Code for "Single-cell image-based genetic screens systematically identify regulators of Ebola virus subcellular infection dynamics". Optical pooled screen analysis code adapted from Feldman, D., Funk, L. et al Nature Protocols (2022).
To train the unsupervised autoencoder, first clone the repository, then enter this directory, and run the genome_wide.py script. Note that this script will train a model for a specified number of epochs then save the model, losses, and reproducing shell command in a directory named models:
# Clone repository
git clone https://github.com/beccajcarlson/EBOVOpticalPooledScreen.git EBOVOpticalPooledScreen
# Enter directory
cd EBOVOpticalPooledScreen/DeepLearning/
# Install dependencies (preferably in new virtual environment)
pip install -r requirements.txt
# (Optional) View flag options on script
python modeling/genome_wide.py -h
# Train Unsupervised Autoencoder Sample
python modeling/genome_wide.py -n
This package is supported for macOS and Linux. The package has been tested on the following systems:
- macOS: Somona (14.4)
- Linux: Debian 5.10.197-1
Installation should take no more than a minute or two on similar systems.