ecDNA FISH bladder cancer analysis

This repository contains analysis code for fluorescence in situ hybridization (FISH) microscopy data used to detect and quantify extrachromosomal DNA (ecDNA) in bladder cancer samples.

Overview

The analysis is implemented as a Jupyter notebook and performs image-based quantification of FISH signals (e.g. MDM2) at the single-cell level. It integrates segmentation masks with fluorescence intensity measurements to assess ecDNA-associated signal distributions across nuclei.

The workflow includes:

• Parsing and organization of microscopy image files
• Matching of DAPI and fluorescence channels (e.g. Cy3)
• Loading and handling of segmentation masks
• Pixel-level and nucleus-level signal quantification
• Thresholding and quality control of signal detection
• Visualization and summary plots used in the manuscript

Input data

The analysis expects:

• Multi-channel FISH microscopy images (e.g. DAPI, Cy3)
• Corresponding segmentation masks (nuclei-level)

File naming conventions are partially encoded in the notebook and support multiple acquisition formats.

Segmentation

Nuclei segmentation was performed using Cellpose.

The script used for segmentation is provided:

scripts/run_cellpose.sh

Key parameters:

• model: cyto
• diameter: 100
• cellprob_threshold: 0.2
• flow_threshold: 0.3

Output

The notebook produces:

• Quantitative measurements of fluorescence signal per nucleus
• Quality control visualizations
• Summary plots (including the final figure used in the manuscript)

Usage

The analysis is provided as a single Jupyter notebook:

ecdna_fish_analysis.ipynb

To run:

1. Install required Python dependencies (see below)
2. Adjust input paths to your local data
3. Run all cells in the notebook

Dependencies

The analysis uses standard Python scientific libraries, including:

• numpy
• pandas
• matplotlib
• seaborn
• scikit-image (for image handling)
• colorcet

A full environment file will be provided for reproducibility.

Data availability

The raw microscopy data is available at:
[TO BE ADDED – Zenodo or other repository DOI]

Author

Anne Rademacher
DKFZ / RippeLab

Notes

This repository is part of the analysis for a collaborative study on ecDNA in bladder cancer. The code is provided in a minimally processed form to ensure transparency and reproducibility of the analysis.