computeWeigtedDiffNetwork: Find differential networks between two classes
In AEBilgrau/Bmisc: Biostatistical miscellaneous functions

Description Usage Arguments Details Value Note

This function attempts to find differential networks between two classes. First, it computes the partial correlation (either unconditioned, the smallest first order conditioned, or full conditioned) for each group. Second it tests each partial correlation for differential correlation using Fishers Z transform and the normal distribution. If wanted, the topological overlap matrix is computed from the resulting matrix of P-values.

1
2
3

computeWeigtedDiffNetwork(x, class, gene.universe, name = "datasetname",
  pcor.type = c("marginal", "partial", "fullpartial"), use.TOM = TRUE,
  rho = 0)

`x`	is the expression matrix where columns correspond to samples/subjects and rows to genes.
`class`	is the factor encoding ABC and GCB samples
`gene.universe`	character vector giving the names of the features to be used. I.e. a subset of `colnames(x)`.
`name`	A character giving the name of the dataset.
`pcor.type`	character of length one. The type of correlation to be used. Equal to one of "marginal", "partial", or "fullpartial".
`use.TOM`	logical. Should the topological overlap matrix be computed? of the P-values of the Fisher transformed tests?
`rho`	Non-negative numeric regularization parameter for lasso. rho = 0 means no regularization. Only used if pcor.type == "fullpartial"

If pcor.type is "partial" then all 1 order partial correlations are computed (i.e. the partial correlation given a third feature) and the minimal partial correlation is used.

A list of length 3:

`Z.scores`	A matrix of Z-scores for differential (partial) correlation.
`adjacency`	A matrix of 1 - P-values for the hypothesis of no differential (partial) correlation
`tom.dissimilarity`	A matrix of topological overlap dissimilarities.