6-99bp indels with BWA/GATK

[gavin.oliver]

Hi,

I am using BWA and GATK to detect mutations in BRCA1. The BRCA1 sequences have been Sanger validated and contain known mutations. I am achieving a fair degree of accuracy so far, successfully detecting 99% of SNPs and over 90% of Indels. The majority of false negatives are for Indels over 5 bp in size. These range from 6-99bp in length. Can anyone recommend what command line parameters/values could be used to get the aligner to pick up some of the larger indels?

Thanks in advance.

[gavin.oliver]

I am now getting all Indels up to 29bp in length. I achieved this by increasing the maximum number of permitted gap extensions with bwa aln -e 50.

I will continue to experiment in order to get the larger indels.

[adaptivegenome]

Do you perform a base recalibration step with GATK before calling indels?

[gavin.oliver]

Do you perform a base recalibration step with GATK before calling indels?

Indeed I do.

[oiiio]

I have been trying to call indels with GATK UnifiedGenotyper from BWA-mapped BAMs for some time now, but with no success.

Did you have to use anything outside of the default parameters with UnifiedGenotyper or COuntCovariates/TableRecalibration? Others with this problem have found that it could be sequencing error rates in the sample were too high.

If you dont mind, could you post a couple command lines from your pipeline? I'm particularly interested in your UnifiedGenotyper and base recalibration commands. It would be an immense help.

[gavin.oliver]

I have been trying to call indels with GATK UnifiedGenotyper from BWA-mapped BAMs for some time now, but with no success.

Did you have to use anything outside of the default parameters with UnifiedGenotyper or COuntCovariates/TableRecalibration? Others with this problem have found that it could be sequencing error rates in the sample were too high.

If you dont mind, could you post a couple command lines from your pipeline? I'm particularly interested in your UnifiedGenotyper and base recalibration commands. It would be an immense help.

I am pretty sure my commands are very standard. Nonetheless, you are welcome to have a look!

for file in *fastq; do bwa aln -e 50 -f ${file%%.fastq}.sai chr17hg19 ${file}; done

for file in *sai; do bwa samse chr17hg19 ${file} ${file%%.sai}.fastq > ${file%%.sai}.sam; done

for file in *bam; do java -Xmx3g -jar /home/goliver/ngs_software/picard-tools-1.53/SortSam.jar I=${file} O=${file%%.bam}_sorted.bam SO=coordinate; done

for file in *_sorted.bam; do java -Xmx3g -jar /home/goliver/ngs_software/picard-tools-1.53/MarkDuplicates.jar I=${file} O=${file%%.bam}_ndup.bam M=metric TMP_DIR=./tmp REMOVE_DUPLICATES=TRUE VALIDATION_STRINGENCY=LENIENT; done

for file in *ndup.bam; do java -jar /home/goliver/ngs_software/picard-tools-1.53/AddOrReplaceReadGroups.jar I=${file} O=${file%%.bam}_rg.bam SO=coordinate ID=1 LB=Z PL=illumina PU=Z SM=Z; done

for file in *rg.bam; do java -Xmx3g -jar /home/goliver/ngs_software/picard-tools-1.53/BuildBamIndex.jar I=${file} O=${file}.bai; done

for file in *rg.bam; do java -Xmx3g -jar /home/goliver/ngs_software/GenomeAnalysisTK-1.2-24-g6478681/GenomeAnalysisTK.jar -T RealignerTargetCreator -R ../ref_chr17.hg19.fa -o ${file%%.bam}.intervals -I ${file}; done

for file in *rg.bam; do java -Xmx3g -jar /home/goliver/ngs_software/GenomeAnalysisTK-1.2-24-g6478681/GenomeAnalysisTK.jar -I ${file} -R ../ref_chr17.hg19.fa -T IndelRealigner -o ${file%%.bam}_2.bam -targetIntervals ${file%%.bam}.intervals --known ../GATK/dbsnp_132.b37.vcf; done

for file in *_2.bam; do java -Xmx20g -jar /home/goliver/ngs_software/GenomeAnalysisTK-1.2-24-g6478681/GenomeAnalysisTK.jar -R ../ref_chr17.hg19.fa -knownSites ../GATK/dbsnp_132.b37.vcf -I ${file} -T CountCovariates -cov QualityScoreCovariate -cov DinucCovariate -cov ReadGroupCovariate -cov CycleCovariate -recalFile ${file%%.bam}.recal.csv --default_read_group 1 --default_platform illumina -nt 4; done

for file in *_2.bam; do java -Xmx3g -jar /home/goliver/ngs_software/GenomeAnalysisTK-1.2-24-g6478681/GenomeAnalysisTK.jar -l INFO -R ../ref_chr17.hg19.fa -T TableRecalibration -I ${file} -o ${file%%.bam}.final.bam -recalFile ${file%%.bam}.recal.csv --default_read_group 1 --default_platform illumina; done

for file in *final.bam; do java -Xmx3g -jar /home/goliver/ngs_software/GenomeAnalysisTK-1.2-24-g6478681/GenomeAnalysisTK.jar -T UnifiedGenotyper -glm BOTH -I ${file} -R ../ref_chr17.hg19.fa -o ${file%%.bam}.vcf; done

[Jon_Keats]

Do you have any paired-end data as opposed to single-ended as you methods suggest? The indel alignment should be better with paired-ends than single ends

[gavin.oliver]

Do you have any paired-end data as opposed to single-ended as you methods suggest? The indel alignment should be better with paired-ends than single ends

This particular dataset is all single end. I am pretty certain the larger indels can still be detected though...