I'm learning WGCNA analysis. I've worked through the first part of WGCNA tutorial by Peter Langfelder and read related posts on both Biostars and Bioconductor Forums.
My attention catched question Can I apply WGCNA on unmatched time point microarray data and clinical data?
I have similar problem, I have RNA-seq data from four maize varieties, I also have phenotypic measurements made on those lines but on different individuals (4 to 5 measurements on separate individual plants). Speaking differently: my RNA-seq and phenotypic data are not from the same individuals. So I wonder if using such phenotypic measurements (means from measurements) as traits for correlation with expression modules make sense?
Just depends on your analysis. If you're comparing 4 maize varieties, as populations it should be just fine. Consider the case where you had RNA expression from members of each of the 4 varieties, but only population statistics (average flowering time of each variety, average root density, etc). You'd be fine looking at modules that "correlate" to these population statistics (which, necessarily, would be replicated for each member of the 4 varieties).
In your scenario, the average of your measurements (per variety) is a replacement for the population average in the above discussion.
Individual-level expression/phenotype analysis is still, quite clearly, statistically invalid.
Thank you, yes I'm comparing the varieties as populations. I was in doubt because the example data from tutorial has both RNA-seq and trait data for each mouse.
Log in to answer this question.
