AMBER Archive (2008)

Subject: RE: AMBER: RESP GAUSSIAN AMBER

From: Piotr Cieplak (cieplak_at_cgl.ucsf.edu)
Date: Fri Dec 12 2008 - 14:12:20 CST


Hi Ross
Yes, you are right, I am the villain behind all of that.
I think I could evenetually come up with some sort of tutorial.
Different story is with nucleic acids and different with amino acids.
As I remember NME and ACE comes from a subset of AA fitting and is used
afterwards on for the rest of cases, but should find my old notes to be
more clear about that. The AA fitting for ff94 was done by Wendy Cornell,
and NA - done by me. One of the problem is that I don't have the original
espots from that fittings. Most probably I could have the original files
with espots on some old dat4 tapes (the smallest ones) that are not easily
readable anywhere now (I mean data for NA, I didn't have data for AA (e.g.
those used by Wendy)). But I should have inputs for polarizable force
field (ff02) instead. I did fitting for both AA and NA in this case.

As far as R.E.D. developments concerns - the approach behind is to
a) make fitting process that is easier (?), and more automatic, b) get the
charges that are reproducible. Reproducible - means that special
orientation of the molecule is forced in QM calculations, that can be
reproduced by anyone and anywhere, and/or forcing multiorientation
(i.e. use the same optimized geometry, but reorient the molecule many
times in space, in order to average out those "small" inconsistencies in
generating grids for electrostatic potential calculations).
In original ff94 and ff02 we did fitting based on single orientation of
the molecules in space in QM calculation, and because of that the
charges could probably not be fully reproducible, although my
feeling is that any orientation used should produce close to each other
set of charges. Unfortunately, additional constraints in RESP
(multiconformational or multimolecular) are modifying charges more than
that "orientation" problem.

Piotr Cieplak

On Fri, 12 Dec 2008, Ross Walker wrote:

> From: Ross Walker <ross_at_rosswalker.co.uk>
> Sender: owner-amber_at_scripps.edu
> To: amber_at_scripps.edu
> Date: Fri, 12 Dec 2008 11:04:49 -0800
> Subject: RE: AMBER: RESP GAUSSIAN AMBER
> Reply-To: amber_at_scripps.edu
>
> Hi Francois,
>
>> The link is @
>> http://www.teokem.lu.se/~ulf/Methods/resp.html#How_to_determine_RESP
>>
>> This means the tutorial @
>> http://ambermd.org/tutorials/advanced/tutorial1/section1.htm
>> demonstrates the way used by Pr Ryde's lab to derive RESP charge, &
>> not that used in the building of the AMBER force field topology
>> database.
>
> Right, but the tutorial shows how to build the charges that Prof. Krueger and his students used for their research into Fluorescein dye's attached to DNA's. At no point is the tutorial actually attempting to show how the AMBER force field RESP charges were derived.
>
> Note the very first paragraph of the tutorial says:
>
> "In this tutorial we will work through an advanced scenario for preparing a system, for simulation with sander, that contains several non-standard residues where the use of Antechamber and the GAFF force field may not be appropriate (see http://www.ambermd.org/tutorials/basic/tutorial4/index.htm for an example of using Antechamber). To illustrate this process we will choose a real world research problem rather than an 'artificially' simple example. This of course comes with several caveats. While it provides a more realistic tutorial it does of course present issues inherent in any research project. That is that there are often multiple ways to accomplish something and one does not necessarily know in advance the best option to choose. This is true here in the case of deriving certain RESP charges as you will see later on. However, this approach should provide you with a solid foundation for preparing your own systems. You just need to be aware that when things get complex!
  t!
> he!
> re is no such thing as a step by step guide and therefore you should use this tutorial merely as a basis for approaching a similar problem. However, substantial thought and literature searching will likely be required before you tackle your own system."
>
> In terms of providing a tutorial on how the AMBER force fields were originally derived I agree that this is indeed desperately needed, if only to provide a definitive document somewhere of how it was actually done. I encourage someone (probably someone like Piotr Cieplak who actually did the original calculations) to write their own tutorial / howto for doing this. Similarly I would encourage you to contribute to the tutorials as well. Send me a username and ssh key and I can give you access to the amber_web CVS repository.
>
>> In your tutorial (dedicated to charge derivation),
>
> There is no "dedicated to charge derivation" in this tutorial. The tutorial is about setting up an advanced system for simulation. Charge derivation forms a small part of this.
>
>> it is
>> important to explain this choice, but also the reason of this choice.
>> Thus, new users will understand which options to use, & will have the
>> opportunity to tests the difference cases. I think this would be
>> really less confusing...
>
> I will update the tutorial to make it clearer that these choices were made for this particular research project and should not be considered Gospel. I'll add some text suggesting that the literature be consulted and also include some links to RED etc.
>
>>> me that, for example, upping the number of points per layer and the
>> number
>>
>>> of layers should simply result in the charges generated converging to a
>>
>>> specific value as a function of the number of points used. If one gets
>>
>>> different charges at the default settings then the default options are
>>
>>> surely not large enough to obtain convergence in the results and will
>> just
>>
>>> lead to large error bars.
>>
>>
>> May be you should do some tests here, & explain this point in your
>> tutorial as the use of a fine grid for organic molecules is reported.
>
> Sure, if my job was to write tutorials!!!, I wish I had time to do this sort of thing I really do but I am already overloaded just with my real job + research etc. By all means if you think such a thing would be beneficial please feel free to do it and modify the tutorial as necessary.
>
> As it stands right now these tutorials are here, freely donated, to provide some examples of how one might setup and run simulations. They are not in any way "The Gospel" although unfortunately many people read them like that. :'(.
>
>>> what is going on and avoid use of any black box approaches as much as
>>
>>> possible. They don't attempt to discuss the complexities of MD
>> simulations
>>
>>> which make MD such a dynamic area of research.
>>
>>
>> Yes I agree, although convenient ;-)
>
> Right although the reason it is "convenient" is because I don't have hours per day to devote to writing such tutorials, even if I did get them perfect for one person you can guarantee they will just upset someone else which is why I have given up trying to get them that way and instead just have them as a way of lowering the energy barrier for someone entering the field of MD and using AMBER. These are not a comprehensive peer reviewed discussion of the field hence I don't attempt to make any guarantee of their accuracy or applicability. As I say feel free to update them and add new ones. I would hope that most people (at least those wanting to go into science) would be sufficiently curious to question everything. The idea of being spoon fed how to do something and simply believing that that is the way should have vanished when graduating high school.
>
>> A new user generally follows what is written (if there is no clear
>> explanation) because she/he gets lost between all the possible
>> options. This tutorial is about "Charge derivation",
>
> No. The purpose of the tutorial is "This tutorial centers on creating a prmtop and inpcrd file for a fluorescein dye bound to a poly-AT DNA decamer."
>
>> and as it is a
>> complex topic it is crucial to well define the conditions in which the
>> computations are done. This is why in a R.E.DD.B. project all the
>> computation conditions are stored. See for instance the W-46 project @
>> http://q4md-forcefieldtools.org/REDDB/up/W-46/ & the "Information
>> regarding Quantum Calculations" section.
>
> Exactly!, there is no clear explanation of what should be done hence why it was done this way in the tutorial. The simplest solution would be for the AMBER website to contain a clear tutorial covering charge derivation that introduces things from the beginning, highlights how it was done for the amber force fields and then covers what is the currently accepted practice for determining charges for new residues within each of the AMBER force field frameworks. Feel free to put one together and you can upload it to the amber_web cvs.
>
>> I have a last comment about this tutorial; the worst one:
>> Usually, one uses (an) intra-molecular charge constraint(s) for charge
>> value(s) for the atom(s) you _keep_ in the target fragment; NOT for
>> atoms that are removed. The tutorial reports the opposite (I think):
>> This arbitrary increases the RRMS for nothing. Usually, one also uses
>> a global charge constraint for the chemical group to be removed, and
>> not individual constraints otherwise one might observe artifacts in
>> the fit and/or an increase of the RRMS value.
>> http://pubs.acs.org/doi/abs/10.1021/ja964372c is one interesting
>> reference illustrating this point...
>
> And here we see why there is such confusion. As you can see I am not an expert in RESP fits, you are, and yet most of this information is obfuscated away in obscure locations where people would not be able to find it. For example my understanding was that the charges for ACE and NME were derived and then they were fixed at these values while each capped residue's charges were derived. However you seem to be suggesting that this should not be the case and ACE and NME when capping should just be locked to have zero charge. Is that true?
>
> The analogous procedure to fixing the ACE and NME charges was used for the rest of the charge fit here but you seem to suggest that the opposite it what should be done. That makes sense I guess if my understanding of the ACE and NME approach is also wrong.
>
> Feel free to update this tutorial to reflect that.
>
>> Whatever the choices made - as I already said there is not a unique
>> answer, explaining those choices in a tutorial dedicated to charge
>> derivation is suitable - in particular if one does not follow the
>> standard procedure.
>
> I would love to see an unambiguous document describing "standard procedure"...
>
> In the meantime I will add further text to the tutorial to make it clear that the choices used here are for the purposes of demonstrating how RESP works and NOT how one should always do charge fits. You have a suggested link (preferably something that is clearly related to AMBER) that people should look at to see how 'standard' calculations should be done?
>
>> Yes, you are right this is difficult to not be lost. However, once
>> again, in the context of writing a tutorial, choices made in that
>> tutorial should be clearly established to guide new users in all the
>> possible options.
>
> Yes well the choices made were based on my, and the students understanding of how RESP fits are done. Clearly these are NOT correct but what this clearly shows is that there is no established guide for how this should be done hence I encourage you to update the tutorial such that it reflects this...
>
>> If you use R.E.D. the charge reproducibility achieved is +/-.0001
>> whatever the method used is, whatever the QM program used, & whatever
>> the initial structure is. Please, use R.E.D. & see. The charge values
>> generated using R.E.D. & available in R.E.DD.B. are highly
>> reproducible: max diff = 0.0001!
>
> So it would be great if this tutorial could be extended to highlight this - i.e. show a RESP fit and then show the same thing being done with R.E.D. This will have to be done in a way that doesn't just confuse people even more though. Plus it won't necessarily be clear to people why RED is needed here but isn't part of AMBER. Note the KEY to this tutorial is showing how to get from a few pdb's of a non-standard residue to a prmtop and inpcrd file so simulations can be run. Any tutorials within this on charge derivation etc should be part of that overall procedure...
>
>> Concerning charge derivation, I think you have more than that @
>> http://q4md-forcefieldtools.org/Tutorial/ - taking the
>> dimethyl-alanine dipeptide as an example. You will find answers about
>> the impact of the molecular orientation/conformation/fitting option on
>> charge values with full R.E.D. jobs/data to be able to re-do the jobs.
>> Charge derivations are available for whole molecules but also for
>> different molecular fragments. The choices of using specific charge
>> equivalencing approaches and specific charge constraints are also
>> described, the goal being not just "how to get charges" but "how to
>> get charges with the lowest RRMS value".
>
> I guess my question would be why one of these tutorials hasn't been incorporated into the main set of amber tutorials on ambermd.org - as part of any actual 'workflow' to get from pdb to prmtop/inpcrd? New users are far more likely to see this sort of thing if it is in with a set of tutorials that they work through rather than off elsewhere on a website that they may not realize is there (even with links from the ones on the ambermd.org page). I.e. something simple with a link saying "there is more comprehensive examples here".
>
> All the best
> Ross
>
>
> /\
> \/
> |\oss Walker
>
> | Assistant Research Professor |
> | San Diego Supercomputer Center |
> | Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
> | http://www.rosswalker.co.uk | PGP Key available on request |
>
> Note: Electronic Mail is not secure, has no guarantee of delivery, may not be read every day, and should not be used for urgent or sensitive issues.
>
>
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
> to majordomo_at_scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
      to majordomo_at_scripps.edu