I have managed to run Rhapsody on a computing node (HPC Slurm cluster). Since it was not a head node, it has no network connection to the outside world, and no Docker.

You will need:

• javascript/node on the PATH, no clue why sometimes it did not work with the container. Maybe toil was more picky
• apptainer (make sure you got unsquashfs. I got mine using conda and squashfs-tools
• in $CWL_SINGULARITY_CACHE : bdgenomics_rhapsody:2.0.sif maybe also node_alpine.sif. You may sftp these from your workstation if the HPC headnode somehow also has restricted connections to the outside .
• cwltools
• TMPDIR set up on some large partition since the pipeline tends to fill in /tmp
• reference file from:

http://bd-rhapsody-public.s3-website-us-east-1.amazonaws.com/Rhapsody-WTA/Pipeline-version2.x_WTA_references/

my script:

#!/usr/bin/bash

#SBATCH --job-name=rhapL1
#SBATCH --nodes=1
#SBATCH --time=10:00:00
#SBATCH --cpus-per-task=32
#SBATCH --mem=64G
#SBATCH --partition=foobar

export CWL_SINGULARITY_CACHE=/path/to/singu_cache/
export PATH=/path/to/soft/progs/node_current/bin:$PATH

cwltool --singularity \
--outdir /some/output/path/rhapsody_lane1  \
--debug \
--cachedir /some/output/path/rhapsody_lane1/cache \
rhapsody_pipeline_2.0.cwl pipeline_inputs_lane1.yml

The above is quite brain dead attempt at executing the whole pipeline sequentially on a one computing node. But at least it did work.

I also was experimenting with toil, but so far without a success with Rhapsody.

Last but not least: if you are curious poke around inside of the container. Compiled QualCLAlign program I guess executing STAR etc.

edit reference file link

2025 update using pipeline version v2.2.1 on a HPC SLURM cluster with Apptainer available as a module. That module contains Node.js.

Most critical: Do not install cwl-runner via conda or anything but strictly use pip. Using anything but pip resulted in obscure pipeline errors I could not debug. Might be some sort of version and/or dependency clash, I did not find out.

mamba create --name bd_rhapsody python
mamba activate bd_rhapsody
pip install cwlref-runner
module load Apptainer

As said, our Apptainer (Singularity) module seems to load Node.js already. If not in your case, maybe load via conda/mamba.

From there I simply executed the commands as instructed in the manual, for example for an AbSeq-only experiment:

# Typical SLURM headers ...
module load Apptainer
mamba activate bd_rhapsody
cwl-runner --cachedir $(pwd)/cache --outdir $(pwd)/out --singularity rhapsody_pipeline_2.2.1.cwl pipeline_inputs.yml

Be sure to set a cachedir, if not the pipeline spams (and maybe overloads?) /tmp. Both cache and output directories should be on partitions with lots of free space, on HPC that typically would be scratch partition.

where the content of pipeline_inputs.yml is:

#!/usr/bin/env cwl-runner
cwl:tool: rhapsody

Reads:

 - class: File
   location: "/scratch/.../foo_S3_L005_R1_001.fastq.gz"

 - class: File
   location: "/scratch/.../foo_S3_L005_R2_001.fastq.gz"

AbSeq_Reference:
 - class: File
   location: "/scratch/.../abseq.fa"

Cell_Calling_Data: AbSeq
Putative_Cell_Call: AbSeq

That worked quite fine "on my machine". No need to manually pull any containers.

Yes, BD Rhapsody can be run locally. In fact it uses a docker so the setup isn't really all that bad: https://www.bdbiosciences.com/content/dam/bdb/marketing-documents/BD_Single_Cell_Multiomics_Analysis_Setup_User_Guide.pdf

If you are deploying somewhere without docker, you will run into trouble as one particular step of the pipeline calls docker from within docker, so no amount of reverse-engineering the environment can get around it.

It should be pointed out that the FAQ you link does not say the software must be used via the seven bridges platform; only that it cannot be downloaded from seven bridges itself:

You cannot download the software from Seven Bridges. Please follow the BD Single-Cell Multiomics Analysis Setup User Guide on installing the pipeline for local use.