Notes from the Biomass

The notion that many networks were scale free was an important discovery and relevant to all the people that studied protein interaction networks, obviously. Many people in my circle of peers, including me never liked the idea and always found it overrated.
Berend Snel and Martijn Huynen, then at the EMBL, noted that the scale does not really span the many orders of magnitude required to actually make the statement that the distribution is in fact independent of it. The most connected protein, JSN1 in yeast had 344 interactions (within a single data set), which can't compare to the millions of links one would find in the topology of the e.g. the internet.

Nevertheless, many bioinformaticians put the notion forward, and many articles included "scale free network" just because it sounded sexy.

Recently, a group around Marc Vidal published a simple, yet revealing experiment, to identify, whether we could acutally observe such properties given our incomplete and patchy sampling of protein-protein interaction networks.
They simply generate random networks, sample from them and were able to identify scale free properties in all of them.

The networks might be in fact scale free but I do not think that the notion will tell us much about biological networks, in particular as we cannot distiguish between sticky proteins that simply bind to a large number of proteins without transmitting any signal and methodological artifacts, both largely contributing to the proteins with many connections.