Weird behaviour - WGCNA hub analysis results in Cytoscape through CytoHubba plugin change every time I run it!

Question

[Raito92]

Hello everyone, I'm using Cytoscape v. 3.9.1 to visualize the results of a previously perfmored WGCNA analysis. The aim is to individuate hub genes through the plugin CytoHubba and show a sub-network of the hubs.

However, I've ran into a very weird problem...

Apparently, if I redo the analysis, following exactly the same steps, the CytoHubba output changes everytime, for NO apparent reason. Am I missing something obvious?? CytoHubba is a widely used and commonly found in literature plugin, so there must be something I'm doing the wrong way.

That's what I do:

1) Import the edge table (through Import Network) into CytoScape (previously obtained from the exportNetworkToCytoscape R function).

2) Analyze the network

3) Run the plugin CytoHubba, calculate node scores and then the top 10 Hubba nodes ranked by MCC

So... here is a subnetwork showing the relationships between hubs... and a hub list itself.

Let's close the software ...

And start an entirely new session, reload the edge table... and repeat the whole thing...

Once...

And twice! The hub list and the ranking changes every time... and apparently in a random way.

What could it be wrong? Thanks in advance for your help!

[scooter]

Greetings,

I'm not that familiar with CytoHubba and their tutorial site seems to be down. Couple of things I note, though:

1. Is there a reason you are analyzing the network first? From my reading of their paper (admittedly very old) they don't rely on any separate analyses.
2. I'm seeing equations errors in the "fromAltName" and "toAltName" columns. If these are necessary, it may explain the issue.
3. Have you checked the weights to see if they are consistent? I'm wondering if the base calculation might be right but the selection of hubs from that base calculation is getting messed up.

Finally, check in your ~/CytoscapeConfiguration/3/framework.log file to see if there are errors coming from CytoHubba that are getting caught and not reported.

If all of that fails, I would suggest send e-mail to the authors....

Sorry I couldn't be of more help!

-- scooter