What's the average run time of Rseg

[Kadir Akdemir]

Hello,

I have been trying to run Rseg on our H3K27me3 human data in an
institutional cluster where I have to specify estimated run time of
the jobs. Currently, I run Rseg with the following parameters and
my .bed file sizes are around 1 gb with 21-26 million unique tags

./rseg-diff -c human-hg18-size.bed.txt -o PWD -i 20 -v -mode 3 -d
deadzones-k36-hg18.bed.txt R2-H3K27ME3-sorted.bed R0-H3K27ME3-
sorted.bed

I submitted the job 12 hours ago with 5 CPU but it's still at the
estimating parameters stage. From experience could you please give me
an approximate running time/memory-needs of Rseg for whole human
histone data?

Thanks,
Kadir

[Song, Qiang]

Hi Kadir,

Evaluating the NBDiff distribution is quite time-consuming: that's the

reason that rseg-diff is quite slow. When I use rseg-diff, I usually train

on a subset of the whole data with the -training-size option, for example, 

-training-size 100000, which takes about 1hr for a typical human ChIP-Seq 

dataset.

Regards,

Song Qiang

[Kadir Akdemir]

Hi Song,

Thanks for your prompt reply. I changed used the training parameter as
you suggested. It took around 25 hours but results looks good. Im
asking this and you may reply it only if you have time:

Do you think another parameter usage might decrease the running time?
I may need to use rseg-diff for several histone modifications and just
wanted to know if it's possible to get 1hr running time as you
described.

Thanks a lot for suggestion and also (to whole team) making Rseq
available to the community...

Kadir

On Apr 13, 12:17 am, "Song, Qiang" <qiang.s...@usc.edu> wrote:
> Hi Kadir,
>
> Evaluating the NBDiff distribution is quite time-consuming: that's the
> reason that rseg-diff is quite slow. When I use rseg-diff, I usually train
> on a subset of the whole data with the -training-size option, for example,
> -training-size 100000, which takes about 1hr for a typical human ChIP-Seq
> dataset.
>
> Regards,
> Song Qiang
>