In principle you can do some MD to your system in explicit solvent, but with caveats:

I would go for this protocol:

1. strip off ions from your system

2. divalent ions are still difficult to treat so I would advice to consider K or Na ions (with relevant counter-ions) and randomize their position after the building of your system.

3. run several independent equilibration runs and see where K or Na ions go.

4. Replace the conserved ones with Ca2+ and see if they last in that position during a simulation

Yes you can do it

I feel stability can also to be determined in terms of energy changes.

More over what are the conformational parameters you are going to measure the changes in pres/abs of calcium?

Hi Atul,

A part of your question is "whether a particular PDB structure has any calcium binding pockets or residues. It is already bound to a calcium ion in PDB"

If its already bound to Ca in the PDB - why do you want to find if your protein of interest has a site for Ca or not?

Moving past that - " so I want to simulate it at a particular conc. of calcium" you dont have to do this. If the PDB structure shows the protein bound to Ca ion, than that site is stable and accepts the Ca ion very well. You dont need to simulate concentration to see if an ion from the surrounding comes in and binds here - an easier way would be to grab this PDB (protein with the bound ion) - and run your simulation on it in explicit solvent - if your structure is stable than over the length of the simulation the Ca should not be seen diffusing outwards but will be held in place.

" I will be looking for stability, sec. structure change and conformation change in the presence and absence of calcium."

These are important questions. If you protein binds to Ca, than it plays a part in a signaling cascade (maybe), in which case there will be considerable change in the structure (between bound and un bound form). Be prepared to run extended simulations to observe these - (more than a couple of ns). The structural changes that you are considering will be visible only at these simulation lengths. Over the order of a few hundred ps you should be able to get correct energy values of interaction between the ions and the residues of the protein in the binding site (usually considered to be a 5-7 Ang area around the ion). If your complex is stable you should get a fairly stable interaction energy value through out the length of your simulation with minor error bars.

Hope this helps.

Best,

/A

Depends on your definition of stability. If this is DeltaG of protein folding then unless you can access something like Anton (through the David Shaw group) then I don't belive that you would sample the transition between folded and unfolded states sufficiently (if at all) to obtain an estimate of this quantity.

However your question is confused/confusing. If the PDB file has Ca2+ sites then the protein can bind Ca2+, by definition. Perhaps, as your last sentence suggests, you are interested in the effect of Ca2+ on conformational stability. One approach might be to set up comparartive MD runs with and without Ca2+ present in the starting starting structure (different Ca2+ concs. will not matter as the off-rate is likely too slow to be captured in normal explicit solvent MD runs. If the cation does leave then you could ignore that simulation run). Then you could do this at different temnperatures to get a feel for the role the metal might play in providing conformational stability. But keep in mind that without significant investment of computing resources you are unlikely to capture protein unfolding under ambient conditions, IMO.]

Dear Atul,

I was reluctant to jump into this stream because all the obvious angles were covered by the previous participants. The answer is YES, of course you can with all the usual caveats of using MD.

However, Paul's remarks above prompted me to to share two more angles that were not touched.

1) Proteins do not have precise preferences for different metal ions. They are very promiscuous in binding metal ions. In my hands all my proteins (usually enzymes in biochemical experiments) had at least two metals with with similar binding constant and similar level of activity. It was not always obvious why this not the other couple of ions but there were always some.

2) Sometimes in proteins the kinetics matter (transient binding but giving the result that is expected) sometimes thermodynamics (free energy stabilization). MD is a very crude tool to answer such questions, even though some people try.

So in my opinion, it would be the first step to go to the data bases or some review papers and confirm the details concerning the precise preferences for binding (coordination number, type of ligands and the distances). Calcium has everything from tetrahedral, through octahedral to hepta coordination. In proteins Ca likes the coordination of 7 but that changes depending on a lot of things (like pH, protein motion, inhibitory versus activated states etc).

Besides MD has inherent problem with describing correctly the movements of exterior atoms versus interior ones. So the residence times of individual ions at the particular site would create a yardstick versus the thermodynamic cycle variations and other measures of biding. Just be judicious. Determination whether the binding is spurious or dominant and specific might be out of reach but it is certainly worth trying.

Dear Atul,

Before that please check the protein sequence whether it has EF motifs which are essential in most of the calcium binding proteins. Secondly, when you want to do simulation in the structure what you want to know whether calcium binding motifs etc calcium gromics is a task which can be carried out if you know the protein really has the affinity to bind to calcium.

Dear Boguslaw,

If i understood correctly you are saying that a set of proteins that you are currently working on binds with the same affinity to different ions. Theoretically from the limited understadning i have, proteins would bind to ions under two circumstances, they help the protein fold - in which case the first closest ion the protein comes into contact with that satisfies a level of interaction would do the job (in which case the energetics would be similar - since the protein needs to achieve a native state), where as in the other case the protein would present an allosteric site of sorts - where and excess of Ca (signalling) ions would be detected - and the protein will be able to accommodate it. In this particular case it is already known that Ca binds - and one only wants to study the conformation as it undergoes changes in the presence or absence of the Ca ion.

Extended MD in explicit solvent should be sufficient to basically answer the later questions posted by Atul. Sould it not?

Best,

/A

The thermodynamic system in question involves at least four 'states':

Pf + Ca2+ Pf:Ca2+

Pu + Ca2+ Pu:Ca2+

Where Pf is the protein in a folded conformation (ensemble), Pu is the protein in an unfolded conformation (ensemble) and ':' indicates complexation. In principle all four states are in equilibrium with each other and the 'stabilities' correspond the associated pairwise microscopic equilibrium constants (two are binding/dissociation constants, two are folding steps. This situation applies, EF hand or not, though the equilibrium population of any state (e.g. Pu:Ca2+) might be small.

Now concerning MD, getting useful numbers out of the calculations is heaviliy dependent on the quality of the parameterisation, conformational space sampling 'coverage' for each state, etc. I doubt whether MD is the right way to go to obtain thermodynamic quantities, or (unless in very long, replicate simulation runs) the effect of Ca2+ removal from the PDB file.

There is a field of MD-based 'free energy calculation' that I am less familar with but usually this is applied to very small perturbations, where the two states are not changed much in structure. Not sure that this strategy applicable in this instance where the difference could be as much as a folded vs. an unfolded polypeptide.