I have the RNA-Seq counts and FPKMs of several samples of breast cancer and adjacent normal tissue (controls).
I performed co-expression analysis of these datasets to identify genes related to my gene of interest.
Now I want to create a network to better visualize how the co-expressed genes are related, but I don't know how to generate a table of edges and vertices. I only have a list with the names of the genes detected in each co-expression module.
I'm planning to use igraph, but I never performed a network analysis before and I don't know if this is the best package to do that.
I can program in R and Python.
How can I get started? I only have the counts and FPKMs from the RNA-Seq datasets and the names of co-expressed genes, without further numeric information about how these genes are related. The list was generated by the enrichR package. How can I generate the datasets that are necessary to run a network analysis?
thank you very much,
Fabiano
igraphis pretty straightforward to use and provides very nice visuals. The edges can be coloured by numeric values, owever, I do not believe this is necessary. The best way to get started is to read tutorials and apply it to your dataset - https://igraph.org/python/tutorial/latest/tutorial.html, https://www.bioss.ac.uk/people/helen/igraphIntro.html. On your other question on network analysis, I am not too sure what you mean.Have you read any papers on coexpression network analysis? If you know how to program you know how to generate a table of edges and vertices. You seam uncertain of what you want the edges in your network to represent, and thus what kinds of questions or interpretation you would bring to a network map if you were able to generate one. If you could clarify this, it would help.
The gene co-expression analysis gives me modules of genes.. I believe all the genes that belong to the module where my gene of interest is are the nodes, but I'm not certain about the links.. I believe it would be a complete network, with all the nodes interconnected.. I'm wondering if it would be appropriate to use the FPKM values from these genes as the weights.. I could represent the weights by the node size, the higher the FPKM value, the greater the node in the gene network..
Again, it's essential to define the questions you think the network can answer, and to be able to write down a basic description of what it represents. What software did you use for the analysis? Did you read the paper behind the software? Usually the edges represent the correlation of expression between the two genes. If two genes have similar profiles across your data set, this would give high correlation and they would be connected by an edge. If they have different profiles, they would not be connected. Many genes with similar profiles, and thus a high degree of interconnectivity, make up modules. Once modules have been defined, one can measure the distance between any gene and a module, etc. etc.