[gmx-users] Representative structure of a long simulation

[Berk Hess]

>the following might be a trivial/basic question, but I would like
>to get as many comments/different ideas/views as possible.
>
>I have a long (10 ns) simulation of a membrane protein system, and
>would like to use one protein structure out of this simulation as
>starting point for some further (simple) studies. So I am somehow
>looking for the most "representative" structure out of these 10 ns.

My suggestion would be to perform a covariance analysis on
the Calpha atoms (g_covar) and then project the trajectory on
the first 2 or 3 (or more) eigenvectors (g_anaeig). The computational
cost does not increase much with the number of frames. In this way
you get a good idea of how the trajectory behaves.
You can then choose a structure which is in the middle of the cloud
of the projections.

For large (membrane) proteins convergence can be a problem.
You could have a look at (sorry for citing my own work):

B. Hess
Similarities between principal components of protein dynamics and
random diffusion
Phys. Rev. E 62, 8438-8448 (2000)

B. Hess
Convergence of sampling in protein simulations
Phys. Rev. E 65, 031910 (2002)


Berk.


_________________________________________________________________
Join the world’s largest e-mail service with MSN Hotmail. 
http://www.hotmail.com