compare structural variation

[bair]

Dear all,

How to compare structural variation in two genomes or with database like DGV?

Software like breakdancer can do structural variation analysis, it outputs INV, DEL, INS and other types of SV, how can I compare it with DGV?

Simply to see if any of the two chromosome coordinates in breakdancer file locate in DGV locus region?


Many thanks!

[Irsan_Kooi]

Hi Bair,

I am having the same problem. Using breakdancer the output consists of DEL, INV, INS, CTX and ITX. Though breakdancer tries to support detection of somatic variants I don't think it is working very well... In the 11th output column it displays the amount of reads supporting the called SV per bam-file. So if you have run:

[your_prompt]$ bam2cfg.pl -g -h tumor.bam normal.bam > config_file.cfg
[your_prompt]$ breakdancer_max -d some_prefix config.txt > tumor_normal.ctx

then the 11the column displays the amount of reads supporting the called SV. But at first, it does not work if you don't have set the read-group information correctly (as is said in the breakdancer documentation). But more importantly, if you filter your results for SVs were this 11th column says that the SV is only supported by reads of your tumor.bam file it does not look very somatic if you view the results in IGV. Not for deletions (which I think are easiest to compare), not even talking about the others.

I think the best way to go is to run breakdancer on the two .bam files seperately and then use secondary tools to compare the results.

For Deletions, you can make 2 bed files for all locations that are deleted in your 2 bam files. Than you can use intersectBed from BEDtools to find the deletions in your tumor results that are not present in your control results:

[your_prompt]intersectBed -v -f 0.5 -a tumor_breakdancer_deletions.bed -b control_breakdancer_deletions.bed

For inversions, insertions, inter- and intra chromosomal SVs the situation looks slightly different to me. Here the breakpoints are of imporantence, not the location from start to end (like when looking at deletions). I think you should make a list of all breakpoints, use a buffer size (say +- 500 bps) and see which breakpoints are present in your tumor and not in your control (maybe the same way using bedtools?)

So far, I have successfully compared breakdancer deletions between samples but am just starting to implement a comparison tool for the other types of SVs

Please feel free to share some thoughts about this.

Cheers,

Irsan