Skip to main content

Folding proteins with ColabFold

Open In Colab

Protein folding in silico​

In this tutorial, we perform protein folding with PLEX.

There are multiple reasons we believe PLEX is a new standard for computational biology 🧫:

  1. With a simple python interface, running containerised tools with your data is only a few commands away
  2. The infrastructure of the compute network is fully open source - use the public network or work with us to set up your own node
  3. Every event on the compute network is tracked - no more results are lost in an interactive compute session. You can base your decisions and publications on fully reproducible results.
  4. We made adding new tools to the network as easy as possible - moving your favorite tool to PLEX is one JSON document away.

We'll walk through an example of how to use PLEX to predict a protein's 3D structure using ColabFold. We will use the sequence of the Streptavidin protein for this demo.

img

Install PLEX​

!pip install PlexLabExchange
Collecting PlexLabExchange
  Downloading PlexLabExchange-0.8.18-py3-none-manylinux2014_x86_64.whl (26.9 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 26.9/26.9 MB 20.1 MB/s eta 0:00:00
[?25hInstalling collected packages: PlexLabExchange
Successfully installed PlexLabExchange-0.8.18

Then, create a directory where we can save our project files.

import os

cwd = os.getcwd()
!mkdir project

dir_path = f"{cwd}/project"

Download protein sequence​

We'll download a .fasta file containing the sequence of the protein we want to fold. Here, we're using the sequence of Streptavidin.

!wget https://rest.uniprot.org/uniprotkb/P22629.fasta -O {dir_path}/P22629.fasta # Streptavidin
--2023-08-01 21:39:21--  https://rest.uniprot.org/uniprotkb/P22629.fasta
Resolving rest.uniprot.org (rest.uniprot.org)... 193.62.193.81
Connecting to rest.uniprot.org (rest.uniprot.org)|193.62.193.81|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 264 [text/plain]
Saving to: ‘/content/project/P22629.fasta’

/content/project/P2 100%[===================>]     264  --.-KB/s    in 0s      

2023-08-01 21:39:21 (144 MB/s) - ‘/content/project/P22629.fasta’ saved [264/264]

Fold the protein​

With the sequence downloaded, we can now use ColabFold to fold the protein.

from plex import CoreTools, plex_create

initial_io_cid = plex_create(CoreTools.COLABFOLD_MINI.value, dir_path)
Plex version (v0.8.4) up to date.
Temporary directory created: /tmp/9ed8c638-c1b0-43da-bf92-7f054517d45c2889128719
Reading tool config:  QmcRH74qfqDBJFku3mEDGxkAf6CSpaHTpdbe1pMkHnbcZD
Creating IO entries from input directory:  /content/project
Initialized IO file at:  /tmp/9ed8c638-c1b0-43da-bf92-7f054517d45c2889128719/io.json
Initial IO JSON file CID:  QmUhysTE4aLZNw2ePRMCxHWko868xmQoXnGP25fKM1aofb

This code initiates the folding process. We'll need to run it to complete the operation.

from plex import plex_run

completed_io_cid, completed_io_filepath = plex_run(initial_io_cid, dir_path)
Plex version (v0.8.4) up to date.
Created working directory:  /content/project/2ef79c16-6f59-4e44-aea7-c39db85280cb
Initialized IO file at:  /content/project/2ef79c16-6f59-4e44-aea7-c39db85280cb/io.json
Processing IO Entries
Starting to process IO entry 0 
Job running...
Bacalhau job id: 476d232b-e1c6-42d6-b1c0-2f4d237244b1 

Computing default go-libp2p Resource Manager limits based on:
    - 'Swarm.ResourceMgr.MaxMemory': "6.8 GB"
    - 'Swarm.ResourceMgr.MaxFileDescriptors': 524288

Applying any user-supplied overrides on top.
Run 'ipfs swarm limit all' to see the resulting limits.

Success processing IO entry 0 
Finished processing, results written to /content/project/2ef79c16-6f59-4e44-aea7-c39db85280cb/io.json
Completed IO JSON CID: QmdnjMsUar6nTqGwgjCwN1Fyjaan4i3zyht9SE9L235YRm

Viewing the results​

After the job is complete, we can retrieve and view the results. The state of each object is written in a JSON object. Every file has a unique content address.

import json

with open(completed_io_filepath, 'r') as f:
  data = json.load(f)
  pretty_data = json.dumps(data, indent=4, sort_keys=True)
  print(pretty_data)
[
    {
        "errMsg": "",
        "inputs": {
            "sequence": {
                "class": "File",
                "filepath": "P22629.fasta",
                "ipfs": "QmR3TRtG1EWszHJTpZWZut6VFqzBPWT5KYVJvaMdXFLWXn"
            }
        },
        "outputs": {
            "all_folded_proteins": {
                "class": "Array",
                "files": [
                    {
                        "class": "File",
                        "filepath": "P22629_unrelaxed_rank_1_model_1.pdb",
                        "ipfs": "QmXZHhB7qP1tnJNyR2TeH7m4gB1R5UF84SzvK94eYB9qdL"
                    },
                    {
                        "class": "File",
                        "filepath": "P22629_unrelaxed_rank_2_model_4.pdb",
                        "ipfs": "QmPWGR36mbm5qptniHxd5KjUQKVn8EFMc57DMJzwcetNnU"
                    },
                    {
                        "class": "File",
                        "filepath": "P22629_unrelaxed_rank_3_model_3.pdb",
                        "ipfs": "QmXQ1F8xD3TP1qDvU1HDhpuR5JDZvxv1G2udJSdTsimKvH"
                    },
                    {
                        "class": "File",
                        "filepath": "P22629_unrelaxed_rank_4_model_2.pdb",
                        "ipfs": "QmV4TZJyWbu4CcmLTvD6nKM8YpzDK4fBsiiA3KQkHjW1RG"
                    },
                    {
                        "class": "File",
                        "filepath": "P22629_unrelaxed_rank_5_model_5.pdb",
                        "ipfs": "QmVHT7nQzmNkxDJsRTJPAFqwqhqEgmD3QBGZpUPneogVqX"
                    }
                ]
            },
            "best_folded_protein": {
                "class": "File",
                "filepath": "P22629_unrelaxed_rank_1_model_1.pdb",
                "ipfs": "QmTxVHTSUr8kLa9W8yM7KUNth2pNn8m3x6M18x8yiaV2SU"
            }
        },
        "state": "completed",
        "tool": {
            "ipfs": "QmcRH74qfqDBJFku3mEDGxkAf6CSpaHTpdbe1pMkHnbcZD",
            "name": "colabfold-mini"
        }
    }
]

The output is a JSON file with information about the folded protein structures. This can be used for further analysis, visualization, and more.

Open In Colab