Single Cell Analysis on Saturn Cloud

Requirements

• A saturn cloud account.
  • The free account is fine for this example.
• Optionally, instead of Saturn Cloud you can use any computer with conda/mamba and internet connectivity.
• Optional - S3 Bucket with Read/Write access.

That's it!

Setup - Saturn Cloud

The Saturn Cloud UI is fairly self explanatory.

Create a Resource

• Create a new Python Server resource type.

Choose your Hardware Specs

• Choose a CPU Instance Type (We're not doing anything here that requires a GPU.)
• Choose your memory requirements
  • From the Size dropdown choose an instance type.

Environment

• Customize your environment

• I went with the default saturn cloud datascience environemnt. If you're using tensorflow or pytorch you may want to choose those environments.

• Add your System Packages of choice. I went for htop tree vim build-essential

• Optional - Add your pip packages. scanpy louvain quilt3[catalog,pyarrow]

• Open a Port for Web Applications

  • Click on the Advanced Options in the Environment tab in order to open a port.
  • This step is necessary if you want to follow along with the visualization portion of the tutorial!

Review and Launch

• Once you've reviewed your settings click Create and you'll be ready to go.

On the next page press the Start button, and then the JupyterLab button.


Optional - ExcellxGene Setup

Excellxgene needs to be installed to it's own conda environment because it doesn't play nice with the existing conda env.

• Grab the excellxgene-environment.yaml from this repo.
• Create the excellxgene environment
• Register the excellxgene environment as a Python Kernel.

Setup is the same whether you are using Saturn Cloud, running locally, or using an HPC system.

wget https://raw.githubusercontent.com/dabble-of-devops-bioanalyze/single-cell-saturn-cloud/main/excellxgene-environment.yaml
mamba env create -f ./excellxgene-environment.yaml -p ~/.conda/envs/excellxgene
source activate conda activate /home/jovyan/.conda/envs/excellxgene

If you are running on a local system or BioAnalyze HPC I recommend using these settings.

Data Analysis

You can grab the scanpy tutorials with:

git clone https://github.com/scverse/scanpy-tutorials
wget https://raw.githubusercontent.com/dabble-of-devops-bioanalyze/single-cell-saturn-cloud/main/excellxgene-environment.yaml
wget https://raw.githubusercontent.com/dabble-of-devops-bioanalyze/single-cell-saturn-cloud/main/prepare.py
wget https://raw.githubusercontent.com/dabble-of-devops-bioanalyze/single-cell-saturn-cloud/main/pbmc3k.ipynb

I'm only going over the pbmc3k.ipynb notebook. It's supplied here because I made one very minor change, but all credit, accolodes, etc should go to the ScanPy team.

From here, open up the pbmc3k.ipynb notebook in your Jupyterhub. Go to Kernel -> Restart Kernel and Run all Cells.

This will run the entire notebook for you.

ExCellxGene

Now that we have some data let's visualize some clusters!

git clone https://github.com/czbiohub/excellxgene
cd excellxgene
cd example-dataset 
wget https://raw.githubusercontent.com/dabble-of-devops-bioanalyze/single-cell-saturn-cloud/main/prepare.py
source activate excellxgene
python ./prepare.py pbmc3k.h5ad --output ./pbmc3k-preprocessed.h5ad --overwrite

Finally, launch the excellxgene browser:

excellxgene launch --host 0.0.0.0 --port 8000 ./pbmc3k-preprocessed.h5ad 

Resources

• Scanpy Tutorials
• Scanpy
• ExCellxGene