GATK downsampling

Hi GATK team,

I am trying to do downsample analysis using gatk PrintReads. The GATK version is 4.0.4.0.
Below is my command:
gatk PrintReads -R path/to/hg19.fa -I LIB.bam -O LIB_downsample10.bam --downsample_to_coverage 10

But it always threw out this error:
A USER ERROR has occurred: downsample_to_coverage is not a recognized option

When I changed the '--downsample_to_coverage' to '-dcov', the error became this:
A USER ERROR has occurred: d is not a recognized option

Could you please help?

Thank you very much.
Li

Tagged:

Answers

  • SheilaSheila Broad InstituteMember, Broadie, Moderator admin

    @zhouli
    Hi Li,

    We are recommending DownsampleSam for downsampling.

    -Sheila

  • zhoulizhouli Member

    Hi Sheila,

    I have tried DownsampleSam before, but it only allows downsample to a certain fraction of original reads. In fact, I wanted to downsample to a certain coverage. Is there any other tool to do this?

    Thanks a lot.
    Li

  • zhoulizhouli Member

    Hi Sheila,

    Following my previous comments.
    In other words, how to use the CommandLineGATK arguments "--downsample_to_coverage" to do the downsampling? Which GATK tool can be used with "--downsample_to_coverage" arguments?

    Thank you very much.
    Li

  • shleeshlee CambridgeMember, Broadie, Moderator admin

    Hi @zhouli,

    These days, if your sequencing center is doing a good job, your coverage should be even throughout your alignments and so fractional downsampling should be sufficient for your needs. Whatever uneven coverage remains is indicative of possible copy number variation. Are you working with very old data? In this case, I think your only recourse is to use GATK3, where the --downsample_to_coverage option was last available.

    Adding to Sheila's offering of DownsampleSam, please also check out PositionBasedDownsampleSam.

  • zhoulizhouli Member

    @shlee said:
    Hi @zhouli,

    These days, if your sequencing center is doing a good job, your coverage should be even throughout your alignments and so fractional downsampling should be sufficient for your needs. Whatever uneven coverage remains is indicative of possible copy number variation. Are you working with very old data? In this case, I think your only recourse is to use GATK3, where the --downsample_to_coverage option was last available.

    Adding to Sheila's offering of DownsampleSam, please also check out PositionBasedDownsampleSam.

    Hi Shlee,

    Thank you for your reply. I will try both methods.

    Cheers,

Sign In or Register to comment.