Re: [ccp4bb]: REFMAC5 maximum likelihood refinement

[Bart Hazes <bhazes@ualberta.ca>]

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


> under which circumstances would maximum likelihood target refinement in
> REFMAC5 produce a dramatic INCREASE in R (from 0.187 to 0.208) and Rfree
> (from 0.232 to 0.257) with 1.8A data? This was not accompanied by
> improved geometry (reduced RMSs). Any idea where I should start looking
> for the underlying problem?

Near the end of a refinement I would call a 2.5% decrease in Rfree a
spectacular drop. Likewise I think it is fair to say that a 2.5% increase
during refinement is dramatic and cause for worry. I just messed up a dual
side-chain conformation by hand editing the PDB and refmac was rightly
unhappy. Scanning the log file often exposes such and other problems by direct
warning messages, lists of bad geometry or strange behaviour of the refinement
statistics.

I also noted that the older refmac5 version was often unstable when applying
SCALE BULK as well as an explicit solvent model and I have seen cases where R
and Rfree decrease nicely during TLS refinement but then increase; either both
R & Rfree or just Rfree. In the latter case you have to think if you are
giving the program too much flexibility to overrefine and may have to tighten
up restraints on geometry or B-facters, or just use TLS parameters.

If both R and Rfree go up significantly you hope you made an error but there
have been occasions where this happened and I was rather certain of doing
things right. Sasha is right in that the ML refinement target is not directly
equivalent to improving the fit between Fc and Fobs, but increases by more
than a few tenth of a % are worrisome. After all, when the model improves so
should the fit between Fobs and Fcalc.

Final advice; make sure you have the latest version of the program, check your
output for clues of errors, read the manual to understand your options and if
you have "an interesting case" contact Garib, he is great and so is refmac (it
did drop my R and R-free by 2.5% of an already highly refined model)

Bart

===============================================================================

Dept. of Medical Microbiology & Immunology
University of Alberta
1-15 Medical Sciences Building
Edmonton, Alberta, T6G 2H7, Canada
phone:	1-780-492-0042
fax:	1-780-492-7521

===============================================================================