are you allocating for each thread?

BTW, did you also try Lumpy and CNVnator?

Leandro