However this is not just a useful depiction of an apposite well-supported statistical model. If we are prepared to allow that the process is driven by a CRG and that the MAP model that we have discovered is indeed generating the idle process, then identifying the disconnected components of the system allows us to immediately make assertions about the impact of various controls we might apply to this regulatory process – just as we can were we to believe the model was a causal extension of a BN. In the context of microarrays, the objective of clustering is to identify patterns among the data and decide which genes to focus on in further, more gene-specific, experiments. It is therefore necessary for the scientist to make such causal conjectures about the effect of controls available to her on the expressions reflecting the underlying regulatory process she studies. These conjectures can be universal or nuanced by evoking ideas of parsimony.
