Author: Bozitao Zhong - zbztzhz@gmail.com
📑 Please cite our paper if you used ParaFold (ParallelFold) in you research.
Recent change: ParaFold now supports AlphaFold 2.3.1
This project is a modified version of DeepMind's AlphaFold2 to achieve high-throughput protein structure prediction.
We have these following modifications to the original AlphaFold pipeline:
- Divide CPU part (MSA and template searching) and GPU part (prediction model)
We recommend to install AlphaFold locally, and not using docker. An install script for ubuntu 20.04 is provided. Make sure to add ENV PATHs to your bashrc.
# install requirements
sudo apt install aria2 rsync git wget tmux tree -y
# clone this repo
git clone https://github.com/ivanpmartell/ParallelFold.git
# move to the scripts directory
cd ParallelFold/scripts
# run GPU installation script
chmod +x local_install_nvidia.sh
sudo ./local_install_nvidia.sh
echo 'export PATH=$PATH:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin' >> ~/.bashrc
# restart terminal
# run alphafold installation script
chmod +x local_install.sh
sudo ./local_install.sh
/opt/conda/bin/conda init
# restart terminal
# make script executable
chmod +x run_alphafold.sh
run_alphafold.py
: modified version of originalrun_alphafold.py
, it has multiple additional functions like skipping featuring steps when existsfeature.pkl
in output folderrun_alphafold.sh
: bash script to runrun_alphafold.py
run_figure.py
: this file can help you make figure for your system
Visit the usage page to know how to run
ParallelFold can help you accelerate AlphaFold when you want to predict multiple sequences. After dividing the CPU part and GPU part, users can finish feature step by multiple processors. Using ParaFold, you can run AlphaFold 2~3 times faster than DeepMind's procedure.
If you have any question, please raise issues