RoseTTAFold2NA
RoseTTAFold2NA is available on all SC clusters.
The software can be loaded as module via:
Which provides the scripts run_RF2NA.sh.
The required databases are stored on the /software filesystem.
The database paths and available CPU and Memory resources have to be set via environment variables.
job script example
#!/bin/bash
#SBATCH --job-name=RoseTTAFold2NA
#SBATCH --output=rosettafold-job-out.%J
#SBATCH --time=01:00:00
#SBATCH --nodes=1
#SBATCH --cpus-per-task=16
#SBATCH --gres=gpu:rtx2080ti:1
#SBATCH --partition=clara
# load the software
module load RoseTTAFold2NA/0.2
# set the database paths
export UNIREF30PATH=/software/databases/rosettafold2na/UniRef30_2020_06/
export BFDPATH=/software/databases/rosettafold2na/bfd/
export PDB100PATH=/software/databases/rosettafold2na/pdb100_2021Mar03/
export RNAPATH=/software/databases/rosettafold2na/RNA
# set hardware constraints
export MEM=$SLURM_MEM_PER_NODE
export CPU=$SLURM_CPUS_PER_TASK
# run_RF2NA.sh <output_dir> <fasta inputs>
run_RF2NA.sh output rna_binding_protein.fa R:RNA.fa
In this example the results, checkpoint files and logs will be stored to the output folder, which will be created within the directory, in which the above command is run. The above input files are the example files from https://github.com/uw-ipd/RoseTTAFold2NA/tree/main/example. See https://github.com/uw-ipd/RoseTTAFold2NA?tab=readme-ov-file#inputs for the input files.
Hints
- if you experience "CUDA out of memory" errors, it might help to run the job on a Tesla v100 or A30 (paula) instead of RTX 2080ti