[gmx-users] Manual refinement of ATB topologies ?

Tue Jan 17 15:10:10 CET 2017

Hi there,

Thanks for the reply ! (and sorry for the delay...)

2016-12-23 19:25 GMT+01:00 Justin Lemkul <jalemkul at vt.edu>:

>
>
> On 12/23/16 3:20 AM, Sim gmx wrote:
>
>> 2016-12-22 14:40 GMT+01:00 Justin Lemkul <jalemkul at vt.edu>:
>>
>>
>>>
>>> On 12/22/16 8:25 AM, Sim gmx wrote:
>>>
>>> Hello,
>>>>
>>>> 2016-12-22 13:56 GMT+01:00 Justin Lemkul <jalemkul at vt.edu>:
>>>>
>>>>
>>>>
>>>>> On 12/22/16 4:28 AM, Sim gmx wrote:
>>>>>
>>>>> Hello,
>>>>>
>>>>>>
>>>>>> Thank you very much for your help!
>>>>>> Good reference indeed! Just to be sure:
>>>>>> - ATB gave me HC and C atom types instead of 'CR1' for the CH groups
>>>>>> involved in double bonds. Would it be right to merge the two atomtypes
>>>>>> (i.e. deleting the HC atoms and changing C atoms to CR1) and add up
>>>>>> their
>>>>>> respective charges to get the charge of each CR1 atoms ?
>>>>>>
>>>>>>
>>>>>> Why do this?  Few GROMOS species even use CR1 any more, as it appears
>>>>> to
>>>>> be a backwards compatibility with old GROMOS parameter sets that used a
>>>>> UA
>>>>> aromatic type.  Phe, Tyr, etc. use C-HC as these are somewhat "polar"
>>>>> C-H
>>>>> bonds so the H is represented explicitly.  I'd leave the parameters
>>>>> alone,
>>>>> mostly because you need the H and associated parameters there to
>>>>> determine
>>>>> the geometry of the double bond.
>>>>>
>>>>>
>>>>> Because in the itp files from the suggested paper above (Poger et al.)
>>>> they
>>>> used CR1 atomtypes for the CH groups involved in double bonds. I would
>>>> be
>>>> a
>>>> bit afraid if I had to justify in a reviewing process why I took their
>>>> parameters for the double bond but did not use the same atom typing, I
>>>> might be a bit paranoid though.
>>>>
>>>>
>>> I just remain skeptical that such a representation is physically valid.
>>> Just representing a double bond as an uncharged entity in the middle of
>>> an
>>> uncharged chain, while consistent with the force field's underlying
>>> theory,
>>> seems quite inadequate to me.  If ATB is suggesting an alternate, it
>>> should
>>> be explored.  I think this is polar character that is not accounted for
>>> in
>>> the simplistic uncharged model.  To what extent that affects the dynamics
>>> needs to be examined.
>>>
>>>
>> I see your point. Indeed it looks a bit strange, especially when comparing
>> with aromatic structures from amino acids. However, such uncharged double
>> bonds are also seen in the widely used lipid parameters from Peter
>> Tieleman
>> for Berger Lipids (even though the forcefields are different, I guess this
>> comparison is acceptable). It does not mean that it is perfect, but
>> hopefully that it is usable.
>>
>>
>>
>>> If I define an improper dihedral with a torsion angle of 180°, wouldn't
>>> it
>>>
>>>> be a trans double bond anyway (no matter the presence or absence of the
>>>> H
>>>> atoms) ?
>>>>
>>>>
>>> Impropers keep planar groups planar; they do not keep bonds in cis or
>>> trans orientation.
>>>
>>>
>> OK. But then I don't get the point of creating a gi_4 in gromos54a7 that
>> is:
>> #define gi_4         180.0   167.42309
>> ; planar groups 40
>>
>> while there is already a gi_1 that is
>>
>> #define gi_1           0.0   167.42309
>> ; planar groups 40
>>
>> Am I wrong if I write that the only difference is the ideal angle that
>> shifts from 0 to 180° ?
>> The use of gi_1 or gi_4 is the only difference I've noticed from the itp
>> files of Poger et al. cited above between their trans and cis double
>> bonds.
>>
>>
> The convention of impropers depends on how the atom order is given.  Both
> are saying "keep the central atom planar" but perhaps the other atoms are
> listed in a different order.
>
> I don't get this one. In the paper, they parametrized two lipids with the
same sequence CH - CI = CJ - CK, but for one the double bond is cis, and
for the other one it's a trans double bond. For the cis, they used a gi_1
for the sequence CH CI CJ CK
and for the trans, they used a gi_4 for the same sequence (same order) CH
CI CJ CK.

If I correctly understand, one could have used a gi_1 for the trans double
bond as well, by changing the order of atoms, instead of using the gi_4.
But since they used the same order for gi_1 and gi_4, defining cis and
trans DB, respectively, doesn't it mean that the difference only comes from
the shift from 0 to 180 degrees between gi_1 and gi_4 ?

>
>>
>>> Do you think I should leave the HC atoms alone and include 2 improper
>>>
>>>> dihedrals for each double bond ? One for H-C=C-C and the other one for
>>>> C-C=C-H (both with a torsion angle of 0, resulting in a trans double
>>>> bond) ?
>>>>
>>>>
>>>> If it's me doing it, I'd test both topologies.  I've done simulations
>>> with, POPC and a simple, uncharged model and didn't notice anything out
>>> of
>>> the ordinary but I've grown skeptical.  You have multiple double bonds in
>>> your compound, separated by only one bond, and their properties may
>>> depend
>>> on a more precise electrostatic representation.
>>>
>>> -Justin
>>>
>>>
>>> To be sure that we are talking about the same thing, "both topologies"
>> means the double bond parameters as they exactly come from ATB server
>> *and*
>> the double bond parameters modified according to Poger et al., right ?
>>
>
> Yes.
>
> How would you do then to select the "best one" ? Except if they exhibit
>> totally different behavior (which is unlikely, I guess), it seems quite
>> complicated to establish. Still I will give it a try! It is true that this
>> proximity between the two double bonds could have a marked effect on
>> electrostatics.
>>
>>
> Understanding whether or not your force field/topology is right requires
> target data.  If you don't have target data, you can't know if your model
> is right.  If you can't know that, you can't know if your results have any
> physical meaning. This is one of the problems with GROMOS parametrization;
> it's highly empirical which means it can be hard to acquire reference
> data.  Other force fields like AMBER, CHARMM, and OPLS have clear
> connections with QM data that make for easy initial checks for dipole
> moments, water interactions, conformational energy scans, etc.  ATB does
> some QM-based optimization, but unless you know what you're explicitly
> targeting, it only does you so much good.
>
> I can get target data, but I'm afraid it will not be precise enough to
allow a clear distinction between those two topologies, except if they lead
to clearly different behaviors, which is not very likely to happen I guess.
I had a first look at CGenFF, validation of the parameters seems really
hard (some parameters have a penalty score between 10 and 50). I've never
done things such as QM calculations, and the tutorial seems to lack details
for me. Without any referee in my lab for such validation, do you think it
would be doable in a reasonable amount of time with sufficient accuracy ?

> -Justin
>
> --
> ==================================================
>
> Justin A. Lemkul, Ph.D.
> Ruth L. Kirschstein NRSA Postdoctoral Fellow
>
> Department of Pharmaceutical Sciences
> School of Pharmacy
> Health Sciences Facility II, Room 629
> University of Maryland, Baltimore
> 20 Penn St.
> Baltimore, MD 21201
>
> jalemkul at outerbanks.umaryland.edu | (410) 706-7441
> http://mackerell.umaryland.edu/~jalemkul
>
> ==================================================
> --
> Gromacs Users mailing list
>
> * Please search the archive at http://www.gromacs.org/Support
> /Mailing_Lists/GMX-Users_List before posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>