Dear All,
Is it not possible to skip that field? that would be a stopgap.
The best solution would be to ask the user for a minimum shift error for
deposition. There is no reliable way of calculating it, I would say, and
there is no place in the data model to put it either. Asking the user you
will still get a pretty meaningless number, but at least it is the users
responsibility.
Ideally, of course, the error would depend on the individual shift. Are
the peaks broad? overlapped? weak? weird? But people are just not going to
estimate those errors one by one by hand, and in the absence of an
algorithm for doing it, the results wouls be all over the place anyway.
How would you do it? In my experience you can get spectrum-to-spectrum
reproducibility of less than a fifth of the point-to-point resolution for
good peaks in badly digitised indirect dimensions. So the point resolution
is not a good estimate. As for the peak width, how would use it to
calculate an uncertainty?
Yours,
Rasmus
---------------------------------------------------------------------------
Dr. Rasmus H. Fogh Email: [log in to unmask]
Dept. of Biochemistry, University of Cambridge,
80 Tennis Court Road, Cambridge CB2 1GA, UK. FAX (01223)766002
On Wed, 2 Mar 2011, Wayne Boucher wrote:
> Hello,
>
> The standard deviation in Analysis is calculated using the actual peak shift
> values, and so 0.0 usually means there is only one peak (or no peaks) for the
> associated resonance. (If by complete luck all the resonance peaks happen to
> have exactly the same shift value then you could still get 0.0 even with more
> than one peak.) (The shift averaging can also be weighted by spectrum, so
> you can say that some spectra are more, or less, accurate than others, for
> example.)
>
> I guess the question now is how best to put in the information that the BMRB
> wants, so where in the code. As Aleks says, the desired value is probably
> the max of the standard deviation and the digital resolution (or maybe it
> should be the sqrt of the sum of the squares, I don't know). Should this be
> in the BMRB export code or somewhere "higher" up inside Analysis somewhere?
>
> Wayne
>
> On Wed, 2 Mar 2011, Aleksandras Gutmanas wrote:
>
>> Hi Catherine,
>>
>> Usually there is some precision with which you determine the chemical shift
>> values based on digital resolution of your spectrum and linewidth of you
>> peaks - basically a function of your spectrometer(s). In addition, CCPN
>> Analysis would provide a non-zero error if your assignment is connected to
>> more than one peak (right, Tim, Wayne?). The larger of the two values for
>> each nucleus is what BMRB would need.
>> In my experience, typical final values are in the range of 0.03-0.05ppm for
>> 1H and 0.3-0.5ppm for 13C and 15N.
>>
>> Best,
>> Aleks
>>> Date: Tue, 1 Mar 2011 13:09:54 -0800
>>> From: Catherine Foo<[log in to unmask]>
>>> Subject: chemical shift error / BMRB deposition
>>>
>>> I tried searching in the archives but could not find the answer
>>> successfully (because "error" is not a very useful search term! :).
>>>
>>> I have submitted chemical shifts to BMRB and some of my error values
>>> are 0.000, and BMRB wants me to provide the correct error information.
>>> I'm not sure what sort of error BMRB is looking for, and I'm hoping
>>> someone here with more experience could help me out.
>>>
>>> Am I correct in my assumption that analysis calculates error simply as
>>> the standard deviation of all the corresponding peaks?
>>> Is that the type of error that BMRB wants?
>>> If I only have one peak picked for that resonance in my project, then
>>> the SD is by definition zero?
>>>
>>> I am using analysis 2.0.7 and I exported the chemical shift list via
>>> Format Converter.
>>>
>>> Cat
>>>
>>
>
|