This repository contains a set of instructions on how-to-install alphafold on a HPC cluster. It is based on the alphafold repository.
There they go over the instructions to install alphafold on a single machine with Docker - however you might want to install it in your local cluster. This repository contains a step-by-step installation procedure as well as a custom-built wrapper CLI to facilitate the execution of alphafold in a small HPC environment.
Note that this is not an in-depth guide and not affiliated with alphafold - please check their documentation for an updated and official instructions. This repo is rather a set of instructions of a case that worked for me. I hope it can be useful for others as well.
Please refer to the INSTALL.md file for a step-by-step installation procedure. These will show how to setup this repository, install the dependencies, configure a self-contained python environment and install alphafold.
Alphafold's CLI is not very user-friendly and is not integrated with the HPC's scheduling system. This repository contains a wrapper CLI that facilitates the execution of alphafold in a small HPC environment.
It aims to expose less parameters and automates the execution of Alphafold via SLURM. It also provides a simple way to configure the execution via environment variables.
It assumes that you followed the installation procedure; it will create a SLURM job file on the fly (in memory) with the correct initialization of the miniconda alphafold environment, the arguments needed by alphafold and the correct paths to the data.
Install go and build from source:
$ cd wrapper
$ go build -o alphafold-wrapper
The wrapper will take the configuration from the system variables, currently it supports the following variables:
PARTITION: The SLURM partition to use.INSTALL_DIR: The installation directory of this repository.DATA_DIR: The path to the data directory.
Either define them in your environment or directly from the command line:
$ export PARTITION=gpu
$ export INSTALL_DIR=/trinity/login/rodrigo/repos/alphafold-wrapper
$ export DATA_DIR=/trinity/login/rodrigo/repos/alphafold-wrapper/dataThe build produces a binary called alphafold-wrapper. It has the following usage:
$ ./alphafold-wrapper -h
Usage: ./alphafold-wrapper -f target.fasta -p <monomer|monomer_casp14|monomer_ptm|multimer> -o /path/to/output
### To fold a multimer, the target.fasta must be a multi-fasta file
### Ex: `https://www.rcsb.org/fasta/entry/2OOB/display`
- Add more customization options to the SLURM job file
- Allow to re-use MSAs
Add automated tests to the wrapper✅- Add a build pipeline
If you have any questions or suggestions, please open an issue.