Computer simulation has become an important tool for the study of biomolecular systems. This thesis deals with molecular dynamics simulations of one-component lipid bilayers, which may serve as models for biological membranes.
The main scientific contributions are:
• It is possible to analyze the electrostatic contribution to the surface tension at a lipid-water interface in terms of dipole-dipole interactions between lipid headgroup shielded by a dielectric medium (water). The interaction can be divided into two parts. The in-plane components of the dipoles give rise to a positive, i.e. contractive
contribution to the surface tension, albeit rather short ranged due to them being fluctuating dipoles. The normal components give rise to a negative, i.e. expansive contribution that will dominate the interaction at large distances.
• Simulated membrane areas are extremely sensitive to details, especially the treatment of long-range electrostatic interactions. When cut-offs are used for the electrostatics, the exact definition of charge groups play an important role. Furthermore, using Ewald summation for the long-range interactions seems to have an overall stabilizing effect, and the area becomes less sensitive to other factors, such as system size and hydration.
• Using atomistic simulations it is possible to study formation and evolution of a hydrophilic trans-membrane pore in detail. Free energy of pore nucleation and expansion can be calculated using potentials of mean constraint force. The resulting free energy profile shows no local maximum between the intact and pre-pore states, contrary to what is suggested by experiments.
• The present force field reproduces even the slowest dynamics in the lipid chains, as reflected in NMR relaxation rates. Furthermore, since the simulated system was relatively small, the experimentally observed variation of relaxation rates with Larmor frequency cannot be explained by large scale collective dynamics, or it would not have shown up in the simulation.
• Lipid lateral diffusion can be studied in detail on all relevant time scales by molecular dynamics. Using simple assumptions, the different diffusion coefficients measured on short and long times respectively can be connected in an analytic expression that fit calculated mean square displacements on timescales ranging from picoseconds to hundreds of nanoseconds.
Stockholm: KTH , 2006. , vi, 61 p.
Marrink, Siewert-Jan, Professor Dr.