Interesting but at first glance it looks a bit paradoxical that they argue for existence of essential modules and at the same time argue against the idea of essential protein interactions. If essential modules are enriched in essential proteins then this favors the idea that proteins are essential because they participate in essential interactions. I have to have a better look at it. - Pedro Beltrao

Pedro, what they find is that the "essential hubs" are in fact members of highly connected modules that contain many essential proteins. These modules are well known protein complexes such as the ribosome and the proteoasome. So the whole story about hubs being essential may simply due to large protein complexes typically being essential. - Lars Juhl Jensen

BTW, I just wrote a short blog post about it (http://larsjuhljensen.wordpress.com/2008...) - Lars Juhl Jensen

Right .. and as you say in the blog post it is not that surprising. I don't think many people in biology ever bought the idea that hubs are more likely essential because they somehow keep the network structure intact ... there is no functional reasoning there. However, at the same time these authors argue (at from what I get at first glance) that hubs are not essential because they are more likely to have essential interactions. This appears to contradict a bit their main finding. - Pedro Beltrao

The "interaction theory" that they disprove is as follows: a certain fraction of protein interactions are essential, hence a protein with more interactions is inherently more likely to be essential. I agree with the authors that this is not the reason for the highly connected proteins being essential. - Lars Juhl Jensen

Agreement here to, but wasn't that all too obvious to begin with? - Roland Krause

There is an alternative hypothesis one could be tested: Essential genes tend to code for abundant proteins (in the TAP data at least), which are common contaminants with a high degree (in the Gavin data, the contaminant regime begins at k > 50). See also http://mbe.oxfordjournals.org/cgi... for an approach to study such features. - Roland Krause

Yes, you might say that their explanation is kind of obvious, but I don't think that it has actually been stated in the literature before (correct me if I am wrong). Considering the amount of attention that the importance of hubs has attracted, I think it is good that someone finally points out that the correlation between connectivity and essentiality may have a completely trivial cause. - Lars Juhl Jensen

Roland, protein abundance likely has a role in it as well, perhaps especially in the TAP data. What find interesting, though, is that the authors find more or less the same set of GO terms no matter which set of interactions they look at: TAP-MS, DIP core, the literature curated set from Mike Tyers' lab, or the Bayesian data integration from Gerstein. To me, this suggests that the signal likely has a biological basis (however trivial). - Lars Juhl Jensen

The point that it is trivial is not easy to make in science nowadays but the altocumulus discussion already refutes much of the hub idea, not only the party/date hub distinction. http://biology.plosjournals.org/perlser... - Roland Krause

It may be that we considered it bit obvious, but it just got a "Research Highlight" commentary in Nature Reviews Genetics (http://dx.doi.org/10...) - Lars Juhl Jensen