Setup

Prerequisites

Install Nextflow

Follow the official Nextflow installation guide.

Nextflow requires:

  • Java 11 or later
  • Linux or macOS (Windows via WSL2)

The quick version is:

# Optional, if you don't have Java 17+
# Check your Java version
# java -version

# Install Java 17+, if you don't have it
# curl -s https://get.sdkman.io | bash
# sdk install java 17.0.10-tem

# Install Nextflow
curl -s https://get.nextflow.io | bash
chmod +x nextflow
mv nextflow $HOME/.local/bin/

# Ensure $HOME/.local/bin/ is on your PATH
# you should also add this line to your ~/.bashrc or ~/.zshrc if required
export PATH="$PATH:$HOME/.local/bin"

# Test it works
nextflow info

Clone the repository using git

For development, it's often more convenient to clone the git repository directly:

git clone https://github.com/Australian-Protein-Design-Initiative/nf-binder-design
cd nf-binder-design

# See the help as a first test
nextflow run main.nf --help

Alternative: pull using Nextflow

To pre-cache the nf-binder-design repository in ~/.nextflow/assets/, run:

nextflow pull Australian-Protein-Design-Initiative/nf-binder-design

you can pull a specific version like:

nextflow pull -r 0.1.4 Australian-Protein-Design-Initiative/nf-binder-design

Test that you have successfully cached the repository by running:

nextflow run Australian-Protein-Design-Initiative/nf-binder-design/main.nf --help

Note: The documentation examples currently assume you have git clone'd the repository, but can be adapted to use this alternative method. In this case, platform-specific config files will be in ~/.nextflow/assets/Australian-Protein-Design-Initiative/nf-binder-design/conf/platforms/ - in the future these will be turned into config profiles to simplify usage.

Container Setup

These workflows use Apptainer containers for reproducibility. You'll typically need Apptainer installed - this is usually already available on modern HPC clusters.

Containers are automatically downloaded when running the pipeline - you will need ~60Gb+ of storage space for the containers.

Model weights / parameters are currently packaged inside the containers - this makes the containers large, but simplifies the workflow setup. In the future we will very likely provide more compact alternative containers without weights.

Platform-Specific Configuration

Local Workstation / single compute node

For local execution of a single compute node, use the -profile local flag when running workflows. See examples/pdl1-rfd/nextflow.dual-gpu.config for a config example using dual-GPUs.

HPC Clusters with SLURM

For running using the SLURM executor, use the -profile slurm flag (this is the default profile if not specified).

Some site-specific configurations for HPC clusters and cloud platforms can be applied via eg -profile m3. Currently defined sites are:

  • m3 - the Monash M3 HPC cluster, generic GPU partitions
  • m3_bdi - the Monash M3 cluster with access to the bdi partitions
  • mlerp - the MLeRP HPC cluster
  • hyperqueue - for running on a generic HyperQueue (hq) cluster

These configuration files are provided in conf/platforms/ and be copied and adapted for your specific setup. Pull requests for additional HPC clusters are welcome!

In addition (on instead of) using a pre-defined -profile, you can use the -c flag to specify the path to a configuration file:

nextflow run main.nf -c conf/platforms/m3.config  # etc ...

HPC Cluster environment variables

Set up your Apptainer cache directory to store downloaded containers:

# Add to your ~/.bashrc or set before running
export APPTAINER_CACHEDIR=${HOME}/.apptainer/cache
export NXF_APPTAINER_CACHEDIR=${APPTAINER_CACHEDIR}

Changing this location is particularly important on HPC clusters where you may have limited space in your home directory.

Ensure your temporary directory is set to a location with sufficient space, eg:

export TMPDIR=/scratch2/myproject/tmp
export NXF_TEMP=$TMPDIR
mkdir -p $TMPDIR

Next Steps

See the Usage page for examples and detailed workflow documentation.