Proteins and nucleic acids, like cells, can only survive in ionic solutions. Every biochemist knows that proteins must be
put into "buffers"  of (typically) 150 mM KCl if they are to survive. Every molecular biologist knows that the properties
of DNA depends on the salt and salt concentration. Every cell biologist knows that ions are important for cell function.
Every physiologist and physician that life does not survive in distailled water.

Ion in water are indeed the solutions of life.  Sfimulations and theory in biology are more useful
if they deal with the central experimental realities. Thus, they should deal with the real properites
of the ionic solutions needed to maintain life. Sadly, this is not always easy and thus not always the case.

Ionic solutions are almost never ideal in the theermodynamic sense. The fundamental thermodynamic variable 
of ionic solutions is the free energy per mole, which has a role equivalent to that of height in gravity, or electrical
potential in electricity, or concentration in biochemical reactions. In an ideal solution the free energy per mole
(also called 'the activity': note the activity is an experimental parameter that can be measured in many different
ways giving quantitatively the same result to several significant figures) follows a simple law proportional to 
the logarithm of the concentration. In ionic solutions, this law is (essentially) never followed because of shielding:
ions can move easily in solutions and rearrange in an ionic atmosphere around a charge (i.e., around any ion 
in the solution) to balance the central charge. After a few nsec, the ionic atmosphere balances the central charge
'perfectly' and the complex of ion and its atmosphere is uncharged. Shielding phenomena produce a square root 
dependence in properties of ionic solutions that make them nonideal, as has been understood for a very long time.
Thus, any treatment of ionic solutions in biological systems must get shielding right.

Ionic solutions in biological concentrations are not well characterized, however, by simple theories of shielding.
Debye Huckel for example fails at low concentrations (mM) of many salts and does poorly even with NaCl, clearly
missing the main properties of a graph of activity vs. concentration. Such graphs describe the fundamental properties
of ionic solutions as importantly as a graph of gravitational potential energy vs. height, or electrical potential energy 
vs. charge. In my view, a theory or simulation must get the plot of free energy per mole vs concentration roughly right
if it is to have any hope of dealing with proteins and nucleic acids quantiatively, just as any theory must get
the osmolarity of solutions right if it is to have any hope of dealing with the volume of cells.

The activity of ionic solutions depends on the chemical nature of the ions in the solution, even if the ions do not share
electrons with water, and in that sense are quantum chemically inert. Na and K for example have different diameters
and so have quite different free energy per mole in concentrated solutions even though they are hard spheres
and do not make covalent bonds of any time with water. The chemically specific properties of many ionic solutions 
can be understood to a good approximation by models that simply treat the water implicitly as a dielectric (with friction if
nonequilibrium properties are involved) with a dielectric constant that depends on the content of the solution, as 
determined by measurements of the dielectric properties of solutions. This primitive model of ionic solutions does
surprisingly well. In my view, any more sophisticated model or simulation of ionic solutions must do at least as well
in dealing with ions if it is to replace the primitive model. Explicit models are better than implicit models in computations
of protein function only if they are (nearly) as accurate.

The properties of concentrated salt solutions may at first seem to be unimportant in most biological contexts, since
the solutions bathing cells---the plasmas of life---are typically 150 mM in concentration. However, most of the 
functions of biological systems are determined by tiny parts of proteins, the active sites of enzymes, the binding
sites of proteins, the selectivity filters of ion channels. These special sites typically are made of ionized acidic and
basic side chains of proteins, and these charged moities are at high number density ('concentration'). Biological 
systems are not able to support electrical potentials of more than a few hundred millivots, and so there are large
number densities of mobile ions in active sites of enzymes, proteins, and channels. In favorable cases these ions
can be resolved in structures determined by x ray crystallography. In all cases, electrical neutrality is likely to be
obeyed, since it is a fundamental property of the electric field, and ion concentrations  (i.e., number densities) 
of many molar are typical.

At these concentrations, ionic solutions are non ideal and indeed may prove to be more akin to ionic liquids of
mobile anions, cations, and side chains of the protein, than
to the ideal solutions implicitly assumed in biochemistry textbooks. 

Nonideal solutions have complex properties and it is my view that these must be computed with reasonable 
fidelity if simulations or theories are to deal with the properties of enzymes, proteins, channels, and nucleic
acids that are measured in the laboratory.

Specifically, the free energy of a particular type of ion depends on the concentration and type of every other 
ion in a nonideal solution, and this dependence is large and important at high concentrations. This dependence
on the type of ion provides an easy way to tell how realistic a simulation of ions is needed in a particular Biological
situation. If a biological system is found experimentally to change properties as the concentration of salt is
changed, simulations or theories of that system must compute that dependence, in particular they must deal with shielding
with reasonable fidelity. If a biological system is found to behave differently when NaCl is replaced with KCl for example,
simualtions or theories of that system must compute the free energy per mole of NaCl and KCl solutions reasonably
well.

These phenomena are important in chemical problems too. If for example buffers like EDTA are used to control 
the concentration of calcium, the local concentration of calcium near the buffer molecule is very large and nonideal
effects are to be expected. The activity of calcium is expected to be different if the background concentration of
(for example) KCl is changed or if NaCl replaces KCl. The nonideal properties of the ionoc solutions near the EDTA guarantees
chemical specificity even if covalent bonds are not invovled at all. Failure to deal with these phenomena in standard
treatments of calcium buffering can lead to strange experimental interpretations (XXX) of some importance.

Thus, I write to advocate the calibration of simualtions and theories of proteins and nucleic acids against 
measurements of the real properties of concentrated salt solutions.

In on case, a model developed in this spirit works surprisingly well, even though in fact the model gets the 
salt dependence in bulk only approximately right. A model of selectivityh phenomena in the calcium channel of 
the heart (L type channel) deals successfully with essentially all the experimental properties of the channel,
in many types of solutions over a wide range of conditions, although the model has only two parameters that
are never changed as conditions or solutions are changed. The model uses crystal radii of ions, but represents
the channel protein in a skeletal way, in a highly reduced model, in which side chains are represented as spheres
free to move within the active site of the channel but not able to leave it. This model has binding sites for ions
formed by its side chains. The location of the side chains is different in each experimental condition. Thus this
model has a self-organized structure. The structure of the binding site is induced by the ions, and the forces
in the model. There is no preformed binding site. The binding sites are calculated outputs of the simulations.
The only forces in the model are electrostatic and steric repulstion. These forces do quite well in describing
ionic solutions. They are all that is used or necessary in this model of the L type calcium channel. 

The L type calcium channel can be mutated into a very different type of channel, the sodium selective channel
of nerve studied so wonderfully by Hodgkin and Huxley en masse and then one molecule at a time by Bezanilla
et al, for example.  Simply changing side chains from EEEE to DEKA changes the selectivity of the channel. Amazingly,
the same thing happens in simulations. Simply changing the side chains WITHOUT CHANGING ANY parameters
of the model changes a model of a calcium channel into a model of a sodium channel. Indeed, the resulting
model of the sodium channel has the Na vs K selectivity so important to the biological function of the real channel,
and that arises not in binding sites but in a depletion zone which does not contain K at all. 

Thus, the biologically important properties of two important types of channels can be described by a model 
containing only two parameters without changing the values of those parameters, if a primitive representation
of the protein is used akin to the primitive model of ionic solutions.

A similar treatment is even more successful in dealing with the properties of the ryanodine receptor in some
120 combinations of solutions and mutations.

It is obvious that descriptions of channels as primitive as those used to compute selectivity will be improved
if they are made more realistic by including real structures. But the improvement will only occur I believe
if sophisticated  computations of the more realistic structures includes realistic computations of the properites
of the ions in nonideal conditions. If sophisticated simulations cannot reproduce the properties of the ions computed
in primitive simulations, they are unlikely to reproduce the resuls of the primitive simulations. And the primitive
simulations get the biology (mostly) right.

More realistic simulations of proteins, nucleic acids, and channels could have a profound effect on our lives,
as realistic simulations of the semiconductors has on our technology. These require realistic simulations of
nonideal properties of concentrated salt solutions in my opinion. The complex properties of life arise in such 
nonideal systems and our simulations must deal with this complexity if ithey are  to be successful.