AMBER Archive (2007)

Subject: Re: AMBER: RE: FF94 or FF99?

From: Jiri Sponer (sponer_at_ncbr.chemi.muni.cz)
Date: Sat Nov 24 2007 - 04:53:24 CST


Hi Ilyas,

It can be a very complex issue. You probably should test
parmbsc0 Perez et al Biophys. J 2007 if the behavior is ff-
dependent and definitely longer longer longer simulations.
However, to get the ultimate conclusion is very difficult and
even the NMR experiment is not necessarily a solution. I
will describe two benchmark cases where extensive x-ray,
NMR and MD data are available, to illustrate the
complexity.

I. HIV-1 DIS kissing loop complex and its four unpaired
(flanking) bases that are highly conserved and thus bio-
important.
- First X-ray structure by Ennifar and Dumas (Nat. Struct.
Biol. 2001 I guess) shows bulged out unpaired bases
adopting a certain geometry (called open).
- a preceding NMR structure showing bulged in bases
instead, and quite different overall topology.
- MD (Reblova, NAR 2003) predicted bulged out geometry,
but different from X-ray (called closed geometry). The NMR
structure was unstable in the sims.
- There were several other short MD studies that were
mutually contradicting (scale 1-15 ns).
- Then Ennifar solved a new X-ray study (JMB 2005 I
guess) and in this particular case fully confirmed the MD
AMBER prediction (closed bulge out). The reason was that the new
X-ray avoided crystal packing in that region of interest.
-Then, two additional NMR studies appeared, both again in
disagreement with X-ray and showing bulge-in flanking
bases. At the same time, the three available NMR studies are
mutually inconsistent regarding the bulge-in topology, but at
least the global topology of the new NMR structures now
agrees with X-ray and MD.
- Trying to solve it we did close to 1 microsecond total MD
study (AMBER and preliminary CHARMM) including long
LES runs. We suggest that the X-ray new geometry is the
prevalent one in long sims one but we see also multiple
competing bulge in substates, what possibly could link the
X-ray and NMR experiment. All is summarized in BJ paper
Reblova et al available on Biophys. J. webpage.
-So, even after years of effort the situation is far from being
clear though I assume that the new X-ray topology predicted also by
amber is a very significant substate (there are biochemical reasons for that).
In addition sims provide path transititions, set of
substates, delocalised ion binding sites not visible to
experiment, etc.

Case II which is much more clear.
- Diagonal four-thymine loops of dG4T4G4 quadruplex.
- Originally, there was in 1992 Science X-ray study. To
shorten the story, that study is incorrect (I assume method
negligence) and was misleading the field for years.
- Simultaneously, Feigon's lab solved it by NMR and they
(again to shorten) have got perfectly the correct topology,
one of the reasons is that the compact G-DNA is very
suitable for NMR. Beautiful experiment.
- Finally (excellent work mainly by Neidle's lab), there are
several independent X-ray studies, different packing, ions,
even (another lab) protein-DNA complex, and the loop are
as predicted by Feigon. Clearly one topology irrespective of
x-ray, packing, solution, ions.
- Unfortunately, (Fadrna et al BJ 2004) we had to conclude
that using simulations, we are getting wrong topologies as
global minima.
- I personally consider single-stranded hairpin loops
(both DNA and RNA) as the most problematic system
for force fields, where the ffs usually do not work,
despite that for many other systems they are VERY
useful and in certain applications MD is superior to
other methods. But I agree with you ffs also can fail.
And these things are not random, they stem from the
overall balance of forces in a given system and their
interplay with the approximations.

So, to really scrutinize a given system needs usually
enormous work, and all methods have limitations.

Best wishes, Jiri
http://www.ibp.cz/labs/LSDNA/
>
> It is the free energy predictions on 2 systems I have been working on. One
> part of the thermodynamic cycle is the single strand part. I have started
> with 10 conformers and done 1 ns simulation for each of them. The TI
> prediction is way off from the experimental predictions.
>
> Then, I have only used the A-form conformations with backbone
> dihedrals restrained to the initial structures. For both 2 systems, TI
> predictions are pretty good with the experimental results (free energy
> results).
>
> Is 1 ns long enough? Of course it is not. But when the structures are
> started with A-form without restraints, it deforms itself to some other
> conformations. I did not run the sim.s for a long time (let's say 100
> ns) to see how much time is spends on the preferred conformations (which I
> cannot do in TI Approach as it is too long to do).
>
> I do not want to go into detail of these systems, but the current amber
> force field does not look like working good on single strands (at least
> for tetramers). Now, the question is, do these tetramer single strand RNAs
> have A-form conformations? Previous works say that single strands have
> some prefered conformations which look like A-form. I will do some single
> strand tetramer NMR work to see if the conformations look like A-form.
>
> My understanding of force field development is that it tries to predict
> the helical structures as good as it is seen experimentally, which means
> that the predictions are good for helices.
>
> Best regards,
>
> On Fri, 23 Nov 2007, David A. Case wrote:
>
> > On Thu, Nov 22, 2007, Ilyas Yildirim wrote:
> >
> > > Just wanted to comment on the current force field's single strand
> > > predictions. We have done some explicit solvent simulations on single
> > > strand RNAs and saw that the current force field (amber99) is not doing
> > > any good predictions.
> >
> > Just curious: what are the bad predictions? How do you know what the right
> > answer is, and are you sure that your calculations are converged?
> >
> > ....dac
> >
> > -----------------------------------------------------------------------
> > The AMBER Mail Reflector
> > To post, send mail to amber_at_scripps.edu
> > To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
> >
> >
>
> --
> Ilyas Yildirim
> ---------------------------------------------------------------
> = Department of Chemistry - =
> = University of Rochester - =
> = Rochester, NY 14627-0216 - Ph.:(585) 275 67 66 (Office) =
> = http://www.pas.rochester.edu/~yildirim/ =
> ---------------------------------------------------------------
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu