Hi, I’m trying to make a co-occurrence network but I’m not sure if the defaults parameters from sparCC are enough for my dataset (I have around 2500 OTU’s). Which parameters should I take in consideration, should I run the job in parallel or is enough with one processor?
mothur > sparcc(shared=data.1.subsample.shared)
The mothur implementation is much faster than the original python code, but I would still split it over a few processors because it can be a slow analysis.
I would recommend removing some of the rare/noisey OTUs. Two good papers on the subject are Berry & Widder (2014) and Weiss et al. (2016), which go into some detail on the ways you can prepare your data for analysis.