A simple method for modeling transmembrane helix oligomers.

We describe an effective procedure for modeling the structures of simple transmembrane helix homo-oligomers. The method differs from many previous approaches in that the only structural constraint we use to help select the correct model is the oligomerization state of the protein. The method involves the following steps: (1) perform 100-250 independent Monte Carlo energy minimizations of helix pairs to produce a large collection of well-packed structures; (2) filter the minimized structures to find those that are consistent with the expected symmetry of the oligomer; (3) cluster the structures that pass the symmetry filter; and (4) select a representative of the most populous cluster as the final prediction. We applied the method to the transmembrane helices of five proteins and compare our results to the available experimental data. Our predictions of glycophorin A, neu, the M2 channel and phospholamban resulted in a single model for each protein that agreed with the experimental results. In the case of erbB-2, however, we obtained three structurally distinct clusters of approximately equal sizes, so it was not possible to identify a clearly favored structure. This may reflect a real heterogeneity of packing modes for erbB-2, which is known to interact with different receptor subunits. Our method should be useful for obtaining structural models of transmembrane domains, improving our understanding of structure/function relationships for particular membrane proteins.