[phenixbb] refining with f' and f" values

Nathaniel Echols nechols at lbl.gov
Thu Nov 1 08:54:07 PDT 2012


On Thu, Nov 1, 2012 at 8:43 AM, Subhani Bandara <ramssb17 at gmail.com> wrote:
> I have selected "F(+),SIGF(+),F(-),SIGF(-)" as data labels in GUI during
> refinement. If I explain more about what I did, I merged Intensity+-,
> sigma+- from reflection file ( as amplitudes) with Rfree flags, F-obs,
> SIGFobs and HL coefficients of data.mtz file of a refinement. Then I used
> this with a pdb file and cif file (gor ligands) in refinement. Also I
> selected anomalous group in settings and asked to refine f' and f" for heavy
> atoms while giving 0.00 as the value.Then it gives an error message saying "
> use ML as target function" as I was using anomalous data. When I changed
> target to ML it gives error saying "if need to use phase restraints in
> refinement please change the target function to MLHL".

Okay, I wasn't aware that MLHL was limited to non-anomalous data.  Good to know.

>  As I am using experimental phases is it OK to uncheck "use experimental
> phase restraints" option and have ML as target function?

If this is the same structure you sent me data for earlier, I don't
think you need experimental phases - they're more helpful at lower
resolution, or when you're more worried about model bias, but you have
high resolution data and a very good model.

> Then I merged only Intensity+-, sigma+- from reflection file ( as
> amplitudes) with Rfree flags of mtz file and used it with ML as target
> function. Then refinement started  with high Rfree value (41% compared to
> 16% for current model)  and ended up with 48% Rfree.

That doesn't sound right... is this the reprocessed dataset or the
original?  Can you please send me (off-list!) all of the input files
(again, sorry!), including the original reflection file with
intensities, and one of us will take a look.  I suspect there is a
simple explanation but it's impossible to guess without seeing the
exact inputs.

FYI, I would just refine against intensities directly, since
phenix.refine will convert them internally using French-Wilson
treatment to correct negative intensities.  The reflection file editor
is somewhat dumber in its handling of intensities - I have since
changed the code to use French-Wilson instead but this isn't released
yet.

-Nat


More information about the phenixbb mailing list