Question: Correlate WGCNA modules from 2 separate analyses (small and large RNA)
0
gravatar for sebastian.lobentanzer
7 months ago by
sebastian.lobentanzer10 wrote:

Hello everybody, this is a general question based on a practical problem: can something meaningful be derived from any kind of correlation between two separate WGCNA analyses of the same data (samples)?

To illustrate: we have sequenced small and long RNA of patients and controls (separately from the same blood samples) and performed WGCNA analyses on each of those sets. As a supplement to targeting- and network-level analyses, we were wondering if we can get a meaningful statement from some module specific correlate between the small and long RNA WGCNA modules, e.g., can one compare eigengene vectors (or even more principal components) from the two sets, and which statistical and ranking processes would be appropriate in those cases?

Thanks in advance, Sebastian

wgcna • 175 views
ADD COMMENTlink modified 7 months ago by Peter Langfelder2.2k • written 7 months ago by sebastian.lobentanzer10

At most you can say (if they are on the same samples) that a module on small RNA behaves similar or opposed as a module in long RNA

ADD REPLYlink written 7 months ago by LluĂ­s Revilla Sancho510
Answer: Correlate WGCNA modules from 2 separate analyses (small and large RNA)
2
gravatar for Peter Langfelder
7 months ago by
United States
Peter Langfelder2.2k wrote:

Yes, you can correlate module eigengenes from different analyses (across common samples), just as you could correlate the expression levels of single short and long mRNAs. I would not compare subleading principal components without a good rationale to believe that they represent something meaningful and reliable since the subleading components tend to be much less stable than the 1st PC (eigengene).

Standard correlation tests should be applicable. You can always rank directly by the value (or absolute value) of the correlation; for Pearson and biweight mid-correlation, standard Student t-test significance is usually appropriate, with the standard disclaimer that the test assumes normally distributed errors.

ADD COMMENTlink modified 7 months ago • written 7 months ago by Peter Langfelder2.2k

Thank you Peter, it worked perfectly! Your package is a work of art.

ADD REPLYlink written 7 months ago by sebastian.lobentanzer10
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 395 users visited in the last hour