[phenixbb] optimize_wxc and wxu
pafonine at lbl.gov
Mon Jan 26 12:37:00 PST 2009
> In general should one optimize wxc and wxu ONLY as a final polishing
> step or can this be done earlier in the refinement and then the
> optimized values used for subsequent rounds of refinement?
Optimizing weights is time consuming: depending on the data and model
size it can take from many minutes to hours, and this is the only reason
for suggesting it as a final step.
However if it runs quickly or/and you are not in rush, you can do it all
the time: then it will always make sure that the weights are "optimal"
(in terms of giving the best Rfree). Note, that most of the time you
don't really need it since the automatically determined weights are
pretty good (most of the time).
The optimal weight found after running phenix.refine with
optimize_wxc=true (or/and optimize_wxu=true) may substantially vary
between structure determination stages or even between refinement
macro-cycles within one refinement run. This is why it is not generally
advised to use the same weights (determined at the beginning) all the
time, although technically you can: just pick up the best wxc_scale from
the log file and specify it for the next run.
Please let me know if you have any questions!
More information about the phenixbb