Entering edit mode
Is it advisable to create WGCN from RNA-seq data generated on different platforms with different read lengths but processed the same way? Suppose I have 3 datasets with each containing 5 CASE and 5 CONTROL samples.
Before we can apply VST in DESeq2 we have to normalize the data. Now Performing VST separately would mean normalizing data separately. Would that be good?
How do I judge success? Any parameter you would suggest.
VST in DESeq2 includes normalization. Since you will be (or should be) using ComBat after VST to adjust for technical differences, doing normalization separately is fine.
You can judge success by whether the resulting network makes biological sense. Are the modules enriched in terms that make sense for the tissue from which the data come? Do the associations with your trait of interest, if you have one, make sense in light of previously known information? Also, if you have independent but somewhat comparable data, you could tun module preservation to see whether modules are preserved.