Differential Network analysis
1
0
Entering edit mode
7 months ago
hannicho ▴ 20

Hello;

I have a questions regarding differential network analysis. I have created my two separate networks using wgcna. Since the networks are different (have different samples); they also have different soft threshold beta power to fit scale free. I saw that on previous studies; people kept the power the same for two networks as long as they fit scale free. However; I can't do that because these networks do not fit scale free at the same powers. Can I still compare the networks for different centrality measures based on their soft threshold power. I want to be sure that the difference in topology I am seeing is not because of the power but the underlying condition. Any inputs appreciated.

Thanks!

wgcna scalefree networks scale free • 356 views
ADD COMMENT
1
Entering edit mode
ADD REPLY
1
Entering edit mode

Hello Kevin. I can remove one of the posts if needed. I just needed to maximize the chance of it being seen. I hope that is not a problem.

ADD REPLY
0
Entering edit mode

No problem. Just making sure that users on both websites can see all answers provided

ADD REPLY
3
Entering edit mode
7 months ago
eggerj ▴ 40

This is an interesting problem. I've been lucky in the past and have been able to use the same beta for each network where they all reach a similar approximately scale-free fit (>= 0.9) and I've seen others as well just use the same beta if the difference in fit aren't too drastic (say 0.82 and 0.87 for example). Keep in mind, soft-thresholding is more about reducing noisy correlations than trying to force a scale-free network. WGCNA just borrows the assumption of scale-freeness because the threshold to use is unknown.

Also, because you're using WGCNA, I assume you plan to focus the network analysis (or differential network analysis) on a module-level. If that's the case, you're main goal is to identify consensus modules and characterize differences in topology in the module networks. The soft-thresholding step shouldn't influence differences in module topology, but might influence the module identification step after computing a consensus TO matrix. To what extent, I'm not sure. I'd suggest trying it both ways and seeing how different the module assignments are between the two approaches. If the differences aren't too drastic, you could probably make a reasonable justification for either approach.

Otherwise, reaching out to the WGCNA authors is always worth a shot as well. They may have a good suggestion if you have luck reaching them.
'

ADD COMMENT

Login before adding your answer.

Traffic: 941 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6