mutation-predictability

This repository contains the data and code used to generate results and corresponding figures of the recently released preprint "What makes the effect of protein mutations difficult to predict?".

• System requirements
• Project structure
• Install instructions
• Reproducibility

System requirements

Hardware requirements

predictability can be run on a standard computer without extensive hardware configurations. GPU availability is not necessary, but will greatly speed up the rita.ipynb notebook executions.

Software requirements

OS requirements

The predictability package is supported for macOS and Linux and tested on macOS Ventura 13.5.2.

Python requirements

predictability requires python ≥ 3.8. All requirements and the corresponding versions are listed in the requirements.txt file.

Project structure

All experiments and processing of results are organized in notebooks, which can be run by installing the predictability package. A simple demo can be bound under notebooks/demo.ipynb.

Install instructions

Clone the repository and install with

git clone https://github.com/florisvdf/mutation-predictability.git
cd mutation-predictability
pip install .

The Potts Regressor model of the predictability package makes use of gremlin_cpp. To use the Potts Regressor, make sure that gremlin_cpp is installed and is added to $PATH.

Installation on a typical computer should take no longer than 10 minutes.

Reproducibility

Results can be reproduced by simply executing all notebooks under the notebooks directory. Plots can be generated by executing the notebooks/figures_for_publication notebook. Different sample assignment to train and test folds can be achieved by executing the notebooks while changing the variable seed in the second cell.