Skip to content

← Back to main page

Running SeqNado on HPC Clusters

SeqNado is designed to run efficiently on high-performance computing (HPC) clusters using job schedulers. This guide covers setting up and running SeqNado on SLURM-based clusters, which is the recommended and most thoroughly tested configuration.

Quick Start: SLURM Clusters

To run SeqNado on a SLURM cluster with containerized environments:

seqnado pipeline atac --preset ss --queue short --scale-resources 1.0

The ss (SLURM Singularity) preset automatically configures job submission to SLURM with container isolation.

See CLI options for presets, queues, and scaling: seqnado pipeline.

SLURM Configuration

Preset: ss (SLURM Singularity)

The SLURM Singularity preset uses:

  • Executor: SLURM job scheduler
  • Containerization: Singularity/Apptainer containers (recommended for HPC environments)
  • Default partition: short (1 hour runtime, 3 GB memory per job)
  • Max parallel jobs: 100

Key Options

When running with the SLURM preset:

Queue/Partition

Specify the SLURM partition:

seqnado pipeline atac --preset ss --queue short

Common partitions:

  • short - Default, 1 hour runtime, general purpose
  • long - Extended runtime (varies by system)
  • gpu - GPU-enabled partition (for GPU-accelerated steps like MCC)

Resource Scaling

Scale memory and runtime resources by a factor:

seqnado pipeline atac --preset ss --scale-resources 1.5

A factor of 1.5 increases all memory and time allocations by 50%. Use this if jobs are being killed due to resource limits.

Verbose Output

Print the generated Snakemake command before execution:

seqnado pipeline atac --preset ss --print-cmd

For a full list of flags, see seqnado pipeline.

GPU Support

For assays using GPU acceleration (e.g., MCC peak calling):

  1. Submit to a GPU partition: 

    seqnado pipeline atac --preset ss --queue gpu

  2. SeqNado automatically allocates GPU resources for GPU-enabled steps.

Multiomics Workflows

For multiomics experiments combining multiple assays:

seqnado pipeline --preset ss --configfile config.yaml

Ensure your configuration file includes all assays and their respective genomes.

Other Execution Options

While SLURM+Singularity is recommended for HPC, SeqNado provides several other profiles:

Local Execution Presets

These options are useful for development, testing, or non-HPC environments:

le (Local Execution)

Standard local execution with environment isolation:

seqnado pipeline atac --preset le

Suitable for:

  • Development machines
  • Local testing
  • Systems without job schedulers

lc (Local Cluster)

Local execution with job-level parallelization (GNU parallel):

seqnado pipeline atac --preset lc

Useful for:

  • Multi-CPU workstations
  • Quick local runs with parallelization

ls (Local Single-threaded)

Single-threaded local execution:

seqnado pipeline atac --preset ls

Useful for:

  • Debugging
  • Testing without parallelization

Container Options

By default, all presets use Singularity/Apptainer containers for reproducibility. To use a local Conda environment instead:

Modify your workflow configuration YAML:

container: null  # Disable containers

Then run with any preset. The workflow will execute with your current Python environment.

Configuration Files

Genome Configuration

Ensure genomes are configured before running pipelines:

seqnado genomes list atac

More details: seqnado genomes.

Workflow Configuration

Generate an assay-specific configuration:

seqnado config atac --output config_atac.yaml

More details: seqnado config.

Edit the YAML to customize:

  • Input/output paths
  • Reference genomes
  • Analysis parameters
  • Resource allocations

Design Metadata

Generate metadata CSV from FASTQ files:

seqnado design atac /path/to/fastqs/*.fastq.gz

More details: seqnado design.

Troubleshooting

Jobs Terminated Due to Resource Limits

If Snakemake jobs are killed with out of memory or timeout errors:

seqnado pipeline atac --preset ss --scale-resources 2.0 --queue long
  • Increase --scale-resources factor
  • Switch to a partition with longer walltime
  • Check cluster documentation for memory limits

Apptainer/Singularity Not Found

SeqNado prefers Apptainer (modern Singularity fork). If unavailable:

  1. Install Apptainer on your cluster
  2. Or switch to local environment execution (modify config YAML)
  3. Or request Singularity installation from cluster admin

SLURM Submission Errors

Verify SLURM is available:

sinfo  # List SLURM partitions
squeue # Check submitted jobs

Add --print-cmd to see the exact Snakemake command and submission details.

Container Image Download Issues

SeqNado pulls container images on first use. If downloads fail due to network restrictions:

  1. Use cluster's container cache if available
  2. Pre-download and cache images on compute nodes
  3. Contact cluster support for container registry access

Advanced: Custom Snakemake Configuration

For advanced users, profiles can be customized by editing the configuration files in:

~/.config/seqnado/profiles/profile_slurm_singularity/

Refer to Snakemake documentation for configuration options.

Support

For issues specific to cluster execution:

  1. Check cluster documentation for scheduler syntax and available partitions
  2. Verify containers are accessible from compute nodes
  3. Test with --preset le on the login node first
  4. Use --print-cmd to debug generated Snakemake commands