How much memory / how many cores should I request for peak calling?

Question

[jrleary]

I've never run ChIPseq before, so I've been writing a pipeline from scratch to perform all the necessary read trimming, alignment, etc. I've successfully gotten to the peak calling step, which I will be performing using MACS2. My question is, how much memory and how many cores should I request from our HPC cluster for each peak call? I have no idea how computationally intensive the process is, and I'd like it to be completed fairly quickly, while not requesting so many resources that the jobs I submit take forever to get picked up. We use slurm as a job scheduler, if that information makes any difference (doubt it does).

Answer 1

[GenoMax]

It will depend on the peak caller you are planning to use. If the peak caller is not thread-aware/multi-threaded you can just ask for a single core. Asking for more cores is not going to speed the process up or allow the program to use multiple cores, if it is not designed/written to do so.

[jrleary]

Ah yes, I had kind of assumed that MACS2 had multithread support but according to this feature request on their GitHub it isn't supported yet. Guess I'll just request one core and a ton of memory. Thanks!

[jared.andrews07]

You shouldn't need more than 8 GB of memory for MACS.