I have a set of RNA-seq data for 2000 samples with multiple obesity-trait indicators (e.g., BMI, Waist-to-hip ratio, etc). I know WGCNA can handle numeric, ordinal, or binary traits. I wondering if I could include all these obesity-trait indicators into the WGCNA analysis all at once or not, because no single indicator is perfect. In addition, I want to keep my indicators as continuous variables to avoid information loss.
So, my questions is: Can WGCNA handle multiple traits/indicators at once in an analysis? Or, do you recommend that I run each indicators in a separate WGCNA?
Another questions: these indicators are somewhat dependent. I am wondering if I can still do the consensus WGCNA to find common modules across these indicators if I run each indicators separately or not.
Thank you in advance!
It can handle a technically infinite number of traits. With the trait metadata, what we usually do is correlate / regress these against the module 'eigengenes' in order to infer which module(s) is(are) associated with which trait(s).
That to which you may be referring is segregating your analysis based on these traits, i.e., segregating the dataset into BMI < 21, BMI 21-30, BMI >30, and then generating separate networks based on these. I see no issue doing this, either. I am not sure that the consensus approach makes too much sense here - try to think it out yourself about what it would mean a consensus network approach in this case.
Kevin
Kevin, thank you very much!
I need to think more about these and your helpful suggestions.
Log in to answer this question.
