But those answers don't mention how many sample sizes are needed per treatment/group. For example, an experiment with 18 total samples and 2 treatments will have 9 samples per treatment, whereas if 6 treatments are used then there will only be 3 samples per treatment (assuming equal number of samples per treatment).
In my case, I think I will have 20-25 samples. I would prefer to use 3 treatments, but I rather limit my design to 2 treatments if my sample size is not big enough.
You need a recommended minimum of 15 samples to build a network using all samples together regardless of treatment. Intuitively I will say that if groups members are evenly distributed is better.
On a second step, you will correlate modules with treatments and test for significance, the sample size on each group would depend on the effect size of your treatment. I don't remember reading any power study for this step. I will use as many samples as I can afford.