[gmx-developers] [gmx-core] removal of encad and gmx FF in 5.0?

[Justin Lemkul]

On 2/14/14, 11:24 AM, Berk Hess wrote:
> On 02/14/2014 05:16 PM, Justin Lemkul wrote:
>>
>>
>> On 2/14/14, 11:12 AM, Mark Abraham wrote:
>>> Removing encad and gmx sounds good. Probably don't have time to rip out the code
>>> that supported encad.
>>>
>>> We're happy to add forcefields that are used by more than one person, have been
>>> extensively tested, have an associated publication record, etc. Any particular
>>> suggestions, Alexey? Others?
>>>
>>> Off the top of my head, Charmm36 would be an obvious thing to add. My
>>> reservation here is that there's not really ever been a good way to run
>>> CHARMM-style non-bondeds in GROMACS. I suspect this will change in 5.0 with
>>> Verlet-kernel support for switch functions, but perhaps a fully validated
>>> forcefield and workflow port for 5.1 would be a good joint project. Thoughts,
>>> Justin?
>>>
>>
>> Yes, we're working on that.  I'm personally looking into the best ways to run
>> CHARMM force fields within Gromacs.  Couple that with my ongoing
>> implementation of our Drude algorithms and I think 5.1 is a good target.  Our
>> only reservations about including a CHARMM36 force field directory in Gromacs
>> are: (1) we update the force field quite frequently, certainly more often than
>> Gromacs versions are released and (2) we keep detailed statistics about
>> downloads from our lab website to help track usage (and get grants :).
>>
>> At minimum, I am perfectly happy to oversee the coordination between Gromacs
>> and the CHARMM force fields in terms of making sure we have a suitable
>> (recent) distribution ready whenever a new Gromacs release comes out.
> Maybe (but maybe not) we currently need a little extra functionality in Gromacs
> for CHARMM36. But after that, the force-field files should be independent of the
> Gromacs version. Force-field distributions not included into the Gromacs
> distribution can be updated independently from the code/releases.
>

True.  This is pretty much the existing model, but there is certainly a lot of 
demand for CHARMM36; people are using it a lot.  We have done a lot of 
single-point energies to validate the force fields between CHARMM and Gromacs, 
but I'm thinking about doing a more comprehensive test suite, especially with 
the new Verlet-compatible switch modifiers.  Those are an important development 
for us.  I'm currently working on a lot of other things, but these validations 
and comparisons are high on my list of priorities.

> Are parameters of existing common molecules being updated in CHARMM36, or are
> only new molecules being added?
>

In general, the "core" of the force field that is most commonly used (proteins 
and nucleic acids) is pretty static, but there is a lot of development going on 
for carbohydrates and lipids.  We are also continually adding model compounds as 
part of CGenFF.  There are occasional small refinements in the parameters across 
the board, but I'd have to consult with a couple of others in the lab to see 
where things stand at the moment.  I'm working on Drude stuff, so I don't check 
in with the additive team too often.

-Justin

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

Justin A. Lemkul, Ph.D.
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

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