Source code for the analyses accompanying the IBDome paper.

git clone https://github.com/ibdome/ibdome-paper.git

export IBDOME_BASEDIR=$(realpath ibdome-paper)

cd $IBDOME_BASEDIR
mkdir data
cd data

Download the ibdome_v1.0.1.sqlite database from: https://ibdome.org/#!/data_download and store it in the data directory.

wget https://ibdome.org/static/ibdome_v1.0.1.zip

You might need to install ncftp in order to perform recursive FTP downloads.

e.g.

Fedora, RedHat, Rocky Linux and other rpm based Linux distributions

sudo dnf install ncftp

Ubuntu, Debian and other apt based Linux distributions

sudo apt update
sudo apt install ncftp

Get the data

ncftpget -T -R ftp://ftp.ebi.ac.uk/biostudies/fire/S-BIAD/753/S-BIAD1753/Files/imaging
cd ..

Paper (Jiang et al., Nat Methods 2021) and app Script: https://github.com/data2intelligence/CytoSig

conda create -n cytosig.v0.1 -y python=3.8 'numpy>=1.19' 'pandas>=1.1.4' 'gcc>=4.2' 'openpyxl>=3.0.9' gsl=2.6 setuptools scipy xlsxwriter
conda activate cytosig.v0.1

cd external_tools
git clone https://github.com/data2intelligence/ridge_significance.git
git clone https://github.com/data2intelligence/data_significance.git
git clone https://github.com/data2intelligence/CytoSig.git

cd ridge_significance
python setup.py install
python -m unittest tests.regression

cd ../data_significance
python3 setup.py install
python3 -m unittest tests.regression

cd ../CytoSig
python setup.py install
python -m unittest tests.prediction
conda deactivate
cd ..

Create a conda environment for the Python version of MOFA2 (mofapy2):

conda create -n MOFA_env python=3.13.3 pip -y && conda run -n MOFA_env pip install mofapy2==0.7.2

First clone the forked and updated version of STAMP (v1 branch), which includes the latest Foundation Models available for use:

git clone -b v1 https://github.com/sandrocarollo/STAMP.git
cd STAMP

Create a new conda environment:

conda create -n stamp python=3.10
conda activate stamp
conda install -c conda-forge libstdcxx-ng=12

Install the STAMP package:

pip install .

NOTE: To use the UNI2 and Virchow2 models, you must have a Hugging Face account with access to the respective model repositories. Please refer to the UNI2 repository and the Virchow2 repository for licensing, fair use, and access details.

To predict disease activity from WSI embeddings, we use the marugoto pipeline.

First clone the forked and updated version of marugoto, which includes stratified training and attention heatmap generation:

cd ..
git clone -b attmil-regression https://github.com/sandrocarollo/marugoto.git
cd marugoto

Create and activate the dedicated environment:

conda deactivate
conda env create -f env_marugoto.yml
conda activate marugoto
pip install .

NOTE: For additional information regarding the STAMP and marugoto pipeline usage please refer to the additional documentation.

The scripts in this repository are numbered in the order they should be executed to fully reproduce the results of our paper.

The repository contains a mix of R Markdown (.Rmd), Bash (.sh), and Python (.py) scripts.

They must be executed in ascending numerical order, as outputs of one step are often inputs for the next.

Each script type is run differently:

• R Markdown (.Rmd) 2 run possibilities:

  • Interactive: open in RStudio and click Knit or by
  • Command line:

    Rscript -e "rmarkdown::render('0X_markdown_file.Rmd')"

• Bash (.sh)

  • Run directly in the shell:

    bash 0X_bash_script.sh

• Python (.py)

  • Run with Python:

    python 0X_python_script.py

Run the scripts starting from 01_IBDome_overview.Rmd and activate the correct Conda environment if needed.

Rscript -e "rmarkdown::render('01_IBDome_overview.Rmd')"
Rscript -e "rmarkdown::render('02_IPSS.Rmd')"
Rscript -e "rmarkdown::render('03a_gene_signatures.Rmd')"

bash 03b_deseq2.sh

Rscript -e "rmarkdown::render('03c_DE_downstream.Rmd')"

conda activate cytosig.v0.1
bash 03d_runCytoSig.sh
conda deactivate

Rscript -e "rmarkdown::render('03e_cytosig_downstream.Rmd')"
Rscript -e "rmarkdown::render('04_protein_panel.Rmd')"
Rscript -e "rmarkdown::render('05a_Extract_histoscores.Rmd')"

conda activate stamp
bash 05b_Embedding_extraction.sh
bash 05c_Imaging_feature_matrix_extraction.sh
conda deactivate 

Rscript -e "rmarkdown::render('05d_MOFA.Rmd')"

conda activate marugoto
bash 06a_Disease_activity_prediction.sh
bash 06b_Attention_Heatmap_generation.sh

python 06c_correlation_plots_and_figure_maker.py
python 06d_sankey_dash.py
conda deactivate 

conda activate stamp
bash 07a_UC_CD_Classifier.sh
python 07b_Confusion_Matrix_and_figure_maker.py
conda deactivate

Notes:

• Use the right Conda environment for each step (cytosig.v0.1, marugoto, stamp).
• Each script may take a significant amount of time depending on hardware resources.
• Script numbers correspond to the figure numbers in the paper for easier reference.

Please use the issue tracker.

Plattner, C., Sturm, G., Kühl, A.A., Atreya, R., Carollo, S., ... & Becker, C., Siegmund, B., Trajanoski, Z. (2025). IBDome: An integrated molecular, histopathological, and clinical atlas of inflammatory bowel diseases. bioRxiv. doi:10.1101/2025.03.26.645544