Parallelization issue for somatic sv calling pipeline

[CHOWDHURY098]

Dear Tobias,

I am presently executing the somatic structural variant (SV) calling pipeline on data from 173 patients. I've observed that processing one sample typically requires around two days. Therefore, extrapolating for all 173 patients, the analysis is estimated to consume a substantial amount of time. To expedite this process, I have attempted setting 'export OMP_NUM_THREADS=4'.
Here i attached the run script "
#!/bin/bash
export OMP_NUM_THREADS=4
#data_list="/path/173_T_N_bam.list"
ref_fasta="/path/Homo_sapiens_assembly38.fasta"
excl_tsv="/path/human.hg38.excl.tsv"

while read sample_id tumor_bam normal_bam; do
echo "$sample_id"
output_bcf="${sample_id}_sv.bcf"

delly call -x "$excl_tsv" -o "$output_bcf" -g "$ref_fasta" "$tumor_bam" "$normal_bam"

done < $1

echo "Delly call completed for all patients in $data_list""
please help me . I want this high speed . How to do this in short time ?

[tobiasrausch]

Usually we prefer to parallelize by patient, i.e., run multiple jobs with matched tumor-normal samples in the cluster. If you are not interested in sub-clonal mutations, you can also increase -z to 3 or even 5, depending on the input coverage. As a rule of thumb for variant allele frequency >=10% you can set -z equal to 0.1 * coverage. If your fragment size is not normally distributed, you should also increase -s to 11 or even 15.