GPU-accelerated ONT Nanopore K8S basecalling

Run basecalling

Computations are running from bioit proxy (use ssh to connect). Analysis is then deployed to k8s on kuba cluster.

1. Create folders for results, project and temps; add their location to zaloha_nextflow.config

k8s {
...
  launchDir = 'PATH_TO_YOUR_LAUNCHDIR'
  projectDir = 'PATH_TO_LOCAL_CODE_FROM_GITHUB_REPO'
  workDir = 'PATH_TO_YOUR_TEMP'
...
}

2. Enter flowcell and kit specification along with path for fast5 files (parameter --data '../') Run the script as in the example bellow (run.sh)

nextflow kuberun xsvato01/nanopore_k8s/main.nf -r main -pod-image 'cerit.io/nextflow/nextflow:22.05.0' -c zaloha_nextflow.config  //
        --flowcell FLO-FLG001 --kit SQK-LSK110 --data '/mnt/shared/MedGen/ONTdata/k8s_testing/20220614_1150_MN16014_ais387_bfa74e7a/fast5_pass'

Using the GPU-accelerated container in other pipelines

To use the preinstalled Guppy-GPU container in eg. Nextflow pipelines, use the container from

'registry.gitlab.ics.muni.cz:443/450402/nanopore_k8s:latest'

In Nextflow, you can facilitate this container by specifying it in the .config file or the actual script:

process XXX {
     accelerator XY
     container 'registry.gitlab.ics.muni.cz:443/450402/nanopore_k8s:latest'

    script:
      """
      guppy_basecaller ....
      """

}

where XY denotes the number of GPUs to facilitate (maximum 2); local_nextlow.config is already pre-set to use gpu-accelerated basecalling for any process starting with name GUPPY_*