Approaches to sum gene expression into pathways/other gene sets and problem of negatively correlated genes

Question

[mdihcroe]

Hi!

I'm currently searching for the best approach to get a number per pathway based on expression of genes encoding metabolites. I've found, for example, gene set variation analysis which allows to calculate random walk deviation. But I wonder which approaches take into account the fact that expression of some genes in pathway can be negatively correlated and biologically that could probably mean inhibition in pathway or something like that?

(i also looked at GSEA, ASSESS, PLAGE, PROGENy, assigning min expression value, average expression value, SVD, principal components. they either do not consider such things or work on a small number of pathways)

Any comments or links to papers on methods I haven't mentioned would be helpful