Sequenza failed >100 GB RAM

[rkendar]

Hi Francesco,

I have run sequenza 3.0.0 on WGS samples (hg38). On my previous run, I succeed to run the sequenza with 32GB of RAM. But now, it is failed with >100 GB RAM, even I tried running it with 300GB it still failed.

I thought it may be due to a different sample. But then when I tried to rerun the previous sample that succeeds to run with 32GB RAM, it failed in >100GB RAM. Then I found some different R package that loaded by sequenza in my previous one and the current one. Here is the screenshot of the R package.

> Previous run (succeed with 32GB RAM)

> Current run (failed with >100GB RAM)

Do you have any suggestions? Or why this can happen?

Regards,

Raden

[Michael Knudsen]

Did you manage to figure out what caused the excessive memory consumption? I have the same issue.

Thanks,

Michael

[Francesco Favero]

I am sorry, it's hard to keep up with the replies :D.

We have located the issues with a bug in vroom, which is used by readr to read seqz files.

Originally it was more memory efficient to use readr vs the standard R i/o methods. The bug has not being addresses in years, so maybe I should think to add an option to use the readr or not, or find a way to avoid  the vroom memory issue

Best

[Michael Knudsen]

Thanks for the details about what is causing the issue. I'm running Sequenza in a conda environment, and I managed to fix things by pinning R to 3.5 instead of 3.6.

[Davide Cittaro]

Hello, is your solution still working? I'm afraid I'm stumbling in a very similar issue

[Michael Knudsen]

Yup, it works fine. I'm running it in production without any issues. I'm using a conda environment with these two dependencies:

r-base ==3.5.1

r-sequenza ==3.0.0

/Michael