Awesome

This is not an officially supported Google product.

SLIP - Synthetic Landscape Inference for Proteins

SLIP is a sandbox environment for engineering protein sequences with synthetic fitness functions. See our preprint

Installation instructions

Tested on python >= 3.7

We recommend installing into a virtual environment to isolate dependencies.

python3 -m venv env
source env/bin/activate

To install:

pip3 install -q -r requirements.txt

To run the unit tests:

bash -c 'for f in *_test.py; do python3 $f || exit 1; done'

Example landscape usage

See this colab for an example of using a landscape.

Constructing a new landscape

All landscapes were constructed using Mogwai. See that repo's example, which shows how to train a new Potts model and how to (optionally) examine contact accuracy after training. All that is required is an alignment in .a3m format, true contacts are not required (e.g. as in this colab).