Alphafold

Alphafold is a deep learning-based tool developed by DeepMind for predicting protein structures with high accuracy. It is computationally intensive and requires GPU resources for optimal performance.

On the wi-hpc cluster, we provide Alphafold2 and the newer Alphafold3.

Note

Alphafold (both version 2 and 3) are GPU heavy applications, to use Alphafold correctly, you must be using GPUs. Please ensure you are requesting GPUs in the gpu partition. I.e., --partition=gpu and --gres=gpu:x where x is the number of GPUs you would like to use. Please see our Hardware page for GPU specifics.

Alphafold 2

Alphafold2 can be used via the module load command

module load AlphaFold/2.2.2-foss-2021a-CUDA-11.3.1

The DB for alphafold2 is located at /resources/AlphafoldDB

Example script

#!/usr/bin/env bash
#SBATCH --job-name=alphafold2_job
#SBATCH --output=alphafold2_output.log
#SBATCH --partition=gpu
#SBATCH --gres=gpu:1
#SBATCH --cpus-per-task=8
#SBATCH --mem=64G
#SBATCH --time=24:00:00

# Load necessary modules
module load Alphafold2/2.2.2-foss-2021a-CUDA-11.3.1

# Print out AlphaFold2 help
alphafold \
    --json_path=/path/to/json/file \
    --output_dir=/path/to/output \
    --model_dir=/resources/alphafold_models/v2 \
    --db_dir=/resources/AlphaFoldDB/

Alphafold 3

Warning

To be able to read and use the model parameters for Alphafold3, you must submit a ticket to the IT Help Desk to review and sign Wistar's "Alphafold3 Model Parameters Access Terms of Use". The CIO (Chief Information Officer) will need to approve of this ticket for it to complete.

The newest version of Alphafold3 is now available on the wi-hpc cluster via a container. This means you must load and use the apptainer module for this to work.

Note

Please note that you must use the --nv flag with apptainer to ensure that the GPU environment variables are properly passed into the container.

Alphafold3 can be used via the apptainer run command:

module load apptainer
apptainer run --nv /applications/Alphafold/alphafold3 --helpfull

The DB for alphafold3 located at /resources/Alphafold3DB and the model parameters located at /resources/alphafold_models/v3.

Example Script

#!/usr/bin/env bash
#SBATCH --job-name=alphafold3_job
#SBATCH --output=alphafold3_output.log
#SBATCH --partition=gpu
#SBATCH --gres=gpu:1
#SBATCH --cpus-per-task=8
#SBATCH --mem=64G
#SBATCH --time=24:00:00

# Load necessary modules
module load apptainer

# Print out AlphaFold3 help
apptainer run --nv /applications/Alphafold/alphafold3 \
    --json_path=/path/to/json/file \
    --output_dir=/path/to/output \
    --model_dir=/resources/alphafold_models/v3 \
    --db_dir=/resources/AlphaFold3DB/