Skip to content

Latest commit

 

History

History
55 lines (46 loc) · 2.77 KB

drac.md

File metadata and controls

55 lines (46 loc) · 2.77 KB

How to install and run AMOEBAE on Digital Research Alliance of Canada (DRAC) clusters

Note: This is mostly relevant for AMOEBAE users in Canada. However, this approach can be easily adapted to any high performance computing cluster.

AMOEBAE depends on Conda packages, but installing Conda packages directly on DRAC isn't appropriate for the system. One way to circumvent that issue is to install dependencies in an Apptainer container, by performing the following steps.

  1. Log on to a DRAC cluster login node, clone the amoebae repository, and navigate into the amoebae directory. Upload your sequence data, as needed using an appropriate method (see DRAC documentation on data transfers). Also, I recommend doing the subsequent steps in a screen or tmux session so that your work won't be interrupted if you get disconnected (see DRAC documentation on prolonging terminal sessions).

  2. Build the container using the pixi.def definition file provided in the AMOEBAE repository (see DRAC Apptainer documentation):

    module load apptainer
apptainer build --disable-cache pixi.sif pixi.def
    • This will take a few minutes.

  3. To run AMOEBAE workflow steps requiring internet access (download_queries and download_dbs), enter a shell session within the container on the login node (unless using the Cedar cluster specifically):

    • Enter the shell session: 
      APPTAINER_BIND=''
apptainer shell -C -B $PWD:/root --pwd /root pixi.sif
    • Now you are in an environment where all the AMOEBAE dependencies are installed. You can run snakemake commands as described in the workflow protocol, without the need to prefix with pixi run .
    • Run the specific rules in the workflow (if/when needed as part of the workflow protocol): 
      snakemake --cores 1 download_queries
snakemake --cores 1 download_dbs
    • Exit the shell session: 
      exit

  4. Otherwise, start an interactive session, so that your analysis will be run on a compute node. For example, like this (see DRAC documentation):

    salloc --time=1:0:0 --mem-per-cpu=16G --ntasks=1 --account=def-leppard

  5. Then, enter a shell session within the container:

    module load apptainer
APPTAINER_BIND=''
apptainer shell -C -B $PWD:/root --pwd /root pixi.sif

  6. Now run the remaining workflow steps, as described in the workflow protocol.

    snakemake get_ref_seqs 
...