AlphaFold v2.3.1 is now available in CCR's 2023.01 software repository.
To use AlphaFold, ensure you're accessing CCR's 2023.01 software repository (see here):
$ module load foss alphafold
You can find an example slurm job here which demonstrates how run AlphaFold on a GPU node to run a single sequence prediction:
/util/software/examples/alphafold
To run a test job, copy the alpha-fold-test.sh and test sequence T1050.fasta to your home or project directory.
A couple notes on running AlphaFold:
1. AlphaFold does not currently support nvidia H100 cards. So you'll need to use CCR's A100 or V100 GPU nodes.
2. CCR is providing the storage space for the AlphaFold databases so that research groups do not need to storage these in their own project directories. We downloaded the full genetic databases and model parameters and the path is automatically included when running run_alphafold.py. The full path can be found here:
3. Many of the examples you'll find online run AlphaFold in docker. You do not want to do this. Instead just substitue "python3 docker/run_docker.py" with the script provided by our module "run_alphafold.py". They will have the same CLI arguments.
If you have any problems or questions, please contact CCR Help.
Dori Sajdak
/util/software/data/alphafold
If you have any problems or questions, please contact CCR Help.