Quick Guide

Introduction

epistasis is a Python library that includes models to estimate statistical, high-order epistasis in genotype-phenotype maps. Using this library, you can

1. Decompose genotype-phenotype maps into high-order epistatic interactions

2. Find nonlinear scales in the genotype-phenotype map

3. Calculate the contributions of different epistatic orders and

4. Estimate the uncertainty in the epistatic coefficients and

For more information about the epistasis models in this library, see our Genetics paper:

Sailer, Z. R., & Harms, M. J. (2017). “Detecting High-Order Epistasis in Nonlinear Genotype-Phenotype Maps.” Genetics, 205(3), 1079-1088.

Simple Example

Follow these five steps for all epistasis models in this library:

1. Import a model. There many models available in the epistasis.models module. See the full list in the next section.

from epistasis.models import EpistasisLinearRegression

2. Initialize a model. Set the order, choose the type of model (see Anatomy of an epistasis model for more info), and set any other parameters in the model.

model = EpistasisLinearRegression(order=3, model_type='global')

3. Add some data. There are three basic ways to do this. 1. Pass data directly to the epistasis model using the add_data method. 2. Read data from a separate file using one of the read_ methods. 3. (The best option.) load data into a GenotypePhenotypeMap object from the GPMap library and add it to the epistasis model.

from gpmap import GenotypePhenotypeMap

datafile = 'data.csv'
gpm = GenotypePhenotypeMap.read_csv(datafile)

# Add the data.
model.add_gpm(gpm)

# model now has a `gpm` attribute.

4. Fit the model. Each model has a simple fit method. Call this to estimate epistatic coefficients. The results are stored the epistasis attribute.

# Call fit method
model.fit()

# model now has an ``epistasis`` attribute

5. Plot the results. The epistasis library has a pyplot module (powered by matplotlib) with a few builtin plotting functions.

from epistasis.pyplot import plot_coefs

fig, ax = plot_coefs(model.epistasis.sites, model.epistasis.values)

Install and dependencies

For users

This library is now available on PyPi, so it can be installed using pip.

pip install epistasis

For developers

For the latest version of the package, you can also clone from Github and install a development version using pip.

git clone https://github.com/harmslab/epistasis
cd epistasis
pip install -e .

Dependencies

The following dependencies are required for the epistasis package.

• gpmap: Module for constructing powerful genotype-phenotype map python data-structures.

• Scikit-learn: Simple to use machine-learning API.

• Numpy: Python’s array manipulation package.

• Scipy: Efficient scientific array manipulations and fitting.

• Pandas: High-performance, easy-to-use data structures and data analysis tools.

There are also some additional dependencies for extra features included in the package.

• matplotlib: Python plotting API.

• ipython: interactive python kernel.

• jupyter notebook: interactive notebook application for running python kernels interactively.

• ipywidgets: interactive widgets in python.

Running tests

The epistasis package comes with a suite of tests. Running the tests require pytest, so make sure it is installed.

pip install -U pytest

Once pytest is installed, run the tests from the base directory of the epistasis package using the following command.

pytest