need to run a step with FixMateInformation after realignment step ?

Tiphaine

Dear Community and GATK's team,

I have one question about the cleaning step before SNP calling, mainly about local realignment around indels.

I read on some website describing their workflow that alignments may change during the realignment process, it would prefer to fix the mate information and Picard offers this utility to do that for us. is it true? Or are there only any insert sizes that can change? If there are some change of insert sizes, is there a tool that checks that these changes are ok?

What do you use Picard's tool, FixMateInformation.jar, to fix the mate information when using paired-end data ?

Up to now, I have not used in my pipeline. Maybe this is a mistake. If we have to add this step, should we add this step after the realignment step or recalibration step?

Thank you for your help,

Tiphaine

Mark_DePristo

The indel realigner fixes mates itself, so the file is valid after realignment. How this is accomplished is a bit of magic code from Eric Banks and myself from several years ago. A previous version of the realigner did require you to run two passes, but that was before the magic code was written.

mht

Hi Mark,

I have the same question as Tiphaine, except that I'm using GATK-lite. Does the indel realigner in this version also fixes mates or do I need to do it myself?

Thanks.

Geraldine_VdAuwera

The realigner in Lite is the same tool as in the full version, so yes, it also fixes mates.

BerntPopp

Mark_DePristo

@Mark_DePristo said: The indel realigner fixes mates itself, so the file is valid after realignment. How this is accomplished is a bit of magic code from Eric Banks and myself from several years ago. A previous version of the realigner did require you to run two passes, but that was before the magic code was written.

Hi Mark,

So it is alright that I still get these errors from Picard "ValidateSamFile" after running the Indel Realigner (GATK 2.13) on my Paired Files? Or does the code keep the original Information and add the fixed mate information somewhere else?

ebanks

The Indel Realigner only fixes the mates of reads that it moves. If your mates information didn't start out correct for other reads, then they will still be wrong even after running through the IR.

BerntPopp

Alright, i understand. So should one use FixMateInformation (as it adds a extra step to the BAM processing pipeline)? Or asked the other way around: Does the UnifiedGenotyper also handle the malformated mate information?

ebanks

UG doesn't care about the exact locations of the mates.

ocanela

Hi,

In my pipeline, I use the filter 'NotPrimaryAlignment' when I run the IndelRealigner function. As a consequence, when some duplicates are removed, the mate information of the remaining reads becomes (sometimes) wrong and I must fix it with Picard.

ocanela

Hi, sorry for my last comment, I was wrong. It seems that the problem is not with gatk. It arises when using Picard's FixMateInformation function in a previous step.