AMBER Archive (2008)
Subject: Re: AMBER: RESP GAUSSIAN AMBER
From: Brent Krueger (kruegerb_at_hope.edu)
Date: Sun Dec 14 2008 - 20:52:14 CST
My apologies for getting involved in this discussion late. Hopefully things
are ready to wind down on this and my message here won't reignite anything.
First, I'd like to thank Francois for finding that the charges in the
RESP/example directory are out-of-date and actually answering the original
question posted by Silas.
Second, I won't address everything that has come up in the last few days,
but let me give some thoughts on a few of them.
A few notes on the IOps, since they generated quite a bit of discussion:
To my knowledge, none of those references that are in our tutorial, nor the
ones you listed in this discussion actually state what IOp values should be
used. Chris's JPC paper does describe the ESP as having 4 layers of points
and 1 point/A^2, though the Gaussian page doesn't help one know what 6/42
value would actually yield 1 point/A^2. It may even depend on the size of
the system. I apologize that I've not thoroughly investigated the Gaussian
6/42 option. Oh, and the default for 6/41 is 4 according to the Gaussian
page, though I have seen incorrect default values given there before.
However, as Ross says, what we've done here is err on the side of a much
more information-rich ESP. Of course, having two ESP's that are different
in information density could lead to slightly different sets of final RESP
charges. However, if this were the case, my answer would be that you've not
converged your results and, therefore, the less-dense one is giving a poor
result. The next step then would be to refit using an ESP of even higher
density until the resulting RESP charges don't change. Since we are using a
pretty high density here, I think we're being quite safe and that we are, in
fact, being completely consistent with AMBER RESP charges. Have you seen
cases that behave otherwise?
Geometry for generating ESP:
I'm quite curious to know if anyone has solid information about what the
proper way to do the geometry is. I've always believed that the geometries
were determined at the HF/6-31G* level, but Ross pointed out to me some time
ago that it never actually says that in any of the papers from that era.
When I was in Peter's group we were moving on to the next sets of force
fields, and so this detail didn't surface often. Probably I heard it in a
group meeting once, but it didn't get locked in. Anyway, you and I have
also discussed this and decided that we should use a method that we felt
gave a reliable geometry at a reasonable cost for the particular molecule we
were considering. I'm quite happy to go with this on into the future, but
if anyone who is still bothering to read this particular discussion would
like to give some history, I'd love to know the historically correct answer.
Charge constraints in RESP:
Thank you for pointing this out. This is a subtlety that I wasn't aware of
and is something we probably should be mentioning in the tutorial. In this
particular case, we did begin the way you describe, fixing the charge of the
desired atoms to the correct total (well, actually we fixed all the other
atoms to the correct total, but that is the same number of constraints, so
that should be the same). However, we ended up getting charges for the
'connecting' atoms of ACE/NME that were very different than those given in
the Cornell FF94 paper. Would this not lead to artifacts when our
'connecting' atom was joined with something already in the AMBER libraries
with a charge that is consistent with the Cornell FF94 paper? I'm sure it
would. We wanted to avoid this, which is why we changed our procedure,
thinking that we would be more consistent with the existing AMBER residues.
Whether we chose poorly or not, clearly how one specifies the charges has a
much larger effect in this case than any issue about the number of
constraints. Anyway, I'd love to have your thoughts about how to best make
our residue charge consistent with AMBER in this respect.
Is it a charge derivation tutorial:
Of course you are correct that we are covering charge derivation in this
tutorial. However, it is equally correct to say that charge derivation is
just one part of the complete tutorial. If we tried to go into every detail
about every step in the procedure, we'd never have finished the thing. More
importantly, no one would ever attempt to wade through. However, we should
have referenced some more authoritative sites and we will make that change
(if Ross hasn't already).
You are also correct that more pairs of eyes are always better for a
tutorial like this. I would invite anyone to please give us feedback about
how things could be improved--especially Piotr and yourself. (Silas, as you
are obviously going through this carefully, so this means you, too.) Ross
and I started working on this because we have had to go through this
procedure many times ourselves and we wanted to document what we've done
because students in our groups will likely be doing it again soon. We
absolutely want this tutorial to be as complete and accurate as possible.
Finally toward your comment that you don't believe a student did the work.
In fact two students did all of the work contained in that tutorial. In
fact, two undergraduate students did all of that work. Of course, Ross and
I offered guidance and advice along the way; no one works on anything
entirely alone. But absolutely 100 % of the commands were typed and run by
the students. The impetus for the tutorial was that we had a research
project that required us to derive charges for fluorescent dyes as in the
tutorial. It fell on these students to do that work. Since Ross and I have
banged our heads against this a number of times, we had them create a web
page along the way describing what they did.
Again, apologies for any inconsistencies that may be present in the
tutorial. Ross and I would love to get specific feedback about how the
tutorial could be improved.
On Sat, Dec 13, 2008 at 2:23 AM, FyD <fyd_at_q4md-forcefieldtools.org> wrote:
> Hi Ross,
> - If You want a reference for 'standard' charge derivation, you have it in
> your own tutorial ;-)
> Cieplak, P., Cornell, W.D., Bayly, C., Kollman, P.A., J.Comp.Chem., 1995,
> 16(11), 1357-1377 (you even provide the corresponding PDF - even if you do
> not have the right to do it). To the best of my knowledge nobody did better,
> so far. Charge derivation for standard residues is well explained is this
> paper, basis set, constraints etc; although once again it takes time to get
> into it as this is a huge/complex & beautiful work. Well, science is not
> simple ;-)
> - You are the last author of this tutorial; Brent is not. Moreover, the
> title of the 1st section of your tutorial is:
> "1) Charge Derivation for Non-standard residues"
> So please, stop hidding & playing with words, this is clearly written ...
> "Non-standard residues" not "Dye". This tutorial pretends to deal with
> charge derivation for Non-standard residues, or I should go back to the 1st
> grade ;-)
> - Concerning required tutorials for charge derivation at the AMBER web
> site, One might decide to update some web pages allowing the AMBER community
> to easily find the q4md-forcefieldtools.org tutorials (only about charge
> derivation & force field library building) @
> http://q4md-forcefieldtools.org/Tutorial/, and to find R.E.DD.B. which
> does not exist! I have no problems to help/participate/whatever_the_word_is,
> however, may be using what is already available might be a good start.
> - Between You & me, and about the fitting approach itself in this tutorial,
> I do NOT think a student did that by her/himself. Someone that knows well
> RESP has done that, I can tell you. Using the option iqopt=2 + preparing by
> hand the qout file (which is really awful BTW) in the 1st stage is very
> subtil (althought the type of constraint used is suspicious).
> - Finally, yes, writing tutorial takes time, you are right. With Piotr we
> have read & re-read our 2 tutorials @
> http://q4md-forcefieldtools.org/Tutorial/. Other people here have also
> read them to help. Personnally, I prefer releasing nothing than something I
> do not master, & let others more qualified explain what they know.
> regards, Francois
> 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
Brent P. Krueger.....................phone: 616 395 7629
Associate Professor................fax: 616 395 7118
Hope College..........................Schaap Hall 2120
Department of Chemistry
Holland, MI 49423
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)