jongseopark/oligofold repository overview

⁠[OligoFold] AlphaFold_oligomer

Link to related GitHub page⁠

This repository has several scripts that enables oligomer modeling using DeepMind's AlphaFold.

You can model a oligomeric structure comprising up to three types of sequences, with any number of protomers.

DeepMind's AlphaFold⁠

These scripts were developed based on the codes of AlphaFold Colab (community version⁠) made by Dr. Martin Steinegger, Sergey Ovchinnikov, and Milot Mirdita (with the insights of Dr. Minkyoung Baek (RoseTTAFold) and Yoshitaka Moriwaki), which uses API hosted at the Södinglab based on the MMseqs2 server.

ColabFold GitHub⁠

These scripts are designed to work in the system which already has a installed AlphaFold.

You can also use a already prepared docker image of this OligoFold, but it may not work if the system environment is different.

Please refer the description at the bottom.

이 저장소에서는 구글 딥마인드의 알파폴드를 통한 oligomeric structure 모델링을 가능케하는 코드를 공유합니다.

총 세 종류의 단백질 시퀀스를 이용할 수 있고, protomer 의 수를 설정하는 것에는 제한이 없습니다.

이 저장소에 있는 스크립트들은 알파폴드 Colab (community version⁠) 에 있는 코드에 기반하여 제작되었습니다.

스크립트 사용을 위해서는 이미 알파폴드를 설치 및 실행할 수 있는 시스템 환경이 필요합니다.

사용법은 아래와 같습니다.

⁠How to use

⁠1. Set the path for database and result

Open the [run_docker_oligofold.py] using text editor, and modify line 31 and 37 to be the path to the directory containing the databases and result folder, respectively.

line 31: DOWNLOAD_DIR = '/path/to/database'
line 37: output_dir = '/path/to/your/result'

⁠2. Copy and paste those scripts

pipeline.py >> /alphafold/alphafold/data/
run_alphafold.py >> /alphafold/
run_docker_oligofold.py >> /alphafold/docker/

⁠3. Rebuild the docker image

Rebuild the docker image using the Dockerfile already used to install the AlphaFold. At this point, the name of the image is named 'oligofold'.

cd /alphafold/
docker build -f docker/Dockerfile -t oligofold .  

⁠4. Run

Running protocol of the oligofold is almost identical with that of the original AlphaFold, but there are several differences.

You have to set the path of one essential fasta file using the argument, '--fasta_paths'.

You can set the number of protomers of the essential fasta file using the argument, '--oligo1'.

Since the default number of 'oligo1' is 1, it is not necessary to set a value, but if you want to make homooligomer using that fasta file, just set the number of it.

Optionally, you can set up the other two fasta paths using these arguments, '--fasta2_paths' and '--fasta3_paths'.

You are also able to set up the number of protomers you want for the fasta2 and fasta3 using these arguments, '--oligo2' and '--oligo3', respectively.

If you don't want to use the other fasta files, just ignore these options.

Finally, you are also able to choose whether to do the amber relaxation or not, using the argument, '--amberrelax' (y/n).

This is an example command.

python3 docker/run_docker_oligofold.py
--fasta_paths=/path/to/first/fastafile
--fasta2_paths=/path/to/second/fastafile
--max_template_date=2021-07-31
--preset=reduced_dbs
--oligo1=2
--oligo2=1
--amberrelax=n

Then, the modelled structure comprising two identical molecules and one another molecule will be obtained.

If you want to model the complex structure comprising three types of proteins,

just enter the three sequences and the number of protomers of each.

python3 docker/run_docker_oligofold.py
--fasta_paths=/path/to/first/fastafile
--fasta2_paths=/path/to/second/fastafile
--fasta3_paths=/path/to/third/fastafild
--max_template_date=2021-07-31
--preset=reduced_dbs
--oligo1=2
--oligo2=1
--oligo3=2
--amberrelax=n

⁠Already prepared docker image

The already prepared docker image may not work due to the differences of system environment.

This is my system environment

Ubuntu 20.04
i9-10900x
128 GB RAM
RTX 3080 * 2 ea
CUDA version 11.2
NVIDIA driver version 460.27.04

Neverthelss, if you want to try this ...

You can download the docker image as follows.

docker pull jongseopark/oligofold

And then, change the name of docker image from jongseopark/oligofold:latest to oligofold:latest

docker tag jongseopark/oligofold:latest oligofold:latest

Open the [run_docker_oligofold.py] using text editor, and modify line 31 and 37 to be the path to the directory containing the databases and result folder, respectively.

line 31: DOWNLOAD_DIR = '/path/to/database'
line 37: output_dir = '/path/to/your/result'

There is no need to rebuild the docker image, because you already download it.

Just type the running command as described earlier.

⁠And ...

Since I am a structural biologist, but not a computational biologist, those scripts which I wrote might not be fancy or cause error. So, feel free to contact me, and tell or teach about the coding.

Thanks.