Running Bento on HPC

This guide provides a practical baseline for SLURM-based execution in a generic HPC environment.

1. Prepare Workspace

git clone https://github.com/LigandPro/Bento.git
cd Bento
cp .env.example .env

Edit .env with cluster-specific paths:

BENTO_WORKDIR=/path/to/Bento
BENTO_DATABASES_DIR=/path/to/shared/datasets
BENTO_GLOSA_DIR=/path/to/glosa_v2.2

2. Load Cluster Modules

Module names can differ by site configuration. Use local equivalents:

module purge
module load python/3.12
module load java/17
module load gcc

3. Build GLoSA Runtime (One-Time)

In your BENTO_GLOSA_DIR:

cd "$BENTO_GLOSA_DIR"
g++ -c glosa.cpp
g++ -o glosa glosa.o
javac AssignChemicalFeatures.java

4. Sync Environment

uv sync --frozen --no-editable --extra annotation --extra similarity

Validate toolchain:

uv run --extra annotation bento check-env --profile annotation
uv run bento check-env --profile similarity --glosa-dir "$BENTO_GLOSA_DIR"

5. Submit Jobs

Example templates are provided in slurm/:

• slurm/run_annotation.sbatch
• slurm/run_similarity.sbatch

Submit:

sbatch slurm/run_annotation.sbatch
sbatch slurm/run_similarity.sbatch

6. Operational Recommendations

• Keep uv.lock committed and use uv sync --frozen.
• Run environment checks before expensive jobs.
• Write outputs to job-specific directories to avoid collisions.
• Capture git commit hash and full command line in each job log.