IntFold

Generating intermediate representations using AlphaFold2 and ColabFold

Overview

This repo uses Evoformer of AlphaFold2 to generate intermediate representations (MSA and Pair) for proteins, especially enzymes with EC numbers. The code is based on ColabFold and LocalColabFold.

The enzyme dataset is splitted in four files: uniprot-filtered-reviewed_yes.tab.gz.partaa, uniprot-filtered-reviewed_yes.tab.gz.partab, uniprot-filtered-reviewed_yes.tab.gz.partac, uniprot-filtered-reviewed_yes.tab.gz.partad, comes from UniProt.

Install

Only Linux is supported to run IntFold, please install Windows Subsystem for Linux if you are using Windows 10 or later.

1. Install Docker

   • Install nVidia Container Toolkit if you have nVidia GPUs
   • Set up Docker as a non-root user
   • Check if your nVidia Container Toolkit installation is successful by running

   docker run --rm --gpus all nvidia/cuda:11.0-base nvidia-smi
   

   Output lists your available GPUs, if no GPU is listed, check if you have followed the instruction of installing nVidia Container Toolkit and take a look at nVidia docker issues

2. If you don't need to modify the code, you can directly use this built docker image by running

   docker pull yuxin60/intfold
   

3. If you need to modify the code to run your tasks, first clone this repo and cd into it

   git clone https://github.com/yuxin212/intfold.git
   

   And modify the code accordingly.

4. Build docker image

   docker build -f docker/Dockerfile -t intfold .
   

Running IntFold

1. First run

   docker run --gpus <number of gpus> yuxin60/intfold:latest
   

2. Get Container id

   docker ps
   

3. After running, copy generated intermediate representations from docker container to host

   docker cp <container-id>:/app/intermediate/ <path to store results>
   

4. After copying the output, please remove the docker container

   docker stop <container-id>
   docker rm <container-id>
   

IntFold Output

The output will be saved as numpy arrays in docker container, and path is /app/intermediate/. This directory has the following structure:

/app/intermediate/<EC 1st number>/<EC 2nd number>/<EC 3rd number>/<EC 4th number>/
    <Entry>_msa_first_row.npy
    <Entry>_msa.npy
    <Entry>_pair.npy
    <Entry>_single.npy

Content of each output file, where r is number of amino acid residues:

<Entry>_msa_first_row.npy: First row of MSA representation, shape: (512, r, 256)

<Entry>_msa.npy: Full MSA representation, shape: (r, 256)

<Entry>_pair.npy: Pair representation, shape: (r, r, 128)

<Entry>_single.npy: Single Representation, shape: (r, 384)

Citation

If you use this source code for your publication, plase cite

1. Mirdita M, Schütze K, Moriwaki Y, Heo L, Ovchinnikov S and Steinegger M. ColabFold: Making protein folding accessible to all. Nature Methods (2022) doi: 10.1038/s41592-022-01488-1

2. Jumper et al. "Highly accurate protein structure prediction with AlphaFold." Nature (2021) doi: 10.1038/s41586-021-03819-2

3. If you use AlphaFold-multimer, please cite Evans et al. "Protein complex prediction with AlphaFold-Multimer." biorxiv (2021) doi: 10.1101/2021.10.04.463034v1