Estimating RAM needed is hard because it depends on both the size and complexity of your data set. But having about as much RAM as your input data set is a good place to start.
One of the settings of the dada2 denoise-paired plugin is --p-n-reads-learn, which is set to 1 million by default. You could lower that to 100,000 or 10,000 to speed up your processing and reduce RAM usage.
(And you could add --p-n-reads-learn 4 to speed up this process too! )
Thank you!
I am checking with the new setting.
One thing I forgot to mention is that I Run through virtual box, and I set for 20GB RAM. My target is bacterial community, but my dataset for whole genome sequencing, is there any problem for setting lower the reads for learning?