bsv-R1-S8_bqsr_gather_run_parallel.pbs

#!/bin/bash

#########################################################
# 
# Platform: NCI Gadi HPC
# Description: run GATK gatherBQSRReports over parallel tasks
# Details:
# 	BQSR BaseRecalibrator produces one table file per BQSR task. 
#	These tables need to 
#	be gathered into a merged table, so the number of gather tasks
#	 = the number of samples. Each task uses 1 CPU for less than 
#	1 minute, however if there are a large number of samples it 
#	makes sense to parallelise. I like to use 2 samples to 1 CPU 
#	or rounded to 1 node, eg 100 samples - use 48 CPU, 80 samples, 
#	use 40 CPU. For datasets with a small number of samples, use 
# 	the non-parallel (bqsr_gather_serial.pbs). Both the parallel 
#	and the serial method require the script bqsr_make_input.sh 
#	to be run first
#
# Author: Cali Willet
# cali.willet@sydney.edu.au
# Date last modified: 14/10/2020
#
# If you use this script towards a publication, please acknowledge the
# Sydney Informatics Hub (or co-authorship, where appropriate).
#
# Suggested acknowledgement:
# The authors acknowledge the scientific and technical assistance 
# <or e.g. bioinformatics assistance of <PERSON>> of Sydney Informatics
# Hub and resources and services from the National Computational 
# Infrastructure (NCI), which is supported by the Australian Government
# with access facilitated by the University of Sydney.
# 
#########################################################

#PBS -P <project>
#PBS -N bsv-R1-S8
#PBS -l walltime=00:10:00
#PBS -l ncpus=37
#PBS -l mem=148GB
#PBS -q express
#PBS -W umask=022 
#PBS -l wd 
#PBS -o ./Logs/bsv-R1-S8.o 
#PBS -e ./Logs/bsv-R1-S8.e 
#PBS -lstorage=<lstorage>
 
 
module load openmpi/4.0.2
module load nci-parallel/1.0.0


set -e

round=<round>

SCRIPT=./bsv-R1-S8_bqsr_gather.sh 
INPUTS=./Inputs/bqsr_gather.inputs   #just a sample ID list - not specific to any round 


NCPUS=1


#########################################################
# Do not edit below this line  
#########################################################

if [[ $PBS_QUEUE =~ bw-exec ]]; then CPN=28; else CPN=48; fi
M=$(( CPN / NCPUS )) #tasks per node

sed "s|^|${SCRIPT} |" ${INPUTS} > ${PBS_JOBFS}/input-file

mpirun --np $((M * PBS_NCPUS / CPN)) \
        --map-by node:PE=${NCPUS} \
        nci-parallel \
        --verbose \
        --input-file ${PBS_JOBFS}/input-file