What is a biologically meaningful size of gene module?

Question

[druggable]

Hi All,

Based on your experience, what is a meaningful size of a gene module detected by, say biclustering or WGCNA? What can I do if a module has more than 1000 genes, does it still make sense to perform enrichment analysis on such large modules? Should I further break them down?

Thanks!

[andres.firrincieli]

Based on your experience, what is a meaningful size of a gene module detected by, say biclustering or WGCNA?

Based on my experience with WGCNA, the size does not tell anything about the biological meaningfulness of the module. Depending on the settings, the overall size of the moduels can change. For example, in cutreeDynamic if you change the cutHeight argument to 0.9 you will get smaller modules; if you leave the cutHeight argument with the default setting (0.99) you will get larger modules but also include within each cluster the genes that populate the upper region of the gene dendrogram (low connected genes). Ultimately, these settings do not really matter because one typically focuses on the tip of the branches (highly connected hub genes). This answer to your next question:

What can I do if a module has more than 1000 genes, does it still make sense to perform enrichment analysis on such large modules?

If you have very large modules, with thousands of genes, you could perform enrichment analysis on the hub genes. With such large modules, most of the categories that are significantly enriched could be easily represented by low-connected genes.