AlphaFold¶

Description¶

This package provides a container that can be used to run AlphaFold 3.0. Due to licensing restrictions, it does not provide a complete installation. Individual researchers must obtain approval from Google to download the model parameters required by AlphFold 3.

AlphaFold 3 results may not always match the alphafoldserver.com outputs; users are referred to the corresponding GitHub issue for ongoing updates: https://github.com/google-deepmind/alphafold3/issues/492

Environment Modules¶

Run module spider alphafold3 to find out what environment modules are available for this application.

Environment Variables¶

HPC_ALPHAFOLD_IMG - container image
HPC_ALPHAFOLD_REF - reference directory

Additional Usage Information¶

Obtaining model parameters¶

Prior to running AlphaFold 3, individual researchers must obtain the model parameters. To request access to the AlphaFold 3 model parameters, please complete this form.

Genetic Databases¶

AlphaFold3 also requires a writable directory of genetic database files. The files can be found in /data/reference/alphafold/3.0.0 but will need to be copied or linked to a user writable directory.

Running AlphaFold¶

The following commands are adapted from the AlphaFold documentation:

ml alphafold3
export AF_WORKDIR = /blue/mygroup/username/alphafold

apptainer exec \
  --nv \
  --bind /data/reference/alphafold/3.0.0:/root/databases \
  $HPC_ALPHAFOLD_IMG \
  python /app/alphafold/run_alphafold.py \
  --json_path=$AF_WORKDIR/af_input/fold_input.json \
  --model_dir=$AF_WORKDIR/models \
  --db_dir=/root/databases \
  --output_dir=$AF_WORKDIR/af_output

Categories¶

protein, structure, machine_learning