Single-cell spatial transcriptomics of inherited retinal degeneration reveals broad and focal organization of disease responses

This repository contains the analysis and plotting code for Figs. 1–4 and Supplementary Figs. S1–S5 for this manuscript.

Repository organization

Directory	Contents
figure_1/code/ through figure_4/code/	Main figure scripts
figure_1_supp/code/, figure_2_supp/code/, figure_3_supp/code/	Supplementary Figs. S1–S3 scripts
figure_4_supp_1/code/, figure_4_supp_2/code/	Supplementary Figs. S4-S5 scripts
_shared/code/	Code shared by multiple scripts
preprocessing/	Code for preprocessing data

Instructions to reproduce analyses and plots

1. Setup

Create a Python environment, then install the required dependencies:

python -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt

2. Download data

Raw and preprocessed data is deposited on Zenodo* at the following link: https://doi.org/10.5281/zenodo.22767514.

Download and unzip the Zenodo data deposit, then set DATA_DIR:

export DATA_DIR=/path/to/retina_spatial_paper_data

*See the Zenodo README for more detailed information on the contents of the deposit.

3. Run data preprocessing scripts

From preprocessing/, run the scripts in this order:

python build_cell_typing.py
python build_arc_columns.py
python build_per_cell_gene_counts.py
python build_all_rods.py
python export_transcript_subsets.py
python build_ds8_cache.py
python ../figure_1/code/Fig1b_00_compute_harmony.py

Notes: The cell segmentation masks and the trained U-Net model and layer predictions are already provided in the Zenodo deposit for convenience. To regenerate them:

1. Run build_segmentation.py for regenerating the cell segmentation
2. Run the scripts in preprocessing/unet/ for regenerating the U-Net model (predict_unet_layers.py applies the existing trained model)

4. Run analysis and figure plotting scripts

Run each script from its own code/ directory:

cd figure_1/code
python Fig1a.py

Notes on the order to run scripts:

1. Run Fig. 3 scripts before Fig. 4 and Supplementary Fig. 3, as some scripts require Fig. 3 outputs.
2. Within Fig. 4, run Fig4a.py first, as it produces the rod injury response score used by the subsequent scripts.