Introduction

This repositiory is a guide for installing and making sequence encodings using the protein language model (PLM) ESM2.

Installation

To install ESM2 you will need a python installation. For easy GPU utilizaiton, we will use python within conda. Conda can be installed from here: https://docs.anaconda.com/miniconda/install/

Create conda environment

$ conda create -n esm2 # create conda virtual environment
$ conda activate esm2
$ conda install numpy
$ conda install pytorch pytorch-cuda=11.8 -c pytorch -c nvidia # install pytorch with gpu support
$ pip install fair-esm # install esm source with pip

Clone this repo

$ git clone https://github.com/mnielLab/esm2_utilities.git

Creating sequence encodings

Source code for encoding sequence batches is included in ./utils/esm2_encode.py A demo script for creating ESM2 encodings for sequences in an example fasta file is included in ./demo.poy To run this example,

$ python demo.py

This will create per residue encodings for all sequences in ./example_proteins/example_proteins.fasta. And save them in a .pickle formatted file, ./example_proteins_esm2enc.pickle