Thanks for your suggestions!
The molecular weight of the protein is 14kD.
Another thing I want to make it clear is that though the mad data and the
native data have the same/similar spacegroup C222 and packing, they are
differed in cell parameters which were grown at different temperatures.
With the mad data (62.6,98.4,40.1)1molecule/AU and 2molecules/AU in the
native data(108.2, 118.6, 63.6).
I wonder that anything else have effect on the f' or f"?
Another thing Edwin mentioned is that Se-Met crystal may have some
defects. So what is the sign of defect of a crystal? I remember someone
said that the R_merge of the low resolution shell (50-6.0)could be
considered. For those R_merge of the low resolution shell >0.05, the
crystal is somwhat non-optimal. Actually the data for my protein R=0.046
with native one and 0.056 with peak data.
> You dont say the size of the protein but in fact errors in f' or f" will
> have very little effect. You can correct the SE formator if you like,
> but even so in most cases if you call the MET MSE then you get holes
> over the Se in difference maps which probably indicate partial conversion.
>
> Frankly any model refined at 2.7A is not going to be optimal - it is
> quite hard to model alternate conformations (which certainly will be
> there for 5-10% of the residues) and the solvent is also hard to fix -
> there are no very good geometric restraints to stabilise its refinement..
>
> So if you have 2.2A data then unless there is something seriously wrong
> with it, then that is the set to use..
>
> If you transfer the model refined against one data set to another then
> you need to preserve the Free R definitions.
>
> Easy if you have an mtzfile
> Merge the data sets - then run uniqueify -f FreeRflag merged.mtz
>
> That will keep the existing flags and generate new ones for the new
> reflections..
>
> Then if you want to you can convert back to the CNS format.
> Eleanor
>
>
> Peng Zhang wrote:
>> Hi, Peter,
>>
>> Does the radition damage have such a large effect? There are many
>> positive
>> peaks seeing from the difference map, seems abnormal. But I have seen
>> such difference maps before which are OK for modeling and refining.
>>
>> Actually when I try to modelling the Se-met into the positive peaks, the
>> R
>> and free_R factor goes up about 2%. Does the occupancy of the Se affect
>> the R factor, or should I change the occupancy?
>>
>> Another problem is f' andf f" prime, which I refine them when phasing
>> with Solve.
>>
>> Actually, I did not use the same free set when refining with the native
>> data and peak data.I try to compare them because I want to make it clear
>> that the difference may be caused by the mad data while not the model
>> itself.
>>
>> Thanks.
>>
>>
>>
>>> Hi Peng Zhang,
>>>
>>> The presence of radiation damage might cause some problems.
>>> Do so see any obvious features in the difference map?
>>>
>>> Another problem (although I doubt it would cause such a big difference)
>>> might be the fact that f' andf f" prime are incorrect.
>>> Try and refine them (CNS or phenix.refine) maybe.
>>>
>>> Does your native data have the same free set as the peak data? if not,
>>> you
>>> are in trouble and have to start from scratch with your native data to
>>> be
>>> able to make a fair comparison. The 22/25 for 2.7 A seems awfully close
>>> together.
>>>
>>> procheck has not very up to date standards of what is good and what is
>>> not. Better use molprobity, available from:
>>>
>>> http://molprobity.biochem.duke.edu/
>>>
>>> HTH
>>>
>>> Peter
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> ----- Original Message -----
>>> From: Peng Zhang <[log in to unmask]>
>>> Date: Tuesday, March 20, 2007 6:04 pm
>>> Subject: Re: [ccp4bb] modelling with sad/mad data
>>> To: [log in to unmask]
>>>
>>>
>>>> Maybe I did not make the questions clear, which leading to the
>>>> misleadings.Firstly, I have collected the mad data and get the
>>>> phase at synchrotron,
>>>> the phased Se is quite good for modelling, and get over 70% of the
>>>> molecule run with resolve autobuilding.The density seems also good for
>>>> building. But when finally refining the model, the gap between the
>>>> R(0.22)and free_R(0.32 )is big, even though modelled the Se-
>>>> methionine. Before
>>>> collecting the mad data at synchrotron, I already have another
>>>> native data
>>>> set collected at home diffractometer (rigku, with R-axis IV++). To my
>>>> surprise, when I using the same model(first model) and run with
>>>> this data
>>>> set, it is quite good( with 2.7A resolution, R=0.24 and Rf=0.28 and
>>>> further refine to R=0.22 and Free_R=0.25), and I got the final
>>>> model.Thegeometry of the first model and final model(actually no
>>>> big difference of
>>>> the two models)is quite good with procheck.The omit map says good
>>>> enoughwith both of the two models.
>>>>
>>>> So I wondering what happened with the peak data? Did the anomolous
>>>> signalhave much effect on the data? and anyone have the similar
>>>> experience?
>>>>
>>>>
>>>>> First it is always best to refine your model against the highest
>>>>> resolution good quality data that you have available. There was
>>>>> correspondence about the geometric weighting - could you have
>>>>>
>>>> weighted> the Xray data too high and have bad geometry - see
>>>> previous Emails!
>>>>
>>>>> And the Free R seems rather low for the Se data.
>>>>> Did you transfer the same Free R set from the native to the Se data?
>>>>> Eleanor
>>>>>
>>>>>
>>>>> Peng Zhang wrote:
>>>>>
>>>>>> Dear friends,
>>>>>>
>>>>>> Recently, I have solved a structure using mad method. When
>>>>>>
>>>> using the
>>>>
>>>>>> peak
>>>>>> data(2.3A) as the native for structure refinement, the gap
>>>>>>
>>>> between R
>>>>
>>>>>> factor and R free is big, about 0.1(0.22 and 0.32). I modelled the
>>>>>> selenomethionine but the gap still exists. When I changed the
>>>>>>
>>>> data for a
>>>>
>>>>>> real native one(2.7A),it seems OK with R=0.24 and Rf=0.28.
>>>>>>
>>>>>> Does anyone have the similar experience?
>>>>>> what should I pay attention to when using the sad/mad data as
>>>>>>
>>>> the native
>>>>
>>>>>> one for modelling and refinement?
>>>>>>
>>>>>> Thanks in advance.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>> --
>>>> Peng Zhang, Ph.D. Student
>>>> Institute of Biochemistry and Cell Biology
>>>> Shanghai Institutes for Biological Sciences
>>>> Chinese Academy of Sciences
>>>>
>>>> 320 Yue-Yang Road
>>>> Shanghai 20031
>>>> P.R. China
>>>>
>>>>
>>>>
>>>> Tel: 021-5492-1117
>>>> Fax: 021-5492-1116
>>>> Email: [log in to unmask]
>>>>
>>>>
>>>
>>
>>
>>
>
>
--
Peng Zhang, Ph.D. Student
Institute of Biochemistry and Cell Biology
Shanghai Institutes for Biological Sciences
Chinese Academy of Sciences
320 Yue-Yang Road
Shanghai 20031
P.R. China
Tel: 021-5492-1117
Fax: 021-5492-1116
Email: [log in to unmask]
|
