Well, the problem seems to occur when Intensities are input to PHENIX
instead of structure factors. I repeated a run using Fobs, SigFobs as
input instead of IMEAN, SIGIMEAN and the Density Fit Analysis graph
looks "normal".
At 16:57 17-01-2013, Nathaniel Echols wrote:
>Actually, after some experimenting, I have just discovered that the
>map coefficients produced by the very latest code are actually on the
>same scale now, regardless of the scale of the input F(obs). This was
>not what I expected at all. However, I should add that I still don't
>know whether this scale will be what Coot expects. It is very
>difficult for me to tell what is happening internally, but when the
>"isotropize" option is selected (it's on by default) the map
>coefficient amplitudes are being multiplied by the overall anisotropic
>scale matrix. I'm not sure how this is calculated.
>
>On Thu, Jan 17, 2013 at 8:29 AM, Nathaniel Echols <[log in to unmask]> wrote:
> > On Thu, Jan 17, 2013 at 8:21 AM, Pedro M. Matias
> <[log in to unmask]> wrote:
> >> The funny thing is that I also have MTZ files from PHENIX that work
> >> perfectly in COOT.
> >
> > I think it will be dependent on having F(obs) already on an
> > approximately absolute scale. This can be accomplished by scaling
> > them during conversion from intensities, based on the Wilson plot (I
> > think). I should note that while the CCP4 conversion tools usually
> > take care of this, if you supply Phenix with intensities it will
> > simply take the square root (the French-Wilson correction doesn't
> > affect the overall scale). It would probably be useful to do the same
> > thing in our code; I think it's pretty straightforward but I'm not
> > sure about the math.
> >
> > My guess would be that some of the MTZ files you used were either a)
> > already on an absolute scale, or b) accidentally ended up being
> > approximately correct. However, if you're willing to send me the
> > inputs and outputs for Phenix and Refmac for the structures you've had
> > a problem with, I can confirm my guesses.
> >
> >> And although I cannot be 100% sure, since the run was in
> >> Dec-2012 I believe the PHENIX version is the same in both cases
> (it would be
> >> nice if PHENIX wrote the version number in the PDB header).
> >
> > It should already do this - but you need to look further down in the
> > remarks, where you will see something like this:
> >
> > REMARK 3 PROGRAM : PHENIX (phenix.refine: 1.7.3_928)
> >
> > Probably not a bad idea to display it at the very top too, however.
> >
> >> Bernhard has suggested tweaking some parameter in COOT
> >>
> >> Extensions->Refinement->Set Density Fit Graph Weight
> >>
> >> But the question is: which value should be used ?
> >
> > I suspect that this will also be completely arbitrary - btw, the fact
> > that this option even exists seems to be confirmation that the graphs
> > are not displaying a CC. I think the way Coot is doing this is
> > completely broken, for what it's worth - but I also think it would be
> > helpful if the maps from Phenix were at least on approximately the
> > correct scale, instead of being (potentially) orders of magnitude off.
> >
> > -Nat
Industry and Medicine Applied Crystallography
Macromolecular Crystallography Unit
___________________________________
Phones : (351-21) 446-9100 Ext. 1669
(351-21) 446-9669 (direct)
Fax : (351-21) 441-1277 or 443-3644
email : [log in to unmask]
http://www.itqb.unl.pt/research/biological-chemistry/industry-and-medicine-applied-crystallography
http://www.itqb.unl.pt/labs/macromolecular-crystallography-unit
Mailing address :
Instituto de Tecnologia Quimica e Biologica
Apartado 127
2781-901 OEIRAS
Portugal
|