[gmx-users] rmsf question

[Andrei Neamtu]

Hi,
thank you all.

I have asked the question about RMSD because of the following reason:

I simulate a protein whose structure was determined by NMR only.
My simulation is quite long (2 microseconds) in the absence of a
ligand, present in the experimental structure file. During the course
of the simulation the protein simulated is subjected to large
conformational tranitions.
To validate the results I also have made a simulation (also 2
microseconds) with the ligand present which I expected to preserve the
original protein conformation.
This is indeed happening for 300 nanoseconds after which a
reorganisation of secondary  structure elements takes place. This is
happening in my opinion due to the fact that my simulation is at the
physiological temperature of 310K while the NMR structure was resolved
at 293K.
Now I am repeating the second simulation at 293K.
To evaluate the accuracy of the simulations I want to compare the data
in the simulation with ligand bound, to as much experimental data as
possible. NMR experiments provide data that can be used to bencmark
the accuracy of simulations.
So, I thought that locally (on short time scale, under 1 nanosecond)
the RMS fluctuations of the C-alpha of the backbone can be correlated
to RMSF of the bundle of conformations in the NMR file. This now seems
to me that it doesn't make sense. So I wil try to use the backbone
order parameters for comparison.

Andrei

On Tue, Mar 16, 2010 at 11:50 AM, Tsjerk Wassenaar <tsjerkw at gmail.com> wrote:
> Hi Andrei,
>
>> If I correctly uderstand the RMSF computed on C-alpha in PDB NMR
>> structure, is a measure of the uncertanty in resolving the structure.
>
> No, that's not what I said. You're saying that there's one structure
> (the structure), but there is uncertainty in resolving it. That's not
> the case. There is a distribution of structures that gives rise to a
> set of signals, from which time and ensemble averaged distances
> between certain pairs of atoms can be deduced. But this set of
> distances is typically leaves a number of degrees of freedom, and on
> top of that the time averaging may actually yield restraints that
> cannot be satisfied simultaneously. So, structures have to be fit to
> the set of distances, which is typically repeated a large number of
> times using simulated annealing with different starting conditions.
> >From the resulting set the ones best fitting the data are selected and
> put forward as a representative set of structures. The least defined
> regions, in terms of (distance) restraints are likely to show most
> variation. But the structures are not weighted according to their
> probabilities and therefore neither the average nor the fluctuations
> can be expected to coincide with the moments of the Boltzmann
> distribution.
>
>> Compared to this the RMSF of C-alpha computed from a MD trajectory
>> reflects the geometric fluctuations of the backbone in the Boltzmann
>> distribution of states in the generated ensemble.
>
> Assuming you have sampled long enough and not just trying to explain
> drift as fluctuation :)
>
>> The uncertanty in the NMR structure can be influenced by the thermal
>> fluctuations in the backbone geometry.
>
> Well, that's part of it, but the way the models are derived contributes a lot.
>
>> Only in this respect the comparison of the two makes sense.
>> Is this correct?
>
> Comparisons can be made, but you have to formulate and justify your assumptions.
>
> Cheers,
>
> Tsjerk
>
>
> --
> Tsjerk A. Wassenaar, Ph.D.
>
> Computational Chemist
> Medicinal Chemist
> Neuropharmacologist
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>