GATK indel raligner after bfast alignement

[david.tamborero]

Hello,

i'm attempting to perform a pipeline to find out somatic mutations of tumor-normal samples sequenced by Illumina paired end reads (100 + 100). After the alignement, I want to do some additional steps (according for instance to http://www.broadinstitute.org/gsa/wi...th_the_GATK_v2), as RemoveDuplicates, Recalibration and Indel_Realignement.

The point is that I want to try the alignement by using bfast, which includes a localign step, in which as far as I know each read is aligned by Smith_Waterman algorithm (allowing gaps) for each CAL. So my question is: when using bfast alignement, is it necessary the GATK_indel_realigner step? I do not know if it is a non wise question, but I want to be sure of performing such GATK step since it's quite demanding and my cluster is not awesome at all.

Thank you for your feedback!

cheers
david

[Heisman]

When aligning you will only look at one read at a time, while with IndelRealignment all reads around a predicted indel will be looked at simultaneously. I would go ahead and try both and see what differences you get in your variant calls. If there's say a 1-2% difference, then maybe that wouldn't be enough to care.

[david.tamborero]

thank you for your explanation, Heisman.

Then, I guess I have to try it. If the cluster and I are alive to get some results about this issue, I will post them in case anyone more is interested.

Thank you again!
cheers
David

[adaptivegenome]

Using the most recent GATK you do not have to do the complete SW realignment if your mapper already produced gapped alignments. See the documentation, there is a shortcut option that saves a little time.

[david.tamborero]

I've just seen it, I will try the indel_realignement command of the latest GATK by using the "--consensusDeterminationModel USE_READS" argument, it seems exactly what I was looking for.

Many thanks!