Re: I did not used relax to get r1,r2,noe -- November 26, 2012 - 15:59

Thank you for your explanation.

I use my results and format them either as .out or .txt and using GUI i import them into relax and use free format to tell relax where data and error columns are. in dauvergene protocol, i also change things like pdb, frequency e.g. in your original script, you have 500 and 600MHz but i have 500 and 700MHz so i simply use my exact SF01 values and either .out or .txt data files.

If i run model free with my  data but using dauvergne protocol without these modifications, relax dont run.

i can run either on command line or GUI without problems but not sure of final output. In modelfree4 mfout is easy to get and extract necessary results but i have to learn this good new software "relax" also.

PS, how can i convert my data with relax to .out files? i simply use my editor to convert the extension from txt to out as simple as that since Mac has this capability.

Kind regards

Kozo

---

From: Edward d'Auvergne <edward@xxxxxxxxxxxxx>
To: Kozo Keisho <crystalrlx@xxxxxxxxxxx>
Cc: "relax-users@xxxxxxx" <relax-users@xxxxxxx>
Sent: Monday, 26 November 2012, 23:03
Subject: Re: I did not used relax to get r1,r2,noe

Hi,

I'll answer below:


> Values are S2 values just for a sample.

Ok, so this must be the s2.txt file relax produces at the end of the
auto-analysis.  I'm guessing you opened this file, and not the s2f.txt
and s2s.txt file in MS Windows using the results viewer window of the
relax GUI.  Are these assumptions correct?


> I did not use relax to get r1,r2 and noe but used NMRDraw text output, so
> all data is in text.

Is relax reading these rates correctly, i.e. do you see any warnings
when loading the relaxation data?  Do the errors you get from NMRDraw
match the errors of other programs (using relax for example)?


> I copied dauvergne protocol and modified it with my own data set and then on
> "gui", just followed each step, no problems with "gui" following.

Please keep these separate.  A bug or problem in the GUI is not a bug
or problem in relax's scripting mode.  If you have a problem with the
script, please say that you are running the script and the problem is
with the script.  If you have a problem with the GUI, then state that
you are using the GUI and not the scripting mode.  The switching
between both or not saying which you are using is confusing me.


> Rex values seem to correlate a bit with Palmers modelfree4 i executed
> perfectly but relax is more robust and can perform complete anisotropic
> which modelfree from Palmer cannot.

The Rex correlation depends on the protocol used not the program.
Always remember that relax can produce identical results to
Modelfree4, and that if you modify the source code and recompile
Modelfree4 it can produce results similar to relax (not identical
because there is no choice of high quality optimisation algorithms in
Modelfree4).  The major difference is that relax supports the
ellipsoidal diffusion tensor whereas Modelfree4 does not.


> so i used relax to gain more
> understanding.

For an even deeper understanding, you should look at Sebastien Morin's
consistency testing code in relax
(http://www.nmr-relax.com/manual/Consistency_testing.html) as well as
the reduced spectral density mapping
(http://www.nmr-relax.com/manual/Reduced_spectral_density_mapping.html).
I would also recommend looking at all the references in the
'Citations' chapter of the relax manual
(http://www.nmr-relax.com/manual/Citations.html, though the links are
currently broken so the PDF manual
http://download.gna.org/relax/manual/relax.pdf would be better).


> Relax result of Rex values also correlates with cpmg results
> but i want to gain more information how to extract data for presentation
> like i want to know distribution of residues per model.

This is a good sign that the model-free Rex values are not due to a
problem with the diffusion tensor!  If the diffusion tensor is
incorrect, you can see either artificial Rex values or artificial
nanosecond motions (see http://dx.doi.org/10.1039/b702202f for a
detailed description of this problem).

Note that because you have multiple nested and non-nested, yet
related, models in a model-free analysis, comparing which model has
been selected is physically meaningless.  It has more to do with your
experimental errors and the quality of your data!  The model selected
is only interesting from a statistical point of view and not from a
dynamics point of view.  To understand the motions of your system, it
is much better to look at the dynamics itself independent of the
models selected.


> my structure looks
> "prolate" but how to tell with relax? i saw it chooses "prolate" for which
> am happy.

When using the auto-analysis which implements the model-free protocol
that I derived (see my 2007 and 2008 papers,
http://www.nmr-relax.com/refs.html), then you will see the file:

final/tensor.pdb

If you are using the GUI, you can see it in the results viewer window.
Just open this file from the results viewer and it will open up in
PyMOL with your molecule.  If you are using the scripting mode, then
in the prompt or script after the analysis, you can see the same by
typing:

pymol.view()
pymol.cartoon()
pymol.tensor_pdb(file="tensor.pdb", dir="final")

I hope some of this helps.

Regards,

Edward