Re: About model-free analysis

[Edward d'A. <ed...@do...>]

Hi Mengjun,

Thank you for responding to the original message.  This preserves the
message threads in the archives:

http://thread.gmane.org/gmane.science.nmr.relax.user/1276
http://www.mail-archive.com/rel...@do...
https://mail.gna.org/public/relax-users/2012-09/msg00009.html

For your "T1 demo.txt" file, I have now made relax throw a RelaxError
saying that this file is not supported.  Unfortunately you cannot use
this file with relax.  There is simply nothing I can do to support
this file.  The new Bruker DC files are supported, but this is an old
formatted file which relax cannot handle.

The model-free analysis tab in the GUI uses the protocol which I
developed for simultaneously solving all of the artificial motion and
diffusion tensor estimate problems from my 2007 and 2008b papers
(http://dx.doi.org/10.1039/b702202f and
http://dx.doi.org/10.1007/s10858-007-9213-3).  Note that the
model-free optimisation is simple, but the full model-free protocols
are incredibly complex.  If you would like to handle single field
strength data, you will need to write your own relax script for this
reimplementing the Mandel et al., 1995 protocol.  See figure 8.1 (the
first figure of the model-free chapter) of the relax manual for how
you will need to implement the massively iterative protocol.  It would
be better though to implement the protocol from the second figure
(8.2) which is unpublished.  This figure numbering is for the new
relax manual (the file with the biggest number at
http://download.gna.org/relax/manual/).  Implementing such a protocol
is complex, but you could look at the file
auto_analysis/dauvergne_protocol.py for ideas (this is essentially a
massive relax script implementing the protocol I developed).  The
easiest solution, and probably only publishable solution, would be to
simply collect data are 2 or more magnetic fields.

Regards,

Edward



On 7 September 2012 11:24,  <men...@do...> wrote:
> Dear Dr. Edward d'Auvergne,
>
> Thank you very much for your detailed explanations. There are some bugs
> when I try to import T1 demo.txt from bruker dynamic center to Relax, and I
> have submit it online.
>
> I remembered you said Relax Gui can not perfom model free analysis with
> only single field strength nmr data, how about the  Relax script mode ? I
> have
> just read the Relax user messages, the similar questions about single
> field strength data have been asked before.
>
> Thank you very much.
>
> With best regards,
>
> Mengjun Xue
>
>
>
>
> Quoting Edward d'Auvergne <ed...@do...>:
>
>> Hi Mengjun,
>>
>> I've been looking for the relevant links as we have discussed this
>> topic a number of times on the relax-users mailing list.  One good
>> reference would be to read all the messages sent by myself, Sebastien
>> Morin, Alex Hansen, Michael Marlow, and Gary Thompson in the thread
>> called "CSA & bond length" starting at:
>>
>> http://thread.gmane.org/gmane.science.nmr.relax.user/192
>>
>> and continued on the relax-devel mailing list at:
>>
>> http://thread.gmane.org/gmane.science.nmr.relax.devel/1115
>> http://thread.gmane.org/gmane.science.nmr.relax.devel/1116
>> http://thread.gmane.org/gmane.science.nmr.relax.devel/1119
>>
>> I'll keep looking for the other threads which discuss this, as this
>> was recently talked about, but I cannot find the messages right now.
>> In summary, for protein backbone NH data you should really use 1.02
>> Angstrom with -172 ppm CSA.  The value of -160 ppm is from solid state
>> studies of peptides and is an underestimation for solution NMR.
>>
>> There is CSA variability but the amount of variability is highly
>> debated.  Some of Nico Tjandra's estimates of variability might be on
>> the high side, David Fushman's might be more reasonable.  In reality,
>> no one really knows what the real CSA and bond length values are and
>> no one has reliably de-convoluted the internal ps-ns dynamics from
>> these parameters to be able to properly answer this question.
>> Therefore everyone takes the compromise of 1.02 A with roughly -172
>> ppm (the Hall and Fushman average value).  There is another school of
>> thought from the NIH direction of Washington DC which says we should
>> use 1.04 A with -160 ppm.  However this will generally shift the order
>> parameters higher and closer to one - causing the optimisation
>> algorithms to have severe problems and failures due to the convolution
>> of the optimisation space that this results in  So it is not
>> recommended for a model-free analysis - the removal of zero-point
>> motions concept.  As long as you are consistent and know that the
>> order parameters are a relative concept rather than absolute - then
>> you can compare different states.  I hope this helps.
>>
>> Regards,
>>
>> Edward
>>
>>
>> P. S.  If you do solve this problem reliably, you will probably
>> receive quite a few citations and make a name for yourself.  However
>> the NMR field has being trying to solve this unsuccessfully for the
>> last who knows how many decades.  So I wouldn't recommend trying
>> unless you have an incredible amount of spare time and don't mind
>> diving into the deepest depths of NMR and physics theory.
>>
>>
>>
>>
>> On 5 September 2012 18:09,  <men...@do...> wrote:
>>>
>>> Hi,
>>>
>>> When doing model-free analysis, NH bond length set to 1.02 angstrom
>>> normally, but it is a little bit different for different amino acid
>>> residue;
>>> and
>>> also the N chemical shift anisotropy set to say -160 ppm normally, but
>>> CSA
>>> is changing from residue to residue, How do Relax solve such kind of
>>> problem when doing model-free analysis?
>>>
>>> Regards,
>>>
>>> Mengjun Xue
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> relax (http://www.nmr-relax.com)
>>>
>>> This is the relax-users mailing list
>>> rel...@do...
>>>
>>> To unsubscribe from this list, get a password
>>> reminder, or change your subscription options,
>>> visit the list information page at
>>> https://mail.gna.org/listinfo/relax-users
>>
>>
>>
>
>
>