Prediction of membrane protein structures by replica-exchange Monte Carlo simulations: case of two helices

We test our prediction method of membrane protein structures with glycophorin A transmembrane dimer and analyze the predicted structures in detail. Our method consists of two parts. In the first part, we obtain the amino-acid sequences of the transmembrane helix regions from one of existing WWW servers and use them as an input for the second part of our method. In the second part, we perform a replica-exchange Monte Carlo simulation of these transmembrane helices with some constraints that indirectly represent surrounding lipid and water effects and identify the predicted structure as the global-minimum-energy state. The structure obtained in the case for the dielectric constant epsilon=1.0 is very close to that from the nuclear magnetic resonance experiments, while that for epsilon=4.0 is more packed than the native one. Our results imply that the helix-helix interaction is the main driving force for the native structure formation and that the stability of the native structure is determined by the balance of the electrostatic term, van der Waals term, and torsion term, and the contribution of electrostatic energy is indeed important for correct predictions. The inclusion of atomistic details of side chains is essential for estimating this balance accurately because helices are tightly packed.