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 | Cardinal | Model Parameters |
|---|---|---|---|---|
| 2.0.0 | X | 2021-07-14 | ||
| 2.1.0 | X | X | X | 2021-10-27 |
| 2.1.2 | X* | X | X | 2022-01-19 |
| 2.2.2 | X | X* | 2022-03-02; Multimer model weights: v2 | |
| 2.2.4 | 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 | X | 2022-12-06; Multimer model weights: v3 |
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.
