Hello to everyone,
I have question regarding to network analysis. Currently, I'm conducting this kind of analysis using data retrieved from STRING db to obtain highly central nodes. I have re-constructed the network in R using the igraph package to subsequently proceed with the topological analysis. However, I'm a little bit confused since my question is how to decide what is the best centrality measure related to my network. A few weeks ago, I found a tutorial in which it was suggested to calculate distinct centrality measures (i.e., closeness, degree, betweenness, ...) and then perform a PCA to distinguish the most informative centrality measure. However, in a recent article I read that authors analyzed their network to obtain highly central nodes by using a related method. They calculated four centrality measures and then proceeded to combine them using MDS to get a weighted centrality score.
What of these methods is the most appropriate to select/get a centrality measure? or does exist a better way to conduct these kind of analysis?
I'm new in this topic, so I acknowledge you for some advices and suggestions.
You should take a step back and think about what the questions you're trying to address with this are. A lot of people seem to be computing centrality measures just because they can. Each measure says something different about the graph so think about what information they provide and whether it's relevant to your questions. I've given a bit more details in a previous answer here. Also consider that centrality measures are heavily influenced by how the graph is built and in particular the most studied proteins tend to have a strong effect.
Thanks Jean, your answer gives me more light about how should I use these centrality measures with respect to my question.