[gmx-users] Directed MD - Transition Paths in GMX

Anthony Ivetac ivetac at biop.ox.ac.uk
Wed Feb 11 13:01:01 CET 2004


The algorithm I am trying to implement specifically is the Path
Exploration with Distance Contraints (PEDC) method, used by Guilbert et
al. in CHARMM.

I am trying to find code in Gromacs which would allow me to add a
distance constraint potential to the existing potential energy function.
The strength of the distance constraint would depend on the direction of
the simulated structure which would be measured by RMSD to a 'reference'
structure - ie. if the RMSD is getting bigger then we would apply a
force proportional to the deviation from where we want it to be at that
time point. If the RMSD is getting smaller then we apply no constraint
and let it run.

If anyone has any ideas on where I could add a subroutine which adds to
the potential energy function, please let me know!

Many thanks,


Kay Gottschalk wrote:
> Can't you use NOE-like restraints with increasing force constants to
> drive the system where you want it? There are quite a few papers by
> Schulten and Karplus about 'steered MD', I think. I don't know how they
> implemented it, though.
> K.
> On Feb 9, 2004, at 12:25 PM, Anthony Ivetac wrote:
>> Thanks David. However, rather than interpolating the intermediates
>> between 2 end-point structures, I am trying to get to the 'open' state
>> from the 'closed' state through MD, because I am interested in seeing
>> the impact of this 'opening' motion on an adjacent subunit.
>> I have no experience of playing with the Gromacs code - does anyone
>> know
>> how easy it would be to modify the potential energy function to include
>> some distance contraints? For example, for every step, I would want to
>> calculate the RMSD between the current structure in the simulation and
>> the 'end-point' structure -> RMSDs closer to the end-point would be
>> encouraged and RMSDs away from the end-point would be penalised.
>> Many thanks!
>> David van der Spoel wrote:
>>>> I'm wondering if anyone has implemented a modified version of Gromacs
>>>> which will handle the directed simulation of a protein moving from
>>>> one
>>>> conformation to another?
>>> **********************************************************************
>>> **
>>> Not really, but you can use the g_morph or g_dyndom program to
>>> generate
>>> a trajectory connecting two structures. Then you can take these
>>> structures and minimize them. The latter program assumes that you have
>>> used the dyndom program beforehand (which is not part of gromacs, but
>>> is
>>> in CCP4.)
>> >**********************************************************************
>> *
>>>> For example, given 2 crystal structures of a protein in an 'open' and
>>>> 'closed' state - we would start from the closed state and direct the
>>>> protein towards the open state. This could be done by measuring the
>>>> RMSD
>>>> between the simulated structure and the 'target' structure at each
>>>> step
>>>> - encouraging progressively smaller RMSD values until the simulated
>>>> structure superimposes with the target structure (using some sort of
>>>> penalty function).
>>>> This is basically the strategy used by Guilbert et al. in their Path
>>>> Exploration with Distance Constraints (PEDC) method.
>>>> If anyone knows of any other techniques to direct a protein from one
>>>> conformation to another, with GMX, please let me know!
>>>> Many thanks.

More information about the gromacs.org_gmx-users mailing list