Entering edit mode
8.1 years ago
jiwpark00
▴
230
Hi,
So I was able to generate networks based on existing data sets and visualize them through Cytoscape. My question is - to more experienced bioinformaticians and whatnot - what do you do with the network? I know that obviously, people are interested in "hubs" (i.e. high degree of nodes) as well as "shortest distance" (are there plugins for Cytoscape for that?). But what else? I would be really intrigued to find which genes are important and whatnot, and maybe WGCNA would be helpful but I'm just really throwing big names at this point...
Thank you.
It depends on what your biological question is. In my experience, hubs are not that interesting because they tend to fall into two categories: either they are general protein modifiers e.g. ubiquitin or they are social artefacts i.e. genes that everybody works on like the classical oncogenes P53, SRC, EGFR ... Making sense of hubs would require removing genes falling into these categories.
Very good point. I guess my question is that, can you use hubs as a way to calculate functional importance for the non-hub genes, for example?
Like here: http://www.nature.com/nature/journal/v493/n7432/images/nature11862-f2.2.jpg
We have TLR4 for inflammation and glycolysis for metabolism. Those two are distinct processes. I guess my question is, if we somehow find genes that are linked to TLR4 and genes linked to metabolism - but no direct link between them (it would be PFKFB3 in this case), is there any way to predict/estimate that? Would removing hub genes help?
This may be too late to be useful but I'll answer anyway.
Algorithms I can think of that would be of use here rely on analysis of network topology so removing hubs would certainly have an effect. However, if hub removal mostly removes noise then the results should be more relevant without hubs than with them.
As to the question of finding indirect links between two processes, I would look into the limited k-walks approach. It's available in my Algorithms::Graph perl module.