Strange strand bias and unusual depths

[Turner Conrad]

Hi all,

I'm seeing some very strange stuff with a resequenced prokaryotic genome coming from an Illumina HiSeq 2000. The design is 50bp single end and workflow is BWA-mem, GATK indel realignment, and visualization with the MATLAB NGS browser. There is a complete negative strand bias with a small island of positive strand bias, an 6pb segment of depth that is 5x normal for the rest of the genome (2300- vs 480-fold), nearly completely mapq 0, and distinct breakpoints. See the attached figure for a visualization.

Any ideas what kind of phenomenon could be causing this?

[Brian Bushnell]

Probably incorrect alignment around a structural variation. But, I'm used to IGV so I don't quite understand the picture.

[Turner Conrad]

Well, it's centered around a duplicate rDNA region, but is there anything about BWA that would introduce such a strong strand bias when trying to pick which copy to map to?

[Brian Bushnell]

Not that I've noticed. Be sure you're using the latest version. Also, try a different aligner like BBMap or Bowtie2 to see if you get the same results. If you do, it's not an algorithmic problem, but something in the data.