I am currently working on a time course data which has three genotypes and six timepoints with three replicates each. I used the vst function to normalize and then transform the raw data. Further, I performed weighted gene co-expression network analysis and got good biological signals. However, later I realized that I have not set any design formula and executed the vst function without "Blind= FALSE". I set the design formula and then executed with "Blind=FALSE". The PCA and hierarchical clustering showed very similar results. Even so, the co-expression network showed very little variation. So, is it fine even if the design formula is not set? or is it very important to set the design formula? It would be really helpful for my thesis if someone could give their opinion on this.

This was asked before: Variance stabilization transformation (VST), blind=TRUE

Take-home message is that influence is minor, as you see yourself.