Working on the HPC

Login to the HPC:

ssh {YOUR_VSC_USERNAME}@login.hpc.kuleuven.be

Then clone the repository in a folder inside /staging/leuven/stg_00143:

cd /staging/leuven/stg_00143
mkdir my_directory
cd my_directory
git clone https://github.com/vibspatial/targeted_transcriptomics_training
cd targeted_transcriptomics_training

Optional: build conda environment on the VSC

1. Install miniconda in $VSC_DATA or $VSC_DATA_VO_USER if a VO is available:

cd $VSC_DATA
mkdir -p miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda3/miniconda.sh
bash miniconda3/miniconda.sh -b -u -p miniconda3
rm -rf miniconda3/miniconda.sh
miniconda3/bin/conda init bash 

Set libmamba as the solver:

conda install -n base conda-libmamba-solver
conda config --set solver libmamba

2. Build the Conda environment:

Change the PATH_CONDA_ENV parameter in the script build_conda_slurm.sh to the path where you want to store the conda environment. Then build the conda environment:

sbatch build_conda_slurm.sh

Use the (prebuild) conda environment and submit jobs

1. Test the conda environment

Start an interactive session:

srun --account=lp_edu_ont2024 --clusters=genius --partition=interactive --ntasks 1 --cpus-per-task 8 --mem=16G --time=00:30:00 --pty bash

Now activate the conda environment (no need to have a conda installation, we will use the conda installation from the VSC to activate our environment):

source /apps/leuven/rocky8/skylake/2021a/software/Miniconda3/4.12.0/bin/activate $PATH_CONDA_ENV

With PATH_CONDA_ENV the path to your custom build conda environment, or the path to the prebuild conda environment we provide at:

/staging/leuven/stg_00143/spatial_data_training/conda_environments/spatial_data_training_env

Now run following commands:

python

and then in the python shell:

import harpy

2. Submit jobs

Go to the login nodes if you are still in an interactive session (you can check if you are on the login nodes by typing in the command line hostname):

exit

In the Slurm script run_slurm.sh, change the output_dir argument in the python run_merscope.py command to specify your desired output directory. The resulting SpatialData .zarr store and some figures will be saved there.

Submit the job:

sbatch run_slurm.sh

This job will approximatly take half an hour to finish.

3. Submit job to explore the data

In the Slurm script run_plot_slurm.sh, change the sdata_path and output_dir argument in the python plot_leiden.py command to respectively the SpatialData .zarr store and the desired directory where figures will be saved.

On a login node, submit the job:

sbatch run_plot_slurm.sh

You can also explore the data using an interactive session:

srun --account=lp_edu_ont2024 --clusters=genius --partition=interactive --ntasks 1 --cpus-per-task 8 --mem=16G --time=02:00:00 --pty bash

Tip:

If your connection was closed, and you still have an interactive session running, you can connect back to it via:

srun --account=lp_edu_ont2024 --clusters=genius --partition=interactive  --pty --overlap --jobid YOUR-JOBID bash

You can get the jobid via the command squeue.

Activate your conda environment:

source /apps/leuven/rocky8/skylake/2021a/software/Miniconda3/4.12.0/bin/activate $PATH_CONDA_ENV

Run the python script:

python plot_leiden.py \
 --sdata_path /staging/leuven/stg_00143/spatial_data_training/merscope/sdata_merscope_crop.zarr \
 --output_dir /staging/leuven/stg_00143/spatial_data_training/merscope

Excercises

For the following excercises, use the SpatialData .zarr store obtained using the run_merscope.py Python script and accompanying slurm script run_slurm.sh.

• Create a Python script that plots the leiden clusters for x_min=30000, x_max=40000, y_min=30000, y_max=40000. Submit to the HPC using a slurm script.
• Create a Python script that plots the expression level of the gene Alas2 in the region x_min=30000, x_max=40000, y_min=30000, y_max=40000. Submit to the HPC using a slurm script.
• Create a Python script that only does leiden clustering and visualization of the leiden clusters. Modify the parameters of hp.tb.leiden.
• Create a notebook that does leiden clustering, and visualizes the leiden clusters. Run the notebook. Please make sure not to run the notebook on one of the login nodes of the HPC.

Moving data to the VSC using Globus

See here for a tutorial on how to move data to and from the VSC using Globus.

Use VS Code on the VSC

Make sure to install miniconda to be able to use the prebuild conda environment in a VS Code session:

cd $VSC_DATA
mkdir -p miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda3/miniconda.sh
bash miniconda3/miniconda.sh -b -u -p miniconda3
rm -rf miniconda3/miniconda.sh
miniconda3/bin/conda init bash

Make a symlink to the prebuild conda environment.

cd $VSC_DATA/miniconda3/envs
ln -s /staging/leuven/stg_00143/spatial_data_training/conda_environments/spatial_data_training_env

After launching VS Code (see below), please make sure to install the Python and Jupyter extensions, otherwise VS Code will not be able to find the Python environments.

VSC KU Leuven

Go to Open OnDemand

Choose Visual Studio Code.

Make sure to change the number of hours, and set the number of processes per node to 8.

Select Launch and then connect.

Select the kernel when you want to run Jupyter Notebooks, e.g. /staging/leuven/stg_00143/spatial_data_training/conda_environments/spatial_data_training_env.

VSC Ugent

Work with a VS Code tunnel.

Follow instruction here

VIB Compute

Work with a VS code tunnel.

Follow instructions here

Or via Open OnDemand.

Connect to a remote Cluster with VS Code Remote - SSH

(not recommended on the login nodes of the VSC)

Install VS Code and the Remote - SSH Extension:

- Open VS Code.

- Go to the Extensions view by clicking on the Extensions icon in the Activity Bar on the side of the window or pressing Ctrl+Shift+X.

- Search for "Remote - SSH" and install it.

Configure SSH in VS Code:

- Press F1 to open the Command Palette in VS Code.

- Type Remote-SSH: Open Configuration File... and select it.

- Choose the SSH configuration file to edit. It’s usually located at ~/.ssh/config.

- Add the following configuration to the file:

Host {NAME}
    HostName {YOUR_HOSTNAME}
    User {YOUR_USERNAME}
    ForwardAgent yes

{YOUR_HOSTNAME} is the name of the cluster you want to connect to (e.g. via ssh {YOUR_USERNAME}@{YOUR_HOSTNAME}), and {NAME} can be a name of choice.

Connect to the Remote Server:

- Press F1 to open the Command Palette.

- Type Remote-SSH: Connect to Host... and select it.

- You should see `{NAME}` in the list. Select it.

- VS Code will open a new window and start connecting to the remote server.

- You might be prompted for your SSH key passphrase or password.