With this kind of thing, the devil is always in the details. He's looking for genes whose expression levels correlate with those of genes known to be involved in the process of interest. This is not a new idea. The trouble is that spurious correlations can arise in all kinds of surprising ways, and swamp the interesting ones.
