Tutorial:PBS setting combination for bismark alignment to BS-seq or RRBS data

Question

[Shicheng Guo]

I am thinking how to set the appropriate paramters for running bismark in PBS system.The following is the pattern I am appling now. Any suggestion to make the settting more perfect? I mean to use less computational source while with less time (human: hg19).

Thanks

Setting of bismark alignment in PBS system (DNA methylation, BS-seq, RRBS)

• multicore=1, ppn=8, memory=3.2G*6=19.2G
• multicore=2, ppn=16, memory=3.2G*12=38.4G

For majority queue, ppn=16, memory= 64G, therefore, for condo, multicore=2 is okay. don't set too larger multicore, no help for more multicore setting.

For bismark Alignment

note: '-p 1' will already use

• 4 threads/cores for Bowtie2 (3.2G/per) plus 1 additional core for Bismark itself

• 2 threads/cores for Perl (3.2G/per)

• 2 threads/cores for SAMTOOLS

note: '-p 2' will already use

• 8 threads/cores for Bowtie2 (3.2G/per) plus 1 additional core for Bismark itself

• 4 threads/cores for Perl (3.2G/per)

• 4 threads/cores for SAMTOOLS

For bismark_methylation_extractor

if you set ppn=6, then --multicore=2 || ppn=12, then --multicore=4

*--multicore=1 , require 3 threads

• 1 threads/cores for bismark_methylation_extractor (? G/per)

• 1 threads/cores for Gzip (? G/per)

• 1 threads/cores for SAMTOOLS (? G/per)

*--multicore=2, require 6 threads

• 2 threads/cores for bismark_methylation_extractor (? G/per)

• 2 threads/cores for Gzip (? G/per)

• 2 threads/cores for SAMTOOLS (? G/per)