Can anyone help me with normal-only calling for panel of normals(PON) with MuTect2.

Question

[ShahiRB]

Can anyone help me with normal-only calling for panel of normals with MuTect2.

The code below does not spit any error as well as variants in vcf, except vcf header.

Here is the code:

java -Xmx45g -XX:+AggressiveOpts -jar /nexus/GenomeAnalysisTK-3.5/GenomeAnalysisTK.jar \
-T MuTect2 \
-R /home/ngs/data/ucsc.hg19.fasta \
-I:tumor /nexus/normal.bam \
--dbsnp /home/ngs/data/dbsnp_137.hg19.vcf \
--cosmic /nexus/rshahi/data/CosmicCodingMuts.vcf5 \
--artifact_detection_mode \
--intervals /nexus/NimbV3_target.gatk.interval_list \
-o $HOME/jobout/normal_only.vcf

Thanx,
Rajendra
Brussels

Answer 1

[DG]

The most likely source of error is your intervals list. If this is incomplete for instance, or not matching up with the regions you have data for, you would get no variants and also no error messages (unless the intervals list file was actually malformed). Try running the same command but without passing it an intervals file and see what happens.

You should also probably check your BAM file to make sure it is actually containing aligned data as expected. Visualize it with IGV for instance and check some genomic coordinates of your favourite genes as a quick check.

[ShahiRB]

Hi Dan,

1. Even without passing intervals list I got the same results i.e no error, no variants in vcf, header only.
2. The bam file is perfect.

Any further suggestions?

Regards,

[DG]

Maybe try changing your java calling? It's possible something is going on with the JVM. You are passing 45G of RAM and Aggressive Opts to Java. If the BAM file is for a single exome or targeted resequencing experiment (assuming since you're passing a Nimblegen targets list) then 45G is about 10x-20x what is actually needed. If GATK runs out of memory it will throw an error, but if the problem is earlier on with the JVM then it could potentially die silently, especially with the -XX:+AggressiveOpts also being passed. That will turn on a whole suite of options and I'm not sure what effect that may have as I have never used it.

If neither of those seem to be the case it is hard to diagnose without trying on your same BAM file (or a portion of the BAM file) to diagnose. There doesn't seem to be anything wrong with the command line. When you look at the BAM file in IGV or something can you find obvious variants bye eye that should be called? If you throw the BAM at FreeBayes or another variant caller do you get results then?

[ShahiRB]

@Dan, I tried without XX:+AggressiveOpts argument; still the same problem. But the HaplotypeCaller runs perfect with the same bam and intervals list.

[DG]

Well I'm stumped. I would suggest you post with the GATK directly. Since MuTect2 is still in Beta you may have come across some sort of bug or edge-case.

[ShahiRB]

@Dan,hanx for your effort. I will update you with some tweaks from Geraldine_VdAuwera (GATK). Cheers!

Answer 2

[2nelly]

Don t worry your command is correct, it s not your fault. There is a bug in MuTect2 when try to create a PON

http://gatkforums.broadinstitute.org/gatk/discussion/6789/mutect2-panel-of-normals-empty-vcf-fileption

I am facing the same issue.

Some people claim that the problem was solved but I am still getting an empty vcf file.

Try the nightly build

Answer 3

[igor]

It looks like the --artifact_detection_mode flag is for HaplotypeCaller (the GATK recommendation is to use HaplotypeCaller for generating the panel of normals): https://www.broadinstitute.org/gatk/guide/tooldocs/org_broadinstitute_gatk_tools_walkers_cancer_m2_MuTect2.php

[DG]

I believe that is a mistake in the documentation. I don't think HC has the artifact_detection_mode as an option.

[igor]

I haven't tried MuTect2 yet. I did think that was odd, which is why I thought it might be causing the error.

Answer 4

[ShahiRB]

Thanx igor, but that doc abt mutect2 looks misleading . Actually HC does not have that argument.

[DG]

Hi biotechnepal, this should really be added as a comment to igor's answer as opposed to an answer itself.

Answer 5

[winter_li]

Hi, I think maybe you can find GATK4 to have a try, I heard it has been published.

[DG]

Still in development. 3.5 is still the last official release