How to design WGCNA: one analysis over all sample or seperate analysis (using consensus module analysis)?


I have (wild type /mutant) (male / female) (3 developmental stages) * (each 3 replicates ) total 36 samples.
Seeing sexual difference in [WT vs MT] is the main goal.

Samples in each developmental stages (12 samples) were prepared, sequenced, and analyzed seperately.
I currently have DESeq normalized read count data as the result.

I am confused about which is the most appropriate method to do WGCNA.

  1. Use all samples in one adjacency matrix to identify modules once.
  2. Separate by condition to identify modules seperately.
    (If so, should I seperate by sex/ genotype/ time ?)
  3. Separate by condition and identify consensus module.
    Currently, I am thinking of seperating samples by developmental stages and identifying consensus modules to avoid batch effect.

Which analysis best fit the purpose? What critetria should I consider when I choose between these different analysis?
If I seperate by developmental stages, I have 12 samples each. I read that at least 15 samples are
needed. Also, to compare between male vs female / WT vs mutant, I only have 3 replicates each. Is the sample size too small?




Source link