fusion detection for DNA Sequencing data

Question

[J.F.Jiang]

Hi all,

We have designed IDT probe to enrich the target region for tumor samples, including intron regions of known fusion genes.

We have tried several fusion detection tools, such as delly, manta, genefuse to determine the fusion status, however, we found delly and manta were sometimes lack of sensitivity, especially for ctDNA based data. While genefuse was too sensitive, which always gave noise/false fusion results. And genefuse can not determine structure variants such as exon deletion (MET EXON14 SKIP).

Can anyone help?

Thanks, Junfeng

Answer 1

[Kevin Blighe]

Hello again Junfeng,

I have previously used SViCT - it was designed specifically for ctDNA. I recommend that you try it:

• Structural variation and fusion detection using targeted sequencing data from circulating cell free DNA
• https://github.com/vpc-ccg/svict

Kevin

[J.F.Jiang]

will have a try, and response if have any question

[J.F.Jiang]

Hi Kevin,

Any suggestions for the input bam? Since unproper mapped reads might be removed during alignment, what kind of bam file should I used?

I tried raw data mapped file or markdup one, without any result, while delly or manta can give out.

[d-cameron]

Inless you've got a weird aligner, or using non-default settings, unmapped reads will still be in the BAM file - they'll just be marked as unmapped.

Answer 2

[d-cameron]

I've had success with using GRIDSS on WGS as well as cancer panels with tiled introns. Haven't tried it on ctDNA.

AFAIK, GRIDSS2/PURPLE/LINX (https://www.biorxiv.org/content/10.1101/781013v1) is the only toolset that will do compound (multi-breakpoint) fusion prediction from DNA-seq data, but it's designed for WGS, not panel data.

[Ömer An]

Now LINX also supports panel data through WiGiTS, but only TSO-500 is supported by default, reference data must be prepared by the user for custom panels.