behaviour of bwa mapping for homologous regions.

[KevinLam]

Hi
I am using bwa on paired end illumina reads for identifying inter chromosomal translocation.
I have used breakdancer on the data but non of the proposed inter chromosomal translocations match what I have on my list of suspects (garnered from FISH)

I am looking at the alignment of the regions now manually and I find that there is low coverage in this region.

I am suspecting that the translocation are taking place in homologous regions within the genome.
Would this cause reads to be 'dropped out' from the final bam file if bwa finds equally plausible regions which the read might map?

How should I hunt for evidence for the translocation?

[epigen]

BWA just chooses a random best alignment if there are several equally scoring best alignments. Hence I think your suspicion is the most probable explanation for not finding the translocation with NGS data that you've seen with FISH. A colleague had a similar problem with a paralogous gene.

BFAST has an option to keep all possible alignments for a read, I think that also works with paired end data. (I'm sure Nils will be helpful in this aspect.) Then the same read ID will appear multiple times in the SAM file and you have to think of a way for filtering with your own criteria.

Barbara