I am using public expression data to create weighted co-expression networks in WGCNA. One source I was hoping to use (COXPRESdb) has aggregated expression data in the form of Pearson correlation coefficients calculated for all pairwise comparisons of gene-expression values. Would this matrix be valid input into the WGCNA pipeline?
Specifically, I'm not sure if I could use this matrix to choose a soft thresholding power as outlined in the manual, since it is assumed the starting point is a matrix of gene expression across samples. I also don't know whether the adjacency() and TOMsimilarity() functions should be used on a matrix of gene-pair correlation values instead of the sample matrix. Does it make sense to use this kind of processed data in WGCNA?