I am trying to use bumphunter to find differencially methylated regions between two groups of samples. I have read in the user guide and some tutorials that it is recommended to use B=1000 to perform 1000 permutations when you have a high amount of samples. In my case, I am studying about 500 samples.
The problem is that the program is highly resource consuming and so it does not finish even when using it on a cluster. I have tryed to set B=50 and run it on 10 cores, but the job was killed because of memory consumption.
Could anybody explain what is exactly the functionality of the permutations? Would it be acceptable to use less than 50 permutations in order to get confident results?