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

[ccp4bb]: estimated error of occupancy



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

This message is a little lengthy, but I think the problem of getting
standard deviations for refined parameters is pretty common and there is 
no easy way (see how to do it the hard way below) to get standard 
deviations from any of the standard programs that I know of.

The often cited exception is SHELX's matrix inversion method.

The problem with matrix inversion is that you need very high quality 
data and a lot of memory in your computer (several GB for a moderately 
sized structure).

If you do not have at least 3-5
reflections for every parameter you are refining, the inversion routine
of SHELX will not work, or worse give you meaningless results.


DO NOT USE SHELX TO GET STANDARD DEVIATIONS UNLESS YOU HAVE AT LEAST
3-5 STRUCTURE FACTOR AMPLITUDES PER REFINED PARAMETER.

This means, if you do not have better than ~1.3 Angstrom data, matrix 
inversion will give you meaningless results.




The best way for getting information on the uncertainty of refined 
parameters from moderately good diffraction data is the same way you 
could get a standard deviation for any parameter that you are refining 
through a complex refinement routine, be it crystallographic refinement 
or anything else.

The strategy works as follows:

- fix your occupancy and run a refinement of all
   other parameters in X-PLOR or CNS until you get convergence
   (i.e. your R-factor does not change anymore)

- do this for multiple different occupancy values.

- plot your R-factor versus occupancy.

- the plot you get should have a minimum for the optimal occupancy
   value. If the uncertainty is low, you should have one clean,
   symmetrical, narrow, negative, spike in your plot. If the
   uncertainty is high the plot will have the shape of a rough,
   boad depression.

- to get your standard deviation fit the resulting plot to a function
   that consists of a baseline minus a gaussian normal distribution.

- gnuplot or a similar fittig program will perform such a fit for you.
   You will have to fit three parameters:
                the constant (i.e. baseline value)
                the mean of the gaussian
                the amplitude of the gaussian
                the sigma of the gaussian

- since you are fitting 4 parameter to this function,
   make sure that you do your refinement for at least,
   15 to 20 different occupancies and choose the occupancies
   so that the values you get are in the area where the plot is
   varying most quickly.


You are also fighting a different problem, the occupancy and the 
B-factor of your metal will be very highly correlated. In other words if 
you decrease the B-factor your occupancy will go down as well. So make 
sure that you have a reasonable B-factor and keep that one fixed in the 
refinement.

To get a good B-factor for the metal, take the average B-factor
of atoms around the metal for which you know that the occupancy is 1
(i.e. side chain atoms that have only one conformation).



If I would have to place my bets, unless you do not have stellar data or 
a very heavy metal the standard deviation of your occupancy will be 
pretty high and because of the coupling to the B-factors the mean value 
of your distribution is very shaky too. I am not surprised that the 
reviewer wanted to have a measure of the uncertainty in the refined value.


Susan Heffron wrote:

 > ***  For details on how to be removed from this list visit the  ***
 > ***          CCP4 home page http://www.ccp4.ac.uk         ***
 >
 > Hi all.
 >
 > We are looking for a method to calculate the "estimated error" for the
 > occupancy values of the ions in our protein structure.  The occupancy
 > was refined in X-PLOR or CNS.  A reviewer of our paper is requiring an
 > "estimated error" of these values.  All suggestions for how to calculate
 > this are appreciated.
 >
 >                          Sincerely,
 >                          Susan
 >
 >










-- 
Ulrich K. Genick
Assistant Professor
Department of Biochemistry
Brandeis University, MS009
Waltham, MA, 02454

Room   Kosow 108
Phone  781-736 2304
Fax    781-736 2349
Email  genick@brandeis.edu