[gmx-users] surfactants simulation topology generation with Automated Topology Builder and lincs warnings

[Thomas Piggot]

Hi Peter,

I'd imagine that in particular the CG angle parameters between beads may 
well be different if you determined them by mapping on to a single one 
of these surfactants in water, compared to mapping onto an ensemble of 
structures in an aggregated state as the hydrophobic chains would try to 
'fold up' and bury themselves away from the water. I haven't actually 
tried this though, that's just how I think it would probably be. Nice 
and easy to test anyway, if Emiliano really does need to parameterise 
this molecule.

Cheers

Tom

On 18/08/17 09:04, Peter Kroon wrote:
> Hi Tom,
>
>
> thanks for your thoughts :)
>
> I want to respond to your sampling argument: I figured that a surfactant
> in solution is "more free" to sample conformations due to sterics than
> one in an aggregated state, and that sampling would therefore be faster.
> Your point of sampling the bonded conformations from a solvated vs an
> aggregated state is a powerful one though (then again, shouldn't the
> difference in aggregated state vs solvated state come from the
> non-bonded parameters? i.e. the bonded parameters should be the same for
> both).
>
>
> Peter
>
>
> On 17-08-17 21:10, Piggot T. wrote:
>> Hi Peter/Emeliano,
>>
>> I'm not sure I agree with some of what Peter says, but I guess it's probably a matter of taste. If it were me, I'd definitely want my atomistic simulations to behave properly before trying to develop CG parameters based upon these simulations. I know that the coarse-graining will lose some of the detail, but I'd want all of the detail in the atomistic simulations to be as accurate as possible to hopefully develop reasonable CG parameters with the appropriate detail lost but the underlying, correct, behaviour retained. You cannot be sure of this in your case here.
>>
>> As for the sampling in the atomistic simulations, I guess you mean you could run one in a box a lot quicker as the system is smaller? With 50, you obviously have more surfactants in there to give you a lot more data for the parameterisaton and as a larger simulation size should scale better, you probably will get better sampling (in terms of stats) with the 50 in a box setup. Plus you get to also check, as Emeliano said, that the atomistic simulations behave sensibly and aggregate/form micelles, etc. (whatever this surfactant does). You can also look for differences in the CG bonds/angles depending upon what state the molecule is in (solvated, aggregated, etc.).  For this specific case, I guess this may not matter if it's only one bound to a protein though.
>>
>> Anyway, regarding the original post, I would firstly ask is it really necessary to have this molecule in the simulations? I couldn't tell from the post why this was wanted to be included. Is it an important ligand, or is it just in the experimental structure as an artefact of the crystallisation conditions/procedure (which I suspect is quite likely)? If it's the latter, there is no need to go to all this effort. As for the LINCS warnings, it's hard to exactly say without seeing the topology/starting structure. It could well be that the ATB topology for things like the sugars isn't that great (the GROMOS sugar force fields are heavily optimised for things like dihedrals), or it could potentially be an issue with the starting structure of the system. If it were me, I would likely make the atomistic parameters manually through combining the building blocks available within the GROMOS force field.
>>
>> Cheers
>>
>> Tom
>>
>> ________________________________________
>> From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se [gromacs.org_gmx-users-bounces at maillist.sys.kth.se] on behalf of Peter Kroon [p.c.kroon at rug.nl]
>> Sent: 17 August 2017 12:03
>> To: gromacs.org_gmx-users at maillist.sys.kth.se
>> Subject: Re: [gmx-users] surfactants simulation topology generation with Automated Topology Builder and lincs warnings
>>
>> Hi Emeliano,
>>
>>
>> since you're just going to use the atomistic simulation to get some
>> parameters for your CG model, I don't think the differences will be
>> significant --- the approximations your are going to make in CG will be
>> larger anyway. I would even argue you'll be better off if you run just
>> one surfactant in water to get your bonded parameters, rather than 50,
>> since sampling will be better.
>>
>> For further validation of your Martini model, you can (should) look at
>> some more macroscopic properties as well, such as dimerization free
>> energy and partition free energy.
>>
>>
>> Peter
>>
>>
>> On 17-08-17 11:18, edesantis wrote:
>>> dear gromacs users,
>>>
>>> I have a problem in the simulation of a surfactant, Octyl Glucose
>>> Neopentyl Glycol, that is present in protein crystals.
>>>
>>> my goal is to have a coarse grained model for this surfactant with
>>> Martini ff.
>>> to do that I have to generated an all atomistic simulation to use as a
>>> reference to build the Martini topology.
>>>
>>> I've downloaded the pdb file https://www3.rcsb.org/ligand/37X of the
>>> surfactant and since there is not an existent ff for the all atom
>>> simulation, I've generated it from ATB web site
>>> (https://atb.uq.edu.au/) for gromos 53a6 united atoms parameters set.
>>>
>>> I've built a cubic box of 7 nm of side, I've put inside the box 50
>>> surfactant molecules and then I've solvated it with spc water.
>>> then after a minimisation with the sd both in vacuum and in the
>>> presence of the solvent, I proceeded with a md in the NVT ensemble,
>>> with dt=0.002 ps
>>> the problem is that I received several lincs warnings and the
>>> simulation stops.
>>> so I've decreased the dt to 0.0015 ps and the simulation ends without
>>> problem, I've continued it increasing the dt to 0.0018 for 20ns, and
>>> than to 0.00183 ps and there are not any kind problem.
>>> but when I try to increase dt to 0.00186 the lincs warning problems
>>> came again.
>>>
>>>
>>> watching the simulation movie with vmd, I can see that the surfactants
>>> form an aggregate (as they should do), and it seems to me that there
>>> is not an apparent weird behavior.
>>>
>>> should I have to continue to increase the dt with small increments
>>> (i.e. 0.0002ps at each run) or can I just trust to the results I have
>>> (angles and bonds distributions) using the dt=0.00183 ps??
>>> could the problem of the lincs warning arise from the generation of
>>> the topology with ATB??
>>>
>>> thank you in advance
>>> Emiliano
>>>
>
>
>

-- 
Dr Thomas Piggot
Visiting Fellow
University of Southampton, UK.