AMBER Archive (2004)

Subject: Re: AMBER: Test Calculation of atomic charges (RESP) for RNA bases

From: FyD (
Date: Fri Sep 10 2004 - 19:35:09 CDT

Dear Vlad,

> I have performed a test calculation using RED (Multi molecule - multi
> reorientation RESP fit) for the 4 canonical RNA bases: RG, RA, RC and
> RU. The starting point were canonical A-form nucleotides with C3'-endo
> sugar pucker. I used 3 reorientations. The fit was performed as
> described in the original paper of RESP, with all the sugar atoms but
> C1' and H1' intermolecular fitted.

I used a very similar approach but using C2'endo sugar pucker ;-)

> Presumably using this protool I hoped
> to reproduce the atomic charges on RN nucleotides that are currently
> used in amber.

Something wrong about the word 'reproducibility':

The RESP charges in the AMBER topology database are NOT 'reproducible' because
of mainly 2 reasons:

-1 The molecular orientation of the optimized structure is not unique for a
target minimum using GAMESS and/or Gaussian. Indeed, the internal
re-rorientation algorithm available in the 2 QM soft does not generate a
constant molecular orientation for a target minimum. This is the most important
problem. The charges are not reproducible because the molecular orientation is
not perfectly controlled in the QM soft (GAMESS or Gaussian) and different when
we compare it in the two QM software (GAMESS and Gaussian).

-2 The minimum accuracy slightly affect the charge values also (small effect).

Now if you decide to control the molecular orientation of the optimized
structure using the rigid-body re-orientation algorithm implemented in R.E.D.
the charges will be reproducible because the molecular orientation based on the
three atoms you selected are known. This means that your set of charges will be
reproducible by others IF you provide the molecular orientation information
based on three atoms you used. That's it.

Finally, you can use multi-orientation RESP fit to average the diff. in charge
observed for one orientation over several orientations.

> To my surprise there are striking differences sometimes
> (especially on purines). Some time ago Francois told me that acceptable
> differences are in the range of 0.07. I checked the input and I
> calculated the charges for single nucleotides and I concluded that there
> is nothing wrong with my inputs. I will list below the results on the
> base atoms using wonder marks where I found differences (I tolarated a
> difference up to 0.03-0.04).

Yes 0.040 is a VERY common difference... We observed up to 0.07 in one
particular case...

> Now, the question that I have is: Should we
> worry about these differences??? Has somebody of you ever tried to do
> this calculation? If yes ... are your data fitting mine or the Amber
> better?


I do not have the answer. I can only answer to you three things:
- The charge incertainty induced by the molecular orientation is very similar to
the charge difference between the 2 RESP stages
- Such differences are observed for C atom in CH3/CH2 groups but also for N and
O atoms.
- An hydrogen bond can be responsible of a charge difference of 0.2: 0.04
represent ~ 20% of this value...

> I am fighting in getting charges for some base analogues. I tried -CH3
> instead of sugar but it is clear that this is not a good approach
> because the differences between the charges from entire nucleodies
> differ significantly.

I think the limitation of the Me- approach is that a unique conformation can be
selected, in this case multi-orientation is an answer. But in same time it is
quicker. So it is up to the user...
My opinion is that if you can explain what you need with a set of charges used
during MD simulations, it is OK...

> My ideea was now to fit all these new bases with
> the existing RNA bases and that's why I performed this test
> calculations.

So you are going to mixte non-reproducible RESP charges with reproducible RESP

We started to generate a set of reproducible RESP charges applying
the Cieplak et al. approach and the R.E.D. multi-orientation/multi-conformation
RESP fit with R.E.D.-II. The R.E.D. II manual is more or less done. R.E.D. II
should be released very soon now...

> The problem with my analogues is that there is very litte
> experimental information them and their behaviour in RNA. They are
> fluorescent so they might be interesting for the future. I also dont
> have too much time to run MD simulations using these bases.
> I am wondering if there is answer to this questions and also I am
> wondering if these differences are observed...did someone ever thought
> of changing the charges in Amber??

We started to do it... Not finished yet...

> * For the calculation all nucleotides were optimized with G03 using
> standard gaussian inputs (as described in RED)

Regards, Francois

F.-Y. Dupradeau
Faculte de Pharmacie, UPJV, Amiens, France
The Scripps Research Institute, San Diego, USA
The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to