Speed up computation time for metasv

Question

[kmsh410]

HI Biostars,

Currently, I am using metasv to merge SVs from the outputs of BreankDancer, CNVNATOR, and Pindel for a human genome. I was wondering if there are some tricks that I could accelerate the computational time? I also posted the same question[issue #134] to the author on metasv GitHub, but I am not sure if I can receive any reply from the developer. Any suggestions will be appreciated.

I downloaded metasv from anaconda by using the command below:

conda install -c bioconda metasv

The version of metasv:

[ksux 18:11:36 ksux_SVE]$ run_metasv.py --version
run_metasv.py 0.5.4

I performed the run_metasv.py on the example files without any issue, so I moved to my own data. The running time of metasv on our HPC is over 5 days now. Here I listed my bash command.

#!/bin/bash
#SBATCH --qos=long
#SBATCH --time=7-00:00:00
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=20
#SBATCH --mem=64G

module load anaconda/2.5.0 bedtools/2.27.1
module load gcc/4.8.2
module load cmake/3.0.2 ROOT/5.34.36
export CONDA_ENVS_PATH=/lustre/project/ksux_SVE
unset PYTHONPATH
source activate SVE

metaSV_ref=Homo_sapiens_assembly38.fasta
breakdancer_our=/data/BreakDancer_out/Subject_ID.sv.tbl
cnvnator_call=/data/CNVnator_out/Subject_ID.cnv.xls
pindel_out=/data/pindel_out/Sample_dir/Subject_ID/*
sample_idSubject_ID_tbl
alignments_bam=/data/Subject_ID.bam
spades_exe=/ksux_SVE/SVE/bin/spades.py
age_align_exe=/ksux_SVE/SVE/bin/age_align
threads=20
work=/data/metaSV_work2
OUTDIR=/data/metaSV_out2
insert_size_mean=260.04
insert_size_sd=56.34
metaSV_svs_to_assemble={'DEL','INS','INV','DUP'}

run_metasv.py --reference $metaSV_ref
--breakdancer_native $breakdancer_our
--cnvnator_native $cnvnator_call
--pindel_native $pindel_out
--sample $sample_id
--bam $alignments_bam
--spades $spades_exe
--age $age_align_exe
--num_threads $threads
--workdir $work
--outdir $OUTDIR
--isize_mean $insert_size_mean
--isize_sd $insert_size_sd

I didn't find any issues in the log file so far but the running time is over than I expected.

Last, thanks for reading this post.

[kmsh410]

Update: The author suggested to turn off the assemble function in the metasv program, and the running time was reduced to ~10 minutes for a subject. If you want to use the assemble function, that can take quite some time depending on the data. In my case, I have 2x 350bp reads (21x) for my bam file, and the overall computational time will be over 7 days. I like to keep it running, but our server allows the maximum computational time in one week only.

[ATpoint]

Out of curiosity, which platform generates PE 350bp reads? Even the MiSeq is capped at 2x300.