EpiK is a Python library designed to infer sequence-function relationships using Gaussian process models. Built on top of GPyTorch and KeOps, EpiK enables fitting these models to large datasets containing hundreds of thousands to millions of sequence measurements.

You can find more detailed documentation and tutorials here

Scripts to reproduce the analyses and figures from the paper are avaiable in a separate repository

• Juannan Zhou, Carlos Martí-Gómez, Samantha Petti, David M. McCandlish. Learning sequence-function relationships with scalable, interpretable Gaussian processes (2025) In preparation.

We recommend using an new independent environment with python3.8, as used during development and testing of EpiK to minimize problems with dependencies. Create a python3 conda environment and activate it

conda create -n epik python=3.8
conda activate epik

Install with pip

pip install epik

Download the repository using git and cd into it

git clone [email protected]:cmarti/epik.git

Install repository

cd epik
pip install .

Run tests with pytest

pytest test