Visualizing Protein Protein Interaction Clusters

Question

[Sudeep]

Dear All,

Right now I am trying to visualize protein-protein interaction graph and clusters generated from that graph, something like this figure. After playing around with igraph package, I can mark the clusters, but could not figure out how to annotate each cluster. I am not planning to manually mark and annotate each cluster one-by-one. Do you know some graph visualization libraries, softwares that can do this ? (as of now I have tried only igraph and Cytoscape)

Thank you.

[Giovanni M Dall'Olio]

What's the problem with Cytoscape? It should work very well to do what you are describing.

[Neilfws]

"i before e except after c." And in protein. And other words. :)

Answer 1

[SES]

Another tool to try would be gephi, which I learned about from a comment by Pierre Lindenbaum on this site. I am impressed with gephi from my limited experience with that tool, but personally, I would look no further than Cytoscape because creating the different line/arrow types, labels, colors, etc. in that figure is easy using the VizMapper editing functions (the control panel on the left side of the main window in Cytoscape).

[Sudeep]

Thanks but AFAIK, by tweaking options in VizMapper, no group annotation can be added for nodes

[SES]

One way you could create groups manually would be to just color the nodes of interest and apply labels. If you are referring to the dashed clouds in that figure that highlight groups of nodes, I'd bet they did that in something like PowerPoint or Illustrator.

[Sudeep]

> I'd bet they did that in something like PowerPoint or Illustrator.
Well I think there is a way to do that without PowerPoint or Illustrator but I haven't found it out yet.

> One way you could create groups manually would be to just color the nodes of interest and apply labels. For sure, it will work, but only when nodes are present only in one cluster, but I have some nodes that appear in multiple clusters

[Giovanni M Dall'Olio]

That figure is done with Cytoscape. If you look at the paper, they even provide a Cytoscape session file.

[zx8754]

Try Gephi, I think you can run ForceAtlas2, which will try to put together related nodes. Then you can group those nodes into clusters, colour them, then Expand, this should give you shaded background (Show Hulls option) for clustered nodes.