Difficult to know whether your collective varibles are good since you do not explain them and we do not know the structural changes from the open to the closed conformations (is it a rigid rigid-body displacement?). If there is one main collective variable, can you use umbrella sampling?

Difficult to know whether your collective varibles are good since you do not explain them and we do not know the structural changes from the open to the closed conformations (is it a rigid rigid-body displacement?). If there is one main collective variable, can you use umbrella sampling?

Dear Prof. Derreumaux,

Thank you for looking at my problem. Here is the detail of the collective variable,

I choose the distance between center of mass of few residues in the loop and few residues in the active site. The motion of the loop is not like 'regid-body' type, rather the loop is highly flexible.

In the metadynamics trajectory, I saw the opening closing events for several times. Will it be a good idea to find out the intermediate windows and run umbrella sampling simulations for this kind of problem?

Thanks for the suggestions.

Regards,

Tarak

The absence of a specific event in a 300 ns molecular dynamics (MD) simulation does not necessarily mean a high free energy barrier for that event.

Consider this situation: There is a very small hill called "Cow's mountain" in Brno, Czech Republic; it is accessible via multiple paved roads and to reach the top of the hill is an easy walk; an easy walk if, and only if one starts the walk somewhere in Brno rather than in India, especially if one is a foreign tourist without the Brno city map and without any knowledge of the site. So, if you are located in India, you need to use an airplane, Brno city bus/tram and the Brno city map. And a lot of time . . .

---

In general, with MD simulation time "t" one can effectively describe motions occurring at "10^-3 to 10^-4 t" (at most). Hence, in your case, 300 ns MD simulation can cover effectively events up to 30-300 ps only, which is a very short time indeed! So if you did not see the opening/closing event, you just did not have the good luck to win the 1:1,000,000 jackpot. Bad luck.

Hi Martin,

Thanks for the reply.

Indeed, a pretty good example to describe the scenario!

Let's see how quickly I can reach the top of the hill, trying with several biased simulations (airplane, bus and tram). :)

A few suggestions:

1. Sometimes choosing too large width of the deposited Gaussians may lead to lower resolution of your FES. I mean the width must be much smaller than the length scale of free energy variation along the CV of your choice. One rule of thumb is to determine the width from the distribution of same CV(s) at the equilibrium states using unbiased MD simulations. Your Gaussian width should be much smaller (~1/2-1/3) than the width of the local minimum from your unbiased simulation. You need smaller width to capture the steeper FES segments.

2. Do you have crystal structures for both open and closed states? Do you reproduce same configurations in your Metadynamics simulation when you observe multiple opening and closing events? Distance between COM may or may not be the correct CV depending on the actual end states. You may need to introduce multiple CVs with distance between specific residue pairs (depending on the structure of the closed state).

3. Make sure you don't overfill the FES (don't add too many Gaussians at a too high rate). You may check convergence of FES with time, but the accuracy would/should improve at the limiting conditions of (i) lower rate of deposition, (ii) lower width, and (iii) lower height of the Gaussians.

If Martin is correct, then what you have is more of a "search" problem and less of a "barrier crossing".  AMBER has a specific protein loop conformation search tool (LMOD) which you might want to try, caveat I have never used it myself.  http://ambermd.org/doc12/ .  LMOD returns a library of local minimum loop conformations which you can use to construct a map of the landscape projected onto one or more collective variables.

Search problem may also imply an entropic barrier/bottleneck. Imagine ions entering a narrow channel. This barrier should show up in the free energy barrier calculated using Metadynamics as well, unless there is significant enthalpy/entropy cancellation going on.

If the effective free energy barrier is actually low and still the observed transition rate in unbiased simulation is lower than expected (from transition state theory), then it might be "diffusion controlled" (in contrast to activated process). In other words, within transition state theory the transmission coefficient (pre-exponential factor) might be much lower than 1.

Thanks Dr. Berryman for the suggestion.

Dear Sir,

Thanks for the elaborative answer.

Regarding the sigma value, from the unbiased simulations, I see fluctuations of the CV around 0.25 (unitless), and thus, I have chosen to use 0.1 in the metadynamics runs. I am running WTM simulations, with initial hills height of 0.3 kcal/mol.

No, only the open structure of the enzyme is available. This makes the life difficult (otherwise, I could have opted for Path collective variable based methods). In my case, state B, the closed one is completely unknown. I gathered a bit of information from closed ligand-bound crystal structures of homologous enzymes from different species.

I use contact map based collective variable which has the information of local contacts as well as the center of mass between the loop and an almost static region of the protein core.

I plot the hills height as a function of the metadynamics simulation time and see an exponential decay of the hills heights, which it should be in case of the WTM simulations. However, even after 400 ns, it does not decay down to almost zero. Block average of the free energy values is still changing.

Yes, for testing purpose, I will use different sigma value(s), smaller than what I have used.

You may also try running some unbiased simulations starting from the "closed" structure obtained from WTM simulations. Ideally the closed structure should be thermodynamically more stable (downhill FES) with bound ligand. You may test the stability of the closed state (and specific contacts leading to the stability) using multiple normal MD trajectories starting from such structures. It should not go back to the open state.

Yes, I am running the unbiased simulations with the closed state.

Thanks.