Gatk multiSample realignment and recalibration

[seq_GA]

Hi,
I have used gatk for single sample realignment followed by recalibration. I want to try multi sample realignment and recalibration as mentioned in the best practice page http://www.broadinstitute.org/gsa/wi...th_the_GATK_v2
The example doesn't show how to use the commandline.
Has anyone used this feature? If so can you please share the command line for gatk to call this? Thanks.

[fpepin]

Is this what you have in mind? http://www.broadinstitute.org/gsa/wi..._around_indels

Score recalibration can be found at http://www.broadinstitute.org/gsa/wi..._recalibration

I could offer more details, but they're simply saying it better (and more completely) than I could.

[seq_GA]

Thanks for your response. Right now I am using the following commandline to do realignment and recalibration sample by sample.

java -Xmx4g -jar /tools/GenomeAnalysisTK-1.0.4013/GenomeAnalysisTK.jar -T RealignerTargetCreator -I test_PCR_removed_sorted.bam -R /Genome/hg19/hg19.fa -o output.intervals 

echo "GATK run realigner"
java -Xmx4g -jar /tools/GenomeAnalysisTK-1.0.4013/GenomeAnalysisTK.jar -T IndelRealigner -U -S Silent -I test_PCR_removed_sorted.bam -R /Genome/hg19/hg19.fa -targetIntervals output.intervals --output test_reAligned.bam

echo "GATK Count covariates and generate csv"
java -Xmx4g -jar /tools/GenomeAnalysisTK-1.0.4013/GenomeAnalysisTK.jar -R /Genome/hg19/hg19.fa --DBSNP /GATK/latest_GATK/resources/dbsnp_129_b37.rod -I test_reAligned_sorted.bam --max_reads_at_locus 20000 --default_platform illumina -T CountCovariates -cov ReadGroupCovariate -cov CycleCovariate -cov DinucCovariate -recalFile test_recal.csv

echo " GATK re-calibration"
java -Xmx4g -jar /tools/GenomeAnalysisTK-1.0.4013/GenomeAnalysisTK.jar -R /Genome/hg19/hg19.fa -I test_reAligned_sorted.bam -T TableRecalibration --default_platform illumina -outputBam test_reAligned_reCal.bam -recalFile test_recal.csv

But I am interested in Best: multi-sample realignment with known sites and recalibration as mentioned here http://www.broadinstitute.org/gsa/wi..._recalibration

How to implement the following.

for each sample
    lanes.bam <- merged lane.bam's for sample
    dedup.bam <- MarkDuplicates(lanes.bam)

samples.bam <- merged dedup.bam's for all samples
realigned.bam <- realign(samples.bam)
recal.bam <- recal(realigned.bam)

Please let me know whether my intrepretation as below is correct.
1. Merge all lane for each sample bam file.
2. Mark duplicates of sample bam
3. Merge all the sample dedup bams
4. Call realignment
5. Call recalibration.
Thanks.

[fpepin]

I'm not an expert, but that does sound reasonable (and very close to what I have). I also call the indel discovery tool before the realignment to get additional regions to focus on.

I use Picard for the duplicate removal:

java -Xmx8g -jar MarkDuplicates.jar I=bamFile O=bamDupRemFile \ 
M=duplicateMetricsFile VALIDATION_STRINGENCY=SILENT ASSUME_SORTED=true \
REMOVE_DUPLICATES=true

Merging can be done with samtools:

samtools merge mergedBam bam1 bam2 bam3

Unfortunately, you'll need to redo the header as it will only use the header from the first bam file and it will miss the @RG from the others. Hopefully I'm behind in the news and someone can mention an easy tool to do it, I just have a bit of perl code to do it. I can post it here if you want.

[mard]

Unfortunately, you'll need to redo the header as it will only use the header from the first bam file and it will miss the @RG from the others. Hopefully I'm behind in the news and someone can mention an easy tool to do it, I just have a bit of perl code to do it. I can post it here if you want.

Picard's MergeSamFiles will merge multiple BAM files and add all @RGs to the header of the merged BAM.

[seq_GA]

Thanks Mard for the info.