[gmx-users] Problem with constraints in NVT calculations.

[Justin Lemkul]

On 10/6/14 10:35 PM, Kester Wong wrote:
> Dear Justin and all,
>
>
> -------- 원본 메일 ---------
>
>     *보낸사람* : Justin Lemkul <jalemkul at vt.edu>
>     *받는사람* : <gmx-users at gromacs.org>
>     *받은날짜* : 2014년 9월 29일(월) 21:32:13
>     *제목* : Re: [gmx-users] Problem with constraints in NVT calculations.
>
>     On 9/29/14 8:01 AM, Kester Wong wrote:
>     > Dear all,
>     >
>     >
>     > I have been using the SWM4-NDP water model that works well with CHARMM27 force
>     > field.
>     >
>
>     Time out.  SWM4-NDP is a polarizable water model, and CHARMM27 is an additive
>     force field.  The two are not compatible.  If you're using them together, you're
>     basically just gambling that they work.  You shouldn't do this.
>
> Thank you for the information. Actually, I took the data from the
> ff-drude-water-ions.itp file and included them in the CHARMM27 itp files. Now I
> realised that it does not work like that.
>
>
> If the ffnonbonded.itp file contains the appropriate nonbonded parameters for
> the polarisable water model (i.e. the atomtypes and dummies for the Drude water
> and ions model), will it work?
>
> Below is my input topologies:
>
> https://drive.google.com/file/d/0B7ym8d6G9-e2dmdQOHNaU3c3Qkk/view?usp=sharing

No, this is fundamentally wrong.  You can't simply combine some additive 
components and some polarizable components and hope for a sensible physical 
model.  Force fields are derived in self-consistent ways.  You can't patch them 
together.

>
>     > The force field (most recent version) was obtained from virtualchemistry, and
>     > consists of polarisable ions and the SWM4-NDP water model from Lamoureux et al.
>     >
>     > The water on graphene (energy minimisation and NVT) calculations are stable, and
>     > have produced a few NVT runs of up to 10ns.
>     >
>     > I also compared the NVT runs with other previously calculated structures,
>     > namely, the visualised NVT runs are very much comparable to that obtained by
>     > CHARMM27 force field with TIPS3P water model.
>     >
>     >
>     > However, as the polarisable H3O+ and Cl- ions are inserted into the simulation
>     > box, the NVT run tends to crash as a result of too many LINCS warnings.
>     >
>     > I suspect that the constraints in the H3O is the cause of the LINCS warnings,
>     > and after removing the [ constraints ] in the topology file, the calculation
>     > seems to be running.
>     >
>     >
>     > ;[ constraints ]
>     >
>     > ;; i     funct   doh     dhh
>     >
>     > ;1 3      1       0.102
>     >
>     > ;1 4      1       0.102
>     >
>     > ;1 5      1       0.102
>     >
>     > ;3 4      1       0.169124
>     >
>     > ;4 5      1       0.169124
>     >
>     > ;3 5      1       0.169124
>     >
>     >
>     >
>     > Can anyone please tell me if what I am doing is correct?
>     >
>     > As I am not sure if removing the H3O constraints in the NVT run is appropriate.
>     >
>     >
>     > With regard to the massless Drude and standard oxygen mass (thank you Justin for
>     > providing the information), I would like to know why are they not appropriate
>     > (as mentioned in the previous email) for the SWM4-NDP model?
>     >
>
>     I wasn't making a declarative statement that it won't work.  I was asking
>     whether anyone had verified that using a massless Drude in that model actually
>     reproduced all of the physical properties of the SWM4-NDP model.  We do all of
>     our development on the Drude FF with Drudes of mass = 0.4 amu and the extended
>     Lagrangian.  I don't know how a massless Drude behaves in this model.  It may
>     work fine, but you should very rigorously verify this yourself before doing
>     anything by doing simulations of pure water.
>
>     -Justin
>
> I have Read the paper by Jochen Hub et al [Chem. Sci., 5, 1745 2013], that their
> calculated data (water with ions) are in line with experimental results.
> Upon further inspection, I found more papers that reported the use of 0.4 amu
> mass and an extended Lagrangian for the Drude particle. e.g. by Yu et al. [J.
> CHem. Theory Comput. 6, 774-786 2010].
> Could you please advise if there is anything I could do, to make the additive
> and polarisable FFs usable after modification?
>

No.  Either the force field must be entirely polarizable or entirely additive.

> I am more than willing to go through the necessary changes/implementation if it
> can be done as this is the only avenue that I have left.
>

I am working on the code-level implementation of all the necessary algorithms so 
people can use our fully polarizable biomolecular FFs for proteins, DNA, water, 
and ions.  It's not a trivial task.  I have only a few more hang-ups to work 
out, but at this point, if you want to be doing polarizable simulations with our 
FFs, you should look at NAMD.  It supports everything you need.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================