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

Re: freeR biasing..

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

> Thanks for your valuable comment about the 'freeR flagging'
> reflections, in case of switching over from XPLOR to REFMAC.
> I found the following changes:
> when I was using a new 10% set for freeR flagging in REFMAC
> and taking the XPLOR refined coordinates as input, I found:
> _refine_overall_R_factor                0.18452
> _refine_free_R_factor                   0.27801
> _Correlation_coefficients_Fo_to_Fc      0.90104
> _Free_correlation_coeff_Fo_to_Fc        0.87480
> Now, when I use the same identical 10%sets of reflections
> (which was used in XPLOR) in REFMAC, (perhaps now it is
> 9.6% only, due to data completion by uniqueify script), I found:
> _refine_overall_R_factor                0.18777
> _refine_free_R_factor                   0.23879
> _Correlation_coefficients_Fo_to_Fc      0.94658
> _Free_correlation_coeff_Fo_to_Fc        0.91943
> Thus the biasness in choosing freeR sets has eventually
> corrupted my freeR calculations and hence the model..
> So, people who are now doing the above steps can also take care
> of this point.
> Thanks again for all..

Dear Ravi,

One common mistake when switching from X-plor to CCP4 is
using number for freeRflag. CCP4 default is 0 but X-plors is
1. If you could send me your .log file or check following lines
from .log file

_refine_overall_R_factor                0.16578
_refine_ls_number_reflns                 433749
_refine_ls_number_reflns_missing           3764
_refine_free_R_factor                   0.17253
_refine_ls_number_reflns_free             22913
_refine_ls_number_reflns_free_missing       209
_refine_ls_WR_factor                    0.16308

then you could see if there was this mistake or not. In principle
if you use same free reflections there should be no bias
problem. Only problem might be difference between scaling techniques
which different programs use. If there is problem with number of
reflections then you could use following line in REFMAc


to let program to know that free reflections flagged 1 not 0. Otherwise
in f2mtz you could let program to know that your file from X-plor
by giving label type X for freeR flag column

I hope it helps. Otherwise .log files could be helpful to see what might
be reason for disagreement.


Garib N. Murshudov, Chemistry Department, University of York, U.K.
Tel: Home +44 (1904) 43 35 89, work:  +44 (1904) 43 25 65