[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ccp4bb]: Summary: REFMAC and resolution limits

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

 > is there a simple possibility to improve the resolution limits in
 > steps from one refinement cycle to the next using REFMAC (something
 > the STIR instruction in SHELX)?

Kevin Cowtan:
Such an approach should not be required in a maximum likelihood
refinement package.

In least squares refinement, the high resolution terms are a handicap
with a poor starting model becuase they vary much faster with small
shifts to the model.

However, in ML refinement, the level of error in itself is a parameter
which is determined and refined, and acts to weight down the high
resolution terms in proportion to the poorness of the model. No
resolution extension scheme is required, because the likelihood
already provides a better scheme than any which could be determined by
the user.

Roberto A. Steiner:
as far as I know no-STIR like command is available and I don't think
it's necessary.
Just go on in steps with the command files and that's just fine


 > And is there a high resolution limit for the refinement with REFMAC
 > should one use other programs - like SHELX or CNS - for data better
 > let's say 1.1 A?

Kevin Cowtan:
I don't see any reson why there should be a high res limit: refmac
uses the full 5 gaussian model for scattering factors.

Ana Gonzalez:
I have refined a couple of atomic resolution structures with
refmac (as I write this I am refining with 0.98 A data!) 
The biggest advantage of using refmac is that is runs *fast*. I also
believe that at that resolution the maps with the maximum likelihood
coefficients from refmac contain practicaly no bias to the model. So, it
is very useful to do corrections in the initial stages of the refining,
try out double conformations and find quickly all/most the waters in
structure in conjunction with ARP. 
Once that is done I usually apply the "finishing touches" (refining
occupancies, etc) with shelxl.

Roberto A.Steiner:
I've refined with refmac a 0.97A structure and it works fine and FAST!!!
You can calculate esu from inversion of LS matrix with SHELX afterwards.


Joerg Mueller                 Department of Biochemistry
Phone +49-221-4706443         University of Cologne
                 6455         Zuelpicher Strasse 47
Fax   +49-221-4705092         D-50674 Koeln