DEGs vs WGCNA: What am I really looking at?

Question

[christiantd]

Hi all,

I have previously performed DEG-analysis using DESeq2 on a dataset that has 2 groups (let's say group A and group B) and 6 replicates and found many DEGs. A PhD in my lab advised me to also perform Weighted correlation network analysis using WGCNA.

I have performed this analysis and find 2 modules associated with group A. However, upon further inspection I see that the genes in the module are not necessarily the top hits in my DEG analysis. I am having trouble wrapping my had around this concept. Could someone explain me what is the logic behind differences in these results? As I am new to this, it's all a bit confusing!

Thanks in advance!

Kind Regards,
TC

[Ram]

Why do you expect these genes to be "top hits"? Are they part of your significant genes (genes that pass the log2FC and p-adj thresholds) or not?

[christiantd]

I have used all genes for the WGCNA. I dont expect 1:1 results but would expect some overlap? Or is that a thinking error?

[Raag]

You’re performing two different types of analyses. It’s not clear to me there should be any significant overlap.

Answer 1

[LChart]

As should be clear from the "C" in WGCNA - this analysis partitions genes by correlation across your samples. Your "top DE genes" have some expression pattern on your samples which is (on average) different between group A and group B; but the pattern of the expression within A and B (i.e., the ranks of the samples) is in no way constrained by saying the means are different. Therefore: (1) your top genes need not be very strongly correlated with each other; and (2) your top genes may be correlated with genes that show a weaker (possibly not even statistically significant) effect.

[christiantd]

Thank you, this was very insightful!