two questions

[zmz1988]

Hi PengNi,

Thanks for providing the nice tool. Here I got two questions to ask you, if you don't mind.

First, I'm running modification base calling step with command CUDA_VISIBLE_DEVICES=0 deepsignal_plant call_mods. However, our cluster admin doesn't allow me to use CUDA_VISIBLE_DEVICES=0 on our GPU cluster, because the cluster setting already set this number. However, when I delete this CUDA_VISIBLE_DEVICES=0, my job seems not running even if it was assigned to a GPU device. May I ask how I can run on a GPU device if I don't set this option in the command? I mean I submit the job and request a GPU device for the job to run.

Secondly, may I ask when will the new version of deepsignal come? I have Nanopore R9.4.1 data, and I think deepsignal is so far the best option for the old Nanopore data to call methylation in all context. But the large file number generated by tombo limits how many jobs I can run in parallel. I have >20 samples to be called modified bases, and each takes more than a week to finish, if everything goes well. So far I can only run one sample per time due to the number of single read fast5. Will the new deepsignal improve this problem? Or how can I solve this problem with your expertise?

Thanks a lot and looking forward to your reply!

[PengNi]

Hi @zmz1988 , sorry for the delay.

1. without setting CUDA_VISIBLE_DEVICES, deepsignal-plant will most likely use all available GPUs automatically. However, I don't have a cluster like yours, so I can not try in an environment exactly the same as yours.

2. Yes, the time cost issue is something we are now planning to improve. We'd like to skip (or change) the tombo resquiggle step, maybe we can release an new version at the end of this year.

Thank you very much for using deepsignal-plant.

Best,
Peng