I'm investigation the role of my gene of interest (GOI) in two types of cancer.
Since it is a low expressed gene, it is not an important DEG. If I run edgeR, I don't find it, using FDR < 0.5.
I decided, then, to run gene co-expression analyses using CEMiTool.
Using this strategy, I detected modules (one for each kind of cancer) containing my GOI and co-expressed genes.
Now I want to run pathways analyses, to discover the pathways associated with these modules. I'm trying to use fgsea, but fgsea asks me to rank the genes in each module by p-value. That works for DEGs that are normally listed by FDR and p-value, but CEMiToll gives me only the list of genes that are correlated with my GOI, it doesn't give me a value to rank them.
How can I investigate the pathways associated with these modules, if I only have the genes names?
If I put these genes on EnrichR, I can detect the pathways and have all the results I want, but I would like to do these analyses using the command line. How is EnrichR performing these analyses without a rank list? How can I reproduce the analyses performed by EnrichR using the command line and the Reactome database? Is there a package in R for that?
So fgsea is essentially the R version of GSEA. Its ranked based, that means that you cannot subset a set of genes. What you could do is more a of hypergeomtric test instead; which means that you take the list of genes and run the analysis, for R I think limma has an equivalent or just plug into something like cluego, david or panther.
"fgsea asks me to rank the genes in each module by p-value." P-value of what?
edgeR compares the counts of the cancer samples versus controls and gives the results (DEGs) in order, by fold change...with p-values and FDRs.. in fact, I believe the most appropriate is to select a FDR and rank the results below this FDR by logFC..