AMBER Archive (2009)

Subject: [AMBER] follow-up to previous discussion: nmr refinement of dimer

From: Sally Pias (
Date: Tue Feb 24 2009 - 01:50:41 CST


I have a follow-up question related to the use of ambiguous restraints
in NMR refinement of a dimeric protein structure. In addition to
commonly occurring ambiguities, I have introduced an ambiguity for
every distance restraint, representing a scenario where it is possible
that a given restraint could involve two atoms/groups on a single
protomer OR one atom/group on the first protomer and one on the second
protomer. My question is whether it is appropriate to correct for
r**-6 averaging due to these "protomer-protomer" ambiguities. The
correction has already been applied to other ambiguities in the
distance restraints list.

Note that, in this particular dimeric model, there are a relatively
large number of residues making contacts with residues in the opposite

I hope I have given enough information for my question to be clear.

Thank you,
Sally Pias

On Thu, Feb 12, 2009 at 6:08 AM, David A. Case <> wrote:
> On Thu, Feb 12, 2009, Sally Pias wrote:
>> I have already defined every distance restraint as ambiguous, with
>> equivalent atoms in both protomers able to satisfy the restraint.
>> However, I do not understand clearly how group coordinate averaging is
>> used.  Does the coordinate averaged "atom" position actually fall
>> halfway between the two atoms defined in the ambiguity?
> No.  Assuming that ir6 is set (the default) the distance is the r**-6
> average of the two (or more) distances involved in the ambiguity. This
> is really the same as what Aria does, for example.  In this procedure, a
> short distance will be "active" and a longer distance will effective be
> ignored.
>> I am working with a homology-modeled starting structure.  After a few
>> rounds of a simple simulated annealing protocol (based on the manual
>> example and Tutorial A4), first in vacuum and then with GB solvent, I
>> have very few larger violations (between 0.5 and 1.5 A, possibly
>> indicating errors in the NOE list) but many smaller violations
>> (between 0.2 and 0.5 A).  I am not sure yet whether this is an effect
>> of working with a homology modeled starting structure or whether the
>> ambiguities may be introducing a structural distortion.  The
>> structures look reasonable after simulated annealing, but I am still
>> suspicious.
> It's hard to say much from afar.  Having lots of violations between 0.2
> and 0.5 Ang. is indeed suspicious, but you want to fix the assignment
> errors (if they are indeed present) first.  It might be worth setting
> ialtd=1 and re-running the refinement, to see what sort of difference
> that makes.
> ....dac
> _______________________________________________
> AMBER mailing list

AMBER mailing list