PDB-redo reports sigmaRfree as 0.6% so 1.4% is nearly 2.5 sigma away
from the authors'. So is this not a significant difference?
I'm guessing this is down to what Eziquiel said - differences in bulk
solvent models and or missing bits of the model. But it would be good
for users of the data to be looking at the best map the authors could
get.
The 4aw6 authors included measured phases - not sure if those were
part of the refinement target. But that is another case where it would
be nice to have the authors' map - if the final map is a combined
experimental and model phases map. In such cases it would be good to
have the HL coefficients too.
I deposited the HL coefficients for my 1e3h structure as it was
refined against an HLML target - but somehow these have been recently
deleted from my entry as part of a 'format standardization and
re-release'.
Has anyone else lost data from that re-formatting?
cheers
Martyn
On Fri, Jan 22, 2016 at 8:46 AM, Robbie Joosten
<[log in to unmask]> wrote:
> Hi Martyn,
>
> In terms of validation I think it is better if the results can be verified in an 'independent' experiment. Having map coefficients can be informative, but I prefer using more raw data. Keeping the current discussion on questionable structures in mind, we should be able to address both what the depositors say and what others make of the same data.
> Under normal circumstances all refinement programs can get quite close to the depositors R-factor. For the case of 4aw6, the results of the PDB validation server are somewhat disappointing. PDB_REDO, which also uses Refmac but in a slightly different way, gets quite a bit closer to reproducing the depositors' R-values. Being off by 1.4% for R-free is quite reasonable.
>
> Cheers,
> Robbie
>
>
>
>> -----Original Message-----
>> From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of
>> Martyn Symmons
>> Sent: Friday, January 22, 2016 01:58
>> To: [log in to unmask]
>> Subject: Re: [ccp4bb] question about map calculation during deposition
>>
>> Dear Ezequiel,
>> I think it would be good if the wwPDB validation defaulted to the depositors'
>> map coefficients if provided. 4aw6 gives an example of the way those can be
>> included in the deposited SF file. This approach has the advantage of
>> allowing authors to report results from their own choice of methods and be
>> judged on those.
>>
>> Actually 4aw6 was an interesting low resolution case where the R and Rfree
>> of the depositors' TNT-BUSTER refinement could not be reproduced by the
>> wwPDB validation suite. It is not clear to me whether the map used for the
>> validation of this entry's electron density fit was the authors' one or the
>> PHENIX/REFMAC one from the validation programs.
>>
>> 4aw6 SF .cif helpfully shows examples of data for:
>> .....
>> _refln.F_calc
>> _refln.phase_calc
>> _refln.pdbx_FWT
>> _refln.pdbx_PHWT
>> _refln.pdbx_DELFWT
>> _refln.pdbx_DELPHWT
>> _refln.phase_meas
>> ....
>>
>>
>> So I would say you should deposit your map coefficients like the 4aw6
>> authors. Maybe the validation will be clever enough to use them.
>>
>> All the best
>> Martyn
>>
>> Martyn Symmons
>> Cambridge
>>
>> On Thu, Jan 21, 2016 at 7:53 PM, Eze Chivi <[log in to unmask]>
>> wrote:
>> > Hello, for deposition, "Reflection Data Used for Final Structure
>> Refinement"
>> > is submitted, this file corresponds to the input mtz used for
>> > refinement (only F and SIGF). Therefore, the deposition server
>> > calculates the map to evaluate "Fit of model and data". I've noted
>> > that validation runs differ in the number of RSRZ>2 when input or
>> > output mtz files (for refinement) are used, then I deduce that
>> > refinement program and validation server differ in the treatment of some
>> parameters (bulk solvent, overall scaling?).
>> > Additionally, ¿How the map in the EDS server is calculated if map
>> > coefficient are not submitted during deposition? I would appreciate
>> > further information or clarification, thank you.
>> >
>> >
>> > Ezequiel
|