Skip to content

Rapid and efficient image preprocessing pipeline for multiplexed spatial proteomics. Code to compare several techniques to denoise Imaging Mass cytometry

License

Notifications You must be signed in to change notification settings

deMirandaLab/PENGUIN

 
 

Repository files navigation

License: GPL v3 Repository Size Python Version Suggestions Welcome Release

PENGUIN

PENGUIN - Percentile Normalization GUI Image deNoising is a rapid and efficient image preprocessing pipeline for multiplexed spatial proteomics. In comparison to existing approaches, PENGUIN stands out by eliminating the need for manual annotation or machine learning model training. It effectively preserves signal intensity differences and reduces noise.

PENGUIN's simplicity, speed, and user-friendly interface, deployed both as script and as a Jupyter notebook, facilitate parameter testing and image processing.

This repository contains the files for running PENGUIN and the comparison with standard image processing methods and solutions designed specifically for multiplex imaging data.

General view: plot

Table of Contents

Clone the Repository

To clone this repository to your local machine, use the following command:

git clone https://github.com/your_username/your_repository.git

Requirements/Installation

You can create the environment installing the packages or using the ymal file.

To manually create and install packages use:

conda create --name penguin
conda activate penguin
conda install matplotlib pandas panel opencv scikit-image ipywidgets jupyter ipykernel plotly
pip install apeer-ometiff-library --no-deps

Alternatively, you can create the environemtn using the yml file:

conda env create --file penguin_env.yml
conda activate penguin
pip install apeer-ometiff-library --no-deps

After you created the environment, and if you want to use the Jupyter notebooks

add the environment kernel to jupyter

python -m ipykernel install --user --name=penguin

launch the jupyter and be sure that you are running with the penguin kernel If you have problems with widgets appearing in Jupyter notebook try downgrade Ipython version, for example:

python pip install iPython==8.20.0

Getting started

There are 2 Notebooks available.

Use check_th_all_ch_per_image if each FOV is a stack of channels.

Use check_th_one_ch_per_image if each FOV is a directory with multiple tiffs inside ( one per each channel)

Open the notebook and click Kernel -> Restart and Run all For all channels in a stack it should look like this: plot

change the path to the location of your data and click 'Change Path' You can now change the channels to visualize, select different percentiles and thresholds. Compare images tab gives you the comparison between raw and the clean image with the defined settings Compare Zoom plots the images using Plotly library that allows for zooming some areas. You can change the number of images that are displayed.

In the case of stacks of channels, your channel names should be in the page tags. Otherwise, the channel names will be set as index numbers. In the case of directory with multiple files, the channel names should be in the file names.

plot

Once you have the values for percentile and thresholds defined you can save your images by just clicking the save button. In case of a file per channel, you can save all the images of the same channel at once. In case of stacks with multiple channels per FOV, you need to define the values per each channel in the pop up table and click save (see below).

plot

Saving images will mimick your structure and filenames (and pagetags) in the saving directory.

You can also use directly PENGUIN functions. Just check pipeline files.

Credits

If you find this repository useful in your research or for educational purposes please refer to: Sequeira, A. M., Ijsselsteijn, M. E., Rocha, M., & de Miranda, N. F. (2024). PENGUIN: A rapid and efficient image preprocessing tool for multiplexed spatial proteomics. bioRxiv, 2024-07. doi: https://doi.org/10.1101/2024.07.01.601513

License

Developed at the Leiden University Medical Centre, The Netherlands and Centre of Biological Engineering, University of Minho, Portugal

Released under the GNU Public License (version 3.0).

About

Rapid and efficient image preprocessing pipeline for multiplexed spatial proteomics. Code to compare several techniques to denoise Imaging Mass cytometry

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Jupyter Notebook 50.5%
  • HTML 49.4%
  • Other 0.1%