Hi,
If I understand correctly, the Phenix manual suggests two options for selecting the R-free set for twin refinement. One is to do this using the phenix.reflection_file_converter with the parameter
--use-lattice-symmetry-in-r-free-flag-generation
which I suppose does not require input of the twin law. The second is in the actual refinement using the parameter
xray_data.r_free_flags.generate=True
in which case the twin law is part of the input. My first question is whether the two options
are truly equivalent. The second question is whether one can go into either of them with an mtz file
in which R-free set reflections have already been selected (e.g. in scala or XDS) without
any twinning assumption, or should one use a file without any R-free set altogether and let the
phenix script choose it from scratch, under the twinning assumption.
Another (related) question is whether simulated annealing is possible in conjunction with
twin refinement.
Thanks,
Boaz
----- Original Message -----
From: Pavel Afonine <PAfonine@lbl.gov>
Date: Tuesday, May 5, 2009 4:53
Subject: Re: [phenixbb] ncs
To: PHENIX user mailing list <phenixbb@phenix-online.org>
> I see. I suggest to run two refinements:
>
> - with NCS and "optimize_wxc=true optimize_wxu=true" options, and
>
> - without NCS and with "optimize_wxc=true optimize_wxu=true".
>
> This will make it weight-choice independent and so easier to
> understand.
> Otherwise, since NCS term is included in weights calculation,
> using or
> not using NCS may change the X-ray/Restrains weight which may
> have
> larger effects than using/not using NCS itself.
>
> Pavel.
>
>
> On 5/4/09 6:48 PM, Maia Cherney wrote:
> > Hi Pavel,
> > The resolution is 2.15 A. The NCS was always on during the
> refinement
> > until we got low R factors (19.2% and 21.2%). Then the NCS was
> turned
> > off for the final refinement and the R factors increased,
> which is
> > strange as they should be going down when you apply less
> restraints.
> > There are five restraint groups in the asym. unit.
> >
> > Maia
> >
> >
> >
> > Pavel Afonine wrote:
> >
> >> Hi Maia,
> >>
> >> - I'm wondering why this puzzles you?
> >>
> >> - The gap Rfree-Rwork seems suspiciously small, although I
> can't tell
> >> without knowing the resolution.
> >>
> >> - Overall, better R-factors (Rwork, Rfree and Rfree-Rwork)
> mean that
> >> among many possible refinement strategies you have chosen the
> one that
> >> is better than the others.
> >>
> >> What is the resolution?
> >> What are the R-factor before using NCS?
> >> What will be the R-factors after you turn NCS back off?
> >>
> >> Pavel.
> >>
> >>
> >> On 5/4/09 12:27 PM, Maia Cherney wrote:
> >>
> >>
> >>> Hi everybody,
> >>> I am wondering why my final structure refined with NCS=True
> has lower R
> >>> factors (19.2% and 21.2%) than without NCS by approximately 1.5%.
> >>>
> >>> Leo
> >>> _______________________________________________
> >>> phenixbb mailing list
> >>> phenixbb@phenix-online.org
> >>> http://www.phenix-online.org/mailman/listinfo/phenixbb
> >>>
> >>>
> >>>
> >> _______________________________________________
> >> phenixbb mailing list
> >> phenixbb@phenix-online.org
> >> http://www.phenix-online.org/mailman/listinfo/phenixbb
> >>
> >>
> >>
> >
> > _______________________________________________
> > phenixbb mailing list
> > phenixbb@phenix-online.org
> > http://www.phenix-online.org/mailman/listinfo/phenixbb
> >
>
Boaz Shaanan, Ph.D.
Dept. of Life Sciences
Ben-Gurion University of the Negev
Beer-Sheva 84105
Israel
Phone: 972-8-647-2220 ; Fax: 646-1710
Skype: boaz.shaanan