parameter sets

Gerard 'CD' Kleywegt GERARD at XRAY.BMC.UU.SE
Fri Nov 24 13:58:21 EST 1995


From:	XRAY::GERARD       "Gerard 'CD' Kleywegt" 24-NOV-1995 16:55:43.86
To:	SMTP%"breeze at zeneca-ph.co.uk"
CC:	GERARD
Subj:	RE: parameter sets

As Oleg Jardetzky always points out, there is a difference
between accuracy and precision.  Since the accuracy of
NMR structures is probably > 0.5 A for main chain and
> 1 A for side chain atoms, I wouldn't worry too much.
I would guess that errors don't build up in one direction,
but more or less compensate each other (you could find out
by refining using a ridiculous target value for the tau
angle (N-CA-C), for instance, and checking if the overall
shape and fold of your protein is retained).  As for
NOE violations, how accurate are those ?  People who don't
fit buildup curves used to classify NOEs as short, medium
or long (fuzzy science ? ;-) when I was young (not sure
if they still do that, having joined "the other side").
To settle the issue of the force field, you could refine
a structure against the nmr and xray (plus Hs) force fields
and see which one gives the best model (don't use procheck,
since that is heavily biased towards anything refined
against engh & & huber; also don't use noe violations
to assess quality since your refinement aims to minimise
those (just as bad as waving a low R factor around in
xtallography) - instead, use the ramachandran plot and
vriend & sander's directional atomic contact analysis;
these check things not restrained in refinement and
are therefore good validation criteria; you could also
use the nmr-equivalent of the free R factor although
i would be cautious (in xray, one observation contains
information about the entire structure, in nmr it
contains highly local information).

Just a few oere's worth,

--Gerard

Opinions expressed here are not necessarily those
of Pope Kurt ... eh John Paul, I mean ;-)



More information about the X-plor mailing list