rosettafold_all_atom

Rosettafold-All-Atom Example

Below we provide a sample bash script, a fasta file renamed from one of the RFAA examples and a config yaml file.

You must run RFAA outside of the installation directory.

In order to do this you must provide your own config yaml file(s) and set the location of this yaml file. We show an example using the --config-dir option from hydra.

These environment variables must be set:

ROSETTAFOLDAA_X

ROSETTAFOLD_DB_PATH

RFAA_DIR

DGLBACKEND

In addition, the script below uses a project directory that needs to be set or changed:

MY_PROJ_DIR=/path/to/my/project

The example script expects two files in the current directory:

sbgrid_example.yaml

and

sbgrid_example.fasta

We purposely renamed these example files to be sure we're using our own files and not the examples files that come with RFAA.

sbgrid_example.yaml is protein.yaml and base.yaml from the RFAA config/infernec examples directory.

sbgrid_example.fasta contains the 7U7W_1 protein from that same example.

#!/usr/bin/env bash

## SBGrid RosettaFold-All-Atom example
##
## help@sbgrid.org
## May 13, 2024


# SLURM
#SBATCH --mem=64G
#SBATCH -t 2:00:00
#SBATCH -p gpu_quad
#SBATCH --gres=gpu:1

MY_PROJ_DIR=<project directory>
cd ${MY_PROJ_DIR}

module load cuda/11.2

# Start SBGrid environment
source /programs/sbgrid.shrc

# Set version of RFAA
export ROSETTAFOLDAA_X=bf21483

# Set critical env variables
export DGLBACKEND=pytorch
export ROSETTAFOLD_DB_PATH=/n/shared_db/RoseTTAFold
export RFAA_DIR=/programs/x86_64-linux/rosettafoldaa/${ROSETTAFOLDAA_X}/RoseTTAFold-All-Atom

# Call the correct python bundled with RFAA
python.rfaa --version

# Run inference, override hydra parameters from YAML file
python.rfaa -m rf2aa.run_inference \
  --config-name sbgrid_example --config-dir ./ \
  checkpoint_path=${RFAA_DIR}/RFAA_paper_weights.pt \
  database_params.command=${RFAA_DIR}/make_msa.sh \
  database_params.hhdb=${ROSETTAFOLD_DB_PATH}/pdb100_2021Mar03/pdb100_2021Mar03