Rapid Prediction of Lipid/Protein Interactions by Molecular Dynamics Simulation

Membrane proteins (MP) are a class of biomolecules responsible for important biological processes in living organisms; malfunction of MPs triggers pathological development in the human body. Since the properties of MPs are coupled to membranes, efforts have been devoted to study the interaction of MPs with membranes to reveal the solvation, fold, and function of MPs in the native/membrane environment. However, the chemically-heterogenous biological membrane usually hampers the application of biophysical techniques. This is particularly true for computer simulation approaches such as molecular dynamics (MD) simulations, where a large amount of computations are required to derive meaningful information at the atomic level. Here, I present an MD protocol to accurately predict the interaction of peptides and proteins with membranes in short computation times. The protocol is comprised of a physically-reasonable yet computationally-efficient model representation of membranes by using an octane membrane mimetic and efficient conformational sampling based on an ensemble of MD simulations. This approach is shown to predict the solvation, orientation, and self-assembly of helical peptides into bundles, peptide-mediated pore formation, as well as protein-protein interactions in membranes. I first validate the protocol by predicting the interaction of an amphipathic peptide with membranes. This work leads to identifying four orientation states of melittin in the membrane and spontaneous peptide-mediated pore formation on a short time scale, demonstrating the accuracy and efficiency of the protocol. I then use the method to predict the structure of α-helical MPs based on the spontaneous self-assembly of transmembrane (TM) helices in membranes. The structure and aggregation number of a homotetramer, the M2 channel, are correctly predicted. Simulations containing four M2 and two glycophorin A helices show that the approach captures the selectivity of helix-helix association . I finally apply the protocol to predict the structures of two homo-dimers, Rcf1 and Rcf2, in membranes. Results demonstrate unique solvation patterns suggesting that these two proteins adopt the monomeric form when bound to membranes. The aforementioned applications demonstrate the power and domain of application of the method for predicting lipid/protein interactions.Ph.D