[R] wgcna choice for softpower by scale indepence when combining to datasets

[Peter Langfelder]

On Wed, Aug 1, 2012 at 6:30 AM, Ingezz <irholtman at gmail.com> wrote:
> Dear Peter,
>
> I have another question about WGCNA. I am using the package for
> meta-analysis to find modules preserved in several datasets. However, I am
> unsure how to handle the softpower, because each dataset has its own ideal
> scale indepence value. When combining several datasets what should I do?
> - pick the lowest scale indepence value and use this for all datasets?
> - calculate an average scale indepence value of the datasets, and use this
> one?
> - use different scale indepence values for different datasets, all though
> combining them later on?
> - or something else, which I havent thought of?

Hi Inge,

I'm not sure what you mean by "modules preserved in several datasets".
Are you calculating consensus modules? If so, I would choose
soft-thesholding powers that (a) give approximate scale-free topology
in each data set, and (b) give roughly comparable mean or median
connectivities across the data sets. You may choose a different power
for each data set. However, it is also fine to choose the same power
(such that the network topology is approximately scale-free in each
set) since the consensus module calculation includes a step in which
the input networks are roughly calibrated to make them comparable.

If you are calculating module preservation (function
modulePreservation), the function chooses the standard power by
network type since for module preservation the consistency of
soft-thresholding powers is more important.

Best,

Peter