[ccp4bb]: Re:

[Randy Read <rjr27@cam.ac.uk>]

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

On Wednesday 30 April 2003 06:03, N. Gautham wrote:
> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
> Hello all,
>
>   This is a general MR question. Can some one enlighten me on the
> relationship, if any, between the resolution of the data and the
> accuracy of the model used in MR. For example, if I use only low
> resolution data say 6.0 A to calculate the rotn. fn., etc, can I get
> results by using a model that has rmsd of say 6.0 A from the 'true'
> structure? Any references, comments would be welcome. Thanks.

Beast uses an estimate of the rms error of the model to calibrate the 
likelihood function, one effect of which is to weight down the influence of 
high resolution data.  Back of the envelope calculations suggest that data 
with dmin less than twice the expected rms error will have relatively little 
influence.  This is supported by results from some test cases, so the default 
resolution limit in Beast is now set as twice the rms error for the best 
model.

In principle, if the likelihood target is properly calibrated, the main effect 
of using data to too high resolution will simply be to waste computer time.  
However, if one is using a program that doesn't apply any weights that depend 
on model quality, it would be best to apply a resolution cutoff (which you 
can think of as a crude form of weighting with weights suddenly dropping from 
1 to 0).

In your case, a model with 6A rms errors will predict data poorly beyond about 
12A resolution.  At 6A resolution, atoms with 6A rms errors will contribute 
with essentially unknown phase angles because they can be virtually anywhere 
between Bragg planes.

-- 

Randy J. Read
Department of Haematology, University of Cambridge
Cambridge Institute for Medical Research      Tel: + 44 1223 336500
Wellcome Trust/MRC Building                   Fax: + 44 1223 336827
Hills Road                                    E-mail: rjr27@cam.ac.uk
Cambridge CB2 2XY, U.K.                       www-structmed.cimr.cam.ac.uk