Breakdancer inaccurate output.

Question

[Fernando Rossello]

Hi Guys,

I have run Breakdancer on a mouse KO sample and got several lines containing intrachromosomal translocations but described as interchromosomal, i. e., CTX instead of ITX, according to breakdancer's structural variant type descriptions.

Find an example/details below.

Software: BreakDancerMax-1.1.2

Command: breakdancer-max -t -q 20 -d sample.ctx -g sample_ctx_q20.bed sample_readgroup.cfg

Library Statistics:

13-19858-SAMPLE.bam mean:311.26 std:75.69 uppercutoff:642.26 lowercutoff:34.69 readlen:95.97 library:Lib-13-19858-H1C1KO_CGATGT reflen:2661926471 seqcov:13.8981 phycov:22.5379 32:778985

Chr1 Pos1 Orientation1 Chr2 Pos2 Orientation2 Type Size Score num_Reads num_Reads_lib 13-19858-SAMPLE.bam
chr10 81178302 29+12- chr10 81178677 29+12- CTX -310 99 41 13-19858-SAMPLE.bam|41

Does anyone found similar results/issues?

Thanks in advance.

Cheers,
Fernando

[nnutter]

For anyone looking for an answer the question was also asked on SEQanswers (but not answered there yet either).

[Fernando Rossello]

These results where observed when bwa-mem was used as aligner. I couldn't see the same type of behaviour when I used bowtie2 local. I will further investigate why this could happen.

Answer 1

[tgi.tabbott]

This is due to supplemental alignments on the same sequence as the primary alignment sneaking in and confusing breakdancer. I just pushed a fix to github. If you grab the latest, these CTX on the same sequence calls should go away.

[Fernando Rossello]

Excellent! Thanks for this. I will test it and let you know if there are any issues.

Cheers,
Fernando

Answer 2

[Charles Warden]

I've also found BreakDancer to produce yield a fair number of false positives. I think calling structural variants is just hard to do.

I don't really have a solution that I feel is a great alternative. If I recall, I think LUMPY at least yielded a reasonably sized list of candidates when I tried it, so maybe you could also try that program.

[Fernando Rossello]

Yes, totally agree. Calling SVs it's just hard and complex. I am testing the same sample using LUMPY and Socrates. I will see how they compare against Breakdancer.

Cheers,
Fernando

Answer 3

[ernfrid]

My best guess is that this is a bug. There are much, much newer versions of breakdancer available at this point. I would suggest trying the latest available via github.

[Fernando Rossello]

Thanks Ernfrid! Really appreciate it. I will test it and see how it goes.

Cheers,
Fernando

[Fernando Rossello]

Just letting you know that the latest Breakdancer version (1.4.4-unstable-9-e2ceb53) and it produces exactly the same output/bug. Will keep investigating.

Cheers,
Fernando

[ernfrid]

Thanks Fernando. I suspect that this is an issue with BWA mem chimeric alignments. Can you post how you ran BWA mem (what command line arguments)?

[Fernando Rossello]

Sure. These were the arguments (default):

bwa mem -t 12 -R "@RG\tID:$TID\tLB:$TLB\tSM:$TSM\tPL:$TPL" $bwaIndex temp1.fastq temp2.fastq > $SN

In any case, tgi.tabbott posted the reason and solution to this issue below.

Thank you, Charles and tgi.tabbot for the help and interest.