Skip to content

Slurm array: Run multiple BLAST queries in parallel with a single submission script

BLAST finds regions of similarity between biological sequences. The program compares nucleotide or protein sequences to sequence databases and calculates the statistical significance

For the demonstrator
  • Working directory /nesi/project/nesi99999/Dinindu/20230503-pfr-demo/blast
  • Sequences were delivered in a single file /nesi/project/nesi99999/Dinindu/20230503-pfr-demo/blast/parent_file/demo-hic.fa
  • Sequence file was split to 150 separate queries by sequence with faSplit and stored in input-queries 
    $ pwd
/nesi/project/nesi99999/Dinindu/20230503-pfr-demo/blast

$ faSplit sequence parent_file/demo-hic.fa 150 split-queries/demo-hic

$ ls split-queries/

demo-hic000.fa  demo-hic019.fa  demo-hic038.fa  demo-hic057.fa  demo-hic076.fa  demo-hic095.fa  demo-hic114.fa  demo-hic133.fa
demo-hic001.fa  demo..............................
  • Slurm Submission script is /nesi/project/nesi99999/Dinindu/20230503-pfr-demo/blast/scripts/demo-array.slurm
  • Query outputs will be saved to /blast-out and the Slurm StdOut to /slurm-logs

Slurm array script

#!/bin/bash -e

#SBATCH --account        nesi99999
#SBATCH --job-name       blast_fastaSplit
#SBATCH --cpus-per-task  1
#SBATCH --mem            2G
#SBATCH --time           24:00:00
#SBATCH --array          0-149
#SBATCH --output         /nesi/project/nesi99999/Dinindu/20230503-pfr-demo/blast/slurm-logs/%A_%a.out

date;hostname;pwd

module load BLASTDB/2024-01
module load BLAST/2.13.0-GCC-11.3.0

export INPUT_DIR=/nesi/project/nesi99999/Dinindu/20230503-pfr-demo/blast/input-queries
export OUTPUT_DIR=/nesi/project/nesi99999/Dinindu/20230503-pfr-demo/blast/blast-out


RUN_ID=$(( $SLURM_ARRAY_TASK_ID + 1 ))

QUERY_FILE=$( ls ${INPUT_DIR} | sed -n ${RUN_ID}p )
QUERY_NAME="${QUERY_FILE%.*}"

QUERY="${INPUT_DIR}/${QUERY_FILE}"
OUTPUT="${OUTPUT_DIR}/${QUERY_NAME}.out"

echo -e "Command:\nblastn –query ${QUERY} –db nt –out ${OUTPUT} -outfmt 6 -max_target_seqs 1 -num_threads $SLURM_CPUS_PER_TASK"

blastn -query ${QUERY} -db nt -out ${OUTPUT} -outfmt 6 -max_target_seqs 1 -num_threads $SLURM_CPUS_PER_TASK 

date

submit

  • submit the script with sbatch scripts/demo-array.slurm
    • If needed, use array throttling (eeping only a certain number of tasks RUNNING at a time). Let's say we want to run only 20 queries at a time (out of 149), then adding #SBATCH --array 0-149%20 to the submission script or call during submission to sbatch command with sbatch --array 0-149%20 scripts/demo-array.slurm
  • Review the status of submission with squeue -j jobid