There has been a serious problem in Refmac which apparently Garib
has fixed in this new release.  When calculating structure factors
for models in space group P1 Refmac would "lose" some of the atoms.

    I first ran into the problem by looking at the Electron Density
Server map for entry 3LEN.  If you pull up that model and map in Coot
and go to the highest peak you will see a 44 sigma (ugh! substitute
"rms") peak centered on a sulfur atom that is clearly present in the
PDB.  Looking in the neighborhood you will see that other atoms have
been lost as well.  In fact you can line them up and see that all
atoms with Z between about -2.3 and -2.7 A are gone.

    This was not some obscure problem, nearly every model with space
group P1 shows the same problem in the EDS maps.  You can pull up
the map for 2XU1 and it shows the same problem - even the same Z
value band is affected.

    Many of these P1 models are flagged in the EDS as having R values
that are not reproducible.  I guess if you leave out enough atoms
you can cause the R value to degrade.

    The EDS uses Refmac to calculate the structure factors and Fourier
coefficients.  I've ran Refmac alone and reproduced this problem.

    If you have refined a model with space group P1 in Refmac I suggest
you download the new version and see if your stats improve.

Dale Tronrud

On 5/27/2011 3:29 AM, Petr Kolenko wrote:
> Dear colleagues,
>
> Q2 is solved by new installation of Refmac. Many thanks for your time
> and effort. I appreciate all your comments.
>
> Petr
>
>
> 2011/5/27 David Cobessi<[log in to unmask]>:
>> Dear Petr,
>> Before running Refmac, you have to remove the ANISOU line using pdbset for
>> example. In fact you should not add the TLS contribution to the output
>> coordinates when you use Refmac except in the last run before to deposit the
>> file into to PDB. Phenix removes it automatically.
>> Kind regards,
>> David
>>
>> On 05/27/2011 10:57 AM, Petr Kolenko wrote:
>>>
>>> Dear colleagues,
>>>
>>> regarding Q2:
>>>
>>> I do not use TLS parameters, the space group is P1, and yesterday I
>>> tried to refine the structure with anisotropic ADP (60,000 reflections
>>> against 50,000 parameters) - no positive maximums. Then I used the
>>> anisotropic model as input and refined isotropically, the maximums are
>>> there again.
>>> I use refmac5.5.0100, is it really too late version? Seems to be
>>> current version according the websites.
>>>
>>> Many thanks.
>>>
>>> Petr
>>>
>>>
>>> 2011/5/27 Jan Dohnalek<[log in to unmask]>:
>>>>
>>>> In the last months we have seen different versions of Refmac give
>>>> different
>>>> maps when displayed in Coot, i.e.
>>>> one version giving "nicer" agreement and no difference peaks in some
>>>> difficult areas and another version resulting in sharp differences where
>>>> it
>>>> was hard to build the protein. We did not investigate much further as
>>>> most
>>>> of the time use of two or three versions gave us a good picture of what's
>>>> going on ..probably a feature which would disappear with newer versions
>>>> anyway.
>>>>
>>>> What's the version of Refmac you use, Petr?
>>>>
>>>> Jan Dohnalek
>>>>
>>>>
>>>> On Thu, May 26, 2011 at 12:11 PM, Petr Kolenko<[log in to unmask]>
>>>>   wrote:
>>>>>
>>>>> Dear colleagues,
>>>>>
>>>>> I have two problems in two structure refinements using REFMAC5.
>>>>>
>>>>> 1) 1.8A resolution, zinc in the active site. Refinement using work
>>>>> reflections - ADP for Zinc was about 14. Final refinement including
>>>>> all reflections increase ADP to 20 or even higher values - followed by
>>>>> very high positive difference density in position of the zinc. I have
>>>>> tried also PHENIX, the same thing. I changed ADP manually to 14 and
>>>>> only calculated maps (no refinement) look good. May I deposit the
>>>>> structure using "manually" fixed ADP according to the best agreement
>>>>> to the observed and difference electron density? By the way, it is
>>>>> clear that this is zinc.
>>>>>
>>>>> 2) 1.9A resolution, about 600AA, all of them OK in electron density.
>>>>> But, somehow, about ten atoms give very strong positive electron
>>>>> density suggesting they are not taken into account in refinement. On
>>>>> the other hand, ADPs are reasonable and seem to be refined. All of
>>>>> these atoms are fully occupied. I tried to omit whole residues and
>>>>> build them again, but the maxima appeared again. Using of PHENIX
>>>>> resulted in no difference electron density for these atoms. I have
>>>>> also tried to take PHENIX output to REFMAC, but the maxima are there
>>>>> again. It is always one or two atoms from the same residues -
>>>>> sometimes Calpha, sometimes Cbeta, sometimes C, sometimes NH1, but
>>>>> still on the same five residues. Does anyone have any idea how to
>>>>> solve this problem?
>>>>>
>>>>> Many thanks for any response.
>>>>>
>>>>> Petr
>>>>>
>>>>>
>>>>> --
>>>>> Petr Kolenko
>>>>> [log in to unmask]
>>>>> http://kolda.webz.cz
>>>>
>>>>
>>>> --
>>>> Jan Dohnalek, Ph.D
>>>> Institute of Macromolecular Chemistry
>>>> Academy of Sciences of the Czech Republic
>>>> Heyrovskeho nam. 2
>>>> 16206 Praha 6
>>>> Czech Republic
>>>>
>>>> Tel: +420 296 809 390
>>>> Fax: +420 296 809 410
>>>>
>>>
>>>
>>
>>
>> --
>> David Cobessi
>> Institut de Biologie Structurale
>> 41, Rue Jules Horowitz
>> 38027 Grenoble Cedex-1, France
>> Tel:33(0)438789613
>>     33(0)608164340
>> Fax:33(0)438785122
>>
>>
>
>
>