[gmx-users] Re: Hamiltonian replica exchange umbrella sampling with gmx 4.6

[Michael Shirts]

Hi, Dejun-

Right now, the vector of lambda parameters is simply vdw, coul, bonded,
restraint, temperature.  You can't have, say, 2 different coul vectors or
two different restraint vectors for different restraints.  But you can
change any of these components.

You define the vector manually by writing out the list.  So to do 2D in vdw
and restraints, you would define states:

vdw-lambdas        =  0.0 0.0 0.0 0.5 0.5 0.5 1.0 1.0 1.0
restraint-lambdas  =  0.0 0.5 1.0 0.0 0.5 1.0 0.0 0.5 1.0

Multidimensional movement in parameter space with replica exchange can be
done with the md setting '-nex Nswaps'.  This does multiple pair swaps,
instead of just neighbor pair swaps.  I'd suggest something like Nswaps =
N^3, where N is the number of lambda states.  This is a close approximation
to Gibbs sampling (http://jcp.aip.org/resource/1/jcpsa6/v135/i19/p194110_s1),
where _all_ permutations are selected based on their Boltzmann weight.
 It's an approximation because with a finite number of swaps, you don't
quite get uncorrelated moves in state space, but it is rigorously correct
thermodynamically.  It is more powerful than randomly selecting a dimension
to move in, since it allows moves in all dimensions simultaneously with
proper weight.

Note that the number of lambdas states is hardcoded as 1024, but that can
(I think) be edited as desired without causing direct problems (other than
perhaps needing to make the .mdp lines longer).

We are working on more general multistate lambda vector formalism for 5.0.
 If you have suggestions, let me know.

I'd be happy to look over input files or give additional advice on a
specific setup.


On Tue, Apr 2, 2013 at 12:52 PM, Dejun Lin <dejun.lin at gmail.com> wrote:

> Hi Michael,
>
> Do the codes now support walking in multidimensional parameter space? i.e.,
> a state is defined by a set of lambda parameters {l1,l2,l3,...,ln} and a MC
> move is attempted along one of the parameter, which is randomly picked.
>
> Thanks,
> Dejun
>
>
>
> --
> View this message in context:
> http://gromacs.5086.n6.nabble.com/Hamiltonian-replica-exchange-umbrella-sampling-with-gmx-4-6-tp5006187p5006842.html
> Sent from the GROMACS Users Forum mailing list archive at Nabble.com.
> --
> 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/Support/Mailing_Lists/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/Support/Mailing_Lists
>