Detecting low frequency mutations - GATK

[spreeth84]

I am analyzing a targeted enrichment of a Hapmap blend. I have 4 hapmap samples mixed in the ratio: 74:20:5:1
I am looking to detect variants at low frequencies (upto 0.5%).
I am using GATK for the SNP analysis - I carried out sample level local realignment around indels and base quality score recalibration.
With Unified Genotyper, I am getting a very good expected: actual ratio for SNPs above 10% expected frequency. However, I am not picking up any of the SNPs in the 0.5 to 10% range. I do see a lot of these SNPs (on CLC bio Genomics workbench) and pick them up the CLC bio SNP caller (along with 100's of false positives). I understand a lot of them may be poor quality or may have been rejected by GATK due to strand bias or other problems. But the fact that not one of the 20 expected calls were picked up makes me wonder if I am doing something wrong.
The coverage in these positions is very high (~2000). Sequencing was done on Illumina. I used the default quality score settings on GATK. Is there any parameter that I can work around to increase the sensitivity on this particular set and pick up these low frequency SNPs ?
Any help would be appreciated.
Thanks,
Preethi

[swe]

Low frequency SNPs with GATK?

I found this post and have about the same question as spreeth84 so I will send this and see if this thread can come to life.

I have 30 small haploid bacterial genomes where I would like to find SNPs at low frequency. Ive used the HaploType-caller

java -jar GenomeAnalysisTK.jar -T HaplotypeCaller -R reference.fna -I S1.bam -I S2.bam -I S3.bam -I S4.bam .... -I S30.bam -ploidy 1 -stand_call_conf 30 -stand_emit_conf 10 -o variants.vcf -dcov 1000000000

Im using the GenomeAnalysisTK-3.3-0-g37228af version of GATK.

Im doing multiple sample calling and I see SNPs with lower frequencies (say 20%) but only when some other sample have a higher frequency at the same position.

Can I get all SNPs, not only high frequency ones with GATK, or should I use some other software?

[spreeth84]

Hi swe, I have had better luck with Mutect than Haplotype caller or Unified Genotyper for calling low frequency alterations. There are parameters that can be further tweaked in Mutect to achieve higher sensitivity at low minor allele frequencies. I have never worked with Haploid bacterial genomes, so I don't know if there is any other reason this wouldn't work for you - but may be worth a shot.

[Zaag]

We also tried Mutect and it indeed calls more true low frequency mutations.

@spreeth84: What parameters do you tweak? Do you know if there is a proper manual?