Greetings all, I'm trying to classify microarray data I collected from the GEO repository. It's fairly heterogenous, comprising different treatments and different cancer cell lines. But the feature they all share is that all treatments target one cellular pathway in cancer. What I've done so far is calculate the fold change of every gene for every treatment relative to its control. So now I ended up with a merged matrix where the rows represent the genes, the columns the various samples, and the cell bulk all the fold changes (in log 2). So now I'm only supposed to apply consensus partitioning to get the sample-classes. But my question is: should the fold changes coming from different sources be normalized or pre-processed in any way? And am I allowed to use them as they are? And will it be any different if I used the fold change without log 2 applied to it?