***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


On Thu, 31 Aug 2006, Vineet Gaur wrote:

> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
> 
> 
> hi all
> i m refining structure at 3.2A.The problem that i m facing is the
> difference between Rfree and Rcrys is decreasing. right now it is only
> 1.1%. i have tryed to improve the steriochemistry by looking at
> Ramachandran but no success. whenever i increase the no. of
> minimization steps, i always end up with an increase in Rfree.
> plz tell me what other options one can try to improve difference
> between Rfree and Rcrys.
> Thanx in advance
> vineet gaur

Hi,

You work at pretty low resolution, so that your problem should be
undetermined (or hardly determined). For "usual" refinement in cartesian
space, you require AT LEAST one measurement (2 preferred) for each of X,
Y, Z and B, for each atom. At 3.2 A you do not have that. Hence very
careful refinement strategies must be adopted (unless of course you have
90% v/v solvent in the unit cell).

Like rigid body refinement to start with,
simulated annealing in torsional angle space (not cartesian),
very very careful (and limited) energy minimisation, perhaps real space
geometry adjustment (in the electron density map), temperature factor
adjustment (and not refinement).

In any case you must use a starting model that is very well refined
already if the difference in R Rfree is only 1.1%. Else you are refining
an isomorphous structure (same space group) from a previous structure and
you haven't kept the same set of reflections flagged for the Rfree
calculations than those used in the "previous" structure.

I hope that I am clear. Anyway you do not give much information on the
type of problem that you are facing (what stage if refinement you are at,
the current statistics and so forth).

Just one final remark: the aim of refinement is not to minimize the
difference between R and Rfree, but to generate a model that best
agrees with all available data (be it X-ray data, biochemical knowledge,
previous knowledge about structural geometry...).

Fred.

-- 

s-mail: F.M.D. Vellieux (B.Sc., Ph.D.)
        Institut de Biologie Structurale J.-P. Ebel CEA CNRS UJF
        41 rue Jules Horowitz
        38027 Grenoble Cedex 01
        France
Tel:    (+33) (0) 438789605
Fax:    (+33) (0) 438785494
e-mail: [EMAIL PROTECTED]

Reply via email to