Hopf, T.A., et al. Cell advance online publication (10 May 2012).

About one-quarter of all our proteins protrude through a membrane. This makes them important communicators both in cellular compartments and outside the cell, but it also makes their structure more difficult to determine, which in turn makes their precise mode of action harder to understand. Several experimental and computational structure prediction tools have been used to estimate the three-dimensional structure of membrane proteins, but these tools are limited by the number of transmembrane helices they can incorporate. Hopf et al. used maximum-entropy analysis to calculate amino-acid changes in protein families across their evolution, which allowed them to determine the proximity of residues in three dimensions and their fold. The authors' EVfold_membrane algorithm predicts de novo structures of transmembrane proteins. When they benchmarked the tool against known crystal structures, the researchers found very high agreement, particularly in functionally important regions and substrate-binding pockets.