AlphaFold

AlphaFold is a software package that provides an implementation of the inference pipeline of AlphaFold v2.0. This is a completely new model that was entered in CASP14 and pusblished in Nature.

Availability and Restrictions

Versions

Version Pitzer Ascend Model Parameters
2.0.0 X   2021-07-14
2.1.0 X   2021-10-27
2.1.2 X*   2022-01-19
2.2.2 X X* 2022-03-02; Multimer model weights: v2
2.3.1 X X 2022-12-06; Multimer model weights: v3
2.3.2 X X 2022-12-06; Multimer model weights: v3
* Current default version

You can use module spider alphafold to view available modules for a given machine. Feel free to contact OSC Help if you need other versions for your work.

Access

AlphaFold is available for all OSC users

Publisher/Vendor/Repository and License Type

Copyright 2021 DeepMind Technologies Limited

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at https://www.apache.org/licenses/LICENSE-2.0. Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See hte License for the specific language governing permissions and limitations under the License. See the License for specific langauge governing permissions and limitations under the License.

The AlphaFold parameters are made available for non-commercial use only, under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. You can find details at: https://creativecommons.org/licenses/by-nc/4.0/legalcode.

Usage

Usage on Pitzer

Set-up

To load the default version of AlphaFold module, use module load alphafold.

Batch Usage

Below is the example batch script (job.txt) for an alphafold job:

#!/bin/bash
#SBATCH --ntasks=8
#SBATCH --gpus-per-node=1
#SBATCH --gpu_cmode=shared


module reset
module load alphafold/2.1.2

run_alphafold.sh --use_gpu_relax=True --db_preset=reduced_dbs --fasta_paths=rcs_pdb_6VR4.fasta --max_template_date=2020-05-14 --output_dir=$(pwd -P)/output

The control options and presets for model and database:

Option Preset Note
--model_preset monomer
monomer_casp14
monomer_ptm
multimer
Control which AlphaFold model to run
--db_preset full_dbs
reduced_dbs

Control MSA speed/quality tradeoff

To get full-options list

run_alphafold.sh --helpshort

For very large simulations use multiple GPUs and to make sure a job can access all the GPU memory, set this before run_alphafold.sh with alphafold/2.2.2:

export TF_FORCE_UNIFIED_MEMORY=1
run_alphafold.sh ...

Note also that not all models are parallelized over multiple GPUs; see https://github.com/deepmind/alphafold/issues/30

Use custom AlphaFold

From 2.1.2 to 2.2.2, you can use own copy of AlphaFold code with our pre-installed packages and database. For example, you download a copy of AlphaFold 2.2.2 in $HOME/alphafold and make some changes. Modify the ALPHAFOLD_HOME variable before calling run_alphafold.sh, e.g.

module reset
module load alphafold/2.2.2

export ALPHAFOLD_HOME=$HOME/alphafold
run_alphafold.sh --db_preset=reduced_dbs --fasta_paths=rcs_pdb_6VR4.fasta --max_template_date=2020-05-14 --output_dir=$(pwd -P)/output

Batch Usage (2.0.0)

Below is the example batch script (job.txt) for an alphafold job:

#!/bin/bash
#SBATCH --ntasks=8
#SBATCH --gpus-per-node=2

module reset
module load alphafold/2.0.0

run_alphafold.sh --preset=reduced_dbs --fasta_paths=rcs_pdb_6VR4.fasta --max_template_date=2020-05-14 --output_dir=$(pwd -P)/output

Other available job options are:

--preset=recued_dbs, --preset=full_dbs, or --preset=casp14

 

Further Reading

Online documentation is available on the AlphaFold homepage.

Notes on AlphaFold output.

Notes on citing AlphaFold.

Tag: 
Supercomputer: 
Service: 
Fields of Science: