normal MARTINI compatibility with Verlet cut-off scheme

  • kmcallenberg
  • kmcallenberg's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
11 years 2 months ago #1345 by kmcallenberg
The Gromacs MD parameters that are given in the mdp files on this website ( md.chem.rug.nl/cgmartini/index.php/downl...rce-field-parameters ), and as specified in the MARTINI research articles, are not compatible with the Verlet cut-off scheme ( www.gromacs.org/Documentation/Cut-off_schemes ). Verlet is useful, particularly because it is now supported by the latest beta version of GROMACS on GPUs.

Here are the changes I needed to make in order to get Verlet to work:
coulombtype: Shift -> Reaction-Field
rcoulomb: 1.2 -> 1.0
vdw_type: Shift -> Cut-off
rlist: 1.2 -> 1.0

These settings came from this thread:
lists.gromacs.org/pipermail/gmx-develope...12-March/005696.html
lists.gromacs.org/pipermail/gmx-develope...12-March/005710.html

Are these acceptable parameters to use? Reaction Field was tested in "Membrane poration by antimicrobial peptides combining atomistic and coarse-grained descriptions."

Thank you,
Keith Callenberg

Please Log in or Create an account to join the conversation.

More
11 years 2 months ago #1349 by djurre
In Martini, like in any forcefield, if you change the potential form, you will change the interactions and you might break it.

If you change the cut-off and and use reaction-field/cut-off the potential shape will change and you would have to check if the solvation/partitioning energies of the different beads changes (compared to table 3, Marrink, J.PHys.Chem.B, 2007).

Reaction field might well be a proper solution. However, as Berk Hess also mentions in one of the threads you refer to, in case of the Martini forcefield the LJ interactions often have the largest energy contribution. Using a straight cut-off would introduce discontinuities in the force, which would most likely change the behavior.

There is now the mdp option "vdw-modifier". However, this only seems the shift the potential and not influence the force. Hence there will still be discontinuities.

If you say: " changes I needed to make in order to get Verlet to work", do you mean to make it run stable or did you already testing of some systems?

Please Log in or Create an account to join the conversation.

  • kmcallenberg
  • kmcallenberg's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
11 years 2 months ago #1350 by kmcallenberg
Replied by kmcallenberg on topic MARTINI compatibility with Verlet cut-off scheme
Thanks for your detailed reply, Djurre. I completely agree with your words of caution with regard to making changes to these parameters. I have observed a small but noticeable difference in a few calculated quantities when I change to Reaction Field. However this is still very anecdotal; I haven't yet tested this rigorously.

When I say "changes I needed to make in order to get Verlet to work" I don't even mean "to make it run stable" as you suggested. Although the systems I have looked at do happen to be relatively stable, what I mean is that GROMACS grompp will literally give an error without these parameters when trying to use Verlet on GPUs.

Thank you for pointing out the vdw-modifier option. I did not notice that. Do you think these modifiers will eventually handle the forces as well?

I completely agree that rigorous testing is needed. Maybe I will send an email to Berk Hess or the MARTINI folks. I was hoping to find that some testing had already been done, or was already underway. I guess I could do my own testing, but I'd worry that I wasn't testing the right quantities. It would be nice if the MARTINI group would put together a package/tutorial for testing the forcefield for things like solvation and partitioning energies like you brought up.

Please Log in or Create an account to join the conversation.

More
11 years 2 months ago #1354 by djurre

kmcallenberg wrote: Thank you for pointing out the vdw-modifier option. I did not notice that. Do you think these modifiers will eventually handle the forces as well?


I don't know, but based on what they write in the text I would say no.

kmcallenberg wrote: I completely agree that rigorous testing is needed. Maybe I will send an email to Berk Hess or the MARTINI folks. I was hoping to find that some testing had already been done, or was already underway. I guess I could do my own testing, but I'd worry that I wasn't testing the right quantities.


We are definitely planning to do check the partitioning with the new Verlet scheme, but I don't see an obvious best solutions yet.

kmcallenberg wrote: It would be nice if the MARTINI group would put together a package/tutorial for testing the forcefield for things like solvation and partitioning energies like you brought up.


We are working on exactly that tutorial and will probably be online before hell freezes over. But when exactly, I don't know.

One thing I was thinking about, can you use a tabulated, user defined potential in combination with the Verlet-scheme? Then you could use the original potential in tabulated form.

Please Log in or Create an account to join the conversation.

Time to create page: 0.091 seconds