RStudio on GenomeDK Desktop

Scripts to launch RStudio Desktop as a native window on desktop.genome.au.dk, running on a Slurm compute node.

---

How it works

RStudio runs inside a conda environment on a Slurm compute node. The launch script requests the job inside a tmux session (so it survives accidental terminal closure), activates the environment, and opens RStudio as a desktop window.

---

Files

File	Run when
rstudio_setup_desktop.sh	Once per user
rstudio_launch_desktop.sh	Every time you want RStudio

---

Prerequisites

Conda (Miniconda or Anaconda)

Check if you have it:

conda --version

If not, install Miniconda:

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh

Say yes to initialising conda, then close and reopen your terminal.

tmux

Pre-installed on GenomeDK, nothing to do.

A project folder

Put both scripts in a your project folder:

/faststorage/project/YOUR_PROJECT/rstudio/

---

Download the scripts

On the GenomeDK Desktop, open a terminal, go to your shared project folder, and download both scripts with wget:

cd /faststorage/project/YOUR_PROJECT/rstudio
wget https://raw.githubusercontent.com/ghasemis/rstudio_gdk/main/rstudio_setup_desktop.sh
wget https://raw.githubusercontent.com/ghasemis/rstudio_gdk/main/rstudio_launch_desktop.sh

---

Usage

Step 1: Open the GenomeDK Desktop

Go to desktop.genome.au.dk, log in, and open a terminal.

Step 2: Go to the folder where you downloaded the files

cd /faststorage/project/YOUR_PROJECT/rstudio

Step 3: Set your project name (one time only)

Before running anything, open rstudio_launch_desktop.sh in a text editor and change the SLURM_ACCOUNT line to your own GenomeDK project name:

SLURM_ACCOUNT="YOUR_PROJECT"   # ← change this to your project name

Step 4: Run setup (one time only)

bash rstudio_setup_desktop.sh

This creates a conda environment called rstudio_env with:

• R 4.4.2
• RStudio Desktop
• tidyverse, ggplot2, renv

Takes a few minutes. Safe to re-run, skips steps already done.

Step 5: Launch RStudio

bash rstudio_launch_desktop.sh

This will:

• Start a tmux session
• Request a Slurm job (10 hours, 128 GB RAM, 8 cores)
• Activate the conda environment
• Open RStudio as a normal desktop window

RStudio will appear as a window on the GenomeDK desktop once the job is allocated (usually within a minute).

---

Stopping RStudio

Close the RStudio window normally, or attach to the tmux session and press Ctrl+C:

tmux attach -t rstudio

---

Customising resources

Edit the top section of rstudio_launch_desktop.sh:

SLURM_ACCOUNT="YOUR_PROJECT"   # your GenomeDK project name
SLURM_CORES=8                 # CPU cores
SLURM_MEM="128g"              # RAM
SLURM_TIME="10:00:00"         # wall time (hh:mm:ss)
CONDA_ENV_NAME="rstudio_env"  # conda environment name

---

Troubleshooting

Problem	Fix
conda: command not found	Install Miniconda (see Prerequisites) and reopen the terminal
RStudio window doesn't appear	The job may still be queued, run tmux attach -t rstudio to watch progress
RStudio crashes on startup	Try conda update -n rstudio_env rstudio-desktop to get the latest version
Job runs but RStudio won't open	Make sure you are on the desktop (desktop.genome.au.dk) not a plain SSH session, a display is required
Want to check what's running	tmux attach -t rstudio shows the live log
Want a longer session	Change SLURM_TIME in rstudio_launch_desktop.sh

---

Notes

• If you already have a conda environment with R and Rstudio, set CONDA_ENV_NAME to its name in rstudio_launch_desktop.sh and skip the setup script.
• For reproducible package management within R projects, use renv. It is included in the environment.
• For datasets over 200k cells you should rise SLURM_MEM to 256g