Starting out with working on the FH cluster can be a daunting process. In this article, we will point you to the right information you need to get started.

We’ll use the example of aligning multiple FASTA files using BWA.

When You’re First Starting

Here is a visual index of the steps you need to get up and running on the cluster:

graph TD
G[Get Credentials] --> A
A[Logging into the Cluster] --> B
B[Uploading Your Data] --> C
H[Find Your Software] --> C
C[Write a Bash Script] --> D
C --> L[Run Using a Workflow Manager]
L --> F
D[Run your script using SLURM] --> E
E[Check job status] --> F
F[Locate your outputs]
click G "#credentials"
click A "#logging-in"
click B "/scicomputing/store_collaboration/"
click C "/scicomputing/software_linux101/#shell-scripting"
click D "/pathways/path-batch-computing/#steps"
click E "/scicomputing/compute_jobs/#managing-jobs"

Credentials

The first thing you’ll need to do is request credentials for the FH cluster.

https://sciwiki.fredhutch.org/scicomputing/access_credentials/

Logging In

The next step is to log in to the cluster through the rhino machines. These are powerful machines that can run code on their own, but act as controllers for the other machines in the cluster.

https://sciwiki.fredhutch.org/scicomputing/access_credentials/#accessing-slurm-clusters

Upload Data

The next thing to do is to move your files to either:

  • /hpc/temp
  • /fh/working

These file systems have the advantage of being directly connected to the cluster and are faster. They are meant to be working storage, so when you are done processing your files, you will need to transfer them back to either /fh/fast/ or your own system. Note that /hpc/temp/ and /fh/working/

To move files, you can use Motuz, which is the web app, to move files back and forth from - your computer - from other Fred Hutch storage (such as /fh/fast/) - or from an AWS S3 bucket

HutchGO is also available for transferring files, which uses an App called Globus Connect.

You can also use an SFTP (secure FTP) client such as [Mountain Duck to transfer files as well.

Find the Reference Genome

  • There are a number of reference genomes that are available for use.

  • For example, for aligning files using bwa, we’ll need the BWAIndex file that is available here:

    • /shared/biodata/reference/iGenomes/Homo_sapiens/UCSC/hg19/Sequence/BWAIndex

Find Your Software

Most software that you need for bioinformatics work are installed and can be utilized via software modules. How do we find them? We can use module avail bwa to look for the proper module:

module avail bwa

We will get the following response:

------------------------------- /app/modules/all -------------------------------
   BWA/0.7.17-GCC-8.3.0     BWA/0.7.17-GCCcore-11.2.0
   BWA/0.7.17-GCC-10.2.0    BWA/0.7.17-GCCcore-12.2.0 (D)

  Where:
   D:  Default Module

Use `module spider` to find all possible modules and extensions.
Use `module keyword key1 key2 ...` to search for all possible modules matching
any of the "keys"

So now we can load the module up using:

module load BWA/0.7.17-GCCcore-12.2.0

If you can’t find a software module, there is Apptainer and other ways to get the software running in the HPC environment.

Batch Processing Files

At this point, we have two main paths to run the software:

  1. We can use the built in job scheduler, SLURM, to batch process our files.
  2. We can automate our batching using a WDL (Workflow Description Language) file and PROOF

Using SLURM

You can use SLURM directly to run compute jobs, which is the low level task manager for the cluster. There is a guide to SLURM for parallel jobs here.

- Pros: This is the fastest way to get a single set of tasks running.
- Cons: It is difficult to execute SLURM jobs that have multiple steps in a pipeline

Batch Jobs with Workflow Managers

Workflow managers such as Cromwell and Nextflow are made to handle multi-step workflows. For example, you may want to process files with BWA and then apply Mutect to each file.

- Pros: reproducible and can be extended to multiple steps with the WDL file. Handles intermediate outputs, and is highly configurable.
- Cons: a little more of a learning curve to write your own WDL files

Updated: