As we have said, ion channels and membrane proteins in general are poorly understood in comparion with globular proteins. Our ignorance of the structure of the channels follows from the fact that the methods of X-ray diffraction and NMR, which have provided molecular structure at atomic resolution for many hundreds of proteins, do not work well for membrane proteins; at the time of writing only about 10 membrane proteins have known structures, even though gene sequencing suggests that about 30% of the genes of a typical organism encode membrane proteins. For a typical channel protein, then, only the amino acid sequence and indirect experimental evidence about its conformation are available, combined in some cases with crude X-ray data. One might hope that improving experimental techniques will eventually permit the detailed elucidation of the structure of the channels (though this may not happen for many years). However, this still leaves the function of the channels unexplained, and there are many aspects of this that are not experimentally accessible: for example, it is hard to gain more than a qualitative understanding through experiment of the ionic selectivity exhibited by most channels, or of the various components making up the forces that act upon an ion at each stage of its passage across the membrane. An improved understanding of ion channels would be expected to have medical applications (even aside from the problem's intrinsic interest): for example, the control of their behaviour is central to anaesthesia and the pharmaceutical treatment of mental illness, while defects in ion channels are known to be involved in certain diseases such as cystic fibrosis.
Computer simulation is widely used in modern molecular biology but is a particularly useful tool in the study of ion channels where detailed structural information is unavailable. In tackling the structure problem, it provides perhaps the best way of distinguishing between and refining possible strutures suggested by the indirect evidence. Moreover, computer simulation can easily give just that information on the details of the interactions involved in gating or ion passage that is experimentally inaccessible. There are, of course, difficulties in applying the methods. For example, the potentials used to model the interactions between the various atoms can be made to reproduce their real behaviour only approximately, and, (more significantly), the time scales accessible to the simulation are at most a few nanoseconds while processes such as the passage of an ion take about three orders of magnitude longer.
Moving from the general to the particular, let us concentrate again on the nicotinic acetylcholine receptor. The structural data available for this channel, from Nigel Unwin's electron diffraction studies on nAChRs from the electric organ of the Torpedo electric ray, is the best for any ligand-gated channel. Previous work in this group, guided by this data, provided a plausible atomic-scale model for the part of the molecule that lines the channel in both its open and closed configurations. However, this model was obtained without consideration of the water molecules that fill the lumen of the channel. The first part of my work (reported here) has thus been to examine and improve our proposed structures by explicit incorporation of water molecules into the channel.