If I align my bams using bwa-mem, do I still need to do local indel realignment?

hovelsonhovelson Posts: 19Member
edited September 2013 in Ask the GATK team

I have ~2,000 whole-genome-sequencing bams that I need to move through an alignment/deduping/recalibration pipeline. I'll be using bwa-mem for the initial alignment, and I'm wondering if you would still recommend carrying out the local realignment step (downstream) for these bams, given bwa-mem does some local/gapped alignment on its own. This is an important question for our group because the local realignment (using RealignerTargetCreator & IndelRealigner) for these bams will take a significant amount of time / computational resources. If I want to call short indels using UnifiedGenotyper (rather than HC, due to number of samples), does bwa-mem for primary alignment reduce the need for local realignment downstream? Any feedback/thoughts would be much appreciated - thanks!

Post edited by hovelson on


  • Geraldine_VdAuweraGeraldine_VdAuwera Posts: 8,434Administrator, GATK Dev admin

    I'm not familiar enough with BWA mem to give you a definitive answer, but basically the key point is that traditionally mappers like BWA only align one read at a time, which is why realignment around indels is necessary to provide consistent mapping -- the indel realigner sees all the reads in a region, not just one. So it depends if BWA mem has that added capability. I'm not aware that it does, so I would still recommend doing indel realignment according to our best practices, but I could be wrong.

    Geraldine Van der Auwera, PhD

  • hovelsonhovelson Posts: 19Member

    Very helpful - thanks for the feedback!

Sign In or Register to comment.