A novel method to calculate transition pathways between two known protein conformations is presented. It is based on a molecular dynamics simulation starting from one conformational state as initial structure and using the other for a directing constraint. The method is exemplified with the T ++ R transition of insulin. The most striking difference between these conformational states is that in T the 8 N-terminal residues of the B chain are arranged as an extended strand whereas in R they are forming a helix. Both the transition from T to R and from R to T were simulated. The method proves capable of finding a continuous pathway for each direction which are moderately different. The refolding processes are illustrated by a series of transient structures and pairs of a, t angles selected from the time course of the nimutations. In the T + R direction the helix is formed in the tast third of the transition, while in the R + T direction it is preserved during more than half of the simutation period. The results are discussed in comparison with those of an atternative method recently apptied to the T -. R transition of insulin which is based on targeted energy minimisation.

No abstract

A 0.76 ns molecular dynamics simulation has been performed on a self-assembled peptide nanotube in water. The peptide structure is composed of 10 /3-sheet-like hydrogen-bonded stacks of the flat ring-shaped cyclic d,l octapeptide subunit cyclo[-(Gln-D-Ala-Glu-D-Ala-)4]. The synthesis and self-assembly of such nanotubes and their ability to form remarkably transport-efficient transmembrane ion channels and pore structures have been reported previously. During the simulation, the tubular construct retains its structural integrity with only slight distortion at the outer ends. Water molecules in the tube tend to be organized in alternating zones of one water near the plane of the backbone Ca atoms and two waters near the plane midway between two adjacent peptide subunits-the idealized 1-2 water structure. On average, 32.8 water molecules are found in the tube during the data-gathering phase of the simulation. Analysis of the motion of water molecules in the tube gives a diffusion constant of 0.44 x 10-5 cm2 s-', which is approximately one-sixth of the self-diffusion constant of bulk water and much faster than either molecular dynamics or experimental measurements of the diffusion constant of water in the structurally related gramicidin A transmembrane channel. A detailed examination of the tube-water trajectories shows that water molecules can pass by one another in contrast to the single-file diffusion occurring in gramicidin A. We suggest that water diffusion can be understood as a series of "hops" between zones which can cause transient local deviations from the ideal number of water molecules in a zone. A slight excess of the average water population over the "ideal" number of water molecules expected from the alternating 1 -2 structure suggests that diffusion in the tubes may be enhanced in a manner roughly analogous to the enhanced conductivity of n-type semiconductors due to doping.