Dear All,
> RMSD of ensemble of NMR structures can be interpreted as the
> uncertainty of coordinates. Core residues of NMR structures have
> smaller RMSDs in a similar way that core residues of crystal
> structures have relatively smaller B-factors compared to surface
> residues.
The problem with the ensemble RMSD in NMR is that it depends on the
refinement protocol. It is very important that the ensemble is refined to
the highest possible RMSD while still fitting the experimental data
(typically the distance restrants from NOEs and torsion angle restraints
from J-couplings). Unfortunately, people are still looking at the RMSD as
a measure of model quality. That is similar to looking only at the X-ray
resolution to assess the quality of a structure model.
> Both x-ray and NMR approaches can reveal "atomic resolution
> structures" but precision and quality of structures should be varying
> depending on individual examples.
That is why one should use validation scores based on a set of
measurements not used in the refinement like R-free (X-ray) or residual
dipolar couplings (NMR). But even then one should always look at the
structure model in the context of the experimental data. High
resolution/low R-free X-ray structure models can still have serious
problems. There are loads of examples in the EDS (especially around
ligands).
Cheers,
Robbie Joosten
Centre for Molecular and Biomolecular Informatics
|