Department of Chemistry, Machine VP
Faculty of Science
Co-Investigators Matthew Hoyles, Serdar Kuyucak and Andrew Slater
Department of Theoretical Physics, Research School of Physical Sciences and Engineering and Department of Chemistry, Faculty of Science
Permeation of Ions Through Membrane Channels--Molecular Dynamics Studies
Biological ion channels are essential for cell activity and communication. They are the active component of excitable membranes, which in turn allow nerve, muscle and sensory cells to work.
Although the function of channels is largely understood, much less is known about their structure and mechanism of operation. Current experimental techniques cannot investigate these directly. Molecular biology can change the primary structure of the protein, but the effects on its shape are unknown. X-ray crystallography could reveal the structure of a channel if it could be crystallized, but no one has been able to do this.
The objective of the project is to investigate the conductivity of biological ion channels and bulk electrolyte solutions under physiological conditions, using methods of computer simulation. This would allow models of channels to be tested, as the conductivity of single channels can be measured by the patch-clamp technique. The investigation of bulk electrolyte solutions is necessary to provide a control on the accuracy of the simulation.
What are the basic questions addressed?
The basic question addressed is how does the shape and charge distribution of a biological ion channel affect its conductivity and selectivity. An answer to this question would allow theoretical models of channels to make predictions which could be tested against experiment, in turn leading to a better understanding of channel structure and mechanism.
A specific example is the question of the internal diameter and length of the neck region of a channel. It is often suggested that the neck is long and narrow, by analogy with the gramicidin channel, which is known to have a diameter of 4 angstroms and a length of 26 angstroms, narrow enough that ions and water must pass through in single file. However gramicidin, being an antibiotic which kills bacteria by forming uncontrolled ion channels across their membranes, is not a typical biological ion channel, and has a very different overall structure to the more complicated channels which control the flow of ions across cell membranes. The necks of ordinary channels may be shorter and wider than gramicidin: this is the sort of question this project could help answer.
What are the results to date and the future of the work?
We have simulated pure water and obtained results for the autodiffusion of water in reasonable agreement with experiment and other molecular dynamics simulations, both in absolute values and temperature dependence. We intend to simulate water and a single ion to obtain values for the diffusion and conductivity of bulk electrolyte solutions. This will require longer runs than for the autodiffusion of water.
We are developing a program to simulate a whole channel using Brownian dynamics. This means not simulating the water explicitly, instead representing it by random forces on the ions. This method should be fast enough to give information about conductivity in channels, but is too crude to accurately represent short range interactions in the neck region of the channel. Our aim is to use molecular dynamics to simulate the neck region, in combination with Brownian dynamics simulation of the whole channel.
What computational techniques are used and why is a supercomputer required?
Our programs are designed to take advantage of the vector architecture of the VP. The molecular dynamics program runs with a vectorization of over sixty percent. They are written in ANSI C and can be compiled and run on workstations. While groups of workstations provide us with extra power for production runs, we use the VP for development and testing, as well as for storage (via silo) and analysis of results.
Whether simulating the small neck region using molecular dynamics, or the whole channel using Brownian dynamics, we will need very long simulation times to investigate conductivity. This will require access either to a supercomputer or a large network of workstations.