Optimal network construction strategy to investigate interactions between proteins and metabolites in more than one disease

Question

[Giovana]

While conducting a systematic review to identify proteins and metabolites associated with three different diseases, I am planning to build networks to investigate interactions among these proteins/metabolites. I intend to use BioGRID to expand the networks. My question is: should I create separate networks for the genes of each disease and then integrate them, or can I build a single network including all genes found in the three diseases? What would be the best approach when constructing networks from data collected from other studies?

Answer 1

[Scooter]

Hi Giovana,

Well, unfortunately, the answer is "it depends". What are you trying to show? If you are trying to show network rewiring, then you might look at the different between three separate networks. If you are looking for overlapping nodes and edges across diseases, I would consider doing a merged network, but create three columns for each gene that state whether that gene is present in that diseases. Then you could use a pie chart which would be colored for each disease, but white if it's not present in that disease. Generally, when approaching a problem like this, I try to whiteboard it out to visualize what it is I want to show. If you don't have too many nodes, you might also include a "disease node" and link each node to the disease node. A layout should move common proteins and metabolites into the center and specific ones to the periphery.

-- scooter

P.S. If you do decide to use a pie chart, keep in mind that that will only be visually distinguishable if the number of nodes is relatively small...