Re: Missing data for spin system and compatibility in GUIRe: Missing data for spin system and compatibility in GUIHi Nav,
Martin is spot on here. The temperature control and temperature
calibration has been a topic much discussed on the relax mailing
lists. For example, here are some threads where you can find out more
information, if you wish:
http://thread.gmane.org/gmane.science.nmr.relax.user/83
http://thread.gmane.org/gmane.science.nmr.relax.user/273/focus=274
(this is Chris MacRaild's response to a message by Seb Morin)
http://thread.gmane.org/gmane.science.nmr.relax.user/1121
http://thread.gmane.org/gmane.science.nmr.relax.user/1419
http://thread.gmane.org/gmane.science.nmr.relax.user/1368 (this one is
from Martin)
http://thread.gmane.org/gmane.science.nmr.relax.user/1397 (a
continuation of the previous thread)
You need to click on all the messages in the thread to follow them.
There are many more threads on this area, but I don't have the time to
find them all right now. The following message might be of
significant help for you:
http://thread.gmane.org/gmane.science.nmr.relax.user/1419/focus=1423
Firstly note that if you have a temperature problem, this will not be
solved by looking at a single field strength as it is a problem
between experiments on the same spectrometer. So even if you perform
an analysis with data from a single field strength, the R1 data might
be up to 2 degrees warmer or colder than the R2, and the same problem
will occur to a different degree (or direction) on the second
spectrometer. Any analysis using such data will be meaningless, as
this will have a large effect on the diffusion tensor. Considering
that the diffusion tensor is the major contributor to liquid state
relaxation, the internal dynamics can contribute 20% or less (see the
original model-free papers for these numbers), then any internal
dynamics will be severely distorted, possibly hidden, and artificial
motions will appear. I would recommend you look at this section of
the relax manual for more temperature related details:
http://www.nmr-relax.com/manual/Temperature_control_calibration.html
As for relax supporting an analysis at a single field strength, there
is nothing stopping you from performing such analysis. relax is
designed with flexibility in mind, so you can perform your model-free
analysis any way you can imagine. With the minimisation settings, you
can replicate the exact results from Art Palmer's Modelfree, from
Dasha, or from Tensor2. However note that I have not written any
scripts or GUI to handle this situation, as I have no interest to. So
you would need to write the script yourself. I would then recommend
looking at the file 'auto_analyses/dauvergne_protocol.py' to get an
idea of how to implement a full model-free protocol. It is quite big
because of the iterative optimisation of the model-free parameters,
then model elimination, then model selection, and finally diffusion
tensor optimisation, with convergence testing. These steps, even for
single field strength data, need to be iterated until convergence.
This can take up to 15 iterations. You will also need to decide how
to determine your initial diffusion tensor estimate - and relax can
perform this as well.
But note that you should be aware of of the problems discovered by
Schurr et al., 1994 and Tjandra et al., 1996 of the artificial
motions. For example see:
http://thread.gmane.org/gmane.science.nmr.relax.user/326/focus=332
Korzhnev's review (I don't have the reference at hand) and my paper
(http://dx.doi.org/10.1039/b702202f, this has the Korzhnev reference
in it) cover all of the problems you will encounter and hopefully
convince you that an analysis of single field strength data would only
be useful for perfectly isotropic systems (which is never possible due
to water shell differences around the system) which have no
significant internal motion (but note that if you see no motion with
single field strength data, that does not mean that there is no
motion).
I hope this helps,
Edward
On 26 April 2013 14:48, Martin Ballaschk <ballaschk@xxxxxxxxxxxxx> wrote:
> Hi Nav,
>
> I had similar problems in the past.
>
> The inconsistetncy has to come from somewhere. For me, it was the temperature difference between different magnets. After I figured out that we calibrated our spectrometers the wrong way, I finally got consistent data.
>
> Just try to superimpose two high-resolution spectra from your two fields. They should be identical. If you can see peaks that are not 99% on top of each other, I would recommend take a long hard stare at temperature control.
>
> How do you control your temperature? I found neat methanol did not work with our spectrometers with cryoprobes, see also [1]. We use d4-methanol now, the impurities give enough signal for a proper temperature calibration (between magnets, and between experiments).
>
> Another issue may be TROSY-based sequences, used on deuterated systems and cryoprobes. As described by Nils Lakomek et al, that can be the source for a lot of annoying artifacts. [2]
>
> Edward will have additional ideas, I guess.
>
> Cheers
> Martin
>
> [1] Lakomek N-A, Ying J, Bax A (2012) Measurement of 15N relaxation rates in perdeuterated proteins by TROSY-based methods. J Biomol NMR 53: 209–221. doi:10.1007/s10858-012-9626-5.
>
> [2] Findeisen M, Brand T, Berger S (2007) A1H-NMR thermometer suitable for cryoprobes. Magn Reson Chem 45: 175–178. doi:10.1002/mrc.1941.
>
>
> On 26.04.2013, at 14:31, Navratna Vajpai <navratna.vajpai@xxxxxxxxx> wrote:
>
>> Hi Edward and other members in the group,
>>
>> Just briefly mentioning my concern: I have acquired 15N-backbone relaxation data on a protein kinase on two different fields (600 MHz and 800 MHz). In the beginning had some difficulties in running your scripts. Following your suggestions, I looked through the literature and developed some understanding before running all these scripts in Relax. The scripts seem all working for the local_tm model. However, for sphere or the spheroid models, it never converged (the run continued for several days with going upto 64 rounds). On looking through the next chapter about data consistency, i thought of doing consistency tests. Tests with J0 checks, suggests inconsistency as described in the chapter. As i do not have access to the third field, i do not know which data amongst the two is bad. Experimental parameters or the sample used were same at both fields. Is there any way to check this without having data for the third field?
>>
>> Do you or someone else has a script which can use data from only single field and let RELAX do model-free analysis?
>>
>> I looked through the mailing list and have seen that this problem has been asked and discussed several times. I know about TENSOR2 which can do such an model-free analysis using single field but was wondering if some has found a fix for the RELAX.
>>
>> Many thanks,
>> Nav
>>
>>
>>
>>
>>
>>
>> On Tue, Feb 12, 2013 at 2:59 PM, Edward d'Auvergne <edward@xxxxxxxxxxxxx> wrote:
>> Hi Nav,
>>
>> Welcome to the relax mailing lists! Please see below:
>>
>>
>> > The situation:
>> > I have experimental data for R1, R2 and NOE at two fields (600 MHz and 800
>> > MHz) on a large protein kinase. As expected, i do not have data for all the
>> > residues in the protein sequence. on searching through Web, i have found a
>> > X-ray structure, which also have some parts missing, possibly due to poor
>> > electron density in those regions.
>>
>> This will complicate your analysis, as you don't have orientational
>> information about your NH vectors! Such information is essential for
>> the prolate and oblate spheroidal and ellipsoidal diffusion tensors.
>> You will need to read the relevant literature if this is not clear
>> (you can find lots of references in the papers linked at
>> http://www.nmr-relax.com/features.html#primary_refs, especially my
>> 2008a paper at http://www.nmr-relax.com/refs.html#dAuvergneGooley08a).
>>
>>
>> > I learnt from RELAX that one can create
>> > spin system solely based on sequence and then attach protons to it or by
>> > using a pdb structure.
>>
>> relax does not currently have an algorithm to automatically place
>> protons into the 'correct position' in 3D space. This just allows you
>> to say that protons are attached - hence you will have dipole-dipole
>> relaxation present. If you have a 3D structure without protons, you
>> will need to use Molmol, PyMOL, etc to add the missing protons
>> yourself prior to loading the structure into relax.
>>
>>
>> > For model free analysis possibly, i would need a pdb
>> > structure (not entirely sure!); as i can see, an example in the manual
>> > illustrating without the use of the structure (page 103)
>>
>> You really need to read more of the literature to understand the
>> reason why. But you can perform a model-free analysis using the
>> protocol I developed which is hard-coded into the GUI. But you can
>> only use the 'local_tm' and 'sphere' models if no 3D data is present.
>> If this is not clear why, then you have a lot more reading to do ;)
>>
>>
>> > The problem:
>> > When i tried doing it by creating spin systems using amino acid sequence
>> > alone, the system never got executed. However, when i started doing it with
>> > structure as an input., it did run but then gave me an error message for all
>> > the spins as follows:
>> > for spins with all six data parameters:
>> > spin YYY deselected due to absence of any relaxation mechanisms
>>
>> This means that you have not specified the relaxation mechanisms.
>> Note that if you are looking at 15N backbone data - importantly with
>> no 13C labelling - then two major relaxation mechanisms are present.
>> These are the dipole-dipole and CSA interactions. You will need to
>> tell relax that these are active, and what the physics for these
>> interactions should be. The reason why you have to do this is because
>> relax can be used for RNA, DNA, or organic molecules. And even in
>> proteins, this simple 2 mechanism relaxation may not always be the
>> case. For example 15N bb relaxation with 13C labelling, you have 3
>> direct dipole-dipole relaxation mechanisms, and you have to also take
>> interference into account. Or for natural abundance 13C CO relaxation
>> where only CSA relaxation is present. relax allows you to handle
>> these different cases.
>>
>>
>> > and for spins with no data:
>> > spin YYY deselected due to absence of any data.
>> >
>> > the second one is understandable but not sure about the first one .
>>
>> Did you follow the tutorial in the relax manual about using the GUI
>> for model-free analysis, specifically the section on setting up the
>> relaxation interactions
>> (http://www.nmr-relax.com/manual/d_Auvergne_protocol_GUI_mode_relaxation_interactio.html)?
>>
>>
>> > To check whether something is wrong with the complete data sets,
>> > i created new data files for only first two residues with structural
>> > coordinates extracted for these two residues. In this case, the program
>> > worked well.
>>
>> You can perform a full analysis using the protocol I developed. If
>> this is not clear what this protocol is, please see my 2007 and 2008
>> papers:
>>
>> http://www.nmr-relax.com/refs.html#dAuvergneGooley07
>> http://www.nmr-relax.com/refs.html#dAuvergneGooley08a
>> http://www.nmr-relax.com/refs.html#dAuvergneGooley08b
>>
>> For residues which have 3D data, you can perform this analysis. For
>> missing residues, you may have to use the concept of global model
>> hybridisation:
>>
>> http://www.nmr-relax.com/refs.html#Horne07
>>
>> This will allow you to combine the local tm models for residues
>> without 3D data with the results from the analysis with 3D data.
>>
>>
>> > Questions from me:
>> > 1) Does that mean the absence of data for certain spins, loaded either from
>> > sequence or structure, causes this problem?
>>
>> No, this is just an indication that you have not set up your active
>> relaxation mechanisms in relax.
>>
>>
>> > 2) Can i do the whole analysis just by using the sequence.
>>
>> Yes, see above. But it would be much better if you use the 3D info
>> that you already have, assuming that structure is correct.
>>
>>
>> > 3) Does the software actually need minimum six values (R1, R2 and NOE at two
>> > fields) for this analysis or it can work with >= 3 values?
>>
>> Please read my 2007 and 2008b papers about this!
>>
>>
>> > As for some
>> > residues, i have < 6 data values. I am currently ignoring those residues
>> > with < 6 data values as i wasn't sure if model free analysis would be able
>> > to handle that.
>>
>> Again, my publications cover this and what the minimum is and why.
>> But note that model m8, as I have defined it, has 5 parameters.
>> Therefore you will require minimally 5 data points.
>>
>>
>> > 4) I am still unclear with the initialization of diffusion tensor. In the
>> > GUI mode the first row asks for The diffusion tensor parameters:
>> > I have tried to understand what is written in the manual, but i am not sure
>> > if i understood it correctly.
>>
>> This is also discussed in full detail in my 2007 and 2008b papers as
>> to why my protocol, which is what you are using when accessing the
>> relax GUI, requires no initial diffusion tensor. These papers also
>> explain the concept behind this protocol and the inversion of the
>> problem of simultaneously finding the interlinked global diffusion
>> tensor and spin specific internal motions.
>>
>>
>> > Would you be able to guide/suggest me on this. Any suggestions from your end
>> > is highly appreciated.
>>
>> One other very useful reference which contains the answer to all your
>> questions (apart from the missing relaxation interaction setup) is my
>> PhD thesis which you can find at:
>>
>> http://www.nmr-relax.com/features.html#primary_refs
>>
>> I hope some of this information helps, but you do have quite some
>> reading ahead of you!
>>
>> Regards,
>>
>> Edward
>>
>> _______________________________________________
>> relax (http://www.nmr-relax.com)
>>
>> This is the relax-users mailing list
>> relax-users@xxxxxxx
>>
>> 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
>
> --
> Martin Ballaschk
> AG Schmieder
> Leibniz-Institut für Molekulare Pharmakologie
> Robert-Rössle-Str. 10
> 13125 Berlin
> ballaschk@xxxxxxxxxxxxx
> Tel.: +49-30-94793-234/315
> Büro: A 1.26
> Labor: C 1.10
>