We’re moving the GATK website, docs and forum to a new platform. Read the full story and breakdown of key changes on this blog.
If you happen to see a question you know the answer to, please do chime in and help your fellow community members. We encourage our fourm members to be more involved, jump in and help out your fellow researchers with their questions. GATK forum is a community forum and helping each other with using GATK tools and research is the cornerstone of our success as a genomics research community.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.
TCGA WGS PoN for CNV

Hi,
I would like to use a whole genome sequencing TCGA panel of normals for the GATK CNV workflow. I have dbGaP access and can use the controlled access TCGA workspaces. So I was wondering if of such a PoN exist somewhere already or if this is something I have to create with the CNV_Somatic_Panel workflow. I think there are about 1000 WGS normals I could use from the controlled access TCGA workspaces. If I use the CNV_Somatic_Panel workflow do I have to go up on memory in the configuration to process that many samples? What would be good parameters?
Thanks,
F
Best Answer
-
slee ✭✭✭
I'd recommend just taking >50 blood normals from relatively quiet tumor types (we typically use THCA) and creating a PoN with your desired bin length. Typically only a few (<3) principal components are needed to achieve a good denoising result for WGS, so we really just need enough samples to nail down the per-bin medians.
Coverage collection typically takes a few hours per BAM and is done in parallel, and building the PoN itself takes ~10 minutes, so the whole process takes just a few hours. The memory requirement for the last step will depend on the number of samples and the bin length, but since it's so short running you may just want to err on the safe side when setting the memory parameter.
For example, I built a PoN from 50 THCA normals with 250bp bins, using n1-highmem-4s for coverage collection and an n1-highmem-16 for building the PoN (both of which were probably overkill in terms of memory). Using non-preemptible instances (which are ~5x the cost of preemptible instances), the total cost came out to ~$25.
Answers
I'd recommend just taking >50 blood normals from relatively quiet tumor types (we typically use THCA) and creating a PoN with your desired bin length. Typically only a few (<3) principal components are needed to achieve a good denoising result for WGS, so we really just need enough samples to nail down the per-bin medians.
Coverage collection typically takes a few hours per BAM and is done in parallel, and building the PoN itself takes ~10 minutes, so the whole process takes just a few hours. The memory requirement for the last step will depend on the number of samples and the bin length, but since it's so short running you may just want to err on the safe side when setting the memory parameter.
For example, I built a PoN from 50 THCA normals with 250bp bins, using n1-highmem-4s for coverage collection and an n1-highmem-16 for building the PoN (both of which were probably overkill in terms of memory). Using non-preemptible instances (which are ~5x the cost of preemptible instances), the total cost came out to ~$25.
Thank you!