Classification of ALS and Stress in Cultures of Motor Neurons

This project aims to increase our understanding of Amyotrophic Lateral Sclerosis (ALS), which is a neurodegenerative disease. It deals with the classification of cultures of human induced Pluripotent Stem Cells (iPSCs) motor neurons. The neurons are either healthy, or contain the ALS mutation, and are put under different cellular stress (osmotic, oxidative and heat).

The project is described in a paper submitted on BiorXiv which you can find here

Installation

Here are the steps to follow to be able to use this package:

1. Install working environment
2. Download the database
3. Adapt configuration file
4. Download trained models

Working environment

Installation requirements can be found in requirements.txt.

Those can be installed within a conda environment (recommended):

conda create -n als # empty environment
conda activate als
conda env update -f environment.yml # add packages from requirements.txt

Database

The database is separated from the package and can be downloaded here. CAUTION: the dataset is very large (~1TB) so make sure you have enough space to store it. You can either put it in the data folder or choose your own location. In the next step, you will enter the location of the folder you chose in the configuration file.

Note: If the complete dataset is too large for you, a subset of the database can be downloaded here.

Configuration file

Open scripts>config_user.yml. In this file, enter a value for each field (batch_size, n_epochs, learning_rate) and in particular, enter the ABSOLUTE path of the folder where you downloaded the database for the source_directory. The target_directory is used during training and testing to create temporary directories of images. Again, this usually requires a large storage size (several GB) so make sure you have the available space in the directory you choose.

When you use a command and need to choose a config, please choose user. You can configure other config files and use those as well (e.g. to train models separately on a grid).

Models

Models can be quite long to train given the large number of images. If you want to evaluate models which are already trained, you can find them here. Put the downloaded models under models.

Getting Started

See notebooks>user_guide.ipynb

Questions

Refer to the documentation and/or source code. Go to docs > source. Run the following:

make html

Go to docs > build > html and open index.html in your browser.

Tests

You can make sure everything is working correctly with the following command:

pytest -sv src/test/

Contact

For questions or reporting issues to this software package, contact the authors of the paper: colombine.verzat@idiap.ch or raphaelle.luisier@idiap.ch.