I would tell if it is not specifically interacting,
Mg2+ is dispensable on that short time scale and
its exclusion may prevent limits from sampling and
some spurious interactions.
Of course, some referees may vigorously require magnesium
must be included because it is "more biological", which, considering
the force field approximation is quite questionable, or at least
such justification is an oversimplification.
Thus we mostly run monovalent simulations and then some control
Mg2+ simulation. Without testing, it is difficult to tell if new
parameters are better or not, even different or not.
Water model is an open issue...
But I would like to reiterate that I do not want to make
any authoritative statement.
We have some discussions in
Understanding RNA Flexibility Using Explicit Solvent Simulations: The
Ribosomal and Group I Intron Reverse Kink-Turn Motifs
Sklenovsky et al. JCTC 7 Pages: 2963-2980 DOI: 10.1021/ct200204t 2011
and Sponer, Cang, Cheatham Molecular dynamics simulations of G-DNA and
perspectives on the simulation of nucleic acid structures Methods in
press.
Again, there are no "perfect" or "guaranteed" options.
Jiri
On Fri, 18 May 2012, Asmita Gupta wrote:
> Date: Fri, 18 May 2012 16:56:01 +0530
> From: Asmita Gupta <asmita4des.gmail.com>
> Reply-To: AMBER Mailing List <amber.ambermd.org>
> To: AMBER Mailing List <amber.ambermd.org>
> Subject: Re: [AMBER] new Mg parameters
>
> Is it like that the type of Mg+ parameters; whether new or old depends on
> the type of RNA simulation one wants to intend? For as it was discussed
> some time back in amber archives (
> http://archive.ambermd.org/201201/0259.html) that if Magnesium is not
> playing an active role in the functionality of molecule and is there just
> to stabilize the fold by neutralizing negative charge on the phosphate
> backbone, in that case how much difference could the new parameters make,
> especially if the simulations are of 50-100ns long??
>
> Also, as Jesper told that some lines have to be commented in parm10.dat for
> implementing new params, then ion08.lib file which is loaded automatically
> by ff12SB (amber 12) and consists of Magnesium information, does that also
> should be deleted or commented??
>
> Asmi..
>
> On Fri, May 18, 2012 at 1:24 AM, Jiri Sponer <sponer.ncbr.muni.cz> wrote:
>
>> Dear Jesper,
>>
>> not, we did not try. I am in general, if possible,
>> trying to avoid divalents in NA simulations, since
>> their basic force field form is not very realistic
>> due to lack of charge transfer and polarization.
>> I.e., there is problem with the basic LennardJones+pointcharge
>> approximation, which definitely is not a real Mg2+ ion.
>>
>> It is not clear to me how to design a proper test.
>> I think, Tom may comment on that, usual problem
>> with the "standard" parameters has been formation of undesired
>> inner-shell binding
>> contacts during equilibration even when not intended.
>> On the other hand, in ribosome and ribozymes, there
>> are inner-shell Mg2+ interactions, although the
>> experimental uncertainties are also very substantial....
>>
>> So I would tell inclusion of Mg2+ into NA simulations
>> is a difficult issue where we have selection of several
>> bad options.
>>
>> Jiri
>>
>>>
>>> Dear Jiri,
>>>
>>> Have you tried them out already?
>>> Okay, well I am not using them for RNA just yet, but I might sometime
>> later.
>>>
>>> I still have to run some production simulations with them before I will
>> comment on their quality.
>>> But I think the situation is more a result of a lack of magnesium
>> parameters to pick from. Here is what I have found:
>>>
>>> Single atom parameters:
>>> * Charmm27 (developed by B. Roux, no publication of the development)
>>> * Amber9x (developed by J. Aqvist 1990)
>>> * Allner et al. enhancement of the B. Roux parameters for Charmm (2012)
>>>
>>> Dummy atom models (MD62+):
>>> * Oelschlaeger et al. (2007)
>>> * Sept and co-workers (maybe 2012)
>>>
>>> I like the idea of the dummy atom models, but I've tried the
>> Oelschlaeger and they don't return correct ion-water rdfs (in pure water)
>> and I see some things when I use it with proteins, where the octhedral
>> coordination is not great, and waters are replaced by direct-interaction
>> with the protein (this latter thing may not be incorrect). Also, they were
>> not developed for TIP* water models or the AMBER MD setup. Plus they change
>> some parameters in the amberFF when they performed their publications.
>>> The Sept and co-worker parameters are not freely available yet, but
>> hopefully will be sometime soon.
>>>
>>> So for now it will be the single atom parameters I will use and newly
>> developed parameters that match experimental results, seem like a better
>> choice than parameters from 1990. Not saying that older parameters aren't
>> sometimes better than newer ones - it depends on the property you are
>> investigating I guess. That is just my 2 cents...
>>>
>>> All the best,
>>> Jesper
>>>
>>>
>>> On May 17, 2012, at 12:19 PM, Jiri Sponer wrote:
>>>
>>>> Dear Jesper, Tom:
>>>>
>>>> it still does not predict perfmormance in RNA simulations,
>>>> which may be a specific issue, so difficult to tell if they
>>>> are to be recommendd.
>>>>
>>>> Jiri
>>>>
>>>>> Thanks Tom. I'll probably submit these to the "Bryce" parameter
>> database one of the following days. I have talked to the author about this.
>>>>>
>>>>> One way I checked to see if my implementation was correct was to run
>> a single point energy in NAMD of an Mg-Mg system in the gas-phase.
>>>>> I did this using the parameters in charmm (which they were developed
>> for) and in amber (with the files I attached).
>>>>>
>>>>> I then checked for a variety of distances that the energies I got
>> were the same (I was both in the repulsive and attractive regions).
>>>>>
>>>>> I redid the above procedure using the older Roux parameters in charmm
>> as well, just to check that that worked too.
>>>>>
>>>>> Does that sound like a reasonable test?
>>>>>
>>>>> Best,
>>>>> Jesper
>>>>>
>>>>> On May 16, 2012, at 12:03 PM, Thomas Cheatham III wrote:
>>>>>
>>>>>>
>>>>>>> I have recently been working on this and I am still in the testing
>> phase, but these files (attached) should make it possible to use
>>>>>>> the Allner et al. parameters you are referring to.
>>>>>>
>>>>>> I haven't tried these new parameters for Mg2+ yet (but we plan too),
>>>>>> however the approach proposed in the Allner et al. paper to get
>> decent
>>>>>> association times for the Mg2+ ions is clever and compelling. I am
>> fairly
>>>>>> confident that these will be better than the current default
>> AMBER-adapted
>>>>>> Aqvist Mg2+ parameters (where we have observed funny artifacts in DNA
>>>>>> simulation such as ions tunneling into the helix leading to base pair
>>>>>> disruption). Jesper's attachments look correct.
>>>>>>
>>>>>> --tec3
>>>>>>
>>>>>> _______________________________________________
>>>>>> AMBER mailing list
>>>>>> AMBER.ambermd.org
>>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> AMBER mailing list
>>>>> AMBER.ambermd.org
>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> AMBER mailing list
>>>> AMBER.ambermd.org
>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>>>
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri May 18 2012 - 05:30:04 PDT
