Simple physical model for protein–protein interaction networks

Eric Deeds et al. examine the statistical properties of protein– protein interaction networks and propose a physical model based on the fundamental principle of desolvation interactions between proteins. The authors examined the network of protein–protein interactions reported by the yeast two-hybrid system. They demonstrated that two published independent measurements of these interactions produced graphs that are not strongly correlated with each other, despite having similar scale-free topologies. Hypothesizing that these interactions are actually dominated by nonspecific interactions between proteins, Deeds et al. showed that their purely physical model based on desolvation interactions, when combined with an elemental source of experimental noise, can reproduce both the scale-free nature of previously reported graphs and also many higher-order correlations in the networks. Additionally, significant correlation between the surface hydrophobicity of a protein and its number of interacting partners was observed, which agrees with and supports the physical model. The results suggest that a large number of observations in these and other biological networks might contain considerable influence from nonspecific interactions. — R.N.