Wgcna: Exporting Network To Cytoscape For Blockwise Module Construction

Question

[pixie@bioinfo]

Hello, I performed wgcna blockwise as I had 18K probe ids. I was following the tutorial for exporting it to Cytoscape:Exporting to external software Here, this particular command is taking a lot of time to run:

TOM = TOMsimilarityFromExpr(datExpr,power= 8);

If I don't run this command, I am getting an error for this command: modTOM = TOM[inModule , inModule];

Is there a way where I can modify the TOM for the calculations in blockWise ?

[laura.eierman]

Sanchari, I am wondering if you ever came up with a solution as I am currently running into the same exact problem. Thank you!

[pixie@bioinfo]

Hi Laura, sorry for the late reply. I had mailed to one of the authors who replied " if the full TOM is too big to fit into memory, you can use the function vectorTOM to calculate the topological overlap of a smaller number of genes while taking into account the entire set of genes."

However, I didn't have to use this function, I left the code "TOM = TOMsimilarityFromExpr(datExpr,power= 8)" running all night ..and it was done in the morning.

Answer 1

[Sudeep]

The issue here is that you have 18k probes, and that's why

TOM = TOMsimilarityFromExpr(datExpr,power= 8)

is taking a long time to run. You are getting an error for

modTOM = TOM[inModule , inModule] because if you are trying to subset a non existing TOM matrix. I would suggest to look into trimming down your 18k probes based on some threshold. In WGCNA you can use softConnectivity or softConnectivity.fromSimilarity to calculate node connectivity from expression matrix and use an arbitrary threshold to prune probes

[pixie@bioinfo]

Thanks for the input. But since Zhang and Horvath had made provisions for a much larger dataset (using the blockwise module given in the 3rd tutorial in the WGCNA tutorial page), I was hoping there is a way around to modify the TOMs according the blocks that we get and someone might have tried that.

Answer 2

[jleandroj]

Hello,

TOM is a step used to minimize effects of noise and spurious associations. This step is necessary to create Step-by-step network construction. If you want to avoid this step you can develop an Automatic network construction, therefore, you will get a network quickly but There will be noise (It is not recommended).