[gmx-users] constraint pulling

[Thomas Schlesier]

Am 17.03.2014 20:11, schrieb 
gromacs.org_gmx-users-request at maillist.sys.kth.se:
>> Dear gromacs users
>> >
>> >I will use pull = constraint (constraint pulling) for PMF calculations.
>> >
>> >There is a tutorial about PMF from Justin in which he used
>> >pull = umbrella (umbrella sampling).
>> >
>> >In justin tutorial, step are as follows:
>> >
>> >1) preparation of system
>> >
>> >2) minimization & equilibration
>> >
>> >3) generate configuration
>> >
>> >4) umbrella sampling
>> >
>> >5) WHAM which gives PMF curve
>> >
>> >
>> >I conceive in my case (pull = constraint), steps 1, 2 and 3 are similar to
>> >
>> >justin case (pull = umbrella). Since, WHAM analyzes output files
>> >
>> >generated by umbrella sampling, I have not idea about steps 4 and 5.
>> >
>> >I did not know how to obtain PMF curve in my case.
>> >
>> >If my notion is wrong, please give me more explanation such that I do
>> >
>> >this PMF calculation using constraint pulling method. Please tell me steps
>> >needed for
>> >
>> >constraint pulling.
>> >
>> >Is there a tutorial about PMF calculation using constraint pulling method?
>> >
> If you're using a constraint to preserve distances along the reaction
> coordinate, there's nothing to de-bias.  There is no force constant for
> constraint pulling, so you can't run WHAM on its output.  The constraint method
> fixes the distance, whereas the umbrella method is a harmonic bias (which is
> what you need for umbrella sampling, hence the name).  If you did the "umbrella
> sampling" with the constraint method, there's nothing you're going to get out of it.
>
> -Justin

The basic simulations are in both cases similar (as is the idea), but 
point (4) and (5) differ somewhat. For the constraint simulations:

4) Where umbrella sampling restraints the distances, one constrains here 
the distances. That's all... One thing to note (which goes also to point 3):
In umbrella sampling the system can move somewhat along the reaction 
coordinate, with the constraints not -> This mean with the constraint 
pulling one should use a smaller spacing then for umbrella sampling.

5) De-bias / Calculate the PMF
For the constraint pulling it's just integrating the constraint force 
along your reaction coordinate (in principle) - no WHAM as Justin said:
->   V_pmf = - \int (R -> R_max) F_c(r) dr
Tricky part is your reaction coordinate. The above holds only true if 
you have a linear reaction coordinate and the force acts only in one 
direction.
For more complicate reaction coordinates it becomes more complicate.
For a linear reaction coordinate where the force acts along all three 
axis it's:
V_pmf = - \int (R -> R_max) [F_c(r) + 2 k_B T / r ] dr


to 3)
Easiest way to get evenly spaced configuration is using the pull code 
with constraints. If one uses a harmonic (/ umbrella) potential, one has 
fluctuations, which might are annoying in that case.

Greetings
Thomas