Based on the reasoning that transcriptional noise arises from the stochastic relaxation of otherwise closed chromatin, Batada and Hurst argue that essential eukaryotic genes tend to cluster within regions of open chromatin because these regions are relatively free from transcriptional noise. This would minimize the possibility of random fluctuations in expression that would kill the cell and therefore provide a selective advantage to those cells whose essential genes cluster together in less noisy regions. They present bioinformatic analyses of yeast data that seem to support this idea.
I like the idea, but am not totally sold on the notion of "essentialness". Yeast essential genes are based on those which are lethal when mutated in yeast grown under the warm and loving conditions of laboratory culture. However, the yeast genome did not evolve in the lab, it evolved in much harsher and variable natural environments. The the real set of essential genes is likely to be much larger than defined here, in fact it may include the whole genome as it is likely that each gene was essential at some point in yeast evolution. This said, even in natural environments, there maybe a spectrum of gene essential-ness, with some genes essential over a broader range of conditions, or longer evolutionary time-span than others. Thus the lab essential genes may actually reflect the core set of "most essential genes". But can we take this for granted?