Hi,
This information is a relief! In relax I have implemented models with
just a single exchangless R2 value per spin per field strength. But
relax can optimise, through the same code paths, more complex models
with R2A and R2B as separate parameters (which I have labelled as
'full' models). In the relaxation dispersion auto-analysis, the
parameters of the simpler model are used as the starting point for the
more advanced model. However now, after your post, I have disabled
the 'full' models by default. These can be enabled by a user for
experimentation or if their system significantly changes size during
the exchange process. I am now blasting tonnes of data (again that
from Flemming Hansen from the test_suite/shared_data/dispersion/Hansen
relax directory) to make sure all these models are robust. Once I am
happy that everything is running well, I will write another email
summarising the numerical models now present within relax.
Cheers,
Edward
On 17 July 2013 16:06, <schanda@xxxxxx> wrote:
Hi Edward,
I agree.
However, in practice the assumption that R2A=R2B is one that I use always.
Fitting two R2 for the two states is close to impossible, unless the data
are really outstanding.
Having the two different R2 in the matrices is nice from an intellectual and
eductational point of view, but otherwise not of practical relevance.
Likewise, all the things with R1 are practically not useful, as we anywas
assume that longitudinal relaxation does not have any impact on the
dispersions. Again, fitting the R1's is very difficult, and in practice not
relevant.
So, as you suggested in a previous message, you can ignore non-180deg
pulses, ignore R2 differences between the states, and ignore R1.
paul
Quoting Edward d'Auvergne <edward@xxxxxxxxxxxxx>:
Hi,
There does appear to be one difference. It looks like this model uses
the assumption that R20A = R20B = R20 (or in the original notation,
R2G = R2E = R2). The R1G, R1E, and R2E values are not used in the
code. So I assume that it should give the same result as the reduced
models as I have now created in relax from the fitting_main_kex.py
code rather than the full R20A != R20B models. So if the spin-spin
relaxation rates for the two states are significantly different, then
this model should not be able to reproduce the results of the other
models in fitting_main_kex.py. Though maybe I have missed something?
Cheers,
Edward
On 17 July 2013 15:35, Paul Schanda <paul.schanda@xxxxxx> wrote:
Hi Edward,
No, there is nothing published. Nikolai said that anyways it's trivial
:-)
He derived it basically by putting the Bloch-McConnell equations to Maple
and simplifying it there. I guess there is something like a FullSimplify
(that's Mathematica-style, but I guess Maple does it similarly).
I guess the main advantage is speed, and in fact in all cases I have seen
it
does exactly the same as the explicit Bloch-McConnell treatment, so I see
it
as the treatment that one would like to use.
I agree that having something undocumented there is not particularly
nice,
but in the end it's up to the user whether or not to use this faster
treatment (and compare it to the slower Bloch-McConnell one if needed).
paul
On 17.07.13 15:14, Edward d'Auvergne wrote:
Hello,
I am now adding the numerical dispersion model derived by Nikolai
Skrynnikov to relax. This is model 5 from the fitting_main_kex.py
script (for reference attached to https://gna.org/task/?7712). I was
wondering if there was a published reference for this that you know
of? This would be useful for the model name and for the
documentation.
Cheers,
Edward
--
Paul Schanda, Ph.D.
Biomolecular NMR group
Institut de Biologie Structurale Jean-Pierre Ebel (IBS)
41, rue Jules Horowitz
F-38027 Grenoble
France
+33 438 78 95 55
paul.schanda@xxxxxx
http://www.ibs.fr/groups/biomolecular-nmr-spectroscopy?lang=en
Re: Reference for the Skrynnikov derivation?