[xlzhang]

@Alex,

I just ended another round of indels calling against several sets of simulated data. In brief I simulated sets of reads for chromosome 20 at four different coverages (5x, 11x, 22x, 44x).

Then I run GATK UnifiedGenotyper with or without local realignment. I found no difference between the number of indels calls with or without realignment?!

So is it worth doing local realignment steps prior to call for indels ON simulated data?

Or maybe I just did something wrong in the realignement steps?!

Here are my commands:

java -jar GenomeAnalysisTK.jar -R human_g1k_v37_chr20.fasta -T RealignerTargetCreator -o output.intervals --known 1000G_biallelic.indels.b37.vcf

java -jar GenomeAnalysisTK.jar -I out_sorted.bam -R human_g1k_v37_chr20.fasta -T IndelRealigner -targetIntervals output.intervals -o realigned_out_sorted.bam -known 1000G_biallelic.indels.b37.vcf --consensusDeterminationModel KNOWNS_ONLY -LOD 0.4

java -jar GenomeAnalysisTK.jar -R human_g1k_v37_chr20.fasta -T UnifiedGenotyper -I realigned_out_sorted.bam -o gatk_indels.vcf -glm INDEL

Regards.

Hi, pmaugeri

I always don't understand the meaning of -LOD in gatk IndelRealigner. Do you know that?

Appreciate your help!

Regards.

[adaptivegenome]

LOD is used to set threshold for realignment. GATK tries to avoid realigning reads unless the impact is likely to be significant.

[xlzhang]

LOD is used to set threshold for realignment. GATK tries to avoid realigning reads unless the impact is likely to be significant.

Thank you!