mirror of
https://github.com/MillironX/nf-configs.git
synced 2024-11-22 16:29:55 +00:00
1.6 KiB
1.6 KiB
nf-core/configs: CRA HPC Configuration
nfcore pipeline sarek and rnaseq have been tested on the CRA HPC.
Before running the pipeline
- You will need an account to use the CRA HPC cluster in order to run the pipeline.
- Make sure that Singularity and Nextflow are installed.
- Downlode pipeline singularity images to a HPC system using nf-core tools
$ conda install nf-core
$ nf-core download
- You will need to specify a Singularity cache directory in your ~./bashrc. This will store your container images in this cache directory without repeatedly downloading them every time you run a pipeline. Since space on home directory is limited, using lustre file system is recommended.
export NXF_SINGULARITY_CACHEDIR = "/lustre/fs0/storage/yourCRAAccount/cache_dir"
- Download iGenome reference to be used as a local copy.
$ aws s3 --no-sign-request --region eu-west-1 sync s3://ngi-igenomes/igenomes/Homo_sapiens/GATK/GRCh38/ /lustre/fs0/storage/yourCRAAccount/references/Homo_sapiens/GATK/GRCh38/
Running the pipeline using the adcra config profile
- Run the pipeline within a screen or tmux session.
- Specify the config profile with
-profile adcra
. - Using lustre file systems to store results (
--outdir
) and intermediate files (-work-dir
) is recommended.
nextflow run /path/to/nf-core/<pipeline-name> -profile adcra \
--genome GRCh38 \
--igenomes_base /path/to/genome_references/ \
... # the rest of pipeline flags