Date: Wed, 15 May 2002 10:09:27 -0700
From: David Case
Subject: Re: NMR refinement

On Tue, May 14, 2002, Jake Isaacs wrote:
>
> I'd like to get a feel for what you folks think to be the "best" protocol
> for structural refinement using NMR-based restraints, specifically for DNA
> systems. How do in vacuo, implicit solvent, and explicit solvent methods
> compare? How about conventional vs. time-averaged restraints? Restraints
> based directly on NOE volumes vs. calculated upper/lower bounds (as with
> MARDIGRAS)? How are your simulated annealing runs structured? How are
> residual dipolar coupling restraints implemented for DNA?
>

We have a paper in press in J. Biomol. NMR about using GB and explicit water
for refinements of proteins, and another paper (still in preparation) about
results with DNA. Briefly, either GB or explicit water is better than
vacuum refinements. For most purposes, GB is faster and easier to use, and
is what we recommend.

I have never (myself) used time-averaged restraints for DNA. Using
Mardigras (or equivalent) is a common practice for our DNA stuff. We have
done some work with explicit volumes, but that requires a lot of work (and
very good NMR data) to make much difference.

We generally don't heat up DNA to as high a temp. as we do for proteins,
since it can be difficult to return to good structures as one cools (and
since, for duplex DNA, the starting structure will already probably be
pretty close to the final one.) Try annealing schedule that heats to 600K,
then cools.

I should empahsize that there is no "canned" approach; we hope people will
use the Amber tools to explore a variety of options.

...good luck...dac