Alphafold¶

Franklin has a deployment of deepmind's alphafold as well as its databases compiled from source¹. The databases are located at /share/databases/alphafold; this directory is exported as $ALPHAFOLD_DB_ROOT when the module is loaded.

This is not a docker deployment. As such, many of the examples provided online need to be slightly modified. The main script supplied by the alphafold package is run_alphafold.py, which is the script that the docker container calls internally. All the same command line arguments that can be passed to alphafold's run_docker.py script can be passed to run_alphafold.py, but the latter requires all the database locations be supplied:

run_alphafold.py \
--data_dir="$ALPHAFOLD_DB_ROOT" \
--uniref90_database_path=$ALPHAFOLD_DB_ROOT/uniref90/uniref90.fasta \
--mgnify_database_path=$ALPHAFOLD_DB_ROOT/mgnify/mgy_clusters_2022_05.fa \
--template_mmcif_dir=$ALPHAFOLD_DB_ROOT/pdb_mmcif/mmcif_files \
--obsolete_pdbs_path=$ALPHAFOLD_DB_ROOT/pdb_mmcif/obsolete.dat \
--bfd_database_path=$ALPHAFOLD_DB_ROOT/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt \
--uniref30_database_path=$ALPHAFOLD_DB_ROOT/uniref30/UniRef30_2021_03 \
--pdb70_database_path=$ALPHAFOLD_DB_ROOT/pdb70/pdb70 \
[OTHER ARGS]

Because this is annoying, we have supplied a wrapper script named alphafold-wrapped with our module that passes these common options for you. Any of the arguments not passed above will be passed along to the run_alphafold.py script; for example:

alphafold-wrapped \
--output_dir=[OUTPUTS] \
--fasta_paths=[FASTA INPUTS] \
--max_template_date=2021-11-01 \
--use_gpu_relax=true

Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., … Hassabis, D. (2021). Highly accurate protein structure prediction with AlphaFold. Springer Science and Business Media LLC. https://doi.org/10.1038/s41586-021-03819-2 ↩