Notice:
If you happen to see a question you know the answer to, please do chime in and help your fellow community members. We appreciate your help!

Test-drive the GATK tools and Best Practices pipelines on Terra


Check out this blog post to learn how you can get started with GATK and try out the pipelines in preconfigured workspaces (with a user-friendly interface!) without having to install anything.

GATK 4.1 HaplotypeCaller detected the same variant in one technical replicate, but failed in another

yingchen69yingchen69 nanjingMember

Hi,

I tried to use GATK 4.1 HaplotypeCaller to detect EGFR T790M variant in a few technical replicates. It turned out that HaplotypeCaller detected the same variant in one technical replicate, but failed in another replicate. The sample is the commercial EGFR T790M positive control.

The following are the IGV screenshots, the first one HC called T790M while the bottom one HC failed even though we can see T790M in both pre-processed and HC bamout bam files.

I also have a few diluted sample runs with no T790M detection and the bamout bam files have no read at all at that region while the reads are there in the pre-processed bam files.

These are amplicon based data with very high depth ( > 12,000x). I tried to adjust parameters and so far --max-reads-per-alignment-start 0 --disable-read-filter NotDuplicateReadFilter --adaptive-pruning true --kmer-size 10 --kmer-size 15 --kmer-size 20 --kmer-size 25 --kmer-size 30 is the best I can get as I can identify T790M from 4 out of 12 T790M positives.

Any suggestion?

Thanks a lot for the help!

Ying

Answers

  • SkyWarriorSkyWarrior TurkeyMember ✭✭✭

    Are these somatic mutation control samples? If so HC is not the tool to use. You should try Mutect2.

  • yingchen69yingchen69 nanjingMember

    No. These are not for somatic calls. These are positive control samples for testing of a targeted tumor panel. We want to make sure that our analysis pipeline can pick up these variants.

    Thanks,

    Ying

  • SkyWarriorSkyWarrior TurkeyMember ✭✭✭
    These are positive control samples for testing of a targeted tumor panel. 
    

    Yes these are somatic positive control samples.

    Your variant ratio is 11 percent in one sample and 18 percent in another sample. haplotype caller is not suited for somatic calls.

  • yingchen69yingchen69 nanjingMember

    Hi,

    @SkyWarrior, these are not for somatic calls. Maybe I did not put it clear. These are positive control samples for testing the analysis workflow of a targeted tumor panel. The two samples I mentioned are the technical duplicates. In each IGV screenshot, the top track is the bam from GATK pre-process, and the bottom track is from GATK 4.1 HaplotypeCaller bamout.

    Since these are amplicon based data with very high depth, we expected some difficulties, and we just try to find a better solution. We tried to adjust parameters, but have not found any magic bullet yet.

    Best,

    Ying

  • SkyWarriorSkyWarrior TurkeyMember ✭✭✭

    Maybe I am misunderstanding however what is the explanation for the 18% and 11% pileup ratios that you show in the picture? Even with amplicon data germline samples fall somewhere between 40 to 60 percent pileup ratios however these look like what you would expect from an FFPE sample.

  • yingchen69yingchen69 nanjingMember

    @SkyWarrior, we did serial dilutions to the sample trying to find detection limits.

    Best,

    Ying

  • SkyWarriorSkyWarrior TurkeyMember ✭✭✭

    How did you dilute your sample? Simply reducing the DNA concentration or adding more normal DNA into the control DNA?

  • yingchen69yingchen69 nanjingMember

    @SkyWarrior, we did serial dilutions to the sample by adding more normal DNA.

    Best,

    Ying

  • SkyWarriorSkyWarrior TurkeyMember ✭✭✭
    edited March 26

    But that changes the ploidy of the DNA that you have. For each dilution step you need to adjust the ploidy variable for HaplotypeCaller and your experiment will pretty much go in vain since you wouldn't be able to tell the ploidy for your actual samples. That's why Mutect2 is the way to go for this experiment.

    Adding more normal DNA makes this experiment pretty much a simulation of somatic variability in tumors.

Sign In or Register to comment.