Comparing hdWGCNA modules with external gene sets

[gboscagli]

Greetings all,
thanks for delivering such a useful tool.

I'm working with some Visium 10X spatial datasets. I have four gene sets representing different transcriptional subtypes: my aim is to assess if some of these groups could be predictive of the cancer progression. These groups are non-overlapping and their size span from 600-1000 genes each.

At first, I followed your vignette on ST and successfully obtained some modules (very skewed in terms of size, as highlighted by other issues such as #252). I found the intersections using UpSet plot but, after that, I feel like I'm at a dead point in my analysis, I don't really know how to compare the information returned by hdWGCNA with my preexisting gene-sets. I could also try subsetting for the genes of each gene set while running SetupForWGCNA, in order to yield a sort of unsupervised validation.

Any other advice from you would be gold!

[smorabit]

Hi,

Thank you for writing this issue.

I have four gene sets representing different transcriptional subtypes: my aim is to assess if some of these groups could be predictive of the cancer progression. These groups are non-overlapping and their size span from 600-1000 genes each.
...
I feel like I'm at a dead point in my analysis, I don't really know how to compare the information returned by hdWGCNA with my preexisting gene-sets.

It sounds like your goal is to use your pre-defined gene sets analyze cancer progression. I am not sure that hdWGCNA is necessary in this case? hdWGCNA is an unsupervised way of identifying a co-expression network and highly co-expressed modules of genes within this network.

Sometimes we can identify biological functions of modules by comparing them to known gene sets, for example via pathway enrichment analysis. Another example, in Fig 4c of the hdWGCNA paper we did a gene set overlap analysis of our modules with a list of genetic risk genes.

It is hard for me to give you concrete advice on what to do next without knowing more context. If you would rather discuss in more details about your project and would like to keep the details more private you can always send me an email.