Quality control and reconstruction for Multiplexed Assays of Variant Effect (MAVE) data.
MAVEpolish uses dictionary learning to reconstruct variant effect maps (VEMs), providing quality metrics and polished score estimates for every variant.
Requires Python 3.9+.
git clone https://github.com/gkudla/mavepolish.git
cd mavepolish
pip install .
This installs mavepolish, to_vem, and mavepolish-web commands that work in any new terminal window.
Optional: install in a virtual environment
If you prefer to keep MAVEpolish isolated from your system Python:
cd mavepolish
python3 -m venv venv
source venv/bin/activate # run this each time you open a new terminal
pip install .
For developers
Use an editable install so code changes take effect immediately:
pip install -e .
mavepolish-web
Then open http://localhost:8051
Convert raw MAVE data to VEM format:
to_vem -i my_data.csv
Run quality control analysis:
mavepolish -t my_data.VEM.tsv -e my_data.VEM.tsv
Use the pretrained model for faster analysis:
mavepolish -m pretrained_model.pkl -e my_data.VEM.tsv
Run mavepolish --help and to_vem --help for all options.